Abstract: According to the demands of coal mine safety production, the safety management and control technology of key area in coal mine based on image recognition is studied. Machine learning algorithm and computer vision technology are used to analyze the data combined with personnel management data and equipment operation data. Through linkage analysis and data mining, the monitoring and control of underground personnel behavior and coal quantity can be realized, and a knowledge base for target risk pre-control can be formed, which is also verified by underground experiments. The experimental data show that the real-time analysis response time of the system platform is less than 2 s, and the recognition rate is more than 98%. The system can effectively realize the safety control of moving targets such as underground personnel and coal quantity.