基于多分辨率层次分类的机载LiDAR点云滤波方法

针对复杂地形环境滤波的需求,提出一种具有较强抗差稳定性的多分辨率层次分类滤波算法。该方法首先对原始点云数据进行格网化,主要通过设置格网分辨率值将数据分为3个层次,在每一层通过改进的局部最小值法选取初始地面种子点;然后利用薄板样条函数迭代内插生成栅格曲面,根据定义的残差阈值判定方法,完成点云的分类;分离出的地面点作为下一次迭代的种子点用于构建栅格曲面。利用ISPRS提供的15个训练数据对该方法进行实验,将实验结果与8种经典滤波算法进行对比分析,证明该方法在适应不同地形环境方面具有较强的稳健性。

Airborne LiDAR Data Filtering Method Based on Multi-Resolution Hierarchical Classification

A multi-resolution hierarchical classification(MHC)algorithm with strong robustness was put forward in the light of the needs of complex terrain environment filter.The grid transformation of the original point cloud was the first step in this method,then the data was divide into three levels of hierarchy mainly by resolution value.At each level,the initial ground seed points were selected through improved local-minimum method at first.Then,in each layer,the grid surface was iterative interpolation to generate by the thin plate spline function,according to the improved residual threshold determination method,the classification of ground point and the non-ground point was completed;then the classified ground points were used to update the surface in the next iteration.Fifteen groups of benchmark dataset provided by the International Society for Photogrammetry and Remote Sensing (ISPRS) commission were used to compare the performance of MHC with eight kinds of classic filtering algorithms.The results showed that the MHC filtering method had strong robustness in adapting to different terrain environment.