Conference Paper

Radar high resolution range profile recognition via Dual-SVDD classifier

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

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.

... Li et al. studied the problem that hyperspectral images (HSIs) contain large amounts of noise, which hampers the target recognition process. They introduced a method to reconstruct the HSI with noise reduction and contrast enhancement strategies using a matting model [18]. Dung et al. proposed a method to recognize the radar targets from radar range profiles with the help of neural networks [16]. ...
... The total number of the samples in test set is also denoted by M . The recognition rate R rec is then obtained by (18). The above steps are shown in Fig.10. ...
... If we denote N suc as the number of data with Label 1, and the total number of the samples in the test set is M . The recognition rate R rec is obtained according to (18). ...
Article
Full-text available
Chaff cloud is widely used in target covering. In this paper, a jamming performance estimation strategy is presented for the elliptical and cubic chaff clouds. The range profiles of the composite models of the plane with the chaff cloud are experimentally accessed in an anechoic chamber using a stepped-frequency waveform measurement system. Then, dimensions, characteristic length, and distributing entropy of the equivalent scattering centers of the plane and the jammer are extracted from the measured range profiles, respectively. Based on these features, support vector machines, naïve Bayesian classification, and decision tree classification methods are employed to determine the recognition rate of the target under the chaff jamming, and then determine the jamming efficiency of the two kinds of chaff clouds. Results show that two types of chaff clouds are able to reduce the target recognition rate. Besides, when the radar uses the Euclidean distance and decision tree classification methods to complete the recognition task, a good jamming efficiency can be achieved by releasing the elliptical or the cubic chaff cloud. Meanwhile, a low level of jamming performance outcomes using the support vector machines classification method. It is also found that the chaff cloud released from the side of the plane achieves a better jamming performance compared with the other releasing locations. Finally, the influence of environmental noise on the recognition rate of the plane is studied. Estimating results show that the recognition rates decrease with the increasing of the power level of the noise. When the transmitting power level of the noise reaches 1.2 times the reflected power level of the target, it is quite hard to estimate the jamming performance of two kinds of chaff clouds.
... T HE automatic target recognition (ATR) technology is an important subject in the radar surveillance field, which is able to automatically determine the attributes and categories of targets after feature extraction [1]. This technology is mainly applicated on top of radar echo signals, such as 1-D high-resolution range profile (HRRP) [2], [3], [4], [5], [6], [7], [8], [9] and 2-D synthetic aperture radar (SAR) images [10], [11], [12], [13], Manuscript [14], [15], [16], [17], [18], [19], [20]. Compared with HRRP, SAR images provide the target scattering information along with both the range and cross range, thus improving the recognition performance dramatically. ...
Article
Full-text available
For the automatic target recognition (ATR) based on synthetic aperture radar (SAR) images, enough training data are required to effectively characterize target features and obtain good recognition performance. However, in practical applications, it is difficult to collect sufficient training data. To tackle the limitation, a novel end-to-end expansion method, called conditional Wasserstein deep convolutional generative adversarial network with gradient penalty (CWDCGAN), is proposed to achieve SAR image expansion with specified category. To be specific, the CWDCGAN innovatively designed a generative adversarial network architecture based on convolutional and deconvolution networks to improve the quality of generated images. At the same time, conditional information is introduced to control the categories of generated images, and Wasserstein distance and gradient penalty are used to modify the loss function, which makes the network training more stable. Besides, feature extraction and classifier design in a typical ATR system often rely heavily on subjective expert knowledge, which seriously affects its generalization performance. Therefore, a joint recognition method of Resnet18 and support vector machine (Renset18-SVM) is adopted to improve the generalization capacity and the recognition performance. Experimental results with public measured data show that the CWDCGAN can generate higher quality SAR images, and by feeding expanded data to Renset18-SVM, the recognition accuracy is improved under different proportions of training samples.
Chapter
The frequent occurrence of “black flying” and illegal invasion make it necessary to identify radar target tracks. Existing research on the radar flight target mainly focuses on the target identification. Existing research on the recognition of target track types ignores the impact of outliers in the track data and the small number of labeled data on the target track type recognition performance. In this paper, we propose a target Track Recognition algorithm based on a Semi-Supervised approach for few-labeled radar target track data with outliers (TRSS). We first remove the outliers and fill in the missing track points caused by the outlier removal. We then extract five basic flight features (BFFs) and an advanced flight feature (AFF) from the track data to obtain a strong recognition flight feature combination (SRFFC). Finally, a semi-supervised generative adversarial network (SSGAN) model is constructed to classify flight target tracks using SRFFC as the input. Simulation results show that the proposed algorithm TRSS can effectively improve the accuracy, precision, and recall performance of target track type recognition.
Article
Full-text available
In this letter, a reliable, simple, and intuitive approach for hyperspectral imagery (HSI) anomaly detection (AD) is presented. This method, namely, the global iterative principal component analysis (PCA) reconstruction-error-based anomaly detector (GIPREBAD), examines AD by computing errors (residuals) associated with reconstructing the original image using PCA projections. PCA is a linear transformation and feature extraction process commonly used in HSI and frequently appears in operation prior to any AD task. PCA features represent a projection of the original data into lower-dimensional subspace. An iterative approach is used to mitigate outlier influence on background covariance estimates. GIPREBAD results are provided using receiver-operating-characteristic curves for HSI from the hyperspectral digital imagery collection experiment. Results are compared against the Reed-Xiaoli (RX) algorithm, the linear RX (LRX) algorithm, and the support vector data description (SVDD) algorithm. The results show that the proposed GIPREBAD method performs favorably compared with RX, LRX, and SVDD and is both intuitively and computationally simpler than either RX or SVDD.
Article
In this paper, a new multi-class classification framework incorporating geometric data relationships described in both intrinsic and penalty graphs in multi-class Support Vector Machine is proposed. Direct solutions are derived for the proposed optimization problem in both the input and arbitrary-dimensional Hilbert spaces for linear and non-linear multi-class classification, respectively. In addition, it is shown that the proposed approach constitutes a general framework for SVM-based multi-class classification exploiting geometric data relationships, which includes several SVM-based classification schemes as special cases. The power of the proposed approach is demonstrated in the problem of human action recognition in unconstrained environments, as well as in facial image and standard classification problems. Experiments indicate that by exploiting geometric data relationships described in both intrinsic and penalty graphs the SVM classification performance can be enhanced.
Article
Discriminant locality preserving projections based on maximum margin criterion (DLPP/MMC) is a useful feature extraction method since it has shown good performances in pattern recognition. The conventional DLPP/MMC, however, is not robust to noises and outliers since its objective function is based on L2-norm. In this paper, we propose a novel L1-norm and maximum margin criterion based discriminant locality preserving projections via trace Lasso (DLPP/MMC-L1TL). L1-norm rather than L2-norm is used in the formulation of DLPP/MMC-L1TL, which makes it be robust to noises and outliers. Besides, in order to improve the performance of DLPP/MMC-L1TL further, we use trace Lasso to regularize the basis vectors. Trace Lasso, which can balance L1-norm and L2-norm and consider sparsity and correlation of data simultaneously, is a recently proposed norm. An iterative procedure for solving DLPP/MMC-L1TL is also proposed in this paper. The experiment results on some data sets demonstrate the effectiveness of DLPP/MMC-L1TL.
Article
Target rejection using high resolution range profile (HRRP) means reject the unknown target which not belongs to any class of the database while maintaining a high accuracy in recognition. In radar automatic target recognition (RATR), lacking complete training database is a main problem in unknown target rejection. In this paper, a novel method which generates training samples based on Gamma model is proposed. The paper firstly analyzed the statistical characteristics of echoes of range cells. Then it derived the parameters estimation process of Gamma model. Compared with uniform distributed model, the proposed method can achieve better rejection performance. The experimental results of measured data demonstrate its effectiveness.
Article
Since the signal-to-noise ratio (SNR) directly relates to the distance between the target and the radar for a given noise power and radar power, the noise robustness of a recognition algorithm is very important to increase the recognition distance between the target and the radar in the real application. In this paper, a novel noise-robust recognition method for high-resolution range profile (HRRP) data is proposed to enhance its recognition performance under the test condition of low SNR. The target dominant scatterers are first extracted based on the scattering center model of complex HRRP data via the orthogonal matching pursuit algorithm. Then, a scatterer matching recognition algorithm based on Hausdorff distance is developed with the magnitudes and locations of extracted dominant scatterers used as the feature patterns. Here, the noise reduction is accomplished based on the sparse distribution property of dominant scattering centers in a target. Experimental results on the synthetic and measured HRRP data demonstrate that the proposed method can improve the recognition performance under the relatively low SNR condition for both orthogonal and superresolution representations of scattering center model.
Article
Minutiae set is one of the prevalent features used to represent a fingerprint. Many minutiae protection schemes have been proposed in recent literature, but only a few have demonstrated successful conversion from minutiae set to fixed-length bit-string. In this paper, we develop a fixed-length binary cancellable fingerprint template generation scheme based on a minutia descriptor known as the multi-line code (MLC). While retaining the core of MLC algorithm, we transform the unordered and variable-size MLC template into an ordered and fixed-length bit-string using kernel principal components analysis (KPCA) and state-of-the-art binarization techniques. The construction of a proper kernel suited for the scenario was validated using Mercer's Theorem. Evaluation of the proposed scheme was performed over several FVC datasets and the best equal-error rate (EER) obtained for the final bit-string is 1.61%. In addition, extensive analysis was done to justify the non-invertibility and revocability property of the cancellable template.
Article
Sparse representation models based on dictionary learning have led to interesting results in signal restoration and target recognition. However, due to the redundancy defined by overcomplete dictionary atoms in new space, finding sparse representations from inaccurate measurements may cause uncertainty and ambiguity. Especially for radar automatic target recognition using high-resolution range profiles (HRRP), the target-aspect sensitivity, amplitude fluctuation and outliers in HRRPs could result in mismatch among the sparse representations of the same class and thus deteriorate the recognition performance. This article proposes a novel stable dictionary learning method to deal with this problem and improve the pattern recognition performance. The proposed method relies on the constraints that the sparse representations of adjacent HRRPs without scatterers' motion through range cells should have the same support and lower variance. The structured sparse regularisation is then used to automatically select the optimal dictionary basis vectors for stable sparse coding. Experiments based on the measured HRRP dataset validate the performance of the proposed method. Moreover, encouraging results are reported with small training data size and under different signal-to-noise ratio conditions.
Article
Support vector data description (SVDD) is a data description method which gives the target data set a hypersphere-shaped description and can be used for one-class classification or outlier detection. To further improve its performance, a novel SVDD called SVDD+ which introduces the privileged information to the traditional SVDD is proposed. This privileged information, which is ignored by the classical SVDD but often exists in human learning, will optimise the training phase by constructing a set of correcting functions. The performance of SVDD+ on data sets from the UCI machine learning repository and radar emitter recognition is demonstrated. The experimental results indicate the validity and advantage of this method.
Article
One of the radar high-resolution range profile (HRRP) recognition issues is the target-aspect sensitivity. Both theoretical analysis and real-world data show that the HRRP shows a high correlation only in a very small aspect region. To overcome this problem, in traditional methods, the scattering centre model and averaged range profile are utilised. In this study, the authors present a graph-based semi-supervised method, called geodesic weighted sparse representation (GWSR), to overcome the target-aspect sensitivity problem. It is assumed that HRRP from different targets is located on different manifolds and the correlation information is utilised to separate these manifolds. In GWSR, the geodesic distance is calculated firstly and then the labelled HRRP is reconstructed by the geodesic weight. The nonlinear structure of HRRP can be transformed into a linear one through the reconstruction process. Then, the unlabelled HRRP is sparsely reconstructed and the sparse reconstruction weight can be utilised to estimate the label of the unlabelled HRRP from the given labels. Experiments on three kinds of ground target HRRPs with different backgrounds demonstrate the effectiveness of the authors' method.
Article
Novelty detection is a problem with a large number of relevant applications. For some applications, the main interest is the prevention or detection of undesired states. In some cases, these undesired states are not known in advance; in others, such as machine monitoring, for instance, machine breakdown may be very rare and examples of this event may be unavailable. In such cases, the most widely accepted approach is to model the normal behavior of the system in order to subsequently detect unknown events. This is the basic concept of One-Class Classification (OCC).
Conference Paper
One-class classification is an important problem with applications in several different areas such as outlier detection and machine monitoring. In this paper we propose a novel method for one-class classification which also implements prototype reduction. The main feature of the proposed method is to analyze every limit of all the feature dimensions to find the true border which describes the normal class. To this end, the proposed method simulates the novelty class by creating artificial prototypes outside the normal description. The method is able to describe data distributions with complex shapes. Aiming to assess the proposed method, we carried out experiments with synthetic and real datasets to compare it with the Support Vector Domain Description (SVDD), kMeansDD, ParzenDD and kNNDD methods. The experimental results show that our one-class classification approach outperformed the other methods in terms of the area under the receiver operating characteristic (ROC) curve in three out of six data sets. The results also show that the proposed method remarkably outperformed the SVDD regarding training time and reduction of prototypes.
Article
Kernel adaptive filters (KAF) are a class of powerful nonlinear filters developed in Reproducing Kernel Hilbert Space (RKHS). The Gaussian kernel is usually the default kernel in KAF algorithms, but selecting the proper kernel size (bandwidth) is still an open important issue especially for learning with small sample sizes. In previous research, the kernel size was set manually or estimated in advance by Silvermans rule based on the sample distribution. This study aims to develop an online technique for optimizing the kernel size of the kernel least mean square (KLMS) algorithm. A sequential optimization strategy is proposed, and a new algorithm is developed, in which the filter weights and the kernel size are both sequentially updated by stochastic gradient algorithms that minimize the mean square error (MSE). Theoretical results on convergence are also presented. The excellent performance of the new algorithm is confirmed by simulations on static function estimation and short term chaotic time series prediction.
Article
One-class classification (OCC) has received a lot of attention because of its usefulness in the absence of statistically-representative non-target data. In this situation, the objective of OCC is to find the optimal description of the target data in order to better identify outlier or non-target data. An example of OCC, support vector data description (SVDD) is widely used for its flexible description boundaries without the need to make assumptions regarding data distribution. By mapping the target dataset into high-dimensional space, SVDD finds the spherical description boundary for the target data. In this process, SVDD considers only the kernel-based distance between each data point and the spherical description, not the density distribution of the data. Therefore, it may happen that data points in high-density regions are not included in the description, decreasing classification performance. To solve this problem, we propose a new SVDD introducing the notion of density weight, which is the relative density of each data point based on the density distribution of the target data using the k-nearest neighbor (k-NN) approach. Incorporating the new weight into the search for an optimal description using SVDD, this new method prioritizes data points in high-density regions, and eventually the optimal description shifts to these regions. We demonstrate the improved performance of the new SVDD by using various datasets from the UCI repository.
Article
This paper shows the use of a data domain description method, inspired by the support vector machine by Vapnik, called the support vector domain description (SVDD). This data description can be used for novelty or outlier detection. A spherically shaped decision boundary around a set of objects is constructed by a set of support vectors describing the sphere boundary. It has the possibility of transforming the data to new feature spaces without much extra computational cost. By using the transformed data, this SVDD can obtain more flexible and more accurate data descriptions. The error of the first kind, the fraction of the training objects which will be rejected, can be estimated immediately from the description without the use of an independent test set, which makes this method data efficient. The support vector domain description is compared with other outlier detection methods on real data.