基于外接圆方法的点云法向量提取

Extraction of Point Cloud Normal Vector Based on Circumcircle Method

在线结构光测量中,为快速准确地提取点云的法向量,提出了一种外接圆方法。通过图像细化法求得光条纹初值中心点集,然后判定相邻三点是否共线:若共线,该点初始法线垂直于该直线;否则,这三点可以确定一个外接圆,连接外接圆心坐标和该点坐标得到该点的初始法向量。为减震荡,需对相邻三条初始法向量做平均,得到最终法向量。通过与Hessian矩阵法和Sobel梯度法进行对比,表明该方法能够准确提取法向量,且处理一幅320 pixel×240 pixel的图像的时间低于8 ms,能够满足实时性要求。

In linear structured light measurement, in order to extract the point cloud normal vector quickly and accurately. A circumcircle extraction method is presented. Every three consecutive points are collinear or not is determined after the initial light stripe centers have been obtained by refining method. If they are coUinear, the normal slope can be calculated by the points. Else, the three points determine a circle. The normal can be gotten by connecting the circle center and the initial center point. In order to reduce the shock, the average value of each three normals are calculated as the last normal. Experimental results show that the presented method can accurately extract the normal in comparison with Hessian matrix method and Sobel gradient method, and the time to process a 320 pixel x 240 pixel image is less than 8 ms, which can meet the real-time requirements.