The invention discloses a monitoring method capable of identifying belt deviation through an artificial intelligent video. The monitoring method comprises the following steps of selecting not less than two support rollers on two side frames of a belt as two or more tracking objects, acquiring a tracking object video in real time through a mineral high-definition anti-explosion video camera, transmitting the tracking object video to a server in real time through a network by the mineral high-definition anti-explosion video camera, analyzing the tracking target video by the server through an artificial intelligent video identifying module, detecting whether the belt shields a tracking target or not when the tracking target is shielded, determining that the belt deviates if the detected result is that shielding is caused by the belt, and judging that the belt does not deviate if the detected result is that shielding is not caused by the belt. The invention aims to solve the belt operationdetecting problem through an artificial intelligent technology. The monitoring method achieves timely finding and alarming while a belt deviation phenomenon appears in a belt operation process, is suitable for various belt transportation scenes, and reduces manual timed correcting operation of an existing detecting sensor, so that coal mine operation safety is ensured, and waste of labor power and time is reduced.