Satellite target detection algorithm based on YOLO-GRU

To solve the problems of poor robustness and low detection accuracy of satellite target detection algorithms in complex environments, a satellite target detection algorithm based on YOLO-GRU is proposed. The algorithm combines the spatial feature perception of YOLO with the time series analysis of GRU, which breaks through the limitations of previous object detection algorithms. It is verified that the algorithm has excellent stability and generalization ability in model performance based on a large number of experiments. The test results on the benchmark SPEED satellite dataset show that in the constantly changing space environment, the recognition accuracy based on time series is improved by 5.1% compared to YOLOv5, the recall rate is increased by 8.3%, and the mAP index is improved by 9.6%. This paper provides an efficient and accurate solution for the detection of continuously moving satellite targets in complex environments.