Ship Wake Detection Based On YOLOv5

Ships are the main transportation carriers on the sea surface.The ship targets are small and difficult to detect.As an important feature of ship motion,the distribution range is large,so monitoring ships through wakes is an important approach.Taking ship wakes as the main detection object not only locate the position of the ship,but also determine the ship movement direction.Traditional wake detection algorithms are difficult to meet the real-time requirements of ship wake detection.The YOLOv5 model is the latest achievement of the YOLO series.It shows higher speed and accuracy in the target detection task.This paper studies ship wake detection based on the YOLOv5 s model.According to the characteristics of ship wake images and the structure of YOLOv5 s model itself,the model is simplified and optimized in structure.The relevant feature layer of small scale is subtracted from the structure of YOLOv5 s model,and CA attention module is added to improve the detection efficiency of the model.Then,wake extraction is performed on the detected anchor box area,K-Means++ algorithm is used to segment the wake,Radon transform is used to process the segmented wake image,analyze the main direction of ship movement,and then use improved Canny operator to detect the edge of segmented wake,indicating the specific location of wake.Based on the detection algorithm,a ship wake detection system is built,which makes it more convenient to operate the wake detection model and also can more intuitively view the detection results.In order to verify the accuracy of the model,SWIM wake data set and self-made data set are selected for training and detection.According to the experimental results,it can be seen that after applying improved YOLOv5 s algorithm model for detection,its accuracy is 90.1%,which is 1.9% higher than original YOLOv5 s algorithm.The model parameter amount is reduced from 14.4M to 10.2M,which is reduced by 29.16%.The experimental results show that after improving YOLOv5 s model,it can effectively reduce the missed detection rate of ship wakes,enhance the detection performance of small,multiple and occluded targets,and improve the effect of ship wake detection.