Object Detection for SAR Ship Images Based on YOLOv5 Deployed On Jetson Agx Xavier

There are a series of scenarios such as ocean-going ship, nearshore ship, multi-ship in SAR(Synthetic Aperture Radar) imaging. With the development of artificial intelligence, it has become possible to deploy deep learning algorithms to detect ship in SAR images. On the other hand, object detection has requirements of running speed, so this algorithm is deployed in the embedded device Jetson Agx Xavier. This paper comes up with sliding window object detection algorithm based on yolov5 (You Only Look Once version 5). Aiming to little object in large image, the detection method in this paper introduces the big picture to cut the subgraphs, and integrate the prediction results of each subgraph to get the final detection result for the sake of enhancing the detection accuracy. In GPU (Graphics Processing Unit) embedded device, the P (Precision rate) of detection of ocean-going ships is 0.979, the R (recall probability) is 0.975, which proves that the detection of ship in SAR images by yolov5 is feasible. This paper focuses on the introduction of algorithms, and selects public data sets to get results.