针对异型结构全体积质量缺陷检测时需要可靠的结构外表面空间位置信息,提出一种高度还原异型结构真实三维场景的方法.该方法首先提取点云数据中的特征点后通过快速点特征直方图(fast point feature histograms,FPFH)描述特征点,接着采用随机采样一致性(random sample consen-sus,RANSAC)算法对点云进行粗配准,随后采用Iterative Closest Point算法进行精配准,来对采集到的点云数据进行拼接,最终得到完整异型结构承载件外表面点云数据;实验结果显示,所提出的算法和传统ICP配准算法比较精度提升了74.3％,与RANSAC+ICP配准方法提升了29.5％.优化后的算法在点云配准方面表现出很好的鲁棒性,为实际应用提供了极大的指导价值,能够显著提高点云配准的效率.
Research on the Registration Method of Heterogeneous Structure Multi-View Point Cloud
In view of the need for reliable outer surface spatial position information in the total volume quality defect detection of deformed structures,a method for highly restoring the real three-dimensional scene of deformed structures is proposed. Firstly,the feature points in point cloud data are extracted and then the feature points are described by fast point feature histogram ( FPFH) . Then random sample consen-sus ( RANSAC) is adopted algorithm carries out rough registration of Point clouds,and then adopts Itera-tive Closest Point algorithm to carry out precise registration to splice the collected point cloud data,and fi-nally obtain the complete point cloud data on the surface of the heterogeneous structural bearing parts. Ex-perimental results show that the accuracy of the proposed algorithm is improved by 74.3％ compared with the traditional ICP registration algorithm,and by 29.5％ compared with the RANSAC+ICP registration method. The optimized algorithm shows good robustness in point cloud registration,provides great guiding value for practical applications,and can significantly improve the efficiency of high point cloud registration.
heterotypic structuremachine visionpoint cloud registrationfeature pointsmulti-angle of view
万方数据
维普
