Abstract: In orde to solve the problems of low degree of automation, complex calculation and poor precision in similar simulation experiments in mines, a method based on photogrammetric computer vision technology to realize automatic monitoring of mine similar material models is proposed. This method realizes the automatic identification of the model monitoring points by designing the coding marks. Based on the fact that the similar material model of the mine is mostly planar, the automatic deformation observation of the similar simulation model is realized by computer vision homograph matrix transformation and object-based cluster matching. The effectiveness of the method is verified by a similar simulation experiment of a longwall fully-mechanized caving with a thick dip angle. The results show that this method has the advantage of high automation degree. Besides, the measurement result is about 0.001 m different from the plane accuracy of the monitoring point measured by the total station. The sinking and horizontal movement of the measured displacement field is consistent with the actual deformation, which can meet the similar simulation experimental requirement of mine strata and ground movement.