基于霍夫投票的变电站设备三维点云识别算法

Recognition of Three-dimensional Substation Equipment Based on Hough Transform

采取霍夫投票法对激光扫描器获取的变电站设备三维点云数据进行匹配与识别.首先,利用八叉树法对点云进行精简和去噪,得到精简有效的点云数据及数字表面模型深度图像;然后,通过霍夫投票得到物体质心的票数,再与模型库的距离直方图相比求相似度,根据相似度的阈值得到初识别的结果;最后,在初识别的候选设备中通过霍夫投票进行识别.实际数据测试表明,该方法可使得设备的最终识别率达到90.1%,单个设备的平均识别时间为15.6 s,并可有效避免所有元器件特征点的冗长搜索过程.同时能在点云缺失较大情况下将不同设备进行有效分类,达到预期效果.

Based on Hough voting method,the recognizing of substation equipment was achived by using 3D point cloud data obtained by laser scanner.Firstly,the equipment point cloud data was preprocessed to obtain appropriate experimental data,including point cloud simplifying and denoising by octree.Secondly,the number of votes of the mass center of the point cloud was got by Hough voting after obtaining DSM.Thirdly,similarity degrees could be got by comparing the DSM with any model DSM in model set,and initial identification results could be obtained by comparing the above similarity degrees with the threshold.Lastly,the final identification result was got by Hough voting based on initial identification result.The actual test showed that this method could effectively avoid the lengthy searching process for all feature points,and could be effective in recognizing of substation equipment in the situation of a larger points lacking.It turned out that this method was effective.