A Lightweight Algorithm for UAV Aerial Image Objects Detection Based on Improved YOLOX
Aimed at small unmanned aerial vehicle(UAV)patrol applications,a lightweight object detection algorithm is pro-posed to effectively solve the dual constraints of wireless transmission channel and onboard computing resource during the patrol process.Firstly,the Mobilenetv2 network is used as feature extraction network of YOLOX algorithm to reduce the parameters and computation of the model,and improve the speed of object detection.Secondly,the CIOU loss function is introduced to improve the precision of object positioning.Thirdly,a Focal Loss confidence loss function is introduced to balance the positive and negative or dif-ficult and easy samples in training process,improving the performance of the model.Experimental results based on VisDrone2019-DET dataset show that the improved algorithm reduces the model parameters by 56.2％,the calculation by 52.5％,and the inference time of single image by 41.4％,without a significant decrease in detection accuracy.Finally,the improved algorithm is deployed to the Nvidia Jetson Xavier NX,and the model detection frame rate reaches by 22 FPS,which meets the application requirements of pa-trol tasks.
万方数据
维普
