To solve the problems of low accuracy and missed detection of small targets in scenarios by UAV,an improved small target detection algorithm DE_YOLOv5 based on YOLOv5 is proposed.The dataset of DE_YOLOv5 adopts VisDrone2019.The experiment includes two aspects.One is to introduce a decoupling joint mechanism,which reduces the impact of receptive field limitations on small target localization through independent center point prediction.mAP@0.5 of the initial model is 0.334,mAP@0.5 value is 0.344 after adding decoupling joint.The second is to replace the loss function with Focal-EIoU,and mAP@0.5 value is 0.351 after replacing the loss function.The experimental results show that DE_YOLOv5 can effectively improve the detection accuracy of small targets.
国家科技期刊平台
NSTL
万方数据
