This paper builds on the insight that automatic accurate registration of 3D point cloud and CAD model is an important step in the process of industrial 3D automated inspection and introduces a no-vel registration algorithm as an improved alternative to the currently available registration algorithms which suffer from poor inherent stability as when they are used for automating registration of point cloud contai-ning irrelevant data obtained in the industry.The registration algorithm able to reject irrelevant data based on adaptive threshold works by eliminating the background data,in the iteration by calculating the adap-tive threshold by statistically analyzing the Euclidean distance of matching point pairs; and reducing the influence of noise on the registration result by performing the rigid transformation of calculation point cloud by selecting matching points by percentage.The result demonstrates that the proposed algorithm en-ables a more stable registration than Trimmed-ICP,contributing to less time in single iteration time thanks to the reduction of matching point pairs in each iteration, hence an improved registration effectiveness. The algorithm could provide automatic and accurate registration of 3D point cloud and CAD model in in-dustrial 3D automated inspection.%三维点云与CAD模型的自动精确配准是工业制造自动化三维检测过程中的重要环节.针对现有精确配准算法对含有无关背景数据的测量点云进行自动配准时稳定性较差的问题,提出了一种基于自适应阈值剔除无关数据的点云配准算法.该算法在迭代过程中通过统计分析匹配点对的欧氏距离计算出自适应阈值,剔除背景数据,采用按百分比选取匹配点对计算点云刚性变换,降低噪声对配准结果的影响.结果表明,该算法相比Trimmed-ICP算法在迭代过程中每次迭代所需计算的匹配点对减少,单次迭代时间逐渐下降,提高了配准的效率.该算法可以实现工业制造自动化三维检测中三维点云与CAD模型的自动精确配准.
展开▼
