UAV Target Tracking and Detection Based on Faster R-CNN Improved Networks

With the arrival of the computer intelligence era, human requirements for industrial automation and intelligence are becoming more and more urgent, this paper mainly relies on the UAV platform to explore the performance advantages and disadvantages of the optimized and improved hrnet as the backbone network of Faster R-CNN in the field of target tracking. The study provides a new algorithmic idea for the future application of target tracking and detection in the field of industrial automation. The experimental results show that the network has improved mAP, Parameters/M, and FPS compared to Faster R-CNN (RESNET50), SSD (VGG), YOLOv3 (Darknet53), and YOLOv8 (EfficientNet-B0). The improved network can accurately track and segment the target in cases where the UAV flies at variable altitude and the target presents multi-scale features. It can be seen that the Faster R-CNN (hrnet) network has good target tracking and detection ability and high robustness, which makes the UAV more valuable in the application fields of logistics and transportation, film and television shooting, environmental monitoring, and emergency rescue.