Moving And Stationary Target Recognition Research Articles

Learning to Generate SAR Images With Adversarial Autoencoder

Deep learning-based synthetic aperture radar (SAR) target recognition often suffers from sparsely distributed training samples and rapid angular variations due to scattering scintillation. Thus, data-driven SAR target recognition is considered a typical few-shot learning (FSL) task. This article first reviews the key issues of FSL and provides a definition of the FSL task. A novel adversarial autoencoder (AAE) is then proposed as an SAR representation and generation network. It consists of a generator network that decodes target knowledge to SAR images and an adversarial discriminator network that not only learns to discriminate “fake” generated images from real ones but also encodes the input SAR image back to target knowledge. The discriminator employs progressively expanding convolution layers and a corresponding layer-by-layer training strategy. It uses two cyclic loss functions to enforce consistency between the inputs and outputs. Moreover, rotated cropping is introduced as a mechanism to address the challenge of representing the target orientation. The moving and stationary Target recognition (MSTAR) 7-target dataset is used to evaluate the AAE’s performance, and the results demonstrate its ability to generate SAR images with aspect angular diversity. Using only 90 training samples with at least 25° of orientation interval, the trained AAE is able to generate the remaining 1748 samples of other orientation angles with an unprecedented level of fidelity. Thus, it can be used for data augmentation in SAR target recognition FSL tasks. Our experimental results show that the AAE could boost the test accuracy by 5.77%.

An SAR ATR Method Based on Scattering Centre Feature and Bipartite Graph Matching

ABSTRACTAutomatic target recognition (ATR) is a crucial application for synthetic aperture radar (SAR). This paper presents an ATR method with the scattering centre (SC) based on the world view vector (WVV) and the weighted bipartite graph model (WBGM). Targets are separated by the filter-based cluster size insensitive enhanced fuzzy c-mean cluster (csiEnFCM) algorithm before feature extraction. Later, WVV-based feature set is constructed with the extracted SCs. Subsequently, the WBGM is applied to recognize target by matching its feature set with templates. Experiments on the moving and stationary target recognition (MSTAR) dataset demonstrate that the proposed method performs well in ATR.

Statistical Analysis of High-Resolution SAR Ground Clutter Data

This paper deals with the problem of modeling high-resolution synthetic aperture radar clutter data from different vegetated areas. We analyzed moving and stationary target recognition (MSTAR) data sets focusing on histograms, moments, and covariance of clutter amplitude, texture, and speckle. The most celebrated statistical models are tested on real data of grass field or wood and trees to validate the goodness of fit of the compound Gaussian model in different scenarios. The results demonstrate that for grass fields, the compound Gaussian model provides a good data fitting. This is not the case for woods images where the speckle is not more Gaussian distributed. Covariance analysis and concluding remarks complete this paper

Analytical derivation of distortion constraints and their verification in a learning vector quantization-based target recognition system

We obtain a novel analytical derivation for distortion-related constraints in a neural network- (NN)-based automatic target recognition (ATR) system. We obtain two types of constraints for a realistic ATR system implementation involving 4-f correlator architecture. The first constraint determines the relative size between the input objects and input correlation filters. The second constraint dictates the limits on amount of rotation, translation, and scale of input objects for system implementation. We exploit these constraints in recognition of targets varying in rotation, translation, scale, occlusion, and the combination of all of these distortions using a learning vector quantization (LVQ) NN. We present the simulation verification of the constraints using both the gray-scale images and Defense Advanced Research Projects Agency's (DARPA's) Moving and Stationary Target Recognition (MSTAR) synthetic aperture radar (SAR) images with different depression and pose angles.

Harmonic wavelet joint transform correlator: analysis, algorithm, and application to image denoising

We propose analysis and algorithm implementation for a novel harmonic wavelet joint transform correlator (HWJTC) for image noise reduction and automatic target recognition (ATR) applications. We investigate useful properties of the HWJTC and apply our novel correlator in ATR for both noiseless and noisy synthetic aperture radar (SAR) images. Performance is demonstrated in reducing both simulated speckle noise and realistic clutter noise for SAR images. We also compare its performance with that of a well-established image denoising technique such as wavelet packet (WP) processing. The test images used in our simulation are obtained from a CAD simulation model such as Xpatch and Defense Advanced Research Projects Agency (DARPA) Moving and Stationary Target Recognition (MSTAR) program's synthetic aperture radar (SAR) image databases respectively. Simulation results show improved noise tolerance for HWJTC.