Moving Target Detection in Video SAR Based on Improved Faster R-CNN

Video SAR makes it possible to perform reliable moving target detection indirectly by detecting shadows. The key of moving target detection in video SAR is how to utilize the information of multiple adjacent frames to reduce false alarms and missing alarms. This paper presents an approach for moving target detection in video SAR based on the improved Faster R-CNN. The improved Faster R-CNN enhances the per-frame features by incorporating the spatiotemporal information extracted from multiple adjacent frames by a 3D convolution neural network, and thus improves the performance of object detection. The proposed approach gives a convincing results on the real video SAR data.