Abstract: Coal mine safety production is crucial to miners’ lives and the economic benefits of enterprises, and miners’ unsafe behaviors are one of the primary causes of coal mine accidents. With the development of computer vision and deep learning technologies, intelligent monitoring methods based on image processing have provided new solutions for coal mine safety management. This paper, addressing the characteristics of complex underground coal mine environments, proposes a miner hidden danger behavior monitoring method based on video image analysis. The method integrates the YOLOv5 object detection algorithm, OpenPose human pose estimation, and a conditional random field model to achieve intelligent recognition and early warning of miners’ unsafe behaviors. Experimental results show that the system achieves an accuracy of 96.9% in recognizing common unsafe behaviors and 94.9% in identifying posture-related hazardous behaviors, with a response time of less than 2 seconds, meeting the real-time monitoring requirements in underground coal mines. This study provides an effective technical means for coal mine safety management and holds significant importance for preventing coal mine accidents.