几何与Gabor纹理特征的机载LiDAR点云建筑物提取  

Building extraction from airborne LiDAR point cloud based on geometric and texture features

作　　者：彭泊涵 欧阳杨 张良[3] 赵博 王建楠 PENG Bohan;OUYANG Yang;ZHANG Liang;ZHAO Bo;WANG Jiannan(Beijing Key Laboratory of Urban Spatial Information Engineering,Beijing Institute of Surveying and Mapping,Beijing 100038,China;School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China;Hubei Key Laboratory of Regional Development and Environmental Response,Hubei University,Wuhan 430062,China)

机构地区：[1]北京市测绘设计研究院城市空间信息工程北京市重点实验室,北京100038 [2]武汉大学遥感信息工程学院,武汉430079 [3]湖北大学区域开发与环境响应湖北省重点实验室,武汉430062

出　　处：《测绘科学》2022年第6期119-126,共8页Science of Surveying and Mapping

摘　　要：针对仅基于LiDAR点云的几何特征难以区分高程相似地物,且建筑物轮廓边缘信息提取精度不高的问题,该文提出一种基于几何特征与纹理特征的建筑物点云提取方法。利用圆柱体邻域计算16个与强度、高程、平面和密度相关的空间几何特征;通过Gabor滤波器提取24个纹理特征;再通过ReliefF特征选择前10个最优特征训练随机森林分类器,实现建筑物点分类。通过3组机载点云数据试验,对比仅使用点云几何特征(OP)、几何与纹理特征(OP+TE)、几何与纹理特征并特征选择(OP+TE+FS)3种方法的建筑物提取效果。实验结果表明,加入点云纹理特征并进行特征选择,能够进一步减少点云建筑物的漏提取和误提取现象,具有更高的完整率和准确率。Aiming at the problem that it is difficult to distinguish surface features with similar elevations based on the geometric features of light detection and ranging(LiDAR) point clouds and the accuracy of extracting building contour edge information is low, a point cloud extraction method based on point cloud geometry and texture features was proposed in this paper. Firstly, 16 spatial geometric features related to strength, elevation, plane and density were calculated based on the neighborhood of point cloud cylinder. Then, 24 texture features were extracted from the elevation map obtained by point cloud projection by Gabor filter. Finally, top 10 optimal features in ReliefF feature selection set were selected to train the random forest classifier to classify building points. The building extraction effects of OP,OP+TE and OP+TE+FS methods were compared based on three groups of airborne point cloud data. Experimental results showed that adding point cloud texture features and feature selection could further reduce the phenomenon of missing extraction and false extraction of buildings, which had higher integrity and accuracy.

关 键 词：LIDAR点云 建筑物提取 纹理特征 GABOR滤波器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]