面向对象的机载LiDAR数据建筑物提取

基于LiDAR数据,提出一种由粗到细的面向对象的建筑物自动提取方法.首先通过机载LiDAR数据构建出归一化数字表面模型(normalized digital surface model, nDSM),利用首尾两次回波高程计算出归一化差值(normalized difference, ND),并采用形态学运算消除边缘特殊回波点.基于nDSM和ND数据,依据建筑物的高程及穿透性信息,用阈值分割法进行建筑物粗提取.结合nDSM和ND数据以及强度信息,对粗提取得到的备选建筑物采取多尺度分割,合并亮度值相差较小的邻近分割结果对象,达到对分割结果的优化处理.最后利用目标对象的亮度、形状、面积和空间关系等特征,完成建筑物的精提取.实验结果表明,该方法可得到较高精度的建筑物信息,是基于机载LiDAR数据提取建筑物的新思路.

Object-Based Building Extraction from Airborne LiDAR Data

Based on airborne LiDAR data, an object-based method for building extraction with coarse-fine accuracies is proposed. A normalized digital surface model (nDSM) and the normalized difference (ND) are extracted from the LiDAR data. Special edge echo points are removed from the ND data using a morphological operator. Taking into consideration height and penetrability of the buildings, coarse profile of the buildings are extracted from ND and nDSM data using a threshold segmentation algorithm. A multi-resolution segmentation algorithm is then used to segment the candidate buildings by integrating intensity, and the nDSM/ND data. The segmentation result is further optimized by merging the adjacent objects with smaller brightness difference. Finally intensive buildings are extracted based on spectral and geometrical characteristics, and the spatial relations of objects. Experiment results show that the proposed method can obtain buildings with high precision, and provides a means for building extraction from airborne LiDAR data.