Change detection using synthetic aperture radar (SAR) multi-temporal images only detects the change area and generates no information such as change type, which limits its development. This study proposed a new unsupervised application of SAR images that can recognize the change type of the area. First, a regionally restricted principal component analysis k-mean (RRPCA-Kmean) clustering algorithm, combining principal component analysis, k-mean clustering, and mathematical morphology composition, was designed to obtain pre-classification results in combination with change type vectors. Second, a lightweight MobileNet was designed based on the results of the first stage to perform the reclassification of the pre-classification results and obtain the change recognition results of the changed regions. The experimental results using SAR datasets with different resolutions show that the method can guarantee change recognition results with good change detection correctness.

Change detection in synthetic aperture radar (SAR) images has garnered significant research interest. However, to date, only the change area can be obtained, which seriously restricts further development of SAR image applications. To solve this problem, this paper proposes a change type recognition method for SAR images based on a statistical bidirectional long short-term memory (LSTM) network. The proposed method can obtain the changed area and identify the land type change between pairs of SAR images. Considering the statistical characteristics and sample number of SAR images, change detection and type recognition are realized in a unified framework. First, feature extraction is performed through the statistical features of the original SAR image and difference image (DI). Then, a bidirectional LSTM network is used to identify the change types and obtain the change detection results under the small sample deep learning network. On two sets of medium-low resolution SAR image data and one set of GaoFen-3 high-resolution real SAR image data, our method produced high-quality results under small sample conditions.

With the advancement of cognitive radar, applying deep neural networks to radar jamming recognition has become an indispensable research direction. However, as a common anti-jamming measure, the agility of radar waveform parameters degrades the effectiveness of jamming recognition, creating a trade-off between jamming recognition and anti-jamming agility. Specifically, for the same type of jamming, radar agility in frequency, pulse width, and bandwidth alters the profile and scale features of the jamming, posing challenges to conventional CNN-based jamming recognition. To address this challenge, this paper proposes an Adaptive Feature-Focusing CNN (AFF-CNN). A pre-trained AFF module is designed to establish a mapping between agile parameters and adaptive feature scales. Operating on time-domain high-resolution range profiles (HRRP) and time–frequency domain short-time Fourier transform (STFT) data, this module calibrates deviations induced by radar inter-pulse parameter agility and enhances the capability of salient signal feature-focusing. Furthermore, a lightweight 1D-2D feature fusion CNN is designed to process these adaptive features and recognize jamming using single-pulse signals, thereby enhancing the network’s adaptability to inter-pulse parameter agility in radar systems. Simulation results demonstrate superior recognition accuracy and generalization capability compared to five comparative approaches, confirming effective adaptation to inter-pulse agility scenarios.

With the advancement of cognitive radar, the application of deep neural networks for radar jamming recognition has become an indispensable research. However, as a common anti-jamming measure, the agility of radar waveform parameters will degrade the effectiveness of jamming recognition, resulting in a effectiveness dilemma to balance jamming recognition, and anti-jamming agility. Specifically, for the same type of jamming, the agility of the radar in frequency, pulse width and bandwidth will alter the profile and scale features of the jamming, thus posing challenges to jamming recognition based on conventional CNN networks. To address this challenge, this paper proposes an Adaptive Feature-Focusing CNN (AFF-CNN), a pre-trained AFF module is designed to establish the mapping relationship between agile parameters and adaptive feature scales. Operating on time-domain high-resolution range profiles (HRRP) and time-frequency domain short-time Fourier transform (STFT) data, this module contributes to calibrating the deviations induced by radar inter-pulse parameter agility while enhancing the capability of salient signal feature focusing. Furthermore, a lightweight 1D-2D feature fusion CNN is designed to process these adaptive features and recognize jamming using single-pulse signals, thus enhancing the network's adaptability to inter-pulse parameter agility in radar systems. Simulation results demonstrate superior recognition accuracy and generalization capability over four comparative approach, confirming effective adaptation to inter-pulse agility scenarios.

With dynamically simulated false targets generated by airborne miniature simulation platforms becoming increasingly similar to genuine targets, radar discrimination faces mounting challenges. Due to the difficulty of accessing publicly available echo data from such platforms, this paper constructs a simulated dataset. Unlike conventional methods that directly leverage intrinsic features of genuine targets for discrimination, the proposed approach exploits multi-channel statistical discrepancies for identification, without focusing solely on the inherent characteristics of true or false targets, thereby avoiding reliance on real echo data. The method first constructs echo models for both genuine and simulated false targets by leveraging differences in radar cross-section (RCS) under various transmitter–receiver geometries and employing fundamental radar waveforms to build the simulated echo dataset. The resulting simulated false target echoes exhibit high similarity to genuine target echoes. Subsequently, a multi-channel time-frequency correlation statistical feature is proposed. This is achieved by applying Short-Time Fourier Transform (STFT) to echoes from each transmitter–receiver channel to extract time-frequency complex matrices, and then computing the mean of the correlation coefficients among these matrices across all channels. Under multiple observations, genuine and simulated false targets show distinct differences in this feature. Finally, a classifier is designed for training and discrimination. Simulation results demonstrate that, under signal-to-noise ratio (SNR) conditions ranging from –18 dB to 0 dB, the proposed method achieves a discrimination accuracy of at least 83.9% and a Kappa coefficient of at least 0.678 for various types of genuine and simulated targets.