In response to challenges such as missed detections,false alarms,and reduced accuracy in small target detection from the perspective of unmanned aerial vehicles(UAV),this study investigates the adaptation of the YOLOv5 object detection algorithm for UAV scenarios.Firstly,to enhance the model's feature extraction capability while reducing parameters and computational complexity,the lightweight MobileNetV3_Small algorithm is introduced into the backbone network,enabling the network to learn more features.This design facilitates deployment on UAV devices.Secondly,to improve detection accuracy in scenarios with clustered targets and reduce missed detections,the conventional non-maximum suppression(NMS)algorithm is replaced with Soft-NMS.Experimental results demonstrate that the improved model achieves a detection accuracy of 34.7％on the VisDrone2019 dataset.Compared to the YOLOv5s algorithm,this represents a 5.4 percentage point improvement in accuracy.Simultaneously,the model's parameters and floating-point operations are reduced,facilitating deployment on UAV devices.The refined algorithm proves to be more suitable for image object detection tasks from the perspective of UAV.
关键词
无人机小目标检测/YOLOv5s/MobileNetV3/非极大值抑制算法
Key words
UAV small target detection/YOLOv5s/MobileNetV3/Non-Maximum Suppression algorithm
维普
万方数据