基于特征选择的双边滤波点云去噪算法

Bilateral filtering denoise algorithm for point cloud based on feature selection

为了去除与真实点混合在一起的噪声并更好地保留特征,将点云噪声分为3类,将其中与真实点混合在一起的数据点称为第3类噪声点,利用改进的双边滤波算法去除该类噪声点.首先,利用邻域点判断该点属于特征点还是非特征点;然后,根据不同范围的点云来计算特征点和非特征点的双边滤波因子,实现基于特征选择的双边滤波点云去噪.利用该算法对手持三维激光扫描仪获得的盒子及工业构件的激光点云数据进行平滑去噪处理.结果表明,所提算法在去除噪声的同时可以有效保持被扫描物体的特征,避免出现因双边滤波不能兼顾邻域点特征而产生的过度光顺现象.

In order to remove the noise mixed with the real points and retain characteristics, the noise points are divided into three categories, among which the ones mixed with the real points are called as the third category noise points. By using the improved bilateral filtering algorithm, this kind of points can be removed. First, the points are judged to be feature points or non-feature points