基于YOLOv2的船只检测算法及其在横琴环岛监控系统中的应用

Ship Detection Algorithm Based on YOLOv2 and Its Application in Hengqin Island Video Surveillance System

船只的实时检测对于维护海上航运安全十分重要。基于监控视频数据构建了针对船只检测的数据集,将YOLOv2网络架构应用于船只检测,并结合船只宽高比特点,聚类选取初始候选框。对比实验结果表明,该检测方法能在保证检测精度的同时达到90帧/s的检测速度,满足实际应用需求。珠海市横琴新区已将该检测模型应用于环岛视频监控系统中,并与GIS地图和无人机影像进行集成,实现了环岛海域的实时船只监测。

Real-time ship detection is very important for ensuring the safety of maritime navigation. Based on the monitoring video data, we constructed a dataset for ship detection tasks. In this paper, we applied YOLOv2 network architecture to perform ship detection, and clustered the initial proposal box according to the characteristics with large aspect ratio. The comparison experimental results on the proposed dataset show that the detection method can maintain good detection performance and at the same time achieve the detection speed of 90 frames/s, which can meet the practical application requirement. The Hengqin new area in Zhuhai City has applied the detection model into the video surveillance system around the island, and integrates with GIS maps and unmanned aerial vehicle images, which can realize the real-time detection of ships around the island.