Abstract: In view of the environmental detection and emergency rescue missions after coal mine accident, the development and use of coal mine rescue robots is a key way to improve rescue efficiency and reduce rescue hazard coefficients, and the binocular vision technology is the premise for the coal mine rescue robot to obtain the accident site information and achieve the autonomous obstacle avoidance and route planning.Firstly, based on the realization process of binocular vision technology, the mathematical principle of visual distance measurement was introduced.The representative methods in the field of camera calibration were summarized, including traditional calibration methods, active vision calibration methods and self-calibration methods.The latest research results of global matching algorithm, local matching algorithm and sub-global matching algorithm in stereo vision matching were described, and the advantages and disadvantages of three kinds of matching algorithms were compared.Then, based on the analysis of recent research literature on coal mine rescue robots, the application and development of binocular vision technology in coal mine rescue robots were studied.It was pointed out that the research scope of binocular vision technology in the field of coal mine rescue robots mainly covers stereo vision matching algorithm, pattern classification and recognition, visual measurement and 3D reconstruction, combined measurement and positioning, visual servo control and visual algorithm simulation based on virtual reality.Finally, based on the characteristics of high dynamic and strong disturbance in the unstructured environment of coal mines, It was pointed out that the key technology of coal mine rescue robot binocular vision field application is to solve the problems of motion blur and lens pollution, large nonlinear distortion of ultra-wide-angle lens, and weak/zero illumination conditions.According to the requirements of binocular vision of coal mine rescue robot, such as large field of view, high-precision and adaptive perception, the suggestions for future development, including multi-degree-of-freedom measurement, multi-sensor information fusion and adaptive perception based on active vision, were proposed.