Ship Target Detection Algorithm Based on Improved YOLOv5

In order to realize the real-time detection of an unmanned fishing speedboat near a ship ahead, perception platform based on target visual system was established. By controlling depth and width model analyze compare training, it found that 5S had fast speed but low accuracy, which judged be insufficient for detecting small targets. this regard, study improved YOLOv5s algorithm, in initial frame is re-clustered by K-means at data input end, receptive field area expanded output loss function optimized. The results show precision model’s images 98.0%, recall rate 96.2%. Mean average (mAP) reached 98.6%, increase 4.4% compared before improvements, shows can identification multiple types ships, laying foundation subsequent path planning automatic obstacle avoidance ships.