改进区域增长和RANSAC由粗到精的建筑物分割方法

Improved Region Growing and RANSAC from Coarse to Fine Building Segmentation Method

针对目前机载LiDAR建筑物屋顶面分割精度不高问题,提出了一种结合改进区域增长和RANSAC由粗到精的建筑物分割方法。该方法首先基于LRSCPK计算点云法向量,然后利用最小曲率区域增长算法进行屋顶面粗分割,最后利用RANSAC进行小平面的分割和屋顶面的优化。使用5栋不同复杂程度的建筑物数据验证本文方法,同时与其他两种算法进行比较,结果表明,所提出的方法能够有效地分割不同程度的复杂建筑物,且在小面积的屋顶面有着较好地分割效果。以屋顶面为评价单元计算建筑物分割完整性、正确性、质量的平均值,结果分别为:100%、94.6%、94.6%。

Aiming at the problem of low segmentation accuracy of airborne LiDAR building roof,this paper proposes a building segmentation method combining improved region growing and RANSAC from coarse to fine.Firstly,the method calculates the point cloud normal vector based on LRSCPK.Then,the minimum curvature region growing algoRithm is used for rough segmentation of the roof surface.Finally,RANSAC is used for the segmentation of the facet and the optimization of the roof surface.The proposed method is verified by using the data of five buildings with different levels of complexity,and compared with the other two algorithms,the results show that the proposed method can effectively segment complex buildings to different degrees,and has a good segmentation effect on small roofs.We take the roof surface as the evaluation unit to calculate the average values of building segmentation Comp,Corr,and Quality,and the results are 100%,94.6%,and 94.6%,respectively.