基于坐标轴投影勘察的深度图像树型分割

Abstract

A tree-type segmentation method based on the exploration of projection onto coordinate axes was proposed for range images with a feature of spatial connective region.By implementing the projections onto coordinate axes at the optimum angle by the exploration of projection,the segmentation of range images was transformed into the segmentation of 1D point data projected onto coordinate axes.Finally,a tree-type structure was employed to carry out the process of segmentation.The algorithm was applied to the range images acquired from a concrete project to verify the effectiveness and the efficiency,and the results of both ideal and extreme conditions were discussed,respectively.The results show that the algorithm can effectively segment range images with the feature of spatial connective region and has a constant speed of segmentation with regard to different range images in the same hardware platform.For the 3 tested range images containing 19 792,35 186 and 79 618 points,the average speeds of segmentation in ideal and extreme conditions are 0.168 s and 0.494 s every 10 000 points,respectively,which verifies the high effectiveness and efficiency of segmentation.