Airborne radar forward-looking imaging holds significant promise for applications such as autonomous navigation, battlefield reconnaissance, and terrain mapping. However, traditional methods are hindered by complex system design, azimuth ambiguity, and low resolution. This paper introduces a distributed array collaborative, forward-looking imaging approach, where multiple aircraft with linear arrays fly in parallel to achieve coherent imaging. We analyze signal model characteristics and highlight the limitations of conventional algorithms. To address these issues, we propose a high-resolution imaging algorithm that combines an enhanced missing-data iterative adaptive approach with aperture interpolation technique (MIAA-AIT) for effective signal recovery in distributed arrays. Additionally, a novel reference range cell migration correction (reference RCMC) is employed for precise range–azimuth decoupling. The forward-looking algorithm effectively transforms distributed arrays into a virtual long-aperture array, enabling high-resolution, high signal-to-noise ratio imaging with a single snapshot. Simulations and real data tests demonstrate that our method not only improves resolution but also offers flexible array configurations and robust performance in practical applications.
Inverse synthetic aperture radar (ISAR) ship target detection is of great significance and has broad application prospects in scenarios such as protecting marine resources and maintaining maritime order. Existing ship target detection techniques, especially target detection methods and detection models in complex settings, have problems such as long inference time and unstable robustness, meaning that they can easily miss the best time for detecting ship targets and cause intelligence loss. To solve these problems, this study proposes a new ISAR target detection model for ships based on deep learning—Complex ISAR Detection Net (CIDNet). The model is based on the Boundary Box Efficient Transformer (BETR) architecture, which combines super-resolution preprocessing, a deep feature extraction network, a feature fusion technique, and a coordinate maintenance mechanism to improve the detection accuracy and real-time performance of ship targets in complex settings. The CIDNet improves the resolution of the input image via the super-resolution preprocessing technique, which enhances the rendering of details of ship targets in the image. The feature extraction part of the model combines the efficient feature extraction capability of YOLOv10 with the global attention mechanism of BETR. It efficiently combines information from different scales and levels through a feature fusion strategy. In addition, the model integrates a coordinated attention mechanism to enhance the focus on the target region and optimize the detection accuracy. The experimental results show that CIDNet exhibits stable performance on the test dataset. Compared with existing models such as YOLOv10 and Faster R-CNN, CIDNet improves precision, recall, and the F1 score, especially when dealing with smaller targets and complex background conditions. In addition, CIDNet achieves a detection frame rate of 63, demonstrating its fine real-time processing capabilities.