步进式区域生长的城区摄影测量点云分类

Classification of urban photogrammetric point clouds based on step-wise region growth algorithm

针对面向对象点云分类中单一分割算法无法满足复杂场景分割需求的问题,该文提出了一种步进式区域生长的倾斜摄影测量点云分割算法。该算法首先进行粗分割,获取光滑分割面片,然后将点云丰富的纹理信息引入精细分割过程,最后,利用可见光波段差异植被指数(VDVI)约束下的连通成分分析合并分割面片,得到最终的分割结果,并在两景具有较强代表性的城区场景下完成了倾斜摄影测量点云的分割分类实验。实验结果表明,该文算法分割的面片形态规则、数量较少,且分类精度高达96.45%、94.44%,相比于单点和传统单一分割算法的分类结果更优,研究结果适用于复杂场景下倾斜摄影测量点云的分类。

According to the fact that there is not a single segmentation algorithm can meet the needs of complex scene segmentation in object-oriented point cloud classification,this paper presented a segmentation algorithm of oblique photogrammetry point clouds based on step-wise region growth.The algorithm of rude segmentation was first carried out to obtain smooth segments.Then,the rich texture information of point cloud was used in the fine segmentation process.Finally,the connected component analysis with visible-band difference vegetation index(VDVI)constraints was used to merge segments,and the final segmentation result was obtained.The experiment of point cloud segmentation and classification in inclined photogrammetry was completed in two typical urban scenes.The results of experiment indicated that our method obtained a small number of regular segments and achieved a superior classification result with an overall classification accuracy larger than 96.45%,94.44%,which is better than single point and single segmentation algorithm.This method can be applied to the classification of oblique photogrammetric point clouds in complex scenes.