Research on vehicle target detection technology based on UAV aerial images
Wei Liu, Meiling Wang, Shichuan Zhang, Ping Zhou · 2022 IEEE International Conference on Mechatronics and Automation (ICMA) · 2022
This paper investigates the application of deep learning based target detection algorithms for UAV aerial photography to identify vehicle targets. In order to achieve the goal of real-time detection, several optimisations are made on the basis of the YOLOv5 target detection algorithm.1) The feature fusion structure of the algorithm is adjusted to assign weights to the input features, effectively improving the detection accuracy of the target. 2) Varifocal Loss is chosen as the loss function of the algorithm to solve the problem of imbalance between positive and negative samples. 3) Combined with the multi-target tracking technology DeepSORT, the target can be tracked normally even when the target encounters a long period of occlusion, and effectively reduces the number of ID conversions. The experimental results show that the improved target detection algorithm can improve the detection accuracy of UAV aerial images and achieve the demand of real-time target detection for UAV aerial photography.