Long-Strip Target Detection and Tracking with Autonomous Surface Vehicle

As we all know, target detection and tracking are of great significance for marine exploration protection. In this paper, propose one Convolutional-Neural-Network-based method named YOLO-Softer NMS long-strip on the water, which combines You Only Look Once (YOLO) Softer algorithms to improve accuracy. The traditional YOLO network structure is improved, prediction scale increased from threeto four, a softer strategy used select original output method. performance improvement compared totheFaster-RCNN algorithm methodin both mAP speed, proposed YOLO–Softer NMS’s reaches 97.09%while still maintaining same speed as YOLOv3. addition, camera imaging model obtain accurate coordinate information tracking. Finally, using dicyclic loop PID control diagram, Autonomous Surface Vehicle controlled approach with near-optimal path design. actual test results verify that our can achieve gratifying results.