Surface mine personnel object video tracking method based on YOLOv5- Deepsort algorithm

The environment in open-pit mines is inherently challenging for intelligent monitoring technologies due to the reliance on artificial lighting, the absence of color information, and the similarity between object and background colors. Implementing effective personnel tracking measures is crucial for ensuring safe production in these harsh underground conditions. Consequently, this thesis introduces an open-pit mine personnel tracking method that leverages the YOLOv5 model and the Deepsort algorithm. Initially, YOLOv5 is employed as a surveillance tool mounted on cameras to detect miners on the surface. Subsequently, the Deepsort algorithm is utilized to track the target personnel in real time. Experiments conducted on custom datasets demonstrated that the accuracy and mean Average Precision (mAP) for open-pit mine personnel tracking remained consistently around 92%, with an F1 score of 90%. Moreover, the system was capable of maintaining real-time target tracking even under conditions of dim light, obstacles, and glare. The YOLOv5-Deepsort-based object tracking method plays a significant role in achieving precise tracking of open-pit miners, thereby safeguarding their production safety.