Abnormal dynamics awareness of space targets is critical for space surveillance. With the intrinsic range-Doppler projection mechanism, an inverse synthetic aperture radar (ISAR) image sequence naturally has crucial potential for abnormal dynamics interpretation on uncooperative targets. In this paper, we develop an automated deep neural network architecture for abnormal dynamic recognition of space targets from ISAR image sequences. With the accommodation of the stacked sparse autoencoder (SSAE) network joint sparsity constraint, the geometrical feature flow of an ISAR image sequence is learned to represent the target dynamics concisely. Then, the abnormal dynamic recognition is turned into a sequential classification task by exploiting the encoded feature flow through the long-short time memory (LSTM) network. Extensive experiment results confirm the superiority of the proposal in both precision and efficiency aspects.

Dynamic identification is crucial for space target investigation and awareness, which is especially challenging for uncooperative and uncontrolled targets. This article presents a dynamic identification algorithm using the time-variant geometric feature flow extracted from an inverse synthetic aperture radar (ISAR) image sequence. The range-Doppler (RD) projection, along with the sequential radar light of slight of the target component, intrinsically relies on the targets’ dynamic states. With the accommodation of a long short-time memory (LSTM) network, reliable dynamic identification of space targets can be implemented by exploiting the geometrical feature flow from the ISAR sequence. The superiority of the proposed method is demonstrated through comparison experiments.

The attributed scattering center (ASC) can provide a concise physical description of radar targets, which has broad applications for physical-induced radar recognition. However, the ASC extraction is of great difficulty due to its high-dimensional parametric modeling. In this letter, a modified Alternative Direction Method of Multipliers (ADMM) unfolded-net is proposed for ASC decomposition of synthetic aperture radar (SAR) targets. By transforming the ASC decomposition challenge into a sparse reconstruction problem, the modified ADMM-net can fast attain ASC parameters as well as reconstruct a complete signal from sparse observations. Instead of tremendous iterations with fixed hyper-parameters as traditional methods, the modified ADMM-net unrolls several iterations into a few layers with learned hyper-parameters. Therefore, the proposed algorithm can realize fast and precise ASC decomposition. Experimental results confirm its effectiveness and superiority.

For target detection with sparsely sampled over‐the‐horizon radar (OTHR) signal, there are two major hassles puzzling its clutter and target Doppler spectrum reconstruction. On one hand, the OTHR signal is usually contaminated by transient interferences, which usually leads to a significant signal loss in its excision with conventional methods. On the other hand, the sparse OTHR signal will introduce serious grating‐lobes in conventional Doppler analysis methods and thus degrade the performance of target detection. In this study, an integrate clutter and target spectrum reconstruction method for sparse OTHR signal are proposed. The proposed method contains two aspects. First, based on the sparse basis representation, a transient interference excision method is developed, which uses time–frequency modulated rectangle window basis to represent the OTHR signal in a sparse way to identify transient interferences. The second aspect lies in that, based on the compressive sensing, a spectrum reconstruction method is proposed to obtain the integrate clutter and target spectrum with sparse samples. By solving the Doppler sparsity constrained optimisation with greedy algorithms, the complete Doppler spectrum is recovered effectively for target detection. The performance of the authors proposal is validated by both simulation and real‐measured data sets.

The unfolding network, which combines traditional compressive sensing (CS) methods and deep learning, has been widely used in sparse aperture (SA) inverse synthetic aperture radar (ISAR) imaging because of its good interpretability, high reconstruction accuracy, and computational efficiency. However, most unfolding networks view SA ISAR imaging under different sampling matrices as diverse problems, resulting in poor network transferability when the sampling matrix changes. In this article, a complex-valued residual alternating direction method of multipliers-Net (CV-Res-ADMM-Net) is designed to address this issue. Based on the radar signal's complex-valued (CV) characteristics, we designed a residual learning-based CV convolutional neural network (CNN) that includes three modules: feature extraction, nonlinear mapping, and feature accumulation to implement the nonlinear operation to enhance the network's ISAR image feature extraction capability. The residual learning technique can accelerate the training process and smooth network information transmission. Moreover, a multi-sampling matrix generation approach is developed, which can improve the training diversity and maintain more structural information on ISAR images for more satisfactory reconstruction. Simulated and real-measured experiments verify that CV-Res-ADMM-Net has good transferability when the sampling matrix changes. Experimental results show the proposal can achieve fast SA ISAR imaging of maneuvering targets.