Article

Automatic detection of ships in RADARSAT-1 SAR imagery

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

NOAA/NESDIS a initié le programme “Alaska SAR Demonstration” dont l'objectif est de faire la démonstration du potentiel des images RSO en bande C de RADARSAT-1 à fournir une information utile et en temps opportun sur l'environnement et pour la gestion des ressources pour des utilisateurs en Alaska. Un des produits développés dans le cadre du programme est une liste de localisations des navires. Cet article décrit l'algorithme développé pour générer ce produit par le biais de la détection automatique des navires basée sur des changements dans les statistiques locales. À l'aide d'images à basse résolution (100 mètres d'espacement), on démontre que l'on peut détecter des navires de dimension supérieure à 35 mètres (représentant 105 navires sur un total de 272 dans la zone test) avec un taux de fausse alerte de 0,01% pour une seule détection. Avec des images à haute résolution (50 mètres d'espacement), on peut détecter des navires d'une dimension supérieure à 32 mètres (représentant 124 navires sur 272) avec un taux de fausse alerte de 0,002% pour une seule détection. L'algorithme est entièrement automatisé et prend environ 10 minutes de temps-machine pour traiter une image ScanSAR en mode B large.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... For vessels detection in SAR images, lots of automatic methods have been well developed [29], [30]. One of the well-known methods is the constant false alarm rate (CFAR) [30]. ...
... For vessels detection in SAR images, lots of automatic methods have been well developed [29], [30]. One of the well-known methods is the constant false alarm rate (CFAR) [30]. On the one hand, more complex CFAR methods take into account different statistic sea clutters to distinguish vessel from the sea clutter [31], e.g., Gaussian distribution [2], [30], alpha-stable distribution [63], [64], generalized-K distribution [39], [65], and generalized Gamma distribution [66], [67]. ...
... One of the well-known methods is the constant false alarm rate (CFAR) [30]. On the one hand, more complex CFAR methods take into account different statistic sea clutters to distinguish vessel from the sea clutter [31], e.g., Gaussian distribution [2], [30], alpha-stable distribution [63], [64], generalized-K distribution [39], [65], and generalized Gamma distribution [66], [67]. On the other hand, many efforts have been devoted to improve the conventional CFAR method. ...
Article
In this paper, we demonstrate that the spaceborne dual-platform TerraSAR-X (TSX) and TanDEM-X (TDX) pursuit monostatic mode full polarimetric (full-pol) synthetic aperture radar (SAR) data with a time lag can be used to monitor maritime traffic. For single polarization (single-pol) SAR data, the performance of vessel velocity estimation is mainly determined by 2-D cross correlation of SAR intensity data. As the sea clutter is changing dynamically during the TSX/TDX data acquisition, the correlation between two dual-platform images decreases significantly. We may get unstable or incorrect estimations of vessel velocity, especially under a higher wind condition. For solving this problem, we propose an object-oriented polarimetric likelihood ratio test (PolLRT) method based on the complex Wishart distribution. The proposed method makes PolLRT statistics of the detected target pixels for eliminating the effect of varied sea clutter. Two pairs of full-pol SAR data sets covering the Strait of Gibraltar acquired by dual-platform TSX/TDX in pursuit monostatic mode with a time lag of approximately 10 s are selected for the experiments. The experimental results demonstrate that the proposed PolLRT method has a better performance than that of the classical normalized cross correlation (NCC) method with VV polarization SAR data and the mutual information (MI) method with full-pol SAR data. Specifically, under the lower wind condition, the correct estimation rate of the NCC, the MI, and the proposed PolLRT methods are 85.7%, 57.1%, and 100%, respectively; under the relatively higher wind condition, the correct estimation rate of the above three methods are 48.8%, 23.2%, and 90.1%, respectively.
... Ship detection in high-resolution SAR images has attracted much attention due to its broad application prospects. Many traditional methods [8][9][10] have been proposed to detect multi-scale ships in complex environments. For example, [9,10] separated the land from the sea and then detected the object, and identified the object based on artificial features. ...
... For example, [9,10] separated the land from the sea and then detected the object, and identified the object based on artificial features. Wackerman et al. [8] proposed a twoparameter constant false-alarm rate (CFAR) algorithm, which can adaptively adjust the threshold and use the estimated statistical distribution to distinguish objects from the background through the computed threshold. The above methods are highly dependent on human participation, and their generalization ability and detection accuracy are not high. ...
Article
Full-text available
Ship detection in Synthetic Aperture Radar (SAR) is a challenging task due to the random orientation of the ship and discrete appearance caused by radar signal. In this paper, We introduce a novel unsupervised domain adaptation framework for ship detection in SAR images by employing context-preserving region-based contrastive learning. We enhance the ship detection in SAR by learning knowledge from both labeled remote sensing optical image domain and unlabeled SAR image domain. Additionally, we propose a pseudo feature generation network to generate pseudo domain samples for augmenting pseudo-features. Specifically, we refine the pseudo-features by calculating a region-based contrastive loss on the features extracted from the object region and the background region to capture the contextual information for SAR ship detection. Extensive experiments and visualizations show that our method can outperform the state-of-the-art and have good generalization performance.
... Active sensors such as SAR [2,6,[25][26][27][28] -based imaging are more effective due to their all-time imaging and cloud penetration capability. However, these systems suffer from lower swath and infrequent revisit thereby resulting in the need for more satellites to cover the required region of interest. ...
... This phase consists of two independent sub tasks. The first sub-task is identification of ship-like objects, for which a variant of CFAR [25,26,28,35] is applied. The second sub-task is a tradeoff between two strategies viz., min-max threshold and Normalized Difference Water Index (NDWI) [36] for land-water separation. ...
Article
Full-text available
In recent years there has been an increased interest in ocean surveillance. The activity includes control and monitoring of illegal fisheries, manmade ocean pollution and illegal sea traffic surveillance, etc. The key problem is how to identify ships and ship-like objects accurately and in a timely manner. In this context, currently, many solutions have been proposed based on high resolution optical and radar remote sensing systems. Most often, these systems suffer from two major limitations viz., limited swath, thereby requiring multiple satellites to cover the region of interest and huge volumes of data being transmitted to ground, even though effective per-pixel information content is minimal. Another limitation is that the existing systems are either simulated on ground or built using the non-space qualified/Commercial Of-The-Shelf (COTS) components. This paper proposes an efficient on-board ship detection system/package connected with medium resolution wide swath optical camera. The methodology adopted has three major components, viz., onboard data processing for improving the radiometric fidelity, followed by a ship detection using modified Constant False Alarm Rate algorithm (CFAR) and a false alarm suppression module to mask false identifications. Finally, the package outputs only the locations of the ships, which is transmitted to the ground. The proposed system reduces the effective volume of data to be transmitted and processed on ground and also significantly cuts down the turnaround time for achieving the end objective. The system is built on radiation hardened Field Programmable Gate Array (FPGA) devices to meet the various engineering constraints such as real-time performance, limited onboard power, radiation hardness, handling of multiple custom interfaces etc. The system is tested with one of the medium resolution Multispectral Visual and Near Infra-Red (MX-VNIR) sensor having a spatial resolution of around 50 m and swath of around 500 Kms, which would be flown with one of the upcoming satellites. The systems performance is also verified on ground with Indian Remote Sensing (IRS) Satellite’s Resourcesat’s Advanced Wide Field Sensor (AWiFS) data and the results are found to be quite encouraging as well as meeting the mission objectives.
... From Formulas (9) and (10), DS-CNN essentially converts continuous multiplication into continuous addition, so the redundancy of the network gets reduced. As a result, the computational efficiency of the network has been greatly improved. ...
... Scale Anchor boxes (width, height) Detection network-1 L/32 (9,12), (12,25), (17,12) Detection network-2 L/16 (21,45), (27,17), ( (17) where TP is the number of the True Positive and FP is the number of the False Positive. In addition, TP can be understood that real ships are correctly detected, and FP can be understood as missed detection. ...
Article
Full-text available
As an active microwave imaging sensor for the high-resolution earth observation, synthetic aperture radar (SAR) has been extensively applied in military, agriculture, geology, ecology, oceanography, etc., due to its prominent advantages of all-weather and all-time working capacity. Especially, in the marine field, SAR can provide numerous high-quality services for fishery management, traffic control, sea-ice monitoring, marine environmental protection, etc. Among them, ship detection in SAR images has attracted more and more attention on account of the urgent requirements of maritime rescue and military strategy formulation. Nowadays, most researches are focusing on improving the ship detection accuracy, while the detection speed is frequently neglected, regardless of traditional feature extraction methods or modern deep learning (DL) methods. However, the high-speed SAR ship detection is of great practical value, because it can provide real-time maritime disaster rescue and emergency military planning. Therefore, in order to address this problem, we proposed a novel high-speed SAR ship detection approach by mainly using depthwise separable convolution neural network (DS-CNN). In this approach, we integrated multi-scale detection mechanism, concatenation mechanism and anchor box mechanism to establish a brand-new light-weight network architecture for the high-speed SAR ship detection. We used DS-CNN, which consists of a depthwise convolution (D-Conv2D) and a pointwise convolution (P-Conv2D), to substitute for the conventional convolution neural network (C-CNN). In this way, the number of network parameters gets obviously decreased, and the ship detection speed gets dramatically improved. We experimented on an open SAR ship detection dataset (SSDD) to validate the correctness and feasibility of the proposed method. To verify the strong migration capacity of our method, we also carried out actual ship detection on a wide-region large-size Sentinel-1 SAR image. Ultimately, under the same hardware platform with NVIDIA RTX2080Ti GPU, the experimental results indicated that the ship detection speed of our proposed method is faster than other methods, meanwhile the detection accuracy is only lightly sacrificed compared with the state-of-art object detectors. Our method has great application value in real-time maritime disaster rescue and emergency military planning.
... Object detection in ocean remotesensing imagery usually refers to detecting objects (ships, oil rigs, etc.) that are distinguished from the surrounding image backgrounds. A constant false-alarm rate (CFAR) is the most common statistical approach for ship detection in ocean remote-sensing images [10]. The methods work but may not be optimal for a specific end-to-end (data-to-information) problem, since the traditional supervised classification and object-detection approaches do not consider spatial structure features Downloaded from https://academic.oup.com/nsr/article/7/10/1584/5809984 by guest on 26 December 2020 REVIEW or use the features extracted by human-designed operators. ...
... A typical conventional method is threshold-based methods that focus on finding bright pixels operating accurate clutter statistical modeling. Algorithms built on the theory of CFAR filtering [10] and generalized likelihood ratio testing (GLRT) [113] are representations. The main drawback of the conventional methods is that they need prior professional REVIEW Li et al. 1599 knowledge to manually design features, which has been a common challenge faced by most fields in the era of big data [11]. ...
Article
Full-text available
With the continuous development of space and sensor technologies during the recent 40 years, ocean remote sensing has entered into the Big Data era with typical Five-V (volume, variety, value, velocity, and veracity) characteristics. Ocean remote sensing data archives reach several tens of petabytes, and massive satellite data are acquired worldwide daily. To precisely, efficiently and intelligently mining the useful information submerged in such ocean remote sensing data sets is a big challenge. Deep learning, a powerful technology recently emerging in the machine-learning field, has demonstrated its more significant superiority over traditional physical- or statistical-based algorithms for image information extraction in many industrial-field applications and starts to draw interest in ocean remote sensing applications. In this review paper, we first systematically reviewed two deep learning frameworks that carry out ocean remote sensing image classifications and then presented eight typical applications in ocean internal wave/eddy/oil spill/coastal inundation/sea-ice/green algae/ship/coral reef mapping from different types of ocean remote sensing imagery to show how effective of these deep learning frameworks. Researchers can also readily modify these existing frameworks for information mining of other kinds of remote sensing imagery.
... In this field, an important branch is ship detection in SAR images. Although numerous methods have been proposed [10,11], it's still an enduring hot topic because of several tough problems for detecting multi-scale ships in complex surroundings. ...
... t refer to classification and regression losses, respectively. When training the classification network, Cross Entropy loss (CE) is utilized, which is given by equation(10). p and y refer to the probabilities of predicted ...
Article
Full-text available
With the development of Deep Learning (DL) and Synthetic Aperture Radar (SAR) imaging techniques, Synthetic Aperture Radar Automatic Target Recognition (SAR ATR) has come to a breakthrough. Numerous algorithms have been proposed and competitive results have been achieved in detecting different targets. However, due to the influence of various sizes and complex background of ships, detecting multi-scale ships in SAR images is still challenging. To solve the problems, a novel network, called Attention Receptive Pyramid Network (ARPN), is proposed in this paper. ARPN is a two-stage detector and designed to improve the performance of detecting multi-scale ships in SAR images by enhancing the relationships among non-local features and refining information at different feature maps. Specifically, Receptive Fields Block (RFB) and Convolutional Block Attention Module (CBAM) are employed and combined reasonably in Attention Receptive Block (ARB) to build a top-down fine-grained feature pyramid. RFB, composed of several branches of convolutional layers with specifically asymmetric kernel sizes and various dilation rates, is used for grabbing features of ships with large aspect ratios and enhancing local features with their global dependences. CBAM, which consists of channel and spatial attention mechanisms, is utilized to boost significant information and suppress interference caused by surroundings. To evaluate the effectiveness of ARPN, experiments are conducted on SAR Ship Detection Dataset (SSDD) and two large-scene SAR images. The detection results illustrate that competitive performance has been achieved by our method in comparison with several CNN-based algorithms, e.g., Faster-RCNN, RetinaNet, FPN, YOLOv3, DAPN, DS-CNN, HR-SDNet and SER Faster-RCNN.
... Ship detection using SAR image is a complex problem depending upon not only the nature of ship but also the sea state. For single-pol amplitude data analysis, widely used SAR ship detection algorithms use the constant false alarm rate (CFAR) detector [3], [4] based on the statistical model of sea clutter. Wang et al. [5] adopted the Alpha-stable distribution for ship detection. ...
... Equations (3) and (4) show that the phase parts of C 12 , C 13 , and C 23 (i.e., φ C 12 , φ C 13 , and φ C 23 ) are corresponding to three phase differences (i.e., φ HHHV , φ HHVV , and φ HVVV ). This means that the phase information of the elements in [C] can be extracted by (4). Considering this point, in this paper, we utilize (4) to construct the new scheme for calculating the phase information of scattering differences between SP and ISP s . ...
Article
In this paper, we proposed a complete polarimetric covariance difference matrix [CP]-based algorithm for ship detection in polarimetric synthetic aperture radar (PolSAR) imagery. To calculate [CP], we first developed a scheme to reflect the polarimetric scattering differences between ship pixel (SP) and its neighboring pixels (ISPs) and, then, dividedly accumulated the amplitude and phase differences between SP and ISPs. Compared to the polarimetric covariance difference matrix [P] developed in our earlier work, [CP] effectively overcomes the drawback of the lack of the phase information in [P]. To demonstrate the effectiveness of the proposed algorithm, we applied the [CP]-based ship detection algorithm to four PolSAR data sets, including one UAVSAR L-band data set with 21 ships, two AIRSAR L-band data sets with 11 and 22 ships, respectively, and one Radarsat-2 C-band data set with 8 ships. Experimental results show that: 1) the proposed algorithm can effectively detect ships with high target-to-clutter ratio (TCR) values and 2) [CP] has a better performance than the traditional polarimetric covariance matrix [C] and [P] on ship detection. To be more specific, the average TCR value of the proposed algorithm (23.86 dB) is 6.07 and 7.47 dB higher than PNFC (i.e., the geometrical perturbation-polarimetric notch filter) and RSC (i.e., the reflection symmetry method), respectively.
... In fact, many statistical models have been developed to describe SAR image data in the constant false alarm rate (CFAR)-based algorithms [40]. The analysis of a large number of measured data shows that Gamma distribution can be well applied to sea clutter modeling [41][42][43]. ...
... From Figure 8, one can conclude that sea clutter meets Gamma distribution. analysis of a large number of measured data shows that Gamma distribution can be well applied to sea clutter modeling [41][42][43]. Thus, it is necessary to integrate the traditional mature methods with rich expert experience into the preprocessing of a DL detector, otherwise on-board SAR ship detection will be time-consuming and labor intensive. ...
Article
Full-text available
Synthetic aperture radar (SAR) satellites can provide microwave remote sensing images without weather and light constraints, so they are widely applied in the maritime monitoring field. Current SAR ship detection methods based on deep learning (DL) are difficult to deploy on satellites, because these methods usually have complex models and huge calculations. To solve this problem, based on the You Only Look Once version 5 (YOLOv5) algorithm, we propose a lightweight on-board SAR ship detector called Lite-YOLOv5, which (1) reduces the model volume; (2) decreases the floating-point operations (FLOPs); and (3) realizes the on-board ship detection without sacrificing accuracy. First, in order to obtain a lightweight network, we design a lightweight cross stage partial (L-CSP) module to reduce the amount of calculation and we apply network pruning for a more compact detector. Then, in order to ensure the excellent detection performance, we integrate a histogram-based pure backgrounds classification (HPBC) module, a shape distance clustering (SDC) module, a channel and spatial attention (CSA) module, and a hybrid spatial pyramid pooling (H-SPP) module to improve detection performance. To evaluate the on-board SAR ship detection ability of Lite-YOLOv5, we also transplant it to the embedded platform NVIDIA Jetson TX2. Experimental results on the Large-Scale SAR Ship Detection Dataset-v1.0 (LS-SSDD-v1.0) show that Lite-YOLOv5 can realize lightweight architecture with a 2.38 M model volume (14.18% of model size of YOLOv5), on-board ship detection with a low computation cost (26.59% of FLOPs of YOLOv5), and superior detection accuracy (1.51% F1 improvement compared with YOLOv5).
... Daytime optical sensors allow the detection of ships; however, their sensors are usually not sensitive enough for detecting low light levels as emitted at night-time [15]. While SAR images have all-weather and day and night capabilities, this approach for detecting ships at sea requires the processing of large amounts of data, and at the moment, there is no operational product offering vessel detection from SAR data [16][17][18]. Thus, there are still many gaps in the monitoring of ships at daily, monthly and annual time scales. ...
... On the South American continent, on average, there was a downward trend in NTL with (Rs = −0. 18 ...
Article
Full-text available
Accurate information on port shipping activities is critical for monitoring global and local traffic flows and assessing the state of development of the maritime industry. Such information is necessary for managers and analysts to make strategic decisions and monitor the maritime industry in achieving management goals. In this study, we used monthly night light (NTL) images of the Suomi National Polar-Orbiting Partnership (Suomi NPP) Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band, between 2012 and 2020, to study the night lights emitted by ships in ports' anchorage areas, as an indicator for shipping activity in anchorage areas and ports. Using a dataset covering 601 anchorage areas from 97 countries, we found a strong correspondence between NTL data and shipping metrics at the country level (n = 97), such as container port throughput (Rs = 0.84, p < 0.01) and maximum cargo carried by ships (Rs = 0.66, p < 0.01), as well as a strong correlation between the number of anchorage points and the NTL values in anchorage areas across the world (Rs = 0.69, p < 0.01; n = 601). The high correspondence levels of the VIIRS NTL data with various shipping indicators show the potential of using NTL data to analyze the spatio-temporal dynamic changes of the shipping activity in anchorage areas, providing convenient open access and a normalized assessment method for shipping industry parameters that are often lacking.
... SAR systems on larger satellites have been widely used for maritime surveillance; for example, both ERS-1 and Seasat [11]. Images from the Canadian Radarsat-1 have also been subject to various detection studies [12]- [14]. Similarly, its successor Radarsat-2 is used for ship monitoring as part of the Polar Epsilon project [15]. ...
Article
Full-text available
The performance trade-offs of a SmallSat Synthetic Aperture Radar (SAR) system for maritime surveillance in the coastal waters of New Zealand are investigated. The lower costs of SmallSat platforms allow for a constellation of SAR satellites that can be launched from New Zealand using an existing local launch service provider. The minimum SAR image quality necessary for a SmallSat system to achieve a desired detection performance is determined using existing X-band satellite data. The image quality is specified in terms of Noise-Equivalent Sigma Zero (NESZ) and resolution. It was found that for a resolution cell of 4m2 a system NESZ of 1.7dB is sufficient to detect small fishing vessels with a probability of detection of 0.5, while maintaining the Probability of False Alarm below 1010. These requirements are translated into a preliminary SAR system design.
... After detecting ship targets, the next step is to remove the false alarm sources. The ambiguities may arise due to many reasons stated; the one to consider above all is that the objects on land are easily removed by the land mask, apart from this there is still a high probability of false alarms due to other factors (Wackerman, Friedman, Pichel & Clemente-Colón 2001). Some prominent false alarm sources include small islands, off-shore construction sites, unrecognized azimuth ambiguities and range ambiguities from strong scatters on land, such as cities, certain mountain slopes that are not included in the land mask. ...
Article
Full-text available
The Earth’s surface is covered with 72% water. This fact alone emphasizes the importance of proper monitoring and regulation of maritime activities. This monitoring can be useful in an array of applications including illegal transitions, rescue operations, territory regulation among many other applications. In order to achieve the task of “Maritime Surveillance” or simply the marine object detection, we need a structured approach combined with a set of algorithms. The objective of this paper is to study an emerging open source tool- Search for Unidentified Maritime Objects (SUMO) developed for the detection of ships which work regardless of weather conditions and coverage limits. Based on the Synthetic Aperture Radar (SAR) data, this paper aims to process the satellite-borne data provided by the Sentinel-1 satellite. Proposed by the Joint Research Centre, SUMO is a pixel-based algorithm which follows a structured approach in order to identify marine objects and remove false alarms. It is observed that many of the false alarms are caused due to the presence of land. These are reduced by using the buffered coastlines referred to as land masks. A local threshold is calculated using the background clutter for the generation of false alarm rate and the pixels above this threshold are identified and clustered to form targets. A reliability value is computed for the elimination of azimuth ambiguities. Also, various attributes of the detected targets are calculated in order to give an accurate description of ships and its characteristics. With the SAR data being freely available due to the open data policy of the EU’s Copernicus program, it has never been more viable to employ new methods for marine object detection and this paper explores this possibility by analyzing the results obtained. Specifically, the employed data consists of Sentinel-1 fine dual-pol acquisitions over the coastal regions of India.
... However, the linear dual-polarization SAR mode having lower system complexity supports a wider swath width [11] while the compact polarimetric SAR (CP SAR) mode provides a compromise between swath width and scattering information [12]. In [13], the authors have developed a novel algorithm for ships detection using low resolution SAR imagery for ships greater than 35 m length and using high resolution SAR imagery for ships greater than 32 m length. In [14], a segmentation method from CP SAR images is proposed for detection of ships in which pixel-wise detection is based on a fully convolutional network, U-Net. ...
Article
Full-text available
In recent years, computer vision finds wide applications in maritime surveillance with its sophisticated algorithms and advanced architecture. Automatic ship detection with computer vision techniques provide an efficient means to monitor as well as track ships in water bodies. Waterways being an important medium of transport require continuous monitoring for protection of national security. The remote sensing satellite images of ships in harbours and water bodies are the image data that aid the neural network models to localize ships and to facilitate early identification of possible threats at sea. This paper proposes a deep learning based model capable enough to classify between ships and noships as well as to localize ships in the original images using bounding box technique. Furthermore, classified ships are again segmented with deep learning based auto-encoder model. The proposed model, in terms of classification, provides successful results generating 99.5% and 99.2% validation and training accuracy respectively. The auto-encoder model also produces 85.1% and 84.2% validation and training accuracies. Moreover the IoU metric of the segmented images is found to be of 0.77 value. The experimental results reveal that the model is accurate and can be implemented for automatic ship detection in water bodies considering remote sensing satellite images as input to the computer vision system.
... Canadian commercial software OMV used CFAR algorithm for ship target detection in which the clutter is described by K distribution model [10]. Waterman et al. [11] proposed twoparameter CFAR to detect SAR images. ...
Article
Full-text available
Target detection and recognition are two important parts in image processing. As it is known to authors, target detection in synthetic aperture radar (SAR) image is usually processed on two-dimension, which is more complex than one-dimension. Here, the target detection in SAR image is achieved by two levels of one-dimensional target detection. Then, the range profiles of the suspected targets are obtained while the target region is detected. Based on the obtained range profiles, features are extracted and the classifier is used to distinguish the targets and others. Finally, a method of target recognition based on mid-level features is used here. Authors’ method is applied on SAR ship data and the result shows that authors’ proposed method works well on target detection.
... This can be explained by many factors including the size, material, and generally the presence of metallic reflectors on the body of ships (Margarit et al. 2009). This specification has led to the development of several algorithms focused on detecting bright points in a dark background (Crisp 2004;Wackerman et al. 2001;Marino, Cloude and Woodhouse 2010). Moreover, there are some other features in the maritime environment, such as braking or rogue waves, that can produce high backscattering intensity and may lead to false alarms in detectors that are based only on intensity measurements (Marino et al. 2015). ...
Article
Full-text available
Synthetic aperture radar (SAR) reveals a valuable contribution to ship detection due to its all-weather and all-illumination sensing capabilities. In this paper, a new polarimetric SAR (PolSAR) ship detector is proposed that jointly exploits sub-look scattering properties of point-like scatterers (i.e. ships) and polarimetric information. The detector, which is physically based on the concepts of coherent scatterers and polarimetric signatures, consists of four main steps: first, polarimetric signatures are extracted for each sub-look image; then, the correlation coefficient (CC) between the sub-look images is calculated for varying polarization bases; ship detection is performed by setting a threshold on the resulting CC image and finally, morphological filters are applied to delineate the contour of the detected ships and automatically generate the number of targets detected. The exploitation of polarimetric information through the polarimetric signatures, on one side provides more discriminative information about target and makes the detector more robust and accurate, on the other side, it increases the computation time. The detector's performance is verified against actual Land C-band PolSAR datasets. Experimental results demonstrate that the proposed detector outperforms state-of-the-art ones and it results in an area under the receiver operating characteristic (ROC) curve that ranges between 0.93 and 0.95.
... The SAR-based studies employ different versions of the CFAR algorithm, which can be traced back to similar domain of SAR marine application: vessel detections (Ai et al., 2010;Aiello and Gianinetto, 2017;Alpatov et al., 2017;Cui et al., 2011;El-Darymli et al., 2013;Gao and Shi, 2017;Greidanus et al., 2017;Marino et al., 2015;Santamaria et al., 2017;Stasolla et al., 2016;Tian et al., 2015;Wackerman et al., 2001;Wang et al., 2018). These vessel detection algorithms derived from CFAR are extensively used in radar signal detection (Rohling, 1983). ...
Article
Full-text available
Although Land Use and Land Cover (LULC) change is primarily focused on the types, rates, causes, and consequences of land change, increased anthropogenic development on the ocean's surface, such as offshore oil extraction, offshore wind energy, aquaculture, and coral reef conversion to military outposts, suggest that LULC change not only pertains to historically terrestrial space, but also new lands created on top of ocean surfaces. Therefore, similar human disturbance analyses are necessary for these transformed marine environments, but the lack of accurate, accessible, and up-to-date location information about these spatially dispersed changes significantly limits examination of their environmental impacts. Subsequently these dynamic changes across the oceans are poorly documented. Here, we developed a cloud-native geoprocessing algorithm to automatically detect and extract offshore oil platforms in the Gulf of Mexico using synthetic aperture radar and Google Earth Engine. Cross-validated results indicate our top model identified offshore infrastructure with a probability of detection of 98.70%, an overall accuracy of 96.09%, a commission error rate of 2.68%, and an omission error rate of 1.30%. Its generalizability was tested across wind farms in waters of China and the United Kingdom, which resulted in an overall accuracy of 97.00%, a commission error rate of 2.07%, and omission error rate of 0.97%. These generalization capabilities indicate our model can be potentially used to map global offshore infrastructure. Such increased ocean transparency could allow for improved marine environmental management by bringing objectivity, scalability, and accessibility.
... The satellite images captured in coastal areas [47] or nearby harbor region also include small fishing boats with or without nets. Ship candidate selection algorithms employed in major research papers face a challenge to define the shapes for these small targets. ...
Chapter
Detection of maritime object is of greater attention in the field of satellite image processing applications in order to ensure the security and traffic control. Even though several approaches were built in the past few years, still it requires proper revamp in the architecture to focus toward the reduction of barriers to improve the performance of ship identification or appropriate vessel detection. The inference due to cluttered scenes, clouds, and islands in between the ocean is the greater challenge during the classification of ship or vessel. In this paper, we proposed a novel ship detection method called deep neural method which works very faster and based on the concept on deep learning methodology. Experimental results provide the better accuracy, and time complexity also reduces little further when compared to the traditional method.
... In general, pixels corresponding to a ship show larger NRCS values than the ocean pixels in the SAR images ( Wackerman et al. 2001). This originates from the different scattering mechanism for the ship with double-bounce scattering rather than singlebounce scattering for the sea surface ( Yeremy et al. 2001). ...
Article
Full-text available
Environmental factors affecting the errors in wind speed estimates were investigated using the Advanced Land Observing Satellite-2 Phased Array L-band Synthetic Aperture Radar-2 (ALOS-2 PALSAR-2) data for the period from 1 November 2014 to 30 November 2017. In total, 45 ALOS-2 PALSAR-2 Stripmap Fine mode images with horizontal dual-polarization were collected, and the wind speeds were calculated using the L-band Geophysical Model Function (GMF) in 2009. Validation of the SAR-derived wind speeds to in-situ wind, converted to the 10-m neutral wind, resulted in the root-mean-square error of 2.11 ms⁻¹, bias error of −1.16 m s⁻¹, and standard deviation of 1.78 m s⁻¹. Investigation of the wind speed errors revealed the contributions of diverse oceanic environmental factors such as the ship and its side-lobe effect that is negligible at a distance of more than 200% of the ship radius, v-shaped ship wakes are accompanied by the Kelvin waves and turbulent wakes caused by ship movement. The sand ridge, formed due to the interaction of shallow bathymetry and tidal currents, contributed to the wind speed errors of more than 50%. The internal waves caused prominent wind speed errors of more than 100%. The atmospheric gravity waves, generated by the orographic effect, caused wind speed fluctuations up to 13.4 m s⁻¹. This study addressed the importance of understanding the factors that affect the normalized radar cross-section of the SAR image in order to correctly retrieve wind speeds from SAR data.
... Constant false alarm rate (CFAR) [14,15] is the most in-depth and widely-used algorithm among traditional target detection algorithms. Its key concept is modelling the distribution of sea clutter with regional sea clutter statistics. ...
Article
Full-text available
Remote sensing (RS) monitoring of ships has important significance in both military and civilian fields. The RS ship detection aims to locate the position of the ship in the remote sensing image and extract its characteristics. Traditional ship target detection algorithms cannot meet the demands for speed and precision of SAR remote sensing and optical remote sensing data. With the development of artificial intelligence technology, the target detection technology such as deep learning algorithms has made significant progress in RS ship detection. Deep learning has become a heated topic in research. This paper has analyzed and summarized previous researches on the application of deep learning algorithms in ship detection technologies based on SAR and optical remote sensing images in recent years and has provided suggestions for future studies. In the future, deep learning-based technologies for RS ship detection will use more data, such as data from multiple sensors in multiple channels. Deep neural networks will also include more rules and specialized knowledge. Its structure will become more complicated and eventually develop into a neural network like the human brain.
... Traditional ship detection methods mainly rely on statistical analysis of image pixels, and most of them are threshold-based methods [4]- [6]. The threshold-based methods determine the threshold that distinguishes ship targets from the background by modelling sea clutter based on the theory of constant false alarm rate (CFAR) filtering [7], [8], which have become the classic methods for SAR image target detection, and have been widely used in practical ship target detection systems [9]. ...
Article
Full-text available
Thanks to the excellent feature representation capabilities of neural networks, deep learning-based methods perform far better than traditional methods on target detection tasks such as ship detection. Although various network models have been proposed for SAR ship detection such as DRBox-v1, DRBox-v2 and MSR2N, there are still some problems such as mismatch of feature scale, contradictions between different learning tasks and unbalanced distribution of positive samples, which have not been mentioned in these studies. In this paper, an improved one-stage object detection framework based on RetinaNet and rotatable bounding box (RBox), which is referred as R-RetinaNet, is proposed to solve the above problems. The main improvements of R-RetinaNet as well as the contributions of this paper are threefold. First, a scale calibration method is proposed to align the scale distribution of the output backbone feature map with the scale distribution of the targets. Second, a feature fusion network based on task-wise attention feature pyramid network (TA-FPN) is designed to decouple the feature optimization process of different tasks, which alleviates the conflict between different learning goals. Finally, an adaptive Intersection over Union (IoU) threshold training method is proposed for RBox-based model to correct the unbalanced distribution of positive samples caused by the fixed IoU threshold on RBox. Experimental results show that our method obtains 13.26%, 9.49%, 8.92% and 4.55% gains in average precision (AP) under an IoU threshold of 0.5 on the public SAR ship detection data set (SSDD) compared with four state-of-the-art RBox-based methods, respectively.
... Land masking is a pre-required stage for most traditional ship detection system. Registering the SAR images with the existing geographic maps is a common means of landing masking [26], yet the easiest methodologies have many shortages. For instance, tidal ranges are self-evident and some minuscule islands and rocks are easily been overlooked, such is-sues may happen with registration errors. ...
Article
Full-text available
Recently, synthetic aperture radar (SAR) ship detection is used in many applications within the marine field, such as fishery management, traffic control, and urgent rescue operations. Meanwhile, deep learning-based methods have bought new capabilities for ship detection in SAR images on account of high accuracy and robustness. However, several challenges remain to be addressed: 1) the shapes of the ships in SAR images have a relatively extreme aspect ratio comparing to the target objects in the optical images, and 2) complex background and clutter noise result in adverse effects for the network to extract prototypical SAR target features, which limit the ship detection performance. To address these issues, this paper proposes two effective approaches to augment the feature extraction ability of the network. Firstly, IOU (Intersection over Union) K-means is carried out to settle the extreme aspect ratio problem. The IOU K-means, as a preprocessing step, clusters a set of aspect ratios from datasets that are suitable for ship detection. Secondly, we embed a soft thresholding attention module (STA) in the network to suppress the impact of noise and complex background. The comparison results with several state-of-the-art object detection algorithms confirm the efficiency and feasibility of proposed approaches.
... In this paper, we are particularly interested in the contribution of polarimetric SAR (Pol-SAR) algorithms on ship detection. This is a line of research that started in 2000 by many scientists such as Wackerman et al. [1], Sciottien et al. [2], Touzi et al. [3] and continues to this day to be developed by other scientists, including Nunziata and Migliaccio [4], Wang et al. [5], and Marino et al. [6]. The main objective of these works is the improvement of ship detection while reducing the false alarm probability. ...
Article
Full-text available
Remote sensing of vessels is an important tool for ship safety and security at sea. In this work, we are interested in improving ships detection using polarimetric Synthetic Aperture Radar (SAR). To develop the appropriate method, different processing techniques are applied on Pol-SAR images such as fusion and polarimetric decompositions and we use adaptive threshold detectors to assess the performances of the processing techniques. The data exploited in this work were acquired on a port area of the city of Vancouver by using RADARSAT-2 satellite. In this paper it is shown first that when exploiting single polarization, the HH channel provides the highest score of detection probability (PD) of 87.2% for a false alarm probability (PFA) of 0.05%, and this while using the cell averaging constant false alarm rate (CA-CFAR) detector. The result is obtained comparatively with other polarizations (HV, VV) and detection algorithms. Second, the fusion of polarimetric channels achieves its best performances with the CA-CFAR detector, comparatively with the two parameters (2P)-CFAR and generalized likelihood ratio test (GLRT). Third, we find that among the conventional polarimetric techniques, the singular value decomposition (SVD) combined with CA-CFAR detector gives the best results and achieves a detection probability of 91% for a false alarm of 0.05%. This result was obtained by comparing the performances of other combinations of decompositions (Pauli, Freeman, Yamaguchi), fusion and ships detection algorithms. In this paper, we obtain with the proposed approach an increase of 3.8% in detection probability for false alarm probability of 0.05%.
... Polarimetric SAR data has applications in many areas and several studies were carried out earlier on the use of Polarimetric data for land cover identification [2,3], urban area identification [4] etc. Many studies have been done to develop algorithms to detect ships in single channel SAR images automatically, where the amplitude information is used to detect ships [5,6]. However, it was observed that amplitude information is not enough to eliminate false alarms caused by speckle and other ambiguities and insufficient to characterize and classify a ship. ...
... Stasolla and Greidanus [9] employed the constant false alarm rate (CFAR) to do the first step. The CFAR family is [10]- [12] widely used in ship detection on SAR images to separate ship signature and background. Furthermore, for the second step, they employed the mathematical morphology method to refine the signature to extract the MBR of the ship. ...
Article
Full-text available
This study develops a deep learning (DL) model to extract the ship size from Sentinel-1 synthetic aperture radar (SAR) images, named SSENet. We employ a single shot multibox detector (SSD)-based model to generate a rotat-able bounding box (RBB) for the ship. We design a deep-neural-network (DNN)-based regression model to estimate the accurate ship size. The hybrid inputs to the DNN-based model include the initial ship size and orientation angle obtained from the RBB and the abstracted features extracted from the input SAR image. We design a custom loss function named mean scaled square error (MSSE) to optimize the DNN-based model. The DNN-based model is concatenated with the SSD-based model to form the integrated SSENet. We employ a subset of the Open-SARShip, a data set dedicated to Sentinel-1 ship interpretation, to train and test SSENet. The training/testing data set includes 1500/390 ship samples. Experiments show that SSENet is capable of extracting the ship size from SAR images end to end. The mean absolute errors (MAEs) are under 0.8 pixels, and their length and width are 7.88 and 2.23 m, respectively. The hybrid input significantly improves the model performance. The MSSE reduces the MAE of length by nearly 1 m and increases the MAE of width by 0.03m compared to the mean square error (MSE) loss function. Compared with the well-performed gradient boosting regression (GBR) model, SSENet reduces the MAE of length by nearly 2 m (18.68%) and that of width by 0.06 m (2.51%). SSENet shows robustness on different training/testing sets. Index Terms-Custom loss function, deep learning (DL), deep neural network (DNN) regression, ship size extraction, synthetic aperture radar (SAR) image.
... Constant false alarm rate (CFAR) has the most profound influence on ship detection. Before the popularity of deep-learning, various classical algorithms [35]- [38] have been proposed based on CFAR to build ship detection system. Recently, due to the powerful representation capability of the neural network, many scholars have made outstanding contributions [13], [25]- [27], [39], [40] in ship detection. ...
Article
Full-text available
Ship detection and classification in SAR images play a vital role for wide applications. Due to the unique SAR imaging mechanism, ship detection and classification tasks have faced numerous challenges, such as land interference, image defocus, and noise. Many detectors and classifiers have been presented to handle these problems. However, the general deep learning-based detectors and classifiers lack the combination of SAR characteristics, which leads to poor performance. Compared with optical images, SAR images lack the texture information of ships, which brings great difficulties to the recognition task. To address the above issues, a novel deep learning-based ship detection and classification network combined with scattering characteristics is proposed in this paper. First, to accurately locate ships in large-scale SAR images, this paper designs a strong scattering point aware network (SPAN) by capturing the strong scattering points that existed in the ship area. SPAN recognizes the ship category according to their distribution characteristics. Second, to compensate for the feature loss caused by the down-sampling operation, this paper designs a more suitable resolution recovery module (RRM) to replace the bilinear interpolation method. Third, a region of interest automatic generation module (RoI-AG) is proposed to fully utilize the axis-align feature of oriented proposal boxes and the sufficient information of horizontal proposal boxes. Furthermore, the classification encoder module extracts the distribution feature of scattering points to classify SAR ships. Finally, the comprehensive experiments in the large-scale dataset for ship detection and classification in SAR images (LDSD) demonstrate the superior performance of the proposed method.
... CFAR is the deepest, most widely 31 used, and best-effective type of many target detection algorithms. This algorithm was ship target detection investigation [9]. Its core is to establish a distribution model based 34 on local area data, draw the probability density curve of the model, and then calculate 35 the object's pixel segmentation threshold using the false alarm rate. ...
Preprint
Synthetic Aperture Radar (SAR) is an active type of microwave remote sensing. Using the microwave imaging system, remote sensing monitoring of the land and global ocean can be done in any weather conditions around the clock. Detection of SAR image targets is one of the main needs of radar image interpretation applications. In this paper, an improved two-parameter CFAR algorithm based on Rayleigh distribution and morphological processing is proposed to perform ship detection and recognition in high resolution SAR images. Through simulation experiments, comprehensive study of the two algorithms for high resolution SAR image target detection is achieved. Finally, the results of ship detection experiments are compared and analyzed, and the effects of detection are evaluated according to the Rayleigh distribution model and algorithms.
... In the past decades, algorithms of ship detection using SAR images data have been extensively researched, and it has got many credible results. Among the existing models, the constant false alarm rate (CFAR) detector based on the statistical model of the sea clutter is a classical ship detection model and is widely used (Robey et al., 1992;Wackerman et al., 2001). Meanwhile, Gaussian distributions, Weibull (Schleher, 1976), Gamma (Principe et al., 1998), K (Jakeman and Tough, 1987), and Generalized K distribution (Ferrara et al., 2011;Liao et al., 2008) were used to describe the statistical of the sea clutter. ...
Preprint
Synthetic aperture radar (SAR) is considered being a good option for earth observation with its unique advantages. In this paper, we proposed an adaptive ship detector using full-polarization SAR images. First, by thoroughly investigating the scattering characteristics between ships and their background, and the wave polarization anisotropy, a novel ship detector is proposed by jointing the two characteristics, named Scattering-Anisotropy joint (joint-SA). Based on the theoretical analysis, we showed that the joint-SA is an effective physical quantity to show the difference between the ship and its background, and thus joint-SA can be used for ship detection of full-polarization image data. Second, the generalized Gamma distribution was used to characterize the joint-SA statistics of sea clutter with a large range of homogeneity. As a result, an adaptive constant false alarm rate (CFAR) method was implemented based on the joint-SA. Finally, RADARSAT-2 and GF-3 data in C-band and ALOS data in L-band are used for verification. We tested on five datasets, and the experimental results verify the correctness and superiority of the constant false alarm rate (CFAR) method based on the joint-SA. In addition, the experimental results also showed that the signal-clutter ratio (SCR) of the proposed ship detector joint-SA (33.17 dB, 35.98 dB, 57.25 dB) is better than that of DBSP (8.92 dB, 3.43 dB, 25.40 dB) and RsDVH (17.28 dB, 11.17 dB, 54.55 dB). More importantly, the proposed detector joint-SA has higher detection accuracy and a lower false alarm rate.
... Correct, rapid and accurate identification of ship targets under our jurisdiction plays an important role in improving the ability of marine rights enforcement. The current situation for rights protection in China is lack of the monitoring capacity, and the shipborne equipment can't monitor all-weather in a wide range [2]. Therefore, space-based is an important part of the integrated ocean management and control network, which mainly includes remote sensing and ocean satellites. ...
... The proposed TDSI algorithm is based on the CFAR detection results. To know more about CFAR detection algorithms, please refer to target detection or ship detection articles such as [18] and [19]. ...
Article
Full-text available
Stationary marine targets, such as oil rigs and offshore wind turbines, usually show up like bright spots without or with trivial position shifts in multi-temporal SAR imagery. They will trigger considerable false alarms in target detection applications tasked with ship detection, if they are not identified. An algorithm for stationary marine target detection, developed based on the assumption that the apparent positions of a stationary target in multi-temporal imagery are also stationary or nearly so, is proposed in this study. This algorithm requires a strict time-series of SAR images in temporal sequence as input. For each input SAR image, all targets on sea surface are initially detected with an iterative CA-CFAR (Cell-Averaging Constant False-Alarm Rate) detection algorithm, and their longitude and latitude positions are then used to identify whether they are stationary targets. Under this algorithm, a stationariness index with five levels (unknown target, suspected new stationary target, stationary target, suspected removed stationary target, and removed target) is defined for each target and must be iteratively updated with the latest level of identification. The proposed algorithm is promising for monitoring the status of stationary marine targets over a large sea area, because the processing of all input SAR images follows the same procedure and meanwhile the stationariness index is generated and kept updated. Two examples with GF-3 and RADARSAT-2 images are presented to illustrate the effectiveness of the proposed algorithm in detecting both offshore wind turbines and oil rigs.
... The common thread of all past and on-going projects is to detect vessels in Synthetic Aperture Radar (SAR) images and characterize them in terms of class and motion-related parameters when they are visible in images. The most exploited technique for ship detection is based on Constant False Alarm Rate (CFAR) algorithms [2,[10][11][12][13], in which the target is detected as pixels brighter than the background ones, keeping constant the false alarm rate over the image. A different concept is exploited by the multi-look/sub-aperture detection algorithm [14,15]. ...
Article
Full-text available
The recognition of wakes generated by dark vessels is a tremendous and interesting challenge in the field of maritime surveillance by Synthetic Aperture Radar (SAR) images. The paper aims at assessing the detection performance in different scenarios by processing Sentinel-1 SAR images along with ground truth data. Results confirm that the Radon-based approach is an effective technique for wake-based detection of dark vessels, and they lead to a deeper understanding of the effects of different sea and wind conditions. In general, the best applicative scenario is a marine image characterized by homogeneous sea clutter; the presence of natural surface film or strong transition from low wind speed areas to more windy zones worsen the detection performance. Nonetheless, the proposed approach features dark vessel detection capabilities by identifying their wakes, without any a priori knowledge of their positions.
Article
Recently, deep learning-based methods have brought new ideas for ship detection in synthetic aperture radar (SAR) images. However, several challenges still exist: 1) deep models contain millions of parameters, whereas the available annotated samples are not sufficient in number for training. Therefore, most deep detectors have to fine-tune networks pre-trained on ImageNet, which incurs learning bias due to the huge domain mismatch between SAR images and ImageNet images. Furthermore, it has a little flexibility to redesign the network structure; and 2) ships in SAR images are relatively small in size and densely clustered, whereas most deep detectors have poor performance with small objects due to the rough feature map used for detection and the extreme foreground-background imbalance. To address these problems, this paper proposes an effective approach to learn deep ship detector from scratch. First, we design a condensed backbone network, which consists of several dense blocks. Hence, earlier layers can receive additional supervision from the objective function through the dense connections, which makes it easy to train. In addition, feature reuse strategy is adopted to make it highly parameter efficient. Therefore, the backbone network could be freely designed and effectively trained from scratch without using a large amount of annotated samples. Second, we improve the cross-entropy loss to address the foreground-background imbalance and predict multi-scale ship proposals from several intermediate layers to improve the recall rate. Then, position-sensitive score maps are adopted to encode position information into each ship proposal for discrimination. The comparison results on the Sentinel-1 data set show that: 1) learning ship detector from scratch achieved better performance than ImageNet pre-trained model-based detectors and 2) our method is more effective than existing algorithms for detecting the small and densely clustered ships.
Article
The paper considers the problem of using images from SAR satellites for the identification of seagoing vessels. It describes the main functions of software and technological complex of the automated monitoring. The system is operated with utilizing space images of SAR satellites Sentinel 1A (B). The algorithmic part, which implements the detection on the sea surface the marks associated with ships, is described in details. To reduce the impact of speckle-noise, the image is pre-processed with the improved Lee-filter. Further processing lies in using an adaptive threshold algorithm that provides detection for each local background fragment of the image the unusually bright pixels, at the same time the algorithm provides a constant probability of error. By solving a nonlinear equation, for each position of the background window the algorithm finds the threshold brightness value and then all pixels above this value are considered vessels. In advance the evaluation of parameters of statistical distribution of pixels’ brightness is performed for each position of the background window. K-mean is used for such distribution. The selected bright pixels are combined into compact groups and their size and coordinates are being determined. The obtained results are compared with the data of the AIS, Automatic Identification System of ships, and the results are displayed on a cartographic basis.
Article
Synthetic aperture radar (SAR) is an active microwave imaging sensor with the capability of working in all-weather, all-day to provide high-resolution SAR images. Recently, SAR images have been widely used in civilian and military fields, such as ship detection. The scales of different ships vary in SAR images, especially for small-scale ships, which only occupy few pixels and have lower contrast. Compared with large-scale ships, the current ship detection methods are insensitive to small-scale ships. Therefore, the ship detection methods are facing difficulties with multi-scale ship detection in SAR images. A novel multi-scale ship detection method based on a dense attention pyramid network (DAPN) in SAR images is proposed in this paper. The DAPN adopts a pyramid structure, which densely connects convolutional block attention module (CBAM) to each concatenated feature map from top to bottom of the pyramid network. In this way, abundant features containing resolution and semantic information are extracted for multi-scale ship detection while refining concatenated feature maps to highlight salient features for specific scales by CBAM. Then, the salient features are integrated with global unblurred features to improve accuracy effectively in SAR images. Finally, the fused feature maps are fed to the detection network to obtain the final detection results. Experiments on the data set of SAR ship detection data set (SSDD) including multi-scale ships in various SAR images show that the proposed method can detect multi-scale ships in different scenes of SAR images with extremely high accuracy and outperforms other ship detection methods implemented on SSDD.
Conference Paper
This paper describes some experimental results of sea target detection by Sentinel – 1 SAR imagery and in-situ validation by using terrestrial sensors (Radar and AIS) of the Vessel Traffic Management and Information System (VTMIS) of Bulgaria. Images of Western Black Sea region, provided by ESA Sentinels Scientific Data Hub for the period of April 1-st till May 30-th, 2018 are processed for automatic target detection. Detected targets are then compared with Radar and AIS data, collected by the Bulgarian port infrastructure company and its VTMIS. Database includes lists of integrated targets (Radar and AIS), Radar targets only (with no AIS transponder, small non-SOLAS vessels) and AIS targets only (outside coverage area of Radar surveillance subsystem).
Article
Purpose The purpose of this paper is to study various ship detection methodologies. The accuracy of ship detection using satellite images still suffers from disturbances due to cluttered scenes and varying ship sizes. The suitability of the techniques for various applications is explained in this survey. Design/methodology/approach A list of data on the subject was gathered and processed into tables. The test outcomes were then discussed to determine the most effective ship detection technique under various complex environments. Findings In this work, the advantages and disadvantages of different classification techniques of ship detection are highlighted. The suitability of the techniques for various applications is also explained in this survey. Several hybrid approaches can be developed in order to increase the accuracy of ship detection system. This survey also aids in highlighting the significant contributions of satellite images to effective ship detection system. Originality/value In this paper, studying various ship detection methodologies is given specific attention. A survey on ship detection and recognition is clarified with the detailed comparative analysis of various classifier techniques.
Article
Full-text available
Since only the spatial feature information of ship target is utilized, the current deep learning-based synthetic aperture radar (SAR) ship detection approaches cannot achieve a satisfactory performance, especially in the case of multiscale or rotations, and the complex background. To overcome these issues, a novel multidimensional domain deep learning network for SAR ship detection is developed in this work to exploit the spatial and frequency-domain complementary features. The proposed method consists of the following main three steps. First, to learn hierarchical spatial features, the feature pyramid network (FPN) is adopted to produce ship target spatial multiscale characteristics with a top-down structure. Second, with a polar Fourier transform, the rotation-invariant features of SAR ship targets are obtained in the frequency domain. After that, a novel spatial-frequency characteristics fusion network is then presented, which seeks to learn more compact feature representations across different domains by updating the parameters of sub-networks interactively. The detection results are obtained due to utilizing the multidimensional domain information, and we evaluate the effectiveness of the proposed method using the existing SAR ship detection data set (SSDD). The results of the proposed method outperform other convolutional neural network (CNN)-based algorithms, especially for multiscale and rotation ship targets under complex backgrounds.
Article
Full-text available
In coastal region, the overestimation of wind speed are observed in SAR wind speed retrievals and the contamination from land and ships are considered as one of the main reasons. A novel method for calculating the value of normalized radar cross section (NRCS) cell has been developed to improve high-resolution synthetic aperture radar (SAR) wind retrieval in the case of contamination. The ratio of median and mean (Rmmpp) has been proposed to assess the effect of the distribution of NRCS on wind retrievals quantitatively. In this research, Sentinel-1 interferometric wide-swath (IW) dual polarization SAR data were collocated with in situ wind measurements from National Data Buoy Center (NDBC) for comparison. CMOD5, CMOD7 and the C-band Cross-Polarization Ocean (C-2PO) were used for wind retrievals. The statistic validation results show that the proposed method is more robust than the conventional method for wind speed retrievals. In addition, the Rmmpp could be used as an indicator for the reliability of SAR wind and the accuracy of wind speed increase with the Rmmpp increasing from 0.2 to 1.0 in the coastal area.
Article
Autonomous ship technologies have gained considerable interest due to the minimization of the challenging issues faced by the unpredictable errors of manual navigation, and therefore reduces human labor, increasing navigation security and profit margin. On autonomous shipping technologies, small ship detection is vital in ensuring shipping safety. With this motivation, this paper presents an efficient optimal mask regional convolutional neural network (Mask-CNN) technique for small ship detection (OMRCNN-SHD) on autonomous shipping technologies. Primarily, the data augmentation process is performed to resolve the issue of the limited number of real-world samples of small ships and helps to detect small ships in most cases accurately. Besides, the Mask RCNN with SqueezeNet model is used to detect ships and the hyperparameter tuning of the SqueezeNet model takes place by the use of the Adagrad optimizer. Furthermore, the Colliding Body's Optimization (CBO) algorithm with the weighted regularized extreme learning machine (WRELM) technique is employed to classify detected ships effectively. The comparative results analysis demonstrates the betterment of the OMRCNN-SHD technique over the current methods with the maximum accuracy of 98.63%.
Article
Small ship detection is an important topic in autonomous ship technology and plays an essential role in shipping safety. Since traditional object detection techniques based on the shipborne radar are not qualified for the task of near and small ship detection, deep learning-based image recognition methods based on video surveillance systems can be naturally utilized on autonomous vessels to effectively detect near and small ships. However, a limited number of real-world samples of small ships may fail to train a learning method that can accurately detect small ships in most cases. To address this, a novel hybrid deep learning method that combines a modified Generative Adversarial Network (GAN) and a Convolutional Neural Network (CNN)-based detection approach is proposed for small ship detection. Specifically, a Gaussian Mixture Wasserstein GAN with Gradient Penalty is utilized to first directly generate sufficient informative artificial samples of small ships based on the zero-sum game between a generator and a discriminator, and then an improved CNN-based real-time detection method is trained on both the original and the generated data for accurate small ship detection. Experimental results show that the proposed deep learning method (a) is competent to generate sufficient informative small ship samples and (b) can obtain significantly improved and robust results of small ship detection. The results also indicate that the proposed method can be effectively applied to ensuring autonomous ship safety.
Article
Full-text available
Illegal, unreported, and unregulated fishing threatens resource sustainability and equity. A major challenge with such activity is that most fishing vessels do not broadcast their positions and are “dark” in public monitoring systems. Combining four satellite technologies, we identify widespread illegal fishing by dark fleets in the waters between the Koreas, Japan, and Russia. We find >900 vessels of Chinese origin in 2017 and >700 in 2018 fished illegally in North Korean waters, catching an estimated amount of Todarodes pacificus approximating that of Japan and South Korea combined (>164,000 metric tons worth >$440 million). We further find ~3000 small-scale North Korean vessels fished, mostly illegally, in Russian waters. These results can inform independent oversight of transboundary fisheries and foreshadow a new era in satellite monitoring of fisheries.
Article
Full-text available
Constant False‐Alarm Rate (CFAR) algorithm is a mature method in marine target detection at present, the key of which is the choice of probability model. The most classical probability model is Gaussian distribution, which is suitable for modeling the calm ocean surface. Complex ocean conditions generally do not obey to Gaussian distribution, which limits the performance of CFAR algorithm. Therefore, we propose Adjoint Covariance Correction Model (ACCM) to improve the performance of CFAR marine target detection. Analyzing the distribution characteristic of ocean clutter under complex ocean conditions based on experiments, we find that it has long‐tail characteristic. Aiming at this characteristic, we have conducted fitting experiment on 23 kinds of probability density functions (PDFs) and find Loglogistic distribution has a better fitting effect on long‐tail characteristic, therefore, we use it to model ocean clutter under complex ocean conditions for the first time. In order to further improve the fitting goodness, we add a variance correction term to the Loglogistic model to construct ACCM for marine target extraction. The experiment result shows that ACCM effectively fits the long‐tail characteristic caused by Synthetic Aperture Radar backscatter under complex ocean conditions. The goodness of fit improves by 50% compared with Loglogistic model, and the amount of false alarms is 81.42% of that of Loglogistic model. The extraction accuracy of marine target characteristic parameter based on ACCM is 11.58% higher than that of Loglogistic model, and 12.18% higher than that of two‐parameter CFAR model based on standard error function (named SEF‐CFAR).
Article
Full-text available
Most of the recent research in the field of marine target detection has been concentrating on ships with large metallic parts. The focus of this work is on much more challenging targets represented by small rubber inflatables. They are of importance, since in recent years they have largely been used by migrants to cross the Mediterranean Sea between Libya and Europe. The motivation of this research is to mitigate the ongoing humanitarian crisis at Europe’s southern borders. These boats, packed with up to 200 people, are in no way suitable to cross the Mediterranean Sea or any other big water body and are in distress from the moment of departure. The establishment of a satellite-based surveillance infrastructure could considerably support search and rescue missions in the Mediterranean Sea, reduce the number of such boats being missed and mitigate the ongoing death in the open ocean. In this work we describe and analyze data from the InflateSAR acquisition campaign, wherein we gathered multiple-platform SAR imagery of an original refugee inflatable. The test site for this campaign is a lake which provides background clutter that is more predictable. The analysis considered a sum of experiments, enabling investigations of a broad range of scene settings, such as the vessel’s orientation, superstructures and speed. We assess their impact on the detectability of the chosen target under different sensor parameters, such as polarimetry, resolution and incidence angle. Results show that TerraSAR-X Spotlight and Stripmap modes offer good capabilities to potentially detect those types of boats in distress. Low incidence angles and cross-polarization decrease the chance of a successful identification, whereas a fully occupied inflatable, orthogonally oriented to the line of sight, seems to be better visible than an empty one. The polarimetric analyses prove the vessel’s different polarimetric behavior in comparison with the water surface, especially when it comes to entropy. The analysis considered state-of-the-art methodologies with single polarization and dual polarization channels. Finally, different metrics are used to discuss whether and to which extent the results are applicable to other open ocean datasets. This paper does not introduce any vessel detection or classification algorithm from SAR images. Rather, its results aim at paving the way to the design and the development of a specially tailored detection algorithm for small rubber inflatables.
Article
Synthetic aperture radar (SAR) is suitable for earth observation due to its unique advantages. In this study, we constructed an adaptive ship detector using full-polarization SAR images. First, we thoroughly investigated the differences in scattering characteristics between ships, their background, and the wave polarization anisotropy. We then constructed a novel ship detector that incorporates scattering and anisotropy (known as joint scattering–anisotropy [joint-SA]). We found that joint-SA is an effective physical quantity representing the difference between the ship and its background; thus, joint-SA can be used for ship detection using full-polarization image data. Second, we used generalized gamma distribution to characterize joint-SA statistics of sea clutter with a large homogeneity range. Third, an adaptive constant false alarm rate (CFAR) method was implemented based on joint-SA. Finally, RADARSAT-2 and GF-3 data in the C-band and ALOS data in the L-band were used for verification. We tested five datasets, and the experimental results verified the correctness and superiority of the CFAR method based on joint-SA. The results show that the signal–clutter ratio of the proposed ship detector (33.17 dB, 35.98 dB, and 57.25 dB) was higher than that of DBSP (8.92 dB, 3.43 dB, and 25.40 dB) and RsDVH (17.28 dB, 11.17 dB, and 54.55 dB). Furthermore, the proposed detector has a higher detection accuracy and lower false alarm rate than those of the other two other methods.
Article
Monitoring ships at sea is of great importance for both civilian and military purposes. Since ship wakes are more distinct than hulls, wakes are widely employed for ship detection from satellite imagery. In this paper, a novel approach was proposed to detect ships from optical imagery. Candidate wakes were first obtained by the normalized Radon transform of an image with the ship hull in the center. False wakes were then removed by pixel value verification, turbulent wake identification, included angle verification, and contrast verification. Meanwhile, bright edges of turbulent wakes were detected after turbulent wake identification. Detected wakes were verified with wakes delineated through visual inspection. By implementing the proposed technique to optical imagery, reflectance at the near-infrared (NIR) band of GF 1 and Sentinel 2 MSI and the panchromatic band of Landsat 8 OLI showed the best accuracies. Total recalls of turbulent and Kelvin wakes were 93.5% and 91.2%, respectively, while total precisions were 93.7% and 93.8%, respectively. Positions, directions, and endpoints of wakes were also obtained with good accuracy. The categories of wakes were confirmed by the wake features and the spatial relationships between wakes. Factors influencing the accuracy of the developed method and comparison with state-of-the-art wake detection methods were then discussed.
Article
Full-text available
The monitoring of maritime areas with remote sensing is essential for security reasons and also for the conservation of environment. The synthetic aperture radar (SAR) can play an important role in this matter by considering the possibility of acquiring high-resolution images at nighttime and under cloud cover. Recently, the new approaches based on the sub-look analysis for preserving the information of point targets (such as ship) in the spectrum of the SAR image have been proposed. In the sub-look analysis, the correlation of the ships in two sub-look images is preserved. Based on this property, in this paper first by using the second order statistics of polarimetric SAR data and the information of different polarization bases, the complex correlation between sub-look images is calculated. Then, using a criterion dependent on each of the four polarimetric channels and all polarization bases, the identification of ships from the sea is carried out. The proposed ship detection method is implemented on RADARSAT-2 image at C-band of Sanfrancisco area. The experimental results demonstrate that the method can discriminate the ships from background (sea clutter) with optimal contrast and desirable accuracy. The accuracy of the proposed method is about 19 and 17 percent better than other polarimetric methods and about 30, 2 and 35 percent better than the methods based on the spectral analysis.
Article
Full-text available
As marine transportation has increased in coastal regions, maritime accidents associated with vessels have steadily increased. Remotely sensed satellite or airborne images can aid rapid vessel monitoring over wide areas at high resolutions. In this study, airborne hyperspectral experiments were performed to detect marine vessels mainly including fishing boat and yacht by applying pixel-based mixture techniques and to estimate the size of the vessels through an objective ellipse fitting method. Various spectral libraries of marine objects and seawaters were constructed through in-situ experiments for spectral analysis of the internal structures of vessels. The hyperspectral images were dimensionally reduced through principal component analysis. Several hyperspectral mixture algorithms, such as N-FINDR, pixel purity index (PPI), independent component analysis (ICA), and vertex component analysis (VCA), were used for the detection of vessels. The N-FINDR and VCA techniques presented a total of 14 vessels, the ICA technique detected seven vessels, and the PPI technique detected two vessels. The pixel-based probability of detection (POD) and false alarm ratio (FAR) for all 14 vessels were 96.40% and 4.30%, respectively. The sizes of the vessels were estimated by extracting the boundaries of the vessels through a two-dimensional gradient and applying the ellipse fitting method. Compared with the digital mapping camera (DMC) images with resolutions of 0.10 m, the root-mean-square errors of the length and width of the vessels were approximately 1.19 m and 0.81 m, respectively. The application of spectral mixing methods provided a high probability of detecting the objects, as well as the overall structures of the decks of the vessels.
Article
Full-text available
A mathematical model for non-Rayleigh microwave sea echo is developed which describes explicitly the dependence of statistical properties of the radar cross section on the area of sea surface illuminated by the radar. In addition to the first probability distribution of the scattered radiation, its temporal and spatial correlation functions are also considered. It is shown that, in general, these correlation functions decay on at least two scales, the second, non-Rayleigh, contributions being strongly dependent on the properties of a "single scatterer." Predictions of the model are found to be in qualitative agreement with existing experimental data. A new class of probability distributions, the " K -distributions," is introduced, which may prove useful for fitting such data.
Article
The goal of the National Oceanic and Atmospheric Administration (NOAA) CoastWatch Program is to provide satellite and other environmental data and products for near-real-time monitoring of U.S. coastal waters in support of environmental science, management, and hazard response. During the last few years, products available through CoastWatch have expanded beyond the original infrared, visible, and sea surface temperature images to include ocean color, scatterometer wind, and synthetic aperture radar (SAR) images. A NOAA research and development program with partners in government, academia, and industry has endeavored to develop coastal ocean SAR applications for CoastWatch. Some of these applications, in particular wind measurement and hard target (i.e., vessel) detection, were developed for a preoperational demonstration. Users include the NOAA National Weather Service, the Alaska Department of Fish and Game, and the U.S. Coast Guard.
Box filtering involves replacing each pixel of an image with the average in a box. When it is extended in several simple ways, it becomes an efficient general purpose tool for image processing. This paper reviews box-filtering techniques and also describes some useful extensions.
Article
Over 200 SAR images of ship wakes were analyzed. Digital techniques for wake detection and analysis are described. Detection of ships and ship wakes from orbiting SAR seems feasible under a wide range of environmental conditions. The state of wake modeling is not sufficient to utilize all the ship information hidden in the wake, although speed and direction of motion can be determined quite straightforwardly.
Article
A statistical approach to point target detection in a clutter background is used to delineate the expected performance of the RADARSAT SAR (C-band HH polarization) for ship detection, and to compare the expected ship detection performance for the various RADARSAT SAR beam modes. Smaller ships may be detected with decreasing wind speed, increasing incidence angle, and finer resolution. ScanSAR Narrow Far is expected to be a good compromise between spatial coverage and detection probability. We present a quantitative validation of these model predictions based upon data acquired during a RADARSAT SAR ship detection/validation field program, which was held in March/April 1996 off the coast of Halifax, Nova Scotia. The program included buoy measurements of wind and wave conditions and RADARSAT SAR passes with known ships in each scene. We present validation of some of the model's key assumptions, in particular those concerning our hybrid C-band HH polarization ocean cross-section model, image probability density function, and ship radar signatures.
Article
This report considers the probability of detection off a target by a pulsed search radar, when the target has a fluctuating cross section. Formulas for detection probability are derived, and curves off detection probability vs, range are given, for four different target fluctuation models. The investigation shows that, for these fluctuation models, the probability of detection for a fluctuating target is less than that for a non-fluctuating target if the range is sufficiently short, and is greater if the range is sufficiently long. The amount by which the fluctuating and non-fluctuating cases differ depends on the rapidity of fluctuation and on the statistical distribution of the fluctuations. Figure 18, p. 307, shows a comparison between the non-fluctuating case and the four fluctuating cases considered.
Article
This report presents data from which one may obtain the probability that a pulsed-type radar system will detect a given target at any range. This is in contrast to the usual method of obtaining radar range as a single number, which can be taken mathematically to imply that the probability of detection is zero at any range greater than this number, and certainty at any range less than this number. Three variables, which have so far received little quantitative attention in the subject of radar range, are introduced in the theory: l.The time taken to detect the target. 2.The average time interval between false alarms (false indications of targets). 3.The number of pulses integrated. It is shown briefly how the results for pulsed-type systems may be applied in the case of continuous-wave systems. Those concerned with systems analysis problems including radar performance may profitably use this work as one link in a chain involving several probabilities. For instance, the probability that a given aircraft will be detected at least once while flying any given path through a specified model radar network may be calculated using the data presented here as a basis, provided that additional probability data on such things as outage time etc., are available. The theory developed here does not take account of interference such as clutter or man-made static, but assumes only random noise at the receiver input. Also, an ideal type of electronic integrator and detector are assumed. Thus the results are the best that can be obtained under ideal conditions. It is not too difficult, however, to make reasonable assumptions which will permit application of the results to the currently available types of radar. The first part of this report is a restatement of known radar fundamentals and supplies continuity and background for what follows. The mathematical part of the theory is not contained herein, but will be issued subsequently as a Separate report(2a)
Article
An automatic ship and ship wake detection system for spaceborne SAR images is described and assessed. The system is designed for coastal regions with eddies, fronts, waves and swells. The system uses digital terrain models to simulate synthetic SAR images to mask out land areas. Then a search for ship targets is performed followed by wake search around detected ship candidates. Finally, a homogeneity test and wake behavior test are performed which reduces the number of false alarms substantially. The system is demonstrated with ERS-1 SAR images and its performance is assessed using Seasat and ERS-1 images. No other information about the ships was available, hence, the basis for the assessment is through comparison with human visual interpretation of the same data. The number of lost ships (ship-like targets) was 7-8% for both Seasat-A and ERS-1. No false ships were detected. The number of lost or false wakes (wake-like features) was higher in ERS-1 images than in Seasat-A images and was nearly 15%. Taking into account the extremely strong variations in sea state in some of the selected scenes, the automatic detection performance is considered to be very good. In addition, the requirement of analyzing a 3-look ERS-1 scene of 100 km×100 km in less than eight minutes has been achieved on a workstation
Article
The objective of the research was twofold: to automatically detect ship wakes and to differentiate ship wakes from other linear ocean features produced by the underwater topography and existing sea conditions. An ADA (automatic detection algorithm) based on the Radon transform was developed and applied to the Seasat imagery. The basic system performs the Radon transform of the SAR (synthetic aperture radar) image, then detects bright and dark peaks produced in the transform by wakes (or other linear features) in the image. As the Radon transform essentially integrates the image intensity along every straight line through an image, each integral becomes one element in transform space. The integration process averages out the intensity fluctuations due to noise, thereby increasing the signal-to-noise ratio of the feature of interest in the transform space relative to that in the original image. A number of additional processing techniques were developed and tested to improve the PD (probability of detection) and reduce the PFA (probability of false alarm). To date, the use of an ADA, which combines a high-pass filter followed by a normalized Radon transform and a Wiener filter, has been shown to reliably distinguish wake peaks from false alarms.
Localized Radon Transform-Based Detection of Ship Wakes in SAR ImageryAn Automated Ship and Ship Wake Detection System for Spacebome SAR Images in Coastal Regions
  • A C Copeland
  • G Ravichandran
  • M M Trivedi
Copeland, A.C., Ravichandran, G., and Trivedi, M.M. (1990). "Localized Radon Transform-Based Detection of Ship Wakes in SAR Imagery", IEEE Trans. Geosci. Remote Sensing, Vol. 28, pp. 553-560, July. Eldhuset, K. (1996). "An Automated Ship and Ship Wake Detection System for Spacebome SAR Images in Coastal Regions", IEEE Trans. Geosci. Remote Sensing, Vol. 34, pp. 1010-1018, July.
A Simple Model for Satellite SAR Radiometric Discrimination Estimation
  • A L Gray
  • R K Hawkins
  • C E Livingston
Gray, A.L., Hawkins, R.K., and Livingston, C.E. (1984). "A Simple Model for Satellite SAR Radiometric Discrimination Estimation", Proc. 8th Canadian Symposium on Remote Sensing, Montreal, Quebec, 3-6 May 1983, pp. 25-38.