提取建筑物轮廓线是机载激光雷达(LiDAR)点云数据特征提取的热点。为了获得较高精度的建筑物轮廓线,提出了一种基于方向预测规则化算法的机载激光雷达建筑物正交轮廓线提取方法。首先利用α-shape算法提取轮廓点,然后利用改进的Douglas_Peucker算法提取关键点并提出角度检验规则筛选关键点,使用随机抽样一致性算法简化轮廓线,最终用提出的方向预测算法进行轮廓线规则化。通过Vaihingen城区数据对算法进行验证,结果表明:与流行的分类强制正交算法相比,方向预测规则化算法最大偏差平均减小了43.1%,均方根误差平均降低了39.7%,建筑物占地面积相对误差平均降低了7.02%,点云贡献率平均提高了9.32%,有效减小了机载激光雷达点云建筑物正交轮廓线规则化误差。

Extraction of building boundary is a hot issue in airborne light detection and ranging (LiDAR) point cloud data feature extraction. In order to obtain high-precision building boundary, we proposed a building orthogonal boundary regularization algorithm based on directional prediction. First, the boundary points are extracted by α-shape algorithm, then the boundary key points are extracted by the improved Douglas_Peucker algorithm, the key points of angle check rules are proposed to select the right key points, the boundary are simplified by random sample consensus algorithm, and finally the regular boundary is got by the proposed direction prediction algorithm. The algorithm is verified by the Vaihingen data released, and the results show that, comparing with the popular classification forced orthogonal algorithm, the proposed algorithm reduces the maximum absolute deviation by an average of 43.1%, reduces the root mean square error by an average of 39.7%, reduces the relative error of the building area by an average of 7.02%, while increases the point cloud contribution rate by an average of 9.32%, and it can effectively reduce the error of building orthogonal boundary regularization of airborne LiDAR point cloud.