ICP算法在双目结构光系统点云匹配中的应用

ICP algorithm for point cloud data matching in a binocular structured light system

双目结构光系统在测量物体时有更好的效果,测量物体视野是单目测量系统的两倍。该文根据针孔成像理论给出了双目结构光重构的数学表达式,并通过Zhang氏标定法给出了标定结果。在双目系统中,投影仪和摄像机的标定误差、仪器设备的系统误差,都会导致得到的两组三维点云数据不能很好地重合。因此,该文提出将标定获得的两个摄像机关系矩阵做为点云匹配的初值,使用改进的最近迭代点(iterative closest points,ICP)算法,加速点云匹配时间,并对经过初值变换的点云数据进行再次匹配,进一步减小系统在标定过程中的误差,从而达到对标定误差进行补偿的目的。实验结果表明:改进的ICP算法使标定后的点云能够很好地重合,并对标定值进行了修正,点云匹配的时间缩短为0.3s。

Binocular structured light systems provide better measurements of objects since the measurement field of view is twice that of a monocular measurement system. Pinhole imaging theory is used here to develop expressions for the object reconstruction with binocular structured light. Calibration results are given using Zhang＇s calibration method. Errors in the projector and camera calibrations and systematic errors in the binocular system will cause errors in matching the two sets of 3 D point cloud data. To overcome this, the matrix obtained from the calibration that represents the relationship between the two cameras is used as the initial value for the point cloud matching, with an improved iterative closest points （ICP） algorithm used to speed up the matching. The initially transformed point cloud data is then re matched, further reducing the error to compensate for the calibration errors. Tests show that the two sets of 3 D point cloud data can be well correlated and thecalibration values can be corrected with the improved ICP algorithm. The time for point cloud data matching is reduced to 0.3 seconds.