融合LiDAR点云与正射影像的建筑物图割优化提取方法

摘要

提出一种基于图割算法的建筑物LiDAR点云与正射影像融合提取方法.首先,利用LiDAR点云计算3个几何特征:平整度、法向量分布和高程纹理一致性.同时利用航空正射影像计算颜色特征———归一化植被指数(NDVI).然后将两类特征联合构建能量函数数据项,综合数字表面模型(DSM)和NDVI构建平滑项,采用图割算法优化得到初始的建筑物区域.最后利用初始建筑物边缘一定范围内的正射影像颜色信息,采用前后景分割的思想进一步优化建筑物边缘.应用ISPRS V aihingen测试数据进行试验,结果表明本文方法具有较高的建筑物提取精度.%An automatic building extraction method based on graph cuts algorithm fusing LiDAR point cloud and orthoimage is proposed.Firstly,three geometric features are computed from LiDAR points including flatness,distribution of normal vector and GLCM(greylevel co-occurrencematrix)homogeneity of normalized height.NDVI is simultaneously calculated from orthoimage.After that,both kinds of features are combined to construct the data term of energy function,then DSM and NDVI is combined to construct smooth term.Thereafter,graph cuts algorithm is applied to obtain the initial building extraction results.Finally, foreground and background segmentation method is employed to optimize the building boundary based on the orthoimage color information in certain range of the initially detected building boundary.ISPRS Vaihingen dataset is used to evaluate the proposed method.The results reveal that the proposed method can obtain high accuracy of the detection building area.