一种无标记点三维点云自动拼接技术

摘要

基于相位投影和双目的三维光学测量系统已经广泛应用于各领域.受投影光测量系统单次测量范围大小的限制,对大型物体的测量需要在表面粘贴圆形标记点进行多次拼接的缺点,探讨一种基于SIFT的无标记点自动拼接技术.该技术采用SIFT方法获取两次测量的特征点,其次结合RANSAC求出图像特征点的匹配关系,再根据立体匹配中图像特征点与三维点云之间的对应关系,将二维特征点的对应关系映射到三维点云的对应关系上,最后由SVD奇异值分解算法求得旋转和平移矩阵实现拼接.实验证明:该方法可以避免在被测量对象上粘贴标记点,能够快速准确地实现自动拼接.%The phase projection and binocular vision based 3D optical measurement system is widely applied to many fields. Due to the limitation of projection light measurement system's single measurement range size along with there being a drawback that it requires time and again pastes of circular markers on the surface when measuring a large object, an unmarked automatic registration technology based on SIFT is explored. The technology uses SIFT method to gain feature points for twice measurements;next,by combining RANSAC, works out the matching relation among image feature points; then, according to the corresponding relation between the image feature points in the stereo match and the 3D point cloud,maps the 2D feature points' corresponding relation onto the 3D point cloud;finally depends on SVD algorithm to solve for the rotation and translation matrix to realize registration. Experiments prove that the method can avoid pasting markers on the measured object;as a result,it can quickly and accurately achieve automatic registration.