Dense Vehicle Target Detection Algorithm Based on YOLOv5 in Circular Synthetic Aperture Radar Image

Recently, detecting the interesting target accurately in circular synthetic aperture radar (CSAR) images has become one of the most important research directions in the actual CSAR application. Due to the influence of the data acquisition mode, the size of the CSAR image is usually large. When the vehicle targets in the CSAR image are parked tightly, they usually cannot be detected well using the traditional detection methods. Therefore, a vehicle target detection method based on YOLOv5 in the CSAR image is proposed in this paper. Firstly, the image augmentation technique is used to solve the problem of the insufficient training samples, and then the data set is input to the model for training. Secondly, the test image is segmented, and then the trained model is used to detect the vehicle target respectively. Finally, the post-processing is used to optimize the detection results. The CSAR image based on the Gotcha dataset is used to verify the correctness of theoretical analysis and feasibility of the proposed method.