基于改进YOLOv5的海面可见光图像船舶检测技术

Ship Detection Technology Based on Improved YOLOv5 in Sea Surface Visible Light Image

针对无人船感知系统中的可见光图像船舶检测,提出了一种基于YOLOv5的算法,对深度学习网络模型的输入端、损失函数和检测框进行优化,使用迁移学习的策略进行网络模型训练。试验结果表明:该算法海面可见光图像船舶检测的平均精度均值达到98.6%,较YOLOv5提升1.69%,平均检测时间约为45 ms,能够满足不同条件下海面船舶检测的需求。

For the visible light image ship detection in the unmanned ship sensing system, an algorithm based on yolov5 is proposed, which optimizes the input end, loss function and detection frame of the deep learning network model, and uses the transfer learning strategy to train the network model. The experimental results show that the average accuracy of the algorithm for ship detection in visible light images on the sea surface reaches 98.6%, which is 1.69% higher than YOLOv5. The average detection time is about 45 ms, which can meet the needs of ship detection on the sea surface under different conditions.