Abstract: To realize the 3D measurement of natural terrain contour features, a stereo vision 3D reconstruction method based on region segmentation was proposed. First, on the basis of image segmentation by using the watershed transformation method, the contour edges were detected according to the boundaries of the segmented regions. Then the contour edges were grouped according to their corresponding regions for stereo matching. On the basis of a region’s global adjacency characteristics and its local position, area and gray characteristics, the region constraint was established. In addition, the gradient direction constraint of an edge point was established according to the gradient distribution. The new constraints effectively reduced the search area, optimized the matching order and increased the correct matching ratio. By using the stereo matching edge point pairs, their 3D coordinates were estimated according to the binocular stereo vision imaging model. Experimental results show that the proposed method can reconstruct the terrain contour edges denoting the terrain 3D structure characteristics quickly and effectively, and it can yield high stereo matching ratio and 3D reconstruction precision.

Key words: stereo vision, 3D reconstruction, edge matching, region constraint, gradient direction