Abstract
In order to improve the accuracy and real-time performance of ship target detection on the sea surface, an improved ship target detection method based on the YOLOv3 algorithm is proposed. Firstly, add the CBAM attention mechanism layer to the network prediction layer to improve the expression of the correlation between the feature map target information and important channels and space; secondly, the spatial pyramid pooling (SPP) module is introduced in YOLOv3 to solve the problem of information loss and scale inconsistency; thirdly, GIoU is selected as the loss function to improve the accuracy of the positioning information of the target prediction frame; then the adaptive spatial feature fusion (ASFF) method is used in the prediction layer to make full use of features of different scales to improve the accuracy of target detection; finally refer to the PASCAL VOC data setting format, established a sea surface ship dataset containing 6041 pictures, and manually annotated it for network training. The test results show that compared with YOLOv3, the improved algorithm can effectively improve the accuracy and speed of ship detection on the sea surface. The detection accuracy (mAP) of ship targets on the sea surface has increased by 2.97% to 94.17%, and the frame rate has reached 48fps. Meet the requirements of real-time detection of ship targets.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call