Abstract
The Red Palm Weevil (RPW, Rhynchophorus ferrugineus) is a destructive pest of palm plants that can cause the death of the entire plant when infested. To enhance the efficiency of RPW control, a novel detection and tracking algorithm based on the joint YOLOv5-DeepSort algorithm is proposed. Firstly, the original YOLOv5 is improved by adding a small object detection layer and an attention mechanism. At the same time, the detector of the original DeepSort is changed to the improved YOLOv5. Then, a historical frame data module is introduced into DeepSort to reduce the number of target identity (ID) switches while maintaining detection and tracking accuracy. Finally, an experiment is conducted to evaluate the joint YOLOv5-DeepSort detection and tracking algorithm. The experimental results show that, in terms of detectors, the improved YOLOv5 model achieves a mean average precision (mAP@.5) of 90.1% and a precision (P) of 93.8%. In terms of tracking performance, the joint YOLOv5-DeepSort algorithm achieves a Multiple Object Tracking Accuracy (MOTA) of 94.3%, a Multiple Object Tracking Precision (MOTP) of 90.14%, reduces ID switches by 33.3%, and realizes a count accuracy of 94.1%. These results demonstrate that the improved algorithm meets the practical requirements for RPW field detection and tracking.