Study on the application of improved target detection algorithm for UAV in maritime search and rescue

To address the inefficiency of target detection algorithms for UAV in maritime search and rescue, an improved UAV target detection algorithm is proposed. First, the MobileNetV3 network is fused in the backbone network of YOLO-v5s to reduce the computational parameters of the algorithm model and make the algorithm easier to be deployed on the UAV embedded platform. Then, to address the problem of more interference factors in the detection image, CBAM (Convolutional Block Attention Module) attention mechanism is introduced to make the algorithm model pay more attention to the feature information of the trapped person, so as to improve the detection efficiency of the algorithm model. The experimental results show that the precision rate P of the improved algorithm reaches 0.966, the recall rate R reaches 0.937, and the mean average precision mAP_0.5 reaches 0.955, which is of high practical value and is expected to play an active role in the maritime search and rescue work.