基于三维重建数据的双向点云去噪方法研究  

Two-way point cloud denoising method based on three-dimensional reconstruction data

作　　者：刘辉[1] 王伯雄[1] 任怀艺[1] 罗秀芝[1] 

机构地区：[1]清华大学精密仪器与机械学系精密测试技术及仪器国家重点实验室,北京100084

出　　处：《电子测量与仪器学报》2013年第1期1-7,共7页Journal of Electronic Measurement and Instrumentation

摘　　要：逆向工程中,三维点云数据的重构是最终获取物体形貌的关键一环,如何从复杂的点云中去除噪声干扰是人们一直研究的重点。根据测量环境中产生的不同类型噪声数据进行分析,对于物体轮廓外的噪声点,采用人机交互的方式直接去除噪声点,对于物体轮廓内单个明显的噪声点采用相邻点加权平均的方法去除,而对轮廓内连续错误的噪声点,采用了双向去噪的方法,即在点云的垂直方向通过相邻点夹角的大小去除起伏明显的噪声点,在点云的水平方向通过相邻点云水平距离的加权平均来调整水平方向其余点的距离,最后应用最小二乘法拟合得到去除噪声点处的三维坐标。结果表明该种方法不仅有效去除了点云数据中夹杂的噪声,还对物体表面失真的数据进行了有效地调整和修改,获得了连续光滑的测量表面,具有良好的效果。In reverse engineering, reconstruction of the 3D point cloud data is the key step to the final profile of the object. How to remove the noise data from the complex point cloud data is the focus of the research. In this paper, the data ac- cording to the different types of noise in measurement environment are analyzed. The noise points for the outside contours of the object is removed using human-computer interaction, the single obvious noise point for the inside contours of the object is removed using adjacent points weighted average method, and the continuous wrong noise points for the inside contours of the object is removed using two-way denoising method. In the vertical direction of the point cloud, the ups and downs of the obvious noise point is removed using the inclination between the adjacent points. In the horizontal direction of the point cloud, the horizontal distance of the rest of the point cloud is adjusted by weighted averaging of horizontal distance of the adjacent point cloud data. Finally, the three-dimensional coordinates at the denoising points are got using least square method. The results show that this method is not only effective to remove the noise which are included in the point cloud data, but also to adjust and modify the distortion of the data about the surface, and to obtain a continuous smooth surface.

关 键 词：点云去噪 最小二乘法 数据重构 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]