目前基于点云面的三维重建方法中,重建的区域性选择存在着两个问题:重建区域过大会导致目标物体不明确,效果不佳,运行时间长;重建区域过小会导致目标物体不完整,信息丢失。针对重建窗口过大时,采用改进的Snake的区域性重建算法,即通过轮廓提取只对窗口内的目标物进行重建;针对重建窗口过小时,采用基于投影面的点云拼接算法,即通过重建后的点云进行拼接的方法使目标物体恢复完整。以上两点改进弥补了点云三维重建及拼接时出现的应用局限性和不稳定性,减少了重建时间,提高重建有效性,鲁棒性。
At present, there are two problems in the regional choice of the 3D reconstruction methods based on point cloud. When the reconstruction area is chosen too big,it will lead to the target object unclear and influence the effect,which also takes long time to reconstruct. When reconstruction area is chosen too small, the target will not be complete, which will result in the lack of the usefull information. For the first case ,this paper adopted 3D reconstruction method based on the improved Snake algorithm to reconstruct the content in the window after the contour extraction of the target object. For the second case, it proposed point cloud mosaic method based on projection plane ,which made the target object complete through point cloud mosaic after the reconstruction. The above two improved methods made up for the limitation and instability of the application of the 3D reconstruction and the point cloud's mosaieing,as well as reduce the reconstruction time,improve the effectiveness and robustness.