Article

Target detection in SAR imagery by genetic programming

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

Abstract

The automatic detection of ships in low-resolution synthetic aperture radar (SAR) imagery is investigated in this article. The detector design objectives are to maximise detection accuracy across multiple images, to minimise the computational effort during image processing, and to minimise the effort during the design stage. The results of an extensive numerical study show that a novel approach, using genetic programming (GP), successfully evolves detectors which satisfy the earlier objectives. Each detector represents an algebraic formula and thus the principles of detection can be discovered and reused. This is a major advantage over artificial intelligence techniques which use more complicated representations, e.g. neural networks.

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.

... Since the 1990s, many methods have been employed for object recognition. These include different kinds of neural networks [30,31,28,32,33], genetic algorithms [34,35], decision trees [36], statistical methods such as Gaussian models and Naive Bayes [37,36], support vector machines [37,36], genetic programming [38,13,22,39], and hybrid methods [40,41,42]. ...
... In terms of the number of classes in object classification, there are two categories: binary classification problems, where there are only two classes of objects to be classified, and multi-class classification problems, where more than two classes of images are involved. While GP has been widely applied to binary classification problems [38,44,10,45], it has also been applied to multi-class classification problems [46,22,47,16,15,48]. ...
... The use of GP in object/image recognition and detection has also been investigated in a variety of application domains. These domains include military applications [10,45], English letter recognition [24], face/eye detection and recognition [53,22,39], vehicle detection [38,13] and other vision and image processing problems [12,54,9,14,55,56]. ...
Article
Full-text available
This paper describes two innovations that improve the efficiency and effectiveness of a genetic programming approach to object detection problems. The approach uses genetic programming to construct object detection programs that are applied, in a moving window fashion, to the large images to locate the objects of interest. The first innovation is to break the GP search into two phases with the first phase applied to a selected subset of the training data, and a simplified fitness function. The second phase is initialised with the programs from the first phase, and uses the full set of training data with a complete fitness function to construct the final detection programs. The second innovation is to add a program size component to the fitness function. This approach is examined and compared with a neural network approach on three object detection problems of increasing difficulty. The results suggest that the innovations increase both the effectiveness and the efficiency of the genetic programming search, and also that the genetic programming approach outperforms a neural network approach for the most difficult data set in terms of the object detection accuracy.
... Object detection is the task of finding different types of objects belonging to different categories and is a challenging task especially, in the field of image processing and computer vision [111] [112]. In the field of image processing, GP has been used by many researchers [113][114][115][116][117][118][119][120][121][122][123][124][125][126] for accurate and efficient prediction of objects from cluttered and noisy scenes or images. In a review article, Krawiec et al. [127] analyzed the applications of GP in object detection related applications. ...
... Howard et al. [114] utilized GP to evolve detectors to detect ships in Synthetic Aperture Radar (SAR) imagery. Terminal nodes were real numerical values derived from random constants or pixel statistics. ...
Article
Full-text available
Genetic programming (GP) has been primarily used to tackle optimization, classification, and feature selection related tasks. The widespread use of GP is due to its flexible and comprehensible tree-type structure. Similarly, research is also gaining momentum in the field of image processing, because of its promising results over vast areas of applications ranging from medical image processing to multispectral imaging. Image processing is mainly involved in applications such as computer vision, pattern recognition, image compression, storage, and medical diagnostics. This universal nature of images and their associated algorithm, that is, complexities, gave an impetus to the exploration of GP. GP has thus been used in different ways for image processing since its inception. Many interesting GP techniques have been developed and employed in the field of image processing, and consequently, we aim to provide the research community an extensive view of these techniques. This survey thus presents the diverse applications of GP in image processing and provides useful resources for further research. In addition, the comparison of different parameters used in different applications of image processing is summarized in tabular form. Moreover, analysis of the different parameters used in image processing related tasks is carried-out to save the time needed in the future for evaluating the parameters of GP. As more advancement is made in GP methodologies, its success in solving complex tasks, not only in image processing but also in other fields, may increase. In addition, guidelines are provided for applying GP in image processing related tasks, the pros and cons of GP techniques are discussed, and some future directions are also set.
... Object detection is the task of finding different types of objects belonging to different categories and is a challenging task especially, in the field of IP and computer vision. In the field of IP, GP has been used by many researchers [69][70][71][72][73][74][75][76][77][78][79][80][81][82] for accurate and efficient prediction of objects from cluttered and noisy scenes or images. ...
... Howard et al. [70] utilized GP to evolve detectors to detect ships in Synthetic Aperture Radar (SAR) imagery. Terminal nodes were real numerical values derived from random constants or pixel statistics. ...
Preprint
Full-text available
During the last two decades, Genetic Programming (GP) has been largely used to tackle optimization, classification, and automatic features selection related tasks. The widespread use of GP is mainly due to its flexible and comprehensible tree-type structure. Similarly, research is also gaining momentum in the field of Image Processing (IP) because of its promising results over wide areas of applications ranging from medical IP to multispectral imaging. IP is mainly involved in applications such as computer vision, pattern recognition, image compression, storage and transmission, and medical diagnostics. This prevailing nature of images and their associated algorithm i.e complexities gave an impetus to the exploration of GP. GP has thus been used in different ways for IP since its inception. Many interesting GP techniques have been developed and employed in the field of IP. To give the research community an extensive view of these techniques, this paper presents the diverse applications of GP in IP and provides useful resources for further research. Also, comparison of different parameters used in ten different applications of IP are summarized in tabular form. Moreover, analysis of different parameters used in IP related tasks is carried-out to save the time needed in future for evaluating the parameters of GP. As more advancement is made in GP methodologies, its success in solving complex tasks not only related to IP but also in other fields will increase. Additionally, guidelines are provided for applying GP in IP related tasks, pros and cons of GP techniques are discussed, and some future directions are also set. https://onlinelibrary.wiley.com/doi/epdf/10.1111/coin.12459
... Object detection is the task of finding different types of objects belonging to different categories and is a challenging task especially, in the field of IP and computer vision. In the field of IP, GP has been used by many researchers [69][70][71][72][73][74][75][76][77][78][79][80][81][82] for accurate and efficient prediction of objects from cluttered and noisy scenes or images. ...
... Howard et al. [70] utilized GP to evolve detectors to detect ships in Synthetic Aperture Radar (SAR) imagery. Terminal nodes were real numerical values derived from random constants or pixel statistics. ...
Preprint
During the last two decades, Genetic Programming (GP) has been largely used to tackle optimization, classification, and automatic features selection related tasks. The widespread use of GP is mainly due to its flexible and comprehensible tree-type structure. Similarly, research is also gaining momentum in the field of Image Processing (IP) because of its promising results over wide areas of applications ranging from medical IP to multispectral imaging. IP is mainly involved in applications such as computer vision, pattern recognition, image compression, storage and transmission, and medical diagnostics. This prevailing nature of images and their associated algorithm i.e complexities gave an impetus to the exploration of GP. GP has thus been used in different ways for IP since its inception. Many interesting GP techniques have been developed and employed in the field of IP. To give the research community an extensive view of these techniques, this paper presents the diverse applications of GP in IP and provides useful resources for further research. Also, comparison of different parameters used in ten different applications of IP are summarized in tabular form. Moreover, analysis of different parameters used in IP related tasks is carried-out to save the time needed in future for evaluating the parameters of GP. As more advancement is made in GP methodologies, its success in solving complex tasks not only related to IP but also in other fields will increase. Additionally, guidelines are provided for applying GP in IP related tasks, pros and cons of GP techniques are discussed, and some future directions are also set.
... Finding a good fitness function for a particular task is an important but difficult problem in developing a GP system. Various fitness functions have been devised for object detection, with varying success [1,7,6,3,8]. These tend to combine many parameters using scaling factors which specify the relative importance of each parameter, with no obvious indication of what scaling factors are good for a given problem. ...
... These tend to combine many parameters using scaling factors which specify the relative importance of each parameter, with no obvious indication of what scaling factors are good for a given problem. The majority of fitness functions for localisation require clustering to be performed to group multiple localisations of single objects into a single point before the fitness is determined [2,3,6]. Other measures are then incorporated in order to include information about the pre-clustered results (such as how many points have been found for each object). ...
Article
Full-text available
Object detection is an important eld of research in computer vision which genetic programming has been applied to recently. This paper describes two new tness functions in genetic programming for object detection. Both tness functions are based on recall and precision of genetic programs. The rst is a tolerance based tness function and the second is a weighted tness function. The merits and eectiv eness of the two tness function are discussed. The two tness functions are examined and compared on three object detection problems of increasing dicult y. The results suggest that both tness functions perform very well on the relatively easy problem, the weighted tness function outperforms the tolerance based tness function on the relatively dicult problems.
... These features make µ-TS perfectly fit to most GP applications. The tournament size typically used in GP practice varies from three [178,179], through four [70,144,149], five [115,167], six [70] to seven [101,87,133,19]. ...
... These features make µ-TS perfectly fit to most GP applications. The tournament size typically used in GP practice varies from three [178,179], through four [70,144,149], five [115,167], six [70] to seven [101,87,133,19]. ...
Thesis
Full-text available
Genetic Programming (GP) is a machine learning technique for automatic induction of computer programs from examples. The examples typically consist of two parts: program arguments – the input and a target program output. Both input and output may be expressed in terms of numeric or textual variables or even a conglomerate of the above. This problem formulation enables formalizing semantics of a program as a tuple of outputs it returns in effect of execution on the sample inputs. Use of semantics in GP gains an interest in research community, since the semantic methods developed so far in GP proved capable of achieving lower error, better generalization and smaller programs and quicker convergence to an optimum than the contemporary methods. We embrace existing notions of semantics of program, semantic distance, semantic neutrality and effectiveness of genetic operators under the umbrella of common conceptual framework for Semantic Genetic Programming (SGP). Then, we show that if a fitness function is a metric, the fitness landscape spanned over a space of all programs proxied by semantics becomes a cone with the optimal semantics in the apex. This provides justification for use of the recently developed Geometric Semantic Genetic Programming (GSGP), where geometric genetic operators utilize the conic shape of the landscape. We derive properties of progress and progress bounds of geometric operators for different combinations of fitness functions and semantic distances. We present a comprehensive literature review of existing semantic methods, discuss their advantages and disadvantages and for each show how the defined properties apply to it. Next, we propose an algorithm for backpropagating semantics trough program structure and competent algorithms for operators: population initialization, parent selection, mutation and crossover that are approximately geometric, effective and free of certain drawbacks of the existing geometric methods. Then, we experimentally assess the proposed algorithms and compare them with the existing methods in terms of training set fitness, generalization on test set, probability of performing geometric and effective application, size of produced programs and computational costs. We use a suite of nine symbolic regression and nine Boolean program synthesis benchmarks. The analysis shows that the proposed algorithms achieve performance that is consistently better than that offered by other semantic GP methods for symbolic regression domain and not worse than the best other methods for Boolean domain. Finally, we experimentally find the proportions of competent mutation and competent crossover that lead to the optimal results in the above range of benchmarks.
... Synthesizing effective new features from primitive features is equivalent to finding good points in the feature combination space where each point represents a combination of primitive features. The feature combination space and feature subset space are huge and complicated and it is very difficult to find good points in such vast spaces unless one has an efficient search algorithm [5,7]. Genetic programming (GP) is used as search algorithm. ...
Article
Full-text available
This document is devoted to the task of object detection and recognition in digital images by using genetic programming. The goal was to improve and simplify existing approaches. The detection and recognition are achieved by means of extracting the features. A genetic program is used to extract and classify features of objects. Simple features and primitive operators are processed in genetic programming operations. We are trying to detect and to recognize objects in SAR images. Due to the new approach described in this article, five and seven types of objects were recognized with good recognition results.
... These characteristics make image analysis problems appropriate candidates for machine learning and evolutionary approaches such as GP [23,7]. For instance, GP has been used for image classification [21,47], object detection and recognition [14,9,13], feature synthesis [22,41], image segmentation [39,45], and local image description [37,38]. In particular, the proposal made in this paper is related to other works that extract a descriptive value for each image pixel. ...
Article
The regularity of a signal can be numerically expressed using Hölder exponents, which characterize the singular structures a signal contains. In particular, within the domains of image processing and image understanding, regularity-based analysis can be used to describe local image shape and appearance. However, estimating the Hölder exponent is not a trivial task, and current methods tend to be computationally slow and complex. This work presents an approach to automatically synthesize estimators of the pointwise Hölder exponent for digital images. This task is formulated as an optimization problem and Genetic Programming (GP) is used to search for operators that can approximate a traditional estimator, the oscillations method. Experimental results show that GP can generate estimators that achieve a low error and a high correlation with the ground truth estimation. Furthermore, most of the GP estimators are faster than traditional approaches, in some cases their runtime is orders of magnitude smaller. This result allowed us to implement a real-time estimation of the Hölder exponent on a live video signal, the first such implementation in current literature. Moreover, the evolved estimators are used to generate local descriptors of salient image regions, a task for which a stable and robust matching is achieved, comparable with state-of-the-art methods. In conclusion, the evolved estimators produced by GP could help expand the application domain of Hölder regularity within the fields of image analysis and signal processing.
... There is something quite visceral about evolving programs. Genetic Programming has been applied to detecting targets in SAR 6 Imagery [10], multi-class object detection [28][28] and Object Tracking [19] amongst others. Compared to basic parameter optimisation, Genetic Programming has a large scope for evolving interesting and novel solutions. ...
... This is where other classifying techniques such as GP are more versatile and provide different rules from being run individually with concentrated data sets and then merged together to provide a robust classifier for many different conditions and/or different machining processes as seen in previous work [4]. For instance, GP has been used in difficult pattern recognition situations such as the work presented by Howard et al. [7,8,9]. This is where GP is configured to recognise ships and vehicles from satellite or aircraft synthetic aperture radar (SAR) imagery. ...
Article
Full-text available
The capability to generate complex geometry features at tight tolerances and fine surface roughness is a key element in implementation of Creep Feed grinding process in specialist applications such as the aerospace manufacturing environment. Based on the analysis of 3D cutting forces this paper proposes a novel method of predicting the profile deviations of tight geometrical features generated using Creep Feed grinding. In this application, there are several grinding passes made at varying depths providing an incremental geometrical change with the last cut generating the final complex feature. With repeatable results from co-ordinate measurements both the radial and tangential forces can be gauged versus the accuracy of the ground features. The results the tangential force was found more sensitive to the deviation of actual cut depth from the theoretical one. However, to make a more robust prediction on the profile deviation its values were considered as a function of both force components. In addition, the power signals were obtained as these signals are also proportional to force and deviation measurements. Genetic Programming (GP), an evolutionary programming technique, has been used to compute the prediction rules of part profile deviations based on the extracted radial and tangential force and correlated with the initial “gauging” methodology. It was found that using this technique, complex rules can be achieved and used online to dynamically control the geometrical accuracy of the ground features. The GP complex rules are based on the correlation between the measured forces and recorded deviation of the theoretical profile. The mathematical rules are generated from Darwinian evolutionary strategy which provides the mapping between different output classes. GP works from crossover recombination of different rules and the best individual is evaluated in terms of the given ‘best fitness value so far’ which closes on an optimal solution. Once the best rule has been generated this can be further used independently or in combination with other close-to-best rules to control the evolution of output measures of machining processes. The best GP terminal sets will be realised in rule-based embedded coded systems which will finally be implemented into a real-time Simulink simulation. This realisation gives a view of how such a control regime can be utilised within an industrial capacity. Neural Networks were also used for GP rule verification.
... Trujillo and Olague [31] have also used GP to generate feature extractors for computer vision applications. In addition, a GP-based detector was proposed by Howard et al. [7] for detecting ship wakes in synthetic aperture radar (SAR) images. Inspired by the successful applications mentioned above, in this paper we propose evolving a spatio-temporal descriptor for human-action recognition using GP. ...
Article
The potential value of human action recognition has led to it becoming one of the most active research subjects in computer vision. In this paper, we propose a novel method to automatically generate low-level spatio-temporal descriptors showing good performance, for high-level human-action recognition tasks. We address this as an op-timization problem using genetic programming (GP), an evolutionary method, which produces the descriptor by combining a set of primitive 3D operators. As far as we are aware, this is the first report of using GP for evolving spatio-temporal descriptors for action recognition. In our evolutionary architecture, the average cross-validation classi-fication error calculated using the support-vector machine (SVM) classifier is used as the GP fitness function. We run GP on a mixed dataset combining the KTH and the Weiz-mann datasets to obtain a promising feature-descriptor solution for action recognition. To demonstrate generalizability, the best descriptor generated so far by GP has also been tested on the IXMAS dataset leading to better accuracies compared with some previous hand-crafted descriptors.
... The second problem is also known as automatic target detection (ATD) problem. Major techniques for ATD include adaptive boosting [3], extended fractal feature [4], genetic programming [5], multiscale autoregressive (MAR), multiscale autoregressive moving average (MARMA) models, singular value decomposition (SVD) methods [6] and constant false alarm rate (CFAR) processing [7]. CFAR processing is widely used to give a globally applicable threshold for a constant probability of false alarms through estimating and removing the local background statistics. ...
Article
Full-text available
In this paper, we develop a target detection algorithm based on a supervised learning technique that maximizes the margin between two classes, i.e., the target class and the non-target class. Specifically, our target detection algorithm consists of 1) image differencing, 2) maximum-margin classifier, and 3) diversity combining. The image differencing is to enhance and highlight the targets so that the targets are more distinguishable from the background. The maximum-margin classifier is based on a recently developed feature weighting technique called Iterative RELIEF; the objective of the maximum-margin classifier is to achieve robustness against uncertainties and clutter. The diversity combining utilizes multiple images to further improve the performance of detection, and hence it is a type of multi-pass change detection. We evaluate the performance of our proposed detection algorithm, using the CARABAS-II synthetic aperture radar (SAR) image data and the experimental results demonstrate superior performance of our algorithm, compared to the benchmark algorithm.
... The extended fractal feature, sensitive to both contrast and sizes of objects, was proposed to detect ships by Kaplan in [16]. In [17], genetic programming was introduced into target detection for SAR images by taking advantage of the "human vision system" approach to identify regions that stand out from their surroundings. The work of Bhanu et al. [18] focused on selecting discriminative features with a genetic algorithm, but the selected features cannot distinguish target and clutter well when the intensity of the clutter is stronger than targets. ...
Article
Full-text available
The robust detection of ships is one of the key techniques in coastal and marine applications of synthetic aperture radar (SAR). Conventional SAR ship detectors involved multiple parameters, which need to be estimated or determined very carefully. In this paper, we propose a new ship detection approach based on multi-scale heterogeneities under the a contrario decision framework, with a few parameters that can be easily determined. First, multi-scale heterogeneity features are extracted and fused to build a heterogeneity map, in which ships are well highlighted from backgrounds. Second, a set of reference objects are automatically selected by analyzing the saliency of local regions in the heterogeneity map and then are used to construct a null hypothesis model for the final decision. Finally, the detection results are obtained by using an a contrario decision. Experimental results on real SAR images demonstrate that the proposed method not only works more stably for ships with different sizes, but also has better performance than conventional ship detectors.
... This method can detect objects buried in strong speckle noises. Howard et al. [18] presented a detector by exploiting genetic programming. These methods were devised for detecting small target with several or tens of pixels so that they are not suitable for airplane detection in high-resolution SAR images in which the target may contains several hundreds of pixels. ...
Article
This paper proposes a new automatic and adaptive aircraft target detection algorithm in high-resolution synthetic aperture radar (SAR) images of airport. The proposed method is based on gradient textural saliency map under the contextual cues of apron area. Firstly, the candidate regions with the possible existence of airport are detected from the apron area. Secondly, directional local gradient distribution detector is used to obtain a gradient textural saliency map in the favor of the candidate regions. In addition, the final targets will be detected by segmenting the saliency map using CFAR-type algorithm. The real high-resolution airborne SAR image data is used to verify the proposed algorithm. The results demonstrate that this algorithm can detect aircraft targets quickly and accurately, and decrease the false alarm rate.
... Genetic Programming has numerous achievements, it copes well with curve fitting, data modeling and time series prediction [34], control of industrial process [13,41], synthesis of electronic circuits [9,45], robot controlling [52], image and signal processing [31,46,88], lossy [68] and lossless [37] compression and more. ...
... Techniques using adaptive boosting [4], extended fractal feature [5], genetic programming [6], multiscale autoregressive (MAR), multiscale autoregressive moving average (MARMA) models, singular value decomposition (SVD) methods [7] and constant false alarm rate (CFAR) processing [8] were studied. According to Lundberg et al. [1], the main technical challenge in designing an ATD algorithm is not detecting targets but sufficiently suppress the false alarm rate to a useful level. ...
Article
Full-text available
Rotation of targets pose great a challenge for the design of an automatic image-based target detection system. In this paper, we propose a target detection algorithm that is robust to rotation of targets. Our key idea is to use rotation invariant features as the input for the classifier. For an image in Radon transform space, namely R(b,theta), taking the magnitude of 1-D Fourier transform on theta, we get |Ftheta{R(b,theta)}|. It was proved that the coefficients of the combined Radon and 1-D Fourier transform, |Ftheta{R(b,theta)}| is invariant to rotation of the image. These coefficients are used as the input to a maximum-margin classifier based on I-RELIEF feature weighting technique. Its objective is to maximize the margin between two classes and improve the robustness of the classifier against uncertainties. For each pixel of a sample SAR image, a feature vector can be extracted from a sub image centered at that pixel. Then our classifier decides whether the pixel is target or non-target. This produces a binary-valued image. We further improve the detection performance by connectivity analysis, image differencing and diversity combining. We evaluate the performance of our proposed algorithm, using the data set collected by Swedish CARABAS-II systems, and the experimental results show that our proposed algorithm achieves superior performance over the benchmark algorithm.
... Howard et al. [22] described a GP approach, where binary classifiers were evolved with the goal of detecting targets within SAR imagery. Later, Zhang et al. [63] used GP to perform multi-class detection of small objects present in large images, with domain independent pixel statistics as the terminal sets. ...
Article
Full-text available
Nowadays, object recognition based on local invariant features is widely acknowledged as one of the best paradigms for object recognition due to its robustness for solving image matching across different views of a given scene. This paper proposes a new approach for learning invariant region descriptor operators through genetic programming and introduces another optimization method based on a hill-climbing algorithm with multiple re-starts. The approach relies on the synthesis of mathematical expressions that extract information derived from local image patches called local features. These local features have been previously designed by human experts using traditional representations that have a clear and, preferably mathematically, well-founded definition. We propose in this paper that the mathematical principles that are used in the description of such local features could be well optimized using a genetic programming paradigm. Experimental results confirm the validity of our approach using a widely accepted testbed that is used for testing local descriptor algorithms. In addition, we compare our results not only against three state-of-the-art algorithms designed by human experts, but also, against a simpler search method for automatically generating programs such as hill-climber. Furthermore, we provide results that illustrate the performance of our improved SIFT algorithms using an object recognition application for indoor and outdoor scenarios.
... Computational Intelligence (CI) techniques, especially those which are bioinspired, are able to handle such descriptive values in combination (see for example [1]). There are many different types of CI techniques and we are pursuing decision trees and artificial neural networks for the oceanic SAR target detection task. ...
Conference Paper
Full-text available
An artificial neural network whose topology is informed by an Oblique Decision Tree is applied to target detection in maritime Synthetic Aperture Radar. The number of neurons in the first layer is the same as the number of decision tree nodes and the number of nodes in the second hidden layer is the same as the number of leaf nodes. The neural network output are the class labels. Our approach differs from other efforts in the literature in that the Oblique Decision Tree and the Fisher´s Linear Discriminant are used as a decision criterion. Classifier testing and validation were achieved, applying these algorithms to radar images. Initial results are practical with satisfactory training time; generalization capability and a speedy architecture definition.
... Since the early 1990s, there has been a number of reports on applying genetic programming techniques to a range of object recognition problems such as shape classification, face identification, and medical diagnosis [1,8,14,19,21,24,27,26,28]. While showing promise, current GP techniques are limited and frequently do not give satisfactory results on difficult classification tasks. ...
Article
Full-text available
This paper describes a new crossover operator in genetic programming for object recognition particularly object classification problems. In this approach, instead of randomly choosing the crossover points as in the standard crossover operator, we use a measure called looseness to guide the selection of crossover points. Rather than using the genetic beam search only, this approach uses a hy- brid beam-hill climbing search scheme in the evolutionary process. This approach is examined and compared with the standard crossover operator and the headless chicken crossover method on a sequence of object classification problems. The re- sults suggest that this approach outperforms both the headless chicken crossover and the standard crossover on all of these problems.
... Recently, many studies have relied on image-based feature extraction (machine learning) methods, such as the artificial neural network (ANN) and support vector machine (SVM) methods [15,16]. The machine-learning analyzes the input data by calculating weighting factors from the relation between input data and training set. ...
Article
Full-text available
For ship detection, X-band synthetic aperture radar (SAR) imagery provides very useful data, in that ship targets look much brighter than surrounding sea clutter due to the corner-reflection effect. However, there are many phenomena which bring out false detection in the SAR image, such as noise of background, ghost phenomena, side-lobe effects and so on. Therefore, when ship-detection algorithms are carried out, we should consider these effects and mitigate them to acquire a better result. In this paper, we propose an efficient method to detect ship targets from X-band Kompsat-5 SAR imagery using the artificial neural network (ANN). The method produces the ship-probability map using ANN, and then detects ships from the ship-probability map by using a threshold value. For the purpose of getting an improved ship detection, we strived to produce optimal input layers used for ANN. In order to reduce phenomena related to the false detections, the non-local (NL)-means filter and median filter were utilized. The NL-means filter effectively reduced noise on SAR imagery without smoothing edges of the objects, and the median filter was used to remove ship targets in SAR imagery. Through the filtering approaches, we generated two input layers from a Kompsat-5 SAR image, and created a ship-probability map via ANN from the two input layers. When the threshold value of 0.67 was imposed on the ship-probability map, the result of ship detection from the ship-probability map was a 93.9% recall, 98.7% precision and 6.1% false alarm rate. Therefore, the proposed method was successfully applied to the ship detection from the Kompsat-5 SAR image.
... • curve fitting, data modeling and symbolic regression [67][68][69], • image and signal processing [70][71][72], ...
Article
Full-text available
In this paper, the publicly available dataset of condition based maintenance of combined diesel-electric and gas (CODLAG) propulsion system for ships has been utilized to obtain symbolic expressions which could estimate gas turbine shaft torque and fuel flow using genetic programming (GP) algorithm. The entire dataset consists of 11934 samples that was divided into training and testing portions of dataset in an 80:20 ratio. The training dataset used to train the GP algorithm to obtain symbolic expressions for gas turbine shaft torque and fuel flow estimation consisted of 9548 samples. The best symbolic expressions obtained for gas turbine shaft torque and fuel flow estimation were obtained based on their R 2 score generated as a result of the application of the testing portion of the dataset on the aforementioned symbolic expressions. The testing portion of the dataset consisted of 2386 samples. The three best symbolic expressions obtained for gas turbine shaft torque estimation generated R 2 scores of 0.999201, 0.999296, and 0.999374, respectively. The three best symbolic expressions obtained for fuel flow estimation generated R 2 scores of 0.995495, 0.996465, and 0.996487, respectively.
... Moreover, GP is a very flexible technique, a characteristic that has allowed researchers to apply it in various fields and problem domains (Koza et al., 2000(Koza et al., , 2008Koza, 2010). In computer vision, for instance, GP has been used for object recognition (Howard et al., 1999;Ebner, 2009;Hernández et al., 2007), image classification (Krawiec, 2002;Tan 2.2 genetic programming overview et al., 2005), feature synthesis (Krawiec and Bhanu, 2005;Puente et al., 2011), image segmentation (Poli, 1996;Song and Ciesielski, 2008), feature detection (Trujillo et al., 2008c(Trujillo et al., , 2010Olague and Trujillo, 2011) and local image description (Pérez and Olague, 2008;Perez and Olague, 2009). GP has been successfully applied to computer vision problems since it is easy to define concrete performance criteria and the development or acquisition of data-sets is now trivially done. ...
Thesis
Novelty Search (NS) is a unique approach towards search and optimization,where an explicit objective function is replaced by a measureof solution novelty. However, NS has been mostly used in evolutionaryrobotics, its usefulness in classic machine learning problems has beenunexplored. This thesis presents a NS-based Genetic Programming(GP) algorithms for common machine learning problems, with the followingcontributions. It is shown that NS can solve real-world classification,clustering and symbolic regression tasks, validated on realworldbenchmarks and synthetic problems. These results are madepossible by using a domain-specific behavior descriptor, related to theconcept of semantics in GP. Moreover, two new versions of the NS algorithmare proposed, Probabilistic NS (PNS) and a variant of MinimalCriteria NS (MCNS). The former models the behavior of each solutionas a random vector and eliminates all the NS parameters while reducingthe computational overhead of the NS algorithm; the latter uses astandard objective function to constrain and bias the search towardshigh performance solutions. The thesis also discusses the effects of NSon GP search dynamics and code growth. Results show that NS can beused as a realistic alternative for machine learning, and particularly forGP-based classification.
... From the fact that GP is a search and optimization algorithm, it can be utilized as a search algorithm for generating an efficient classifier. Since the early 1990s, several reports (Howard, Roberts, & Brankin, 1999;Song, Ciesielski, & Williams, 2002) for applying GP techniques to a range of classification problems were cited. The most common methods used by many authors for the detection of epileptic seizures are ANN, linear discriminate analysis, LS-SVM, Bayesian classifier, and nearest neighbour classifier. ...
Article
Full-text available
Epilepsy, sometimes called seizure disorder, is a neurological condition that justifies itself as a susceptibility to seizures. A seizure is a sudden burst of rhythmic discharges of electrical activity in the brain that causes an alteration in behaviour, sensation, or consciousness. It is essential to have a method for automatic detection of seizures, as these seizures are arbitrary and unpredictable. A profound study of the electroencephalogram (EEG) recordings is required for the accurate detection of these epileptic seizures. In this study, an Innovative Genetic Programming framework is proposed for classification of EEG signals into seizure and nonseizure. An empirical mode decomposition technique is used for the feature extraction followed by genetic programming for the classification. Moreover, a method for intron deletion, hybrid crossover, and mutation operation is proposed, which are responsible for the increase in classification accuracy and a decrease in time complexity. This suggests that the Innovative Genetic Programming classifier has a potential for accurately predicting the seizures in an EEG signal and hints on the possibility of building a real‐time seizure detection system.
... Furthermore, Daida et al. (1996) devised a GP approach to identify ice-flow ridges from SAR imagery. In the same way, the work made by Howard et al. (1999) automates ship detection from SAR images of the English Channel taken by the European Remote Sensing (ERS) satellite using a two-stage GP process. Also, Ross et al. (2005) used GP to evolve Boolean and general mathematical expressions in order to discriminate among three specific minerals (buddingtonite, alunite, and kaolinite) from hyperspectral images. ...
Article
Full-text available
This work describes a genetic programming (GP) approach that creates vegetation indices (VI’s) to automatically detect the sum of healthy, dry, and dead vegetation. Nowadays, it is acknowledged that VI’s are the most popular method for extracting vegetation information from satellite imagery. In particular, erosion models like the “Revised Universal Soil Loss Equation” (RUSLE) can use VI’s as input to measure the effects of the RUSLE soil cover factor (C). However, the results are generally incomplete, because most indices recognize only healthy vegetation. The aim of this study is to devise a novel approach for designing new VI’s that are better - correlated with C, using field and satellite information. Our approach consists on stating the problem in terms of optimization through GP learning, building novel indices by iteratively recombining a set of numerical operators and spectral channels until the best composite operator is found. Experimental results illustrate the efficiency and reliability of our approach in contrast with traditional indices like those of the NDVI and SAVI family. This study provides evidence that similar problems related to soil erosion assessment could be analyzed with our proposed methodology.
... Since the early 1990s, there has been a number of reports on applying GP techniques to a range of object recognition problems such as shape classification, face identification, and medical diagnosis [10,11,12,13,14,15,16,17]. While showing promise, current GP techniques are limited and frequently do not give satisfactory results on difficult classification tasks. ...
Article
Full-text available
To improve the standard crossover operator in genetic programming (GP), Tackett (19) introduced the brood recombination crossover method. The key pa- rameter in this method is the brood size. This paper investigates a number of new developments of brood size in the brood recombination crossover method in GP. We first investigate the effect of different fixed brood sizes, then construct three dynamic models for setting the brood size. These developments are examined and compared with the standard crossover operator on three object classification problems of increasing difficulty. The results suggest that the brood recombina- tion methods with all the new developments outperforms the standard crossover operator for all the problems. As the brood size increases, the system effective performance can be improved. When it exceeds a certain point, however, the ef- fective performance will not be improved and the system will become less efficient. Investigation of three dynamic models for the brood size reveals that a variable brood size which is dynamically set with the number of generations can further improve the system performance over the fixed brood size.
... In one approach, statistical distribution models are widely used to detect ships in SAR imagery (Hansen, 1973;Rohling, 1983;Armstrong and Griffiths, 1991;Alberola-López et al., 1991;Wang et al., 2008;Wang et al., 2016). Another approach relies on the information extracted by image spectral analysis, decomposition or wavelet transformation (Souyris et al., 2003;Ouchi et al., 2004) while other studies depend on SAR imagebased feature extraction and selection (Kaplan et al., 2001;Howard et al., 1999). ...
Article
Ship detection in synthetic aperture radar(SAR)imagery has long been an active research topic and has many applications. In this paper,we propose an efficient method for detecting ships from SAR imagery using filtering. This method exploits ship masking using a median filter that considers maximum ship sizes and detects ships from the reference image, to which a Non-Local means (NL-means) filter is applied for speckle de-noising and a differential image created from the difference between the reference image and the median filtered image. As the pixels of the ship in the SAR imagery have sufficiently higher values than the surrounding sea, the ship detection process is composed primarily of filtering based on this characteristic. The performance test for this method is validated using KOMPSAT-5 (Korea Multi-Purpose Satellite-5) SAR imagery. According to the accuracy assessment, the overall accuracy of the region that does not include land is 76.79%, and user accuracy is 71.31%. It is demonstrated that the proposed detection method is suitable to detect ships in SAR imagery and enables us to detect ships more easily and efficiently.
... Since the early 1990s, there have been a number of reports on applying genetic programming (GP) techniques to object recognition problems [1,3,6,10,11,12,13]. Typically, these GP systems used either high level or low level image features as the terminal set, arithmetic and conditional operators as the function set, and classification accuracy, error rate or similar measures as the fitness function. ...
Article
Full-text available
This paper describes an approach to the use of gradient descent search in tree based genetic programming for object recognition problems. To learn better par- tial programs, a weight parameter is introduced in each link between every two nodes in a program tree, so that a change of a weight corresponds to a change of the eect of the sub-program tree. Inside a particular generation, weight changes are learnt locally by gradient descent search, but the whole evolution process is still carried out across dieren t generations globally by the genetic beam search. This approach is examined and compared with the basic genetic programming approach without gradient descent on three object classication problems of var- ious dicult y. The results suggest that the new approach outperforms the basic approach on all problems.
... We fix the value "0" for the yellow class, and "1" for the red class. When solving a classification problem with an automatic programming algorithm, the traditional threshold value used in the case of binary classification ( two classes) is "0" [28]. However, the experimentation demonstrated that the value "0.5" is more suitable as a threshold. ...
Article
Full-text available
Nowadays more and more complex railway systems can operate efficiently only if we have tools for planning their maintenance. Train accidents are mainly caused by infrastructure problems, or more specifically by track geometry failures.In this paper, we present a support decision system for forecasting the deterioration of track geometry. Two types of defects can be identified for each railway track segment. If a defect belongs to the first type, it must be repaired immediately; otherwise, the defect can be fixed after a specific time period. For resolving this problem, we first decompose the problem into two stages: prediction and classification. Both phases contain a learning phase and a testing phase. The solution technique for both stages are based on automatic programming field and its recently proposed heuristic, called variable neighborhood programming. Our new method is tested on real world problems. The results show that the proposed approach is a good and reliable tool for the preventive maintenance planning of the railway infrastructure.
... Another type of target detection method relies on the information extracted by image decomposition or transform, rather than statistical distribution models [11]. Genetic programming was also introduced for target detection from SAR images by making use of human vision system approach to identify regions that are distinct from their surroundings [12,13]. Based on the diverse characteristics of ship and sea clutter, heterogeneity based target detection has also been done in literature [14]. ...
Article
Full-text available
We introduce the concept of predetection based on an attention mechanism to improve the efficiency of small-target detection by limiting the image region of detection. According to the characteristics of small-target detection, local contrast is taken as the only feature in predetection and a nonlinear sampling model is adopted to make the predetection adaptive to detect small targets with different area sizes. To simplify the predetection itself and decrease the false alarm probability, neighboring nodes in the sampling grid are used to generate a saliency map, and a short-term memory is adopted to accelerate the `pop-out' of targets. We discuss the fact that the proposed approach is simple enough in computational complexity. In addition, even in a cluttered background, attention can be led to targets in a satisfying few iterations, which ensures that the detection efficiency will not be decreased due to false alarms. Experimental results are presented to demonstrate the applicability of the approach.
Article
The research in this paper is to recognize and locate sand dredges in Changjiang River based on ASAR remote sensing data. Chenglingji is selected to be experimental area. Adopting advanced radar remote sensing image can satisfy the requirement to monitor and locate sand dredges in Changjiang River ignoring the effect of weather condition and night limit. The program is developed with the IDL which property is a visual language facing matrix. Some key technologies in this paper, such as filtering and noise reduction, the process of image based on morphology, edge detection, clustering, extraction of ship's shape feature and so on, are researched. Comparing with the information gathering from land monitoring station, this method can achieve satisfied accurateness.
Article
Full-text available
This paper looks at the multiple characteristics and investigations of two grinding anomalies: grinding burn and grinding chatter. A genetic programming (GP) of multiple classifications was investigated for different machining strategies and associated anomaly phenomena. Such a GP paradigm could evolve rules to provide the correlation between monitored signals and grinding phenomena. The investigation also looks at both Short-Time Fourier Transforms (STFT) and Wavelet Packet Transforms (WPT) to convert the raw acoustic emission (AE) signal into a time based frequency signal, segmented into different frequency bands. A set of encouraging results is presented.
Article
In tree-based Genetic Programming, subtrees which represent potentially useful sub-solutions can be encapsulated in order to protect them and aid their prolifer-ation throughout the population. This paper investigates implementing this as a multi-run method. A two-stage encapsulation scheme based on subtree survival and frequency is compared against Automatically Defined Functions in fixed and evolved architectures and standard Genetic Programming for solving a Parity problem.
Article
This paper wants to explore a key technique of identifying and locating the sand exploitation ships using ASAR images. Chenglingji of Yangtze River is selected as the research region. One software module was developed with the IDL (Interactive Data Language) which is a facing matrix, simple visual language. This paper researches on some key techniques based on ASAR images such as filtering and noise reduction, the process of image based on morphology, edge detection, region growth, extraction of ship's shape feature and so on. Comparing with the field monitoring results, this method can achieve 80%-90% accurateness.
Article
A major factor for an efficient design of solar energy systems is to provide accurate estimations of the solar radiation. Many of the existing studies are focused on the analysis of monthly or annual solar radiation. This is while less attention has been paid to the determination of daily solar radiation. Accordingly, the main goal of this paper is to develop a robust machine learning approach, based on genetic programming (GP), for the estimation of the daily solar radiation. The solar radiation is formulated in terms of daily air temperature, relative humidity, atmospheric pressure, wind speed, and earth temperature. A comprehensive database containing about 7000 records collected for about 20 years (1995-2014) in a nominal city in Iran is used to develop the GP model. The performance of the derived model is verified using different criteria. A multiple linear regression analysis is performed to benchmark the GP model with a classical technique. The influences of the input variables on the solar energy are evaluated through a sensitivity analysis. The proposed model has a very good prediction performance and significantly outperforms the traditional regression model.
Article
A target detection algorithm is developed based on a supervised learning technique that maximises the margin between two classes, that is, the target class and the non-target class. Specifically, the proposed target detection algorithm consists of (i) image differencing, (ii) maximum-margin classifier, and (iii) diversity combining. The image differencing is to enhance and highlight the targets so that the targets are more distinguishable from the background. The maximum-margin classifier is based on a recently developed feature weighting technique called Iterative RELIEF; the objective of the maximum-margin classifier is to achieve robustness against uncertainties and clutter. The diversity combining utilises multiple images to further improve the performance of detection, and hence it is a type of multi-pass change detection. The authors evaluate the performance of the proposed detection algorithm, using the CARABAS-II synthetic aperture radar (SAR) image data and the experimental results demonstrate superior performance of the proposed algorithm, compared to the benchmark algorithm.
Conference Paper
Automatic gesture recognition has received much attention due to its potential in various applications. In this paper, we successfully apply an evolutionary method-genetic programming (GP) to synthesize machine learned spatio-temporal descriptors for automatic gesture recognition instead of using hand-crafted descriptors. In our architecture, a set of primitive low-level 3D operators are first randomly assembled as tree-based combinations, which are further evolved generation-by-generation through the GP system, and finally a well performed combination will be selected as the best descriptor for high-level gesture recognition. To the best of our knowledge, this is the first report of using GP to evolve spatio-temporal descriptors for gesture recognition. We address this as a domain-independent optimization issue and evaluate our proposed method, respectively, on two public dynamic gesture datasets: Cambridge hand gesture dataset and Northwestern University hand gesture dataset to demonstrate its generalizability. The experimental results manifest that our GP-evolved descriptors can achieve better recognition accuracies than state-of-the-art hand-crafted techniques.
Article
In this paper, we present a new method for classification of electroencephalogram (EEG) signals using Genetic Programming (GP). The Empirical Mode Decomposition (EMD) is used to extract the features of EEG signals which served as an input for the GP. In this paper, new constructive crossover and mutation operations are also produced to improve GP. In these constructive crossover and mutation operators hill climbing search is integrated to remove the destructive nature of these operators. To improve GP, we apply constructive crossover on all the individuals which remain after reproduction. A new concept of selecting the global prime off-springs of the generation is also proposed. The constructive mutation approach is applied to poor individuals who are left after selecting globally prime off-springs. Improvement of the method is measured against classification accuracy, training time and the number of generations for EEG signal classification. As we show in the results section, the classification accuracy can be estimated to be 98.69% on the test cases, which is better than classification accuracy of Liang and coworkers method which was published in 2010.
Article
Extracting discriminative and robust features from video sequences is the first and most critical step in human action recognition. In this paper, instead of using handcrafted features, we automatically learn spatio-temporal motion features for action recognition. This is achieved via an evolutionary method, i.e., genetic programming (GP), which evolves the motion feature descriptor on a population of primitive 3D operators (e.g., 3D-Gabor and wavelet). In this way, the scale and shift invariant features can be effectively extracted from both color and optical flow sequences. We intend to learn data adaptive descriptors for different datasets with multiple layers, which makes fully use of the knowledge to mimic the physical structure of the human visual cortex for action recognition and simultaneously reduce the GP searching space to effectively accelerate the convergence of optimal solutions. In our evolutionary architecture, the average cross-validation classification error, which is calculated by an support-vector-machine classifier on the training set, is adopted as the evaluation criterion for the GP fitness function. After the entire evolution procedure finishes, the best-so-far solution selected by GP is regarded as the (near-)optimal action descriptor obtained. The GP-evolving feature extraction method is evaluated on four popular action datasets, namely KTH, HMDB51, UCF YouTube, and Hollywood2. Experimental results show that our method significantly outperforms other types of features, either hand-designed or machine-learned.
Chapter
Automatic programming is an efficient technique that has contributed to an important development in the artificial intelligence and machine learning fields. In this chapter, we introduce the technique called Variable Neighborhood Programming (VNP) that was inspired by the principle of the Variable Neighborhood Search (VNS) algorithm. VNP starts from a single solution presented by a program, and the search for a good quality global solution (program) continues by exploring different neighborhoods. The goal of our algorithm is to generate a good representative program adequate to a selected problem. VNP takes the advantages of the systematic change of neighborhood structures randomly or within a local search algorithm to diversify or intensify search through the solution space. To show its efficiency and usefulness, the VNP method is applied first for solving the symbolic regression problem (VNP-SRP) and tested and compared on usual test instances from the literature. In addition, the VNP-SRP method is tested in finding formulas for life expectancy as a function of some health care economic factors in 18 Russian districts. Finally, the VNP is implemented on prediction and classification problems and tested on real-life maintenance railway problems from the US railway system.
Article
Full-text available
The capability to generate complex geometrical features at tight tolerances and fine surface roughness is a key element in implementation of Creep Feed grinding process in specialist applications such as the aerospace manufacturing environment. Based on the analysis of 3D cutting forces this paper proposes a novel method of predicting the profile deviations of tight geometrical features generated using Creep Feed grinding. In this application, there are several grinding passes made at varying depths providing an incremental geometrical change with the last cut generating the final complex feature. With repeatable results from co-ordinate measurements both the radial and tangential forces can be gauged versus the accuracy of the ground features. The tangential force was found more sensitive to the deviation of actual cut depth from the theoretical one. However, to make a more robust prediction on the profile deviation its values were considered as a function of both force components (proportional to force: power was also included). For multi process, one machining platforms hole making was also investigated in terms of monitoring the force to ensure the mean cylinder was kept within required tolerances and minimise subsequent machining (due to these imposed accuracies this is also considered a complex feature). Genetic Programming (GP), an evolutionary programming technique, has been used to compute the prediction rules of part profile deviations based on the extracted radial and tangential force and correlated with the initial “gauging” methodology (for grinding process). GP was also used to correlate the force and flank wear (VB) for hole deviations. It was found that using this technique, complex rules can be achieved and used online to dynamically control the geometrical accuracy of ground and drilled hole features. The GP complex rules are based on the correlation between the measured forces and recorded deviation of the theoretical profile (both grinding and hole making). The mathematical rules are generated from Darwinian evolutionary strategy which provides the mapping between different output classes. GP works from crossover recombination of different rules and the best individual is evaluated in terms of the given ‘best fitness value so far’ which closes on an optimal solution. The best obtained GP terminal sets were realised in rule-based embedded coded systems which were finally implemented into a real-time Simulink simulation. This realisation gives a view of how such a control regime can be utilised within an industrial capacity. Neural Networks were used for GP decision verification ensuring less sensitivity to possible outliers giving more robustness to the integrated system.
Chapter
Stereo vision is one of the most active research areas in modern computer vision. The objective is to recover 3-D depth information from a pair of 2-D images that capture the same scene. This paper addresses the problem of dense stereo correspondence, where the goal is to determine which image pixels in both images are projections of the same 3-D point from the observed scene. The proposal in this work is to build a non-linear operator that combines three well known methods to derive a correspondence measure that allows us to retrieve a better approximation of the ground truth disparity of stereo image pair. To achieve this, the problem is posed as a search and optimization task and solved with genetic programming (GP), an evolutionary paradigm for automatic program induction. Experimental results on well known benchmark problems show that the combined correspondence measure produced by GP outperforms each standard method, based on the mean error and the percentage of bad pixels. In conclusion, this paper shows that GP can be used to build composite correspondence algorithms that exhibit a strong performance on standard tests.
Article
Full-text available
Global Navigation Satellite System (GNSS) reflected signals can be used to remotely sense the Earth’s surface, known as GNSS reflectometry (GNSS-R). The GNSS-R technique has been applied to numerous areas, such as the retrieval of wind speed, and the detection of Earth surface objects. This work proposes a new application of GNSS-R, namely to detect objects above the Earth’s surface, such as low Earth orbit (LEO) satellites. To discuss its feasibility, 14 delay Doppler maps (DDMs) are first presented which contain unusually bright reflected signals as delays shorter than the specular reflection point over the Earth’s surface. Then, seven possible causes of these anomalies are analysed, reaching the conclusion that the anomalies are likely due to the signals being reflected from objects above the Earth’s surface. Next, the positions of the objects are calculated using the delay and Doppler information, and an appropriate geometry assumption. After that, suspect satellite objects are searched in the satellite database from Union of Concerned Scientists (UCS). Finally, three objects have been found to match the delay and Doppler conditions. In the absence of other reasons for these anomalies, GNSS-R could potentially be used to detect some objects above the Earth’s surface.
Preprint
Full-text available
In this paper, the publicly available dataset of condition based maintenance of combined diesel-electric and gas (CODLAG) propulsion system for ships has been utilized to obtain symbolic expressions which could estimate gas turbine shaft torque and fuel flow using genetic programming (GP) algorithm. The entire dataset consists of 11934 samples that was divided into training and testing portions of dataset in an 80:20 ratio. The training dataset used to train the GP algorithm to obtain symbolic expressions for gas turbine shaft torque and fuel flow estimation consisted of 9548 samples. The best symbolic expressions obtained for gas turbine shaft torque and fuel flow estimation were obtained based on their $R^2$ score generated as a result of the application of the testing portion of the dataset on the aforementioned symbolic expressions. The testing portion of the dataset consisted of 2386 samples. The three best symbolic expressions obtained for gas turbine shaft torque estimation generated $R^2$ scores of 0.999201, 0.999296, and 0.999374, respectively. The three best symbolic expressions obtained for fuel flow estimation generated $R^2$ scores of 0.995495, 0.996465, and 0.996487, respectively.
Article
Full-text available
Target detection is the front-end stage in any automatic target recognition system for synthetic aperture radar (SAR) imagery (SAR-ATR). The efficacy of the detector directly impacts the succeeding stages in the SAR-ATR processing chain. There are numerous methods reported in the literature for implementing the detector. We offer an umbrella under which the various research activities in the field are broadly probed and taxonomized. First, a taxonomy for the various detection methods is proposed. Second, the underlying assumptions for different implementation strategies are overviewed. Third, a tabular comparison between careful selections of representative examples is introduced. Finally, a novel discussion is presented, wherein the issues covered include suitability of SAR data models, understanding the multiplicative SAR data models, and two unique perspectives on constant false alarm rate (CFAR) detection: signal processing and pattern recognition. From a signal processing perspective, CFAR is shown to be a finite impulse response band-pass filter. From a statistical pattern recognition perspective, CFAR is shown to be a suboptimal one-class classifier: a Euclidean distance classifier and a quadratic discriminant with a missing term for one-parameter and two-parameter CFAR, respectively. We make a contribution toward enabling an objective design and implementation for target detection in SAR imagery.
Genetic evolution of automatic ship detection in SAR imagery
  • Howard D Brankin
  • Ruc ), Dera / Cis ( Sec-M ) / Proj0 D Howard
Roberts S, Howard D, Brankin R. May 1998, Genetic evolution of automatic ship detection in SAR imagery, (UC), DERA/CIS(SEC-M)/PROJ/154/ASD/1.0. D. Howard et al. / Advances in Engineering Software 30 (1999) 303–311
Automatic ship detection in 10m resolution SAR imagery: An Investigation
  • R Brankin
Brankin R. May 1998, Automatic ship detection in 10 m resolution SAR imagery: An Investigation, (UC), DERA/CIS(SEC-M)/PROJ/ 154/ASD10/1.0.
Application of genetic programming to target detection and CFD problems
  • D Howard
Genetic evolution of automatic ship detection in SAR imagery
  • S Roberts
  • D Howard
  • R Brankin