Abstract: In the present study, a new algorithm for automatic target detection (ATR) in synthetic aperture radar (SAR) images has been proposed. First, moving and stationary target acquisition and recognition image chips have been segmented and then passed to a number of preprocessing stages such as histogram equalisation, position and size normalisation. Second, the feature extraction based on Zernike moments (ZMs) having linear transformation invariance properties and robustness in the presence of the noise has been introduced for the first time. Third, a genetic algorithm-based feature selection and a support vector machine classifier have been presented to select the optimal feature subset of ZMs for decreasing the computational complexity. Experimental results demonstrate the efficiency of the proposed approach in target recognition of SAR imagery. The authors obtained results show that just a small amount of ZMs features is sufficient to achieve the recognition rates that rival other established methods, and so ZMs features can be regarded as a powerful discriminatory feature for automatic target recognition applications relevant to SAR imagery. Furthermore, it can be observed that the classifier performs fairly well until the signal-to-noise ratio falls beneath 5 dB for noisy images.

Abstract: This paper presents a forward approach to establish parametric scattering center models for known complex radar targets. In this approach, an automatic technique based on ray tracing and clustering is first developed to extract scattering centers directly from the computer-aided design (CAD) model of the targets. Following this, a set of forward methods is developed to determine the physically relevant parameters of two-dimension (2-D) attributed scatterers, such as type, amplitude, position and length. Finally, this approach is validated through the parametric model establishment of two complex targets and good agreement has been demonstrated between the reconstructed and actual radar characteristics. Different from the familiar inverse extraction approaches, the proposed approach provides a new forward way of constructing radar targets' feature database based on 2-D parametric scattering center model, which will ultimately facilitate the feature matching in the synthetic aperture radar (SAR) automatic target recognition (ATR) system.

Abstract: This letter proposes a novel algorithm for focusing moving targets with fast cross-track velocities and estimating their motion parameters with single-antenna synthetic aperture radar. The spectrum of a fast-moving target contains the Doppler ambiguity and the third-order phase error. Hough transform is utilized to estimate the slope of the range walk trajectory, from which the cross-track velocity is obtained and the Doppler ambiguity problem is solved. Then, polynomial Fourier transform is adopted to estimate the second- and third-order Doppler parameters of the moving target. By estimating and compensating for the third-order phase error, the imaging resolution of the fast-moving target is improved and the motion parameters, including the cross-track acceleration, are correctly estimated. Both simulated and real data processing results are provided to demonstrate the effectiveness of the proposed algorithm.

Abstract: Feature extraction from high-dimensional synthetic aperture radar images is one of the key steps for SAR automatic target recognition. In this paper, we propose a new approach to SAR image feature extraction that is named neighborhood geometric center scaling embedding, which is based on manifold learning theory. In our framework, neighborhood geometric center scaling is introduced to construct neighborhood relationships. The samples are endowed with clear clustering directions in dimensionality reduction, and the classification is better conducted in the feature space than in the original space. Moreover, by introducing neighborhood geometric center scaling, the influence of neighbor parameters on recognition performance is reduced effectively. The experiment based on the Moving and Stationary Target Acquisition and Recognition database shows that the proposed method has better recognition performance and higher stability than other methods.

Abstract: In the work presented here we propose a novel bicoherence-based method for the classification of aerial radar targets in automatic target recognition (ATR) systems. The possibility of classifying aerial targets using the micro-Doppler contributions caused by a jet engine or the rotor of a helicopter is studied. The method is based on classification features computed in the form of bicoherence estimates, as well as cepstral coefficients extracted from the micro-Doppler contribution contained in radar returns. The performance of the classification method developed is compared with the performance of common methods using high-resolution radar range profiles (HRRPs). Correct classification probability rates are computed for three different types of aerial targets. The benefits achieved by using bicoherence-based classification features are demonstrated and discussed.

Abstract: A novel inverse synthetic aperture radar imaging algorithm is proposed for application when the maneuverability of an uncooperative target is not too severe, and the Doppler variation of the subechoes from the scatterers can be considered a first-order polynomial. Based on a modified keystone transform, the proposed algorithm can simultaneously transform all multicomponent linear frequency modulation subechoes into multicomponent single-frequency signals. Hence, the fast Fourier transform can be used for cross-range imaging. Simulation results also proved that the proposed algorithm is much more effective and computationally efficient than the previously reported range-instantaneous Doppler algorithms such as the Radon-Wigner transform algorithm.

Abstract: Feature extraction is a key step in synthetic-aperture-radar automatic target recognition. In this letter, we propose a novel feature extraction method named sample discriminant analysis (SDA) that is based on the manifold learning theory. The method directly extracts features from 2-D image matrices rather than vectors. Furthermore, SDA preserves the neighborhood information of the original data in dimension reduction. It also makes within-class samples closer and makes between-class samples father away in a low-dimensional space. Meanwhile, a sample discriminant coefficient is employed in the method to give each sample a weight related to its location and similarity to neighboring samples. Thus, the discriminative ability of the method is improved. Experimental results based on the moving and stationary target acquisition and recognition database show that the proposed method can improve recognition performance.

Abstract: We study automatic target recognition (ATR) in infrared (IR) imagery by applying two recent computer vision techniques, Histogram of Oriented Gradients (HOG) and Bag-of-Words (BoW). We propose the idea of dense HOG features which are extracted from a set of high-overlapped local patches in a small IR chip and we apply a vocabulary tree that is learned from a set of training images to support efficient and scalable BoW-based ATR. We develop a relevance grouping of vocabulary (RGV) technique to improve the ATR performance by additional voting from grouped visual words. Different from traditional word grouping techniques, RGV groups visual words of the same dominant class to enhance the voting confidence in BoW-based classification. The proposed ATR algorithm is evaluated against recent sparse representation-based classification (SRC) approaches that reportedly outperform traditional methods. Experimental results on the COMANCHE IR dataset demonstrate the advantages of the newly proposed algorithm (BoW-RGV) over the recent SRC approaches.

Abstract: Due to the characteristic of the synthetic aperture radar (SAR) image's sensitivity to the target aspect angles, a multiple sparse representation (MSR) method for SAR target configuration recognition is proposed. Making use of the prior information, dictionaries are constructed by using the samples of each configuration to better capture the detail information of the SAR images. The advantage of MSR over sparse representation for detail feature extraction is analyzed. Moreover, to achieve better recognition results, the Dempster-Shafer fusion is carried out to get comprehensive description of the target for configuration recognition. Two mass functions are constructed based on MSR and the sample statistical property. The combined mass function has the advantages of both the detail and global features of the target. Experiments on the moving and stationary target acquisition and recognition data sets validate the effectiveness and superiority of the proposed algorithm.

Abstract: Airborne radar tracking in moving ground vehicle scenarios is impacted by sensor, target, and environmental dynamics. Moving targets can be characterized by 1-D High Range Resolution (HRR) Radar profiles with sufficient Signal-to-Noise Ratio (SNR). The amplitude feature information for each range bin of the HRR profile is used to discern one target from another to help maintain track or to identify a vehicle. Typical radar clutter suppression algorithms developed for processing moving ground target data not only remove the surrounding clutter, but a portion of the target signature. Enhanced clutter suppression can be achieved using a Multi-channel Signal Subspace (MSS) algorithm, which preserves target features. In this paper, we (1) exploit extra feature information from enhanced clutter suppression for Automatic Target Recognition (ATR), (2) present a Decision-Level Fusion (DLF) gain comparison using Displaced Phase Center Antenna (DPCA) and MSS clutter suppressed HRR data; and (3) develop a confusion-matrix identity fusion result for Simultaneous Tracking and Identification (STID). The results show that more channels forMSS increase identification over DPCA, result in a slightly noisier clutter suppressed image, and preserve more target features after clutter cancellation. The paper contributions include extending a two-channel MSS clutter cancellation technique to three channels, verifying the MSS is superior to the DPCA technique for target identification, and a comparison of these techniques in a novel multi-look confusion matrix decision-level fusion process.

Abstract: The capability to exploit multiple sources of information is of fundamental importance in a battlefield scenario. Information obtained from different sources, and separated in space and time, provide the opportunity to exploit diversities in order to mitigate uncertainty. For the specific challenge of Automatic Target Recognition (ATR) from radar platforms, both channel (e.g. polarization) and spatial diversity can provide useful information for such a specific and critical task. In this paper the use of pseudo-Zernike moments applied to multi-channel multi-pass data is presented exploiting diversities and invariant properties leading to high confidence ATR, small computational complexity and data transfer requirements. The effectiveness of the proposed approach, in different configurations and data source availability is demonstrated using real data.

Abstract: Air-traffic controllers cannot identify air vehicles flying with a defective or nonexistent transponder. Primary radar does not help, because it cannot classify air vehicles from echoes. Passive radar offers a potential solution, the main difficulty lying in the analysis of the data. We present a novel, automatic object-target recognition system capable of classifying air vehicles observed with passive radar. We describe the testbed we deployed near Orly Airport, and the data it provided over 10 days. A preliminary experiment involving three broad classes, and using a grid of parallel classifiers, gives a promising correct recognition rate of 71%.

Abstract: In this paper, we study the problem of target recognition based on RCS observation sequence. First, we use discrete wavelet transform of RCS observation sequence and extract five valid statistical features in the transform domain. Second, we establish the model and apply the method of set-valued identification to determine the relationship between these five characteristics and the target, thus giving recognition criteria. Finally, simulate the situation that the true target is frustum and the false target with normally distributed RCS. Comparing the results to the ones of fuzzy classification method and evidential reasoning method, the result shows that the set-valued identification method has a higher recognition rate.

Abstract: This paper presents a novel feature extraction algorithm based on the local binary features for automatic target recognition (ATR) in infrared imagery. Since the inception of the local binary pattern (LBP) and local ternary pattern (LTP) features, many extensions have been proposed to improve their robustness and performance in a variety of applications. However, most attentions were paid to improve local feature extraction with little consideration on the incorporation of global or regional information. In this work, we propose a new concave-convex partition (CCP) strategy to improve LBP and LTP by dividing local features into two distinct groups, i.e., concave and convex, according to the contrast between local and global intensities. Then two separate histograms built from the two categories are concatenated together to form a new LBP/LTP code that is expected to better reflect both global and local information. Experimental results on standard texture images demonstrate the improved discriminability of the proposed features and those on infrared imagery further show that the proposed features can achieve competitive ATR results compared with state-of-the-art methods.

Abstract: A novel algorithm for detection and classification of 3D objects in high-resolution microwave radar images for security screening purposes is proposed. To detect and automatically recognize concealed objects a computer vision approach is applied. A three-step detection procedure based on SIFT, Histogram thresholding and matched filtering is developed for 3D image processing. This novel procedure has proved to be capable of detecting and classifying a set of objects which are likely to be carried by an individual at an airport security checkpoint.

Abstract: In order to correctly identify a remote target, an efficient and robust target signature identification technique is required. Radar target identification based on complex natural resonances (CNRs) has drawn the interest of many researchers following the development of the singularity expansion method (SEM). As evident from the literature, statistical techniques such as the generalized likelihood ratio test (GLRT) have produced a better identification result, in the presence of noise, compared to some other SEM-based identification methods such as the extinction pulse (E-pulse) technique. However, one of the issues related to a resonance based target classifier is that it requires the commencement of the late time period for the unknown target response to be determined accurately in order to avoid false alarms during target classification process. For automatic target recognition (ATR) applications, usually such information is not known a priori. In view of this problem, a modified GLRT technique that utilizes time-frequency analysis is presented in this paper. The improved GLRT method does not require prior knowledge of the beginning of the late time period for the transient response of the unknown target. Simulation results using various targets show that our method is comparable to the original GLRT technique when the commencement of the late time period for the unknown target response is correctly determined and outperforming the original GLRT technique when the commencement of the late time period for the unknown target response is incorrectly determined.

Abstract: A method for target classification in synthetic aperture radar (SAR) images is pro- posed. The samples are first mapped into a high-dimensional feature space in which samples from the same class are assumed to span a linear subspace. Then, any new sample can be uniquely represented by the training samples within given constraint. The conventional methods suggest searching the sparest representations with l1-norm (or l0) minimization constraint. However, these methods are computationally expensive due to optimizing nondifferential objec- tive function. To improve the performance while reducing the computational consumption, a simple yet effective classification scheme called kernel linear representation (KLR) is presented. Different from the previous works, KLR limits the feasible set of representations with a much weaker constraint, l2-norm minimization. Since, KLR can be solved in closed form there is no need to perform the l1-minimization, and hence the calculation burden has been lessened. Meanwhile, the classification accuracy has been improved due to the relaxation of the constraint. Extensive experiments on a real SAR dataset demonstrate that the proposed method outperforms the kernel sparse models as well as the previous works performed on SAR target recognition. © The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original pub- lication, including its DOI. (DOI: 10.1117/1.JRS.8.083613)

Abstract: The feasibility of using Keystone transform based range migration compensation for range-Doppler processing in ultra-wideband (UWB) moving target indication (MTI) radar is investigated. The conventional Fourier transform based approach is outlined, and the challenge of using the conventional method in UWB radar for slow-moving extended target (human) is addressed. Keystone transform is applied to both simulated and experimental human target data. It is shown that Keystone transform based approach outperforms the conventional approach in terms of better Doppler resolution and less range-Doppler image blurring. The result also demonstrates that hypothesis-testing based approach and Keystone Transform based approach have similar performance, but the latter is more suitable for real-time implementation due to its smaller computation burden.

Abstract: Advanced correlation filters (CFs) were introduced over three decades ago to offer distortion-tolerant object recognition and are used in applications such as automatic target recognition (ATR) and biometric recognition. Some of the advances in CF design include minimum average correlation energy (MACE) filters that produce sharp correlations and offer excellent discrimination, optimal tradeoff synthetic discriminant function (OTSDF) filters that allow the filter designer to control the tradeoff between peak sharpness and noise tolerance, maximum average correlation height (MACH) filter that removes correlation peak constraints to reduce filter design complexity and quadratic correlation filters (QCFs) that extend the linear CFs to include second-order nonlinearity. In this paper, we summarize two recent major advances in CF design. First is the introduction of maximum margin correlation filters (MMCFs) that combine the excellent localization properties of CFs with the very good generalization abilities of support vector machines (SVMs). Second is the introduction of zero-aliasing correlation filters (ZACFs) that eliminate the aliasing in CF design due to the circular correlation caused by the use of discrete Fourier transforms (DFTs).

Abstract: We propose a new integrated target tracking, recognition and segmentation algorithm, called ATR-Seg, for infrared imagery. ATR-Seg is formulated in a probabilistic shape-aware level set framework that incorporates a joint view-identity manifold (JVIM) for target shape modeling. As a shape generative model, JVIM features a unified manifold structure in the latent space that is embedded with one view-independent identity manifold and infinite identity-dependent view manifolds. In the ATR-Seg algorithm, the ATR problem formulated as a sequential level-set optimization process over the latent space of JVIM, so that tracking and recognition can be jointly optimized via implicit shape matching where target segmentation is achieved as a by-product without any pre-processing or feature extraction. Experimental results on the recently released SENSIAC ATR database demonstrate the advantages and effectiveness of ATR-Seg over two recent ATR algorithms that involve explicit shape matching.

Abstract: Multiple Input Multiple Output sonar systems offer new perspectives for target detection and underwater surveillance. In this paper we present an unified formulation for sonar MIMO systems and study their properties in terms of target recognition and imaging. Here we are interested in large MIMO systems. The multiplication of the number of transmitters and receivers non only provides a greater variety in term of target view angles but provides also in a single shot meaningful statistics on the target itself. We demonstrate that using large MIMO sonar systems and with a single shot it is possible to perform automatic target recognition and also to achieve super-resolution imaging. Assuming the view independence between the MIMO pairs the speckle can be solved and individual scatterers within one resolution cell decorelate. A realistic 3D MIMO sonar simulator is also presented. The output of this simulator will demonstrate the theoretical results.

Abstract: This paper offers a compacted mechanism to carry out the performance evaluation work for an automatic target recognition (ATR) system: (a) a standard description of the ATR system's output is suggested, a quantity to indicate the operating condition is presented based on the principle of feature extraction in pattern recognition, and a series of indexes to assess the output in different aspects are developed with the application of statistics; (b) performance of the ATR system is interpreted by a quality factor based on knowledge of engineering mathematics; (c) through a novel utility called "context-probability" estimation proposed based on probability, performance prediction for an ATR system is realized. The simulation result shows that the performance of an ATR system can be accounted for and forecasted by the above-mentioned measures. Compared to existing technologies, the novel method can offer more objective performance conclusions for an ATR system. These conclusions may be helpful in knowing the practical capability of the tested ATR system. At the same time, the generalization performance of the proposed method is good.

Abstract: This paper describes the use of local sea-floor characteristics to train a neural network to remove false alarms from an Automatic Target Recognition (ATR) algorithm. We demonstrate that this reduces the Probability of False Alarm (PFA) in difficult areas without impacting the Probability of Detection (PD) in flat areas. The sea-floor characteristics are calculated from the texture and appearance of clutter on the seafloor. Textural characteristics are extracted using a Dual Tree Wavelet (DTW) transform. Highlight and shadow regions are segmented using Markov Random Field (MRF) and graph cuts. Clutter density and height are calculated from the segmented image. The method is tested by training a neural network to filter the detections from a Haar cascade ATR algorithm. The neural network is trained on the ATR response and the seafloor characteristics. On Synthetic Aperture Sonar (SAS) data we report an average reduction of 50% in the false alarm rate over that of the ATR algorithm. The processing time for an 8000×3000 pixel image is approximately 1 second.

Abstract: The problems of multi-threshold image segmentation remain great challenges for image compression, target recognition and computer vision. However, most of them are time-consuming. This paper proposes a cooperative honey bee mating-based algorithm (CHBMA) for image segmentation to save computation time while conquer the curse of dimensionality. CHBMA, based on honey bee mating algorithms (HBMA) and the cooperative learning, greatly enhances the search capability of the algorithm. Moreover, we adopt a new population initialization strategy to make the search more efficient, according to the characters of multilevel thresholding in an image arranged from a low gray level to a high one. Extensive experiments have shown that CHBMA can deliver more effective and efficient results to be applied in complex image processing such as automatic target recognition, compared with state-of-the-art population-based thresholding methods.

Abstract: Automatic Target Recognition (ATR) is one of the most challenging areas of the modern radar signal processing field. In this paper a recognition algorithm for full-polarimetric SAR images, that is robust with respect to rotations and target roll, is presented. It is based on the use of the pseudo-Zernike moments and the Krogager decomposition components, and exploits multiple sources of information such as polarization and spatial diversity. The effectiveness of the proposed approach has been demonstrated with real full polarimetric SAR data.

Abstract: Summary form only given. Inverse Synthetic Aperture Radar (ISAR) is a technique used for reconstructing radar images of targets. Modern highresolution radars implicitly offer the system requirements needed for implementing ISAR imaging. ISAR images are obtained by means of a signal processing that can be enabled both on and off-line. Automatic Target Recognition (ATR) systems are often based on the use of radar images because they provide a 2D e.m. map of the target reflectivity. Therefore, classification features that contain spatial information can be extracted and used to increase the performance of classifiers. This tutorial aims at providing an introduction to ISAR. The lecture is divided in three parts: the first part deals with principles of ISAR, the second part concerns ISAR processing and the third part focuses on advanced ISAR systems, such as bistatic, passive and multistatic ISAR systems. The ISAR system is introduced by defining the radar-target geometry and by considering simple radar concepts. The derivation of the ISAR processor is obtained by defining the signal model and by interpreting it in the Fourier domain. Differences between ISAR and SAR are also highlighted in order to better understand ISAR concepts.

Abstract: The microwave imaging technique, especially for synthetic aperture radar (SAR), has significant advantages in providing high-resolution complex target images, even in darkness or adverse weather conditions. Nevertheless, it is still difficult for human operators to identify targets on SAR images because they are generated using radio signals with wavelengths at the order of cm. To deal with this, various approaches for efficient automatic target recognition (ATR), based on neural networks or support vector machines (SVM), have been developed. Previously we proposed a promising ATR method using a supervised self-organizing map (SOM), where a binarized SAR image is accurately classified by exploiting the unified distance matrix (U-matrix) metric. Although this method enhances ATR performance considerably, even with SAR images heavily contaminated by random noise, the calculation burden is enormous under expansions of scale and then cannot maintain the ATR performance, especially in cases with azimuth angle variations. In this letter, we propose a constrained learning scheme for generating the SOM and introduce the A-star algorithm to handle SOM scale expansion. Experimental investigations demonstrate the effectiveness of our proposed method.

Abstract: Automatic Target Recognition (ATR) methods have been successfully applied to detect possible objects or regions of interest in sonar imagery. It is anticipated that the additional information obtained from additional views of an object should improve the classification performance over single-aspect classification. In this paper the detection of mine-like objects (MLO) on the seabed from multiple side-scan sonar views is considered. We transform the multiple-aspect classification problem into a multiple-instance learning problem and present a framework based upon the concepts of multiple-instance classifiers. Moreover, we present another framework based upon the Dempster-Shafer (DS) concept of fusion from single-view classifiers. Our experimental results indicate that both the presented frameworks can be successfully used in mine-like object classification.