Research on Tibetan Antelope Detection and Tracking Algorithm Based on Improved YOLO11

To achieve the detection of Tibetan antelopes, a detection algorithm for Tibetan antelopes based on the improved YOLO11 is proposed. Firstly, the downsampling convolution of the backbone network is replaced by DSConv, and the original CIoU loss function is replaced by the EIoU loss function, thereby improving the detection speed and accuracy of the model. Finally, the CoordAtt attention mechanism is introduced to improve the model’s detection ability for small-scale Tibetan antelope objects and enhance the model’s generalization ability. Compared with the original YOLO11 model in terms of model size and parameters, the improved YOLO11 algorithm has improved the Precision by 1 percentage points, the recall rate by 2 percentage points, and the mAP 50 value by 1.2 percentage points. It has higher detection accuracy and good robustness. Using the improved YOLO11 as the detector of the ByteTrack object tracking algorithm, the average multi-object tracking accuracy rate (MOTA) was 86.5%, the multi-object tracking accuracy rate (MOTP) was 82.1%, the average ID switching times (IDs) for each test video was 12, and the frame rate (FPS) was 90 frames/s. The Tibetan antelope detection and tracking algorithm based on the improved YOLO11 and ByteTrack proposed in this paper can achieve individual tracking of Tibetan antelopes with relatively high accuracy and real-time performance in the wild scene, providing technical support for the wild monitoring and protection of Tibetan antelopes.