中文摘要：

为改善对LiDAR数据点滤波分类的有效性与准确性,在形态学滤波方法的基础上,提出了基于开重建的LiDAR点云形态学滤波算法。首先对格网化DSM进行腐蚀运算获得标记图像,通过对标记图像反复进行测地膨胀运算实现开重建过程,然后利用白顶帽变换得到nDSM实现地物与地面点的正确分类。使用ISPRS提供的测试数据的实验结果表明,该方法的Ⅰ类、Ⅱ类及总误差均值分别为3.51%、7.20%和4.26%,与同类滤波方法相比,在Ⅱ类误差增幅不显著的情况下,15个样本区的总误差均值和Ⅰ类误差均值均达到最小。

英文摘要：

To improve the data classification effect of LiDAR points filtering,a method of morphological opening by reconstruction is proposed on the basis of traditional morphological filtering algorithms. In this method,marker image will be got from grid DSM by erosion firstly,then opening by reconstruction process is finished through geodesic dilating on the marker image repeatedly. Finally,nDSM is achieved by white top-hat reconstruction to realize correct classification of ground objects and ground points. The experimental results with ISPRS data show that this approach is able to classify ground and non-ground points effectively. Mean error of type Ⅰ is3. 51%,mean error of type Ⅱ is 7. 20%,and total error is 4. 26%. Under the condition of error of typeⅡkeeps the same level,mean error of typeⅠand total error are minimum,compared to traditional filtering methods.