Abstract: Ultra Wide Band (UWB) positioning technology in coal mine underground environments faces limitations including multipath effects, sparse deployment of positioning base stations, and dependence on communication cables. These issues lead to poor positioning stability, low accuracy, and communication interruptions in disaster scenarios, failing to meet the autonomous navigation and positioning requirements of intelligent mobile equipment such as unmanned vehicles, robots, and automated mining devices. To address these problems, a precise underground visual positioning method based on optical beacons was proposed. Optical beacon groups were deployed in pairs on both sides of the roadway, periodically transmitting light signals containing unique codes. Cameras mounted on intelligent mobile equipment captured images containing the optical beacon groups. By decoding the light signals in the images to obtain unique codes and querying the known coordinates of the optical beacons accordingly, distances between the intelligent mobile equipment and the optical beacons were calculated based on the pixel positions of the optical beacons in the images and their actual coordinates. The Perspective-n-Point (PnP) algorithm was used to solve for the three-dimensional coordinates and pose of the intelligent mobile equipment. Positioning experiments conducted in confined spaces and roadways showed that the method achieved three-dimensional precise positioning with errors less than 0.15 meters and high-precision pose perception with Euler angle errors less than 7°, outperforming UWB positioning accuracy. Compared with UWB positioning, this method requires no wireless communication network support and realizes underground three-dimensional positioning and pose perception entirely through machine vision. It is particularly suitable for application in complex electromagnetic environments and areas with high positioning accuracy requirements, such as mining working faces.