Automatic Target Recognition Problem Research Articles

Sparse Signal Models for Data Augmentation in Deep Learning ATR

Automatic target recognition (ATR) algorithms are used to classify a given synthetic aperture radar (SAR) image into one of the known target classes by using the information gleaned from a set of training images that are available for each class. Recently, deep learning methods have been shown to achieve state-of-the-art classification accuracy if abundant training data are available, especially if they are sampled uniformly over the classes and in their poses. In this paper, we consider the ATR problem when a limited set of training images are available. We propose a data-augmentation approach to incorporate SAR domain knowledge and improve the generalization power of a data-intensive learning algorithm, such as a convolutional neural network (CNN). The proposed data-augmentation method employs a physics-inspired limited-persistence sparse modeling approach, which capitalizes on the commonly observed characteristics of wide-angle synthetic aperture radar (SAR) imagery. Specifically, we fit over-parametrized models of scattering to limited training data, and use the estimated models to synthesize new images at poses and sub-pixel translations that are not available in the given data in order to augment the limited training data. We exploit the sparsity of the scattering centers in the spatial domain and the smoothly varying structure of the scattering coefficients in the azimuthal domain to solve the ill-posed problem of the over-parametrized model fitting. The experimental results show that, for the training on the data-starved regions, the proposed method provides significant gains in the resulting ATR algorithm’s generalization performance.

A Global Model Approach to Robust Few-Shot SAR Automatic Target Recognition

In real-world scenarios, it may not always be possible to collect hundreds of labeled samples per class for training deep learning-based SAR Automatic Target Recognition (ATR) models. This work specifically tackles the few-shot SAR ATR problem, where only a handful of labeled samples may be available to support the task of interest. Our approach is composed of two stages. In the first, a global representation model is trained via self-supervised learning on a large pool of diverse and unlabeled SAR data. In the second stage, the global model is used as a fixed feature extractor and a classifier is trained to partition the feature space given the few-shot support samples, while simultaneously being calibrated to detect anomalous inputs. Unlike competing approaches which require a pristine labeled dataset for pretraining via meta-learning, our approach learns highly transferable features from unlabeled data that have little-to-no relation to the downstream task. We evaluate our method in standard and extended MSTAR operating conditions and find it to achieve high accuracy and robust out-of-distribution detection in many different few-shot settings. Our results are particularly significant because they show the merit of a global model approach to SAR ATR, which makes minimal assumptions, and provides many axes for extendability.

Multi-scale feature vector reconstruction for aircraft classification using high range resolution radar signatures

High-Resolution Range Profile (HRRP) is effective in various Radar Automatic Target Recognition problems. Multi-scale techniques have been verified in aircraft target HRRP feature enrichment to achieve aircraft recognition performance optimization. The involvement of multiple training-classification procedures in existing multi-scale methods results in tremendous resource consumption which challenges their real-time classification performance and application significance. This paper introduces a novel method that enables multi-scale HRRP features to be manipulated under a single scale for efficiency enhancement. Numerical analysis indicates that multi-scale intensity variations for particular HRRP scatterers could be modeled as Gaussian noises. Linear Discriminant Analysis and Singular Value Decomposition techniques are applied on reconstructed feature vectors for better separability and noise tolerance capability. Experimental results from dynamic aircraft recognition experiments verify the expected efficiency enhancement of the proposed method while maintaining comparable classification accuracy and better noise tolerance performance.

Automatic Target Recognition for Low Resolution Foliage Penetrating SAR Images Using CNNs and GANs

In recent years, the technological advances leading to the production of high-resolution Synthetic Aperture Radar (SAR) images has enabled more and more effective target recognition capabilities. However, high spatial resolution is not always achievable, and, for some particular sensing modes, such as Foliage Penetrating Radars, low resolution imaging is often the only option. In this paper, the problem of automatic target recognition in Low Resolution Foliage Penetrating (FOPEN) SAR is addressed through the use of Convolutional Neural Networks (CNNs) able to extract both low and high level features of the imaged targets. Additionally, to address the issue of limited dataset size, Generative Adversarial Networks are used to enlarge the training set. Finally, a Receiver Operating Characteristic (ROC)-based post-classification decision approach is used to reduce classification errors and measure the capability of the classifier to provide a reliable output. The effectiveness of the proposed framework is demonstrated through the use of real SAR FOPEN data.

Aircraft Type Identification Using Reinforcement Active Learning

As a subclass of Automatic Target Recognition problem, Automatic Aircraft Recognition plays an important role in air traffic management and modern battlefield for automatic monitoring and detection. The research on Automatic Aircraft Recognition is still in the exploratory stage. Since aircrafts move at high speeds in complex background, it is still a challenging task of fast data processing and accurate aircraft type recognition. Besides, active learning has recently attracted many researcher’s interesting. Based on this, we employ a learning-based approach which combines active learning with reinforcement learning to learn how and when to request labels for the aircraft type recognition problem. The experimental results show that the model can achieve a good prediction accuracy with few label requests.

Target recognition for synthetic aperture radar imagery based on convolutional neural network feature fusion (Erratum)

Driven by the great success of deep convolutional neural networks (CNNs) that are currently used by quite a few computer vision applications, we extend the usability of visual-based CNNs into the synthetic aperture radar (SAR) data domain without employing transfer learning. Our SAR automatic target recognition (ATR) architecture efficiently extends the pretrained Visual Geometry Group CNN from the visual domain into the X-band SAR data domain by clustering its neuron layers, bridging the visual—SAR modality gap by fusing the features extracted from the hidden layers, and by employing a local feature matching scheme. Trials on the moving and stationary target acquisition dataset under various setups and nuisances demonstrate a highly appealing ATR performance gaining 100% and 99.79% in the 3-class and 10-class ATR problem, respectively. We also confirm the validity, robustness, and conceptual coherence of the proposed method by extending it to several state-of-the-art CNNs and commonly used local feature similarity/match metrics.

Target recognition for synthetic aperture radar imagery based on convolutional neural network feature fusion

Driven by the great success of deep convolutional neural networks (CNNs) that are currently used by quite a few computer vision applications, we extend the usability of visual-based CNNs into the synthetic aperture radar (SAR) data domain without employing transfer learning. Our SAR automatic target recognition (ATR) architecture efficiently extends the pretrained Visual Geometry Group CNN from the visual domain into the X-band SAR data domain by clustering its neuron layers, bridging the visual—SAR modality gap by fusing the features extracted from the hidden layers, and by employing a local feature matching scheme. Trials on the moving and stationary target acquisition dataset under various setups and nuisances demonstrate a highly appealing ATR performance gaining 100% and 99.79% in the 3-class and 10-class ATR problem, respectively. We also confirm the validity, robustness, and conceptual coherence of the proposed method by extending it to several state-of-the-art CNNs and commonly used local feature similarity/match metrics.

Combination of global and local filters for robust SAR target recognition under various extended operating conditions

In this paper, a robust synthetic aperture radar (SAR) automatic target recognition (ATR) method is proposed by combining the global and local filters, especially aiming to improve the recognition performance under various extended operating conditions (EOCs). Due to the cleanness of the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset, which is the prevalent benchmark for SAR ATR, extremely high performance has been reported in many literatures. However, in the real-world scenarios, there exist many types of EOCs such as noise corruption, partial occlusion, sensor variation, etc. Therefore, this study focuses on the SAR ATR problem under several typical EOCs. The intensities of the test image to be classified is taken as the global filter. The spatial correlation between the test image and its corresponding template is calculated as the global response, which describes the global consistency. To depict the local properties of the target, the attributed scattering centers (ASCs) are employed as local filters. Each ASC of the test image is matched with its counterpart from the template ASC set. Then, the local response is evaluated based on the attributes’ differences. The global response and local responses are linearly combined using a random weight matrix. Afterwards, a judgment variable is designed as the measure for target recognition. Experiments are conducted on the MSTAR dataset under the standard operating condition (SOC) and various EOCs including configuration variance, depression angle variance, noise corruption, resolution variance, and partial occlusion. Compared with several state-of-the-art SAR ATR methods, the validity and robustness of the proposed method is fully demonstrated.

SAR Automatic Target Recognition Using Joint Low-Rank and Sparse Multiview Denoising

In recent years, many researchers have focused on the automatic target recognition problem for high-resolution synthetic aperture radar (SAR) systems. Most have directly employed the training data as the dictionary, which introduces error from speckle noise. In this letter, a joint low-rank and sparse multiview denoising (JLSMD) dictionary is generated, which combines multiview training samples for denoising. To extract the dictionary, we fully consider the low-rank property of multiview target images and the sparsity of speckle noise for SAR systems. The designed dictionary is more accurate than the training data in representing the target. With the help of the proposed JLSMD dictionary, we develop three algorithms based on the sparse representation classification and the support vector machine approach. We carry out experiments on the moving and stationary target acquisition and recognition public data set to evaluate the excellent performance of the proposed methods against several state-of-the-art methods, including deep learning methods.

Automatic target recognition in missing data cases

Automatic target recognition (ATR) is the process where computer algorithms are used to detect and classify objects or regions of interest in sensor data [1]. ATR algorithms have been developed for a broad range of sensors, including electro-optic (EO), infrared (IR), and microwave sensors (radar). A key advantage of microwave radar is its utility for all weather and time of day compared with other types of sensors. One limitation of many radar systems' ATR performance is the poor spatial resolution in the range and cross-range dimensions. While the range resolution of the microwave radar system can be straightforwardly improved by increasing the system's radio frequency bandwidth, superior cross-range (or azimuth) resolution requires an increase in antenna size. In order to achieve resolution similar to EO or IR systems, an enormous antenna must be used for transmission and reception, hindering microwave radar system's general applicability to ATR problems. However, by operating a synthetic aperture radar (SAR), the large antenna requirement can be overcome, and very fine cross-range resolution can be obtained with a modest real aperture antenna. Indeed, SAR has been used to generate high-resolution 2-D or 3-D object images at a variety of different microwave frequencies. In particular, wideangle SAR images can contain important features of an object from a diverse set of observation angles collected by a radar system orbiting around a target or scene. These feature-rich wide-angle SAR images are useful for ATR due to their high resolution and comprehensive coverage of the target. Interference, jamming, or data dropouts, however, are commonplace in practical SAR environments, and result in an incomplete data set. These missing and corrupted data cause substantial degradation in the generated SAR imagery, hampering their utility for subsequent ATR processing especially when data-independent image formation algorithms are used. A prime example of these difficulties is operating a SAR system in the very high frequency (VHF) or ultrahigh frequency (UHF) bands, where the spectrum is frequently crowded [2].

Infrared variation reduction by simultaneous background suppression and target contrast enhancement for deep convolutional neural network-based automatic target recognition

Automatic target recognition (ATR) is a traditionally challenging problem in military applications because of the wide range of infrared (IR) image variations and the limited number of training images. IR variations are caused by various three-dimensional target poses, noncooperative weather conditions (fog and rain), and difficult target acquisition environments. Recently, deep convolutional neural network-based approaches for RGB images (RGB-CNN) showed breakthrough performance in computer vision problems, such as object detection and classification. The direct use of RGB-CNN to the IR ATR problem fails to work because of the IR database problems (limited database size and IR image variations). An IR variation-reduced deep CNN (IVR-CNN) to cope with the problems is presented. The problem of limited IR database size is solved by a commercial thermal simulator (OKTAL-SE). The second problem of IR variations is mitigated by the proposed shifted ramp function-based intensity transformation. This can suppress the background and enhance the target contrast simultaneously. The experimental results on the synthesized IR images generated by the thermal simulator (OKTAL-SE) validated the feasibility of IVR-CNN for military ATR applications.

Pseudo‐Zernike‐based multi‐pass automatic target recognition from multi‐channel synthetic aperture radar

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, provides the opportunity to exploit diversities to mitigate uncertainty. In this study, the authors address the problem of automatic target recognition (ATR) from synthetic aperture radar platforms. The author's approach exploits both channel (e.g. polarisation) and spatial diversity to obtain suitable information for such a critical task. In particular they use the pseudo-Zernike moments (pZm) to extract features representing commercial vehicles to perform target identification. The proposed approach exploits diversities and invariant properties of pZm leading to high confidence ATR, with limited computational complexity and data transfer requirements. The effectiveness of the proposed method is demonstrated using real data from the Gotcha dataset, in different operational configurations and data source availability.

Minimum Divergence Approaches for Robust Classification of Ground Moving Targets

This work addresses the problem of automatic target recognition (ATR) using micro-Doppler information obtained by a low-resolution ground surveillance radar. Gaussian mixture models (GMMs) are used to represent the prior statistical information of threatening ground moving targets such as walking personnel and tracked or wheeled vehicles. A minimum divergence (MD) classification approach with a variety of distance measures is proposed. The proposed MD classification approach is robust with respect to modeling errors and can be efficiently used in low signal-to-noise (SNR) and training data deficient scenarios. The MD classifier is implemented using a variety of computationally efficient approximations of distance measures between GMMs. Performance of the MD classifier was analyzed using collected radar measurements and the influence of different distance measures and their approximations on classification performance is assessed. The proposed MD classifier outperforms the maximum likelihood (ML) classifier in low-SNR and training data deficient scenarios while providing a computationally efficient implementation.

HRR Radar Signature Database Validation for ATR: An Information Theoretic Approach

The ability to make radar signature databases portable for use within sensor systems is critical to the affordability of future airborne signature exploitation systems. The capability to hybridize measured and synthetic signature database components will maximize the impact of the investment required to build complex radar signature databases. Radar target scattering mechanisms can be modeled and the signature signal model analyzed as a random process to enable portability and hybridization. Modal mutual information is developed as a measure of similarity for future use in the comparison of measured field data to modeled synthetic data. The inherent qualities of mutual information to be used in the context of the automatic target recognition problem are demonstrated using synthetic signature sets comprised of both similar targets and dissimilar targets.

Geometric diffusion as a classifier

Often, the problem of automatic target recognition can be reduced down to two separate but related problems, feature extraction and classifier design. The best classifier only works as well as the input data provided to the system. In this presentation, we will outline a new approach to classification known as geometric diffusion as proposed by Coifman et al, and demonstrate the power of this new metric for classification of imagery.

Clutter metric based on the Cramer-Rao lower bound on automatic target recognition

This is a performance evaluation on the implementation of the Cramer-Rao lower bound (CRLB) for background clutter measurement on automatic target recognition (ATR). In essence the background clutter evaluation problem for ATR is consistent with the deterministic parameter estimation problem. Thus, useful concepts and theories of deterministic parameter estimation can be introduced into the investigation of background clutter. In this paper, the CRLB is employed as a metric for clutter measurement. Requirements needed for this application are analyzed, and the approach for obtaining the CRLB of a scene image is produced. The flexibility of the CRLB metric is analyzed. Discussion and comparison are made on the relationship between the CRLB metric and the Sims signal-to-clutter metric. Finally, we illustrate how this metric defines the potential for false alarms by determining the correspondence level between a target and background through the application of the CRLB.

Decision Fusion With Confidence-Based Weight Assignment for Hyperspectral Target Recognition

Conventional hyperspectral image-based automatic target recognition (ATR) systems project high-dimensional reflectance signatures onto a lower dimensional subspace using techniques such as principal components analysis (PCA), Fisher's linear discriminant analysis (LDA), and stepwise LDA. Typically, these feature space projections are suboptimal. In a typical hyperspectral ATR setup, the number of training signatures (ground truth) is often less than the dimensionality of the signatures. Standard dimensionality reduction tools such as LDA and PCA cannot be applied in such situations. In this paper, we present a divide-and-conquer approach that addresses this problem for robust ATR. We partition the hyperspectral space into contiguous subspaces based on the optimization of a performance metric. We then make local classification decisions in every subspace using a multiclassifier system and employ a decision fusion system for making the final decision on the class label. In this work, we propose a metric that incorporates higher order statistical information for accurate partitioning of the hyperspectral space. We also propose an adaptive weight assignment method in the decision fusion process based on the strengths (as measured by the training accuracies) of individual classifiers that made the local decisions. The proposed methods are tested using hyperspectral data with known ground truth, such that the efficacy can be quantitatively measured in terms of target recognition accuracies. The proposed system was found to significantly outperform conventional approaches. For example, under moderate pixel mixing, the proposed approach resulted in classification accuracies around 90%, where traditional feature fusion resulted in accuracies around 65%.

Divide and conquer approach to improve performance on ATR systems

In this paper different methods applied to the Automatic Target Recognition problem are studied. A database of High Range Resolution radar profiles of six kinds of aircrafts is used to study the performance of four classification methods: k-Nearest Neighbor method, Multilayer Perceptrons, Radial Basis Function Networks, and Support Vector Machines. Results obtained with these classifiers show a high correlation between two of the classes of targets that cause the majority of errors. We propose to split the task into two subtasks. A first one in which the classes of correlated targets are grouped in a single class, and a second one to distinguish between them. Different classifiers are studied to be applied to each subtask. Results demonstrate that Radial Basis Function Networks are very good classifiers for the main subtask, while Support Vector Machines are the best classification method, among the studied, to distinguish between the correlated targets.

Multiscale Connected Operators

Among the major developments in Mathematical Morphology in the last two decades are the interrelated subjects of connectivity classes and connected operators. Braga-Neto and Goutsias have proposed an extension of the theory of connectivity classes to a multiscale setting, whereby one can assign connectivity to an object observed at different scales. In this paper, we study connected operators in the context of multiscale connectivity. We propose the notion of a ?-connected operator, that is, an operator connected at scale ?. We devote some attention to the study of binary ?-grain operators. In particular, we show that families of ?-grain openings and ?-grain closings, indexed by the connectivity scale parameter, are granulometries and anti-granulometries, respectively. We demonstrate the use of multiscale connected operators with image analysis applications. The first is the scale-space representation of grayscale images using multiscale levelings, where the role of scale is played by the connectivity scale. Then we discuss the application of multiscale connected openings in granulometric analysis, where both size and connectivity information are summarized. Finally, we describe an application of multiscale connected operators to an automatic target recognition problem.

Stochastic models for recognition of occluded targets

Recognition of occluded objects in synthetic aperture radar (SAR) images is a significant problem for automatic target recognition. Stochastic models provide some attractive features for pattern matching and recognition under partial occlusion and noise. In this paper, we present a hidden Markov modeling based approach for recognizing objects in SAR images. We identify the peculiar characteristics of SAR sensors and using these characteristics we develop feature based multiple models for a given SAR image of an object. The models exploiting the relative geometry of feature locations or the amplitude of SAR radar return are based on sequentialization of scattering centers extracted from SAR images. In order to improve performance we integrate these models synergistically using their probabilistic estimates for recognition of a particular target at a specific azimuth. Experimental results are presented using both synthetic and real SAR images.