Target detection algorithm for marine radar images based on improved YOLOv8
In order to solve the problems of complex navigation scenes of inland waterway vessels,few shape and color char-acteristics of marine radar images,and the difficulty of anno-tation,an improved YOLOv8 marine radar image target detec-tion method was proposed. Firstly,to alleviate the issues of annotation errors and model overfitting,a label smoothing strategy was introduced during the model training phase. Then,combining the unique positional prior information of ra-dar images,a coordinate based convolution structure was de-signed to simultaneously extract the shape,color,and posi-tional features of the target. To verify the effectiveness and su-periority of the proposed method,comparative experiments were conducted on the collected radar images of the Yangtze River channel under different weather conditions. Results show that the proposed method achieves an accuracy rate of 91.52％ while ensuring real-time object detection,with an av-erage accuracy improvement of 5.17％ compared to the classic YOLOv8,which can provide technical support for improving the modernization and intelligent management level of inland waterway transportation.
inland water transportmaritime radar imagetarget detectionYOLOv8
万方数据
维普
