Abstract: Conventional model monitoring methods have many disadvantages such as complication of device or sensor installation, much hard work, limitation in sampling points etc. Here, we propose a method using computer vision and digital close-range photogrammetry technique to monitor model deformation. Image groups of different simulating excavation status are captured by high resolution CCD camera. In order to determine the coordinates of the target’s points, binary image are acquired by single threshold segmentation Targets points are detected and recognized according to the shape parameters of labeling areas. Then, the coordinates of the target’s points are determined by contour tracking and ellipse fitting algorithm. Approximate values of camera’s extrinsic parameters are decomposed using 2D direct linear transformation and co-linearity equations and bundle adjustment is carried out to determine the target space coordinates and the accuracy is better than ±1mm. The test results also indicate that the displacement field by this technique fit well with the phenomenon occurred in the process of test, which can meet requirements of model deformation monitoring for mine stratum and surface movements.

Key words: strip mining, digital close-range photogrammetry, direct linear transformation, bundle adjustment, camera calibration