面向未知复杂曲面的视点自主规划方法

A Method for Automatic View Planning of Unknown Complex Surfaces

针对未知复杂曲面的视点规划问题,提出一种基于边界检验进行视点自主规划的方法。基于初始视点采集的曲面点云数据建立网格模型,依据边界边与三角形的关联属性,提取当前网格模型的边界;基于边界边之间的连通性和边界的封闭性,聚类闭合边界,进而根据闭合边界的相对长短区分外边界与孔洞边界;将外边界依据边界间方向向量的夹角偏差进行分段,对应不同潜在趋势面,采用二次曲面拟合估计趋势面并求导计算法矢,结合主成分分析法确定边界最大主成分方向进行视点规划;然后基于孔洞的中心、大小与中心的法向量规划孔洞区域的观测视点;最后使用规划的外边界和孔洞的视点观测当前模型边界,进行真伪边界判断自主决策视点规划是否终止。以汽车后保险杠为规划对象,试验结果表明:该方法能够自主实现未知复杂曲面三维测量时的视点规划任务,平均每个视点规划用时67 ms,总体建模时间用时约6 min;累积覆盖率达到100%,在扫描时间和完整性方面具有良好性能。

To solve the problem of viewpoint planning for unknown complex surfaces,an autonomous viewpoint planning method based on boundary test was proposed.A mesh model was built based on the surface point cloud data collected from the initial view⁃point,and the boundary of the current mesh was extracted according to the associated attributes of the boundary edge and the triangle.The closed boundary was clustered based on the connectivity between the boundary edges and the closure of the boundary,then the outer boundary and the hole boundary were distinguished according to the relative length of the closed boundary.The outer boundary was segmented according to the angle deviation of the direction vector between the boundaries and each segment and they were corre⁃sponded to different potential trend surfaces.Quadric surface fitting was used to estimate the trend surface and derive the algorithm vector.Combining the principal component analysis method,the direction of the maximum principal component of the boundary was deter⁃mined for viewpoint planning.The observation points of the hole area were planed based on the center,size and normal vector of the center of the hole.Finally,the outer boundary of the plan and the viewpoints of the hole were used to observe the boundary of the cur⁃rent model,then the authenticity of the boundary was judged and it was decided whether to terminate the viewpoint plan independently.Taking the rear bumper as the planning object,the test results show that the method can be used to realize the viewpoint planning task for the three-dimensional measurement of unknown complex curved surfaces autonomously;the average planning time for each view⁃point is 67 ms,and the total modeling time is about 6 min;the cumulative coverage rate reaches 100%,and it has good performance in scanning time and integrity.