基于超体素的机载LiDAR点云滤波与建筑物提取方法

Super Voxel-based Airborne LiDAR Point Cloud Filtering and Building Extraction Method

为了提高目前机载LiDAR点云数据滤波和建筑物提取精度,本文提出一种基于超体素的点云滤波与建筑物提取方法。对每个超体素进行相似性判断来实现向上一层的区域增长式合并,并通过判断超体素簇与邻接超体素簇的凹凸性实现再向上一层的合并,最终通过计算聚类对象自身几何特征来过滤地面、植被和噪声,从而得到建筑物点云。实验利用ISPRS公布的标准数据集,选择3个地形数据进行实验。实验结果表明本文算法滤波性能更好,建筑物提取结果优于专业点云处理软件的提取结果。

In order to improve the current airborne LiDAR point cloud data filtering and building extraction accuracy,this paper proposes a super voxel-based point cloud filtering and building extraction method.The similarity judgment of each super voxel is performed to achieve the area-growing merging in the upper layer,and the merging in the upper layer is achieved by judging the concavity of super voxel clusters and neighboring super voxel clusters,and finally the ground,vegetation and noise are filtered by calculating the geometric features of the clustered objects themselves to obtain the building point cloud.The experiments are conducted using the standard dataset published by ISPRS,and three terrain data are selected.The experimental results show that the filtering performance of the algorithm in this paper is better than the commonly used filtering algorithms,and the building extraction performance is better than the extraction results of professional point cloud processing software.