Design and Implementation of Object Detection Model for UAV Aerial images Based on TS-YOLOv8

The main task of UAV aerial object detection is to accurately detect and locate objects using rectangular boxes in aerial images or videos. Due to factors such as shooting height and angle, small objects and overlapping objects account for a large proportion of aerial photography objects. Current UAV aerial photography object detection models may encounter missed detection issues when detecting these two types of objects. In response to the above issues, this paper proposes a TS-YOLOv8 UAV aerial photography object detection model. This model adds a tiny object detection layer on the basis of YOLOv8n to solve the problem of missed detection of small objects. Secondly, the Soft-NMS algorithm is used to optimize the candidate boxes of YOLOv8n to solve the problem of missed detection of mutually occluding objects. The experimental results on the VisDrone2019 dataset show that the proposed TS-YOLO unmanned aerial vehicle aerial photography object detection model can effectively solve the problem of missed detection of small and mutually occluding objects.