适应曲线轮廓的机载点云建筑外轮廓提取算法

Adaptive Curve Building Contour Extraction Algorithm of Airborne Point Cloud

针对最小边界矩形(MBR)算法可利用机载LiDAR点云提取正交建筑轮廓,但当建筑轮廓存在斜直线或曲线时该算法会出现锯齿形轮廓的问题,提出了一种可适应曲线的MBR(ACMBR)算法。在改进MBR(IMBR)算法的基础上引入边界点类型划分算法,再利用最小二乘法拟合斜直线轮廓以及限制端点位置的B-Spline拟合曲线轮廓。结果表明,ACMBR算法能保持建筑轮廓的正交性,且能够识别和拟合包含斜直线、曲线的建筑物轮廓,扩展了MBR类型算法的适用范围。对于斜直线和曲线建筑,ACMBR算法的距离均值和均方根误差均小于IMBR算法,提取的建筑轮廓精度也优于IMBR算法。

The minimum bounding rectangle(MBR)algorithm can extract the orthogonal building contour based on airborne LiDAR point cloud.When there are oblique lines or curves in the building contour,MBR algorithm can only represent the building contour with sawtooth segments.In view of this,we proposed a new adaptive curve MBR(ACMBR)algorithm.Based on the improved MBR(IMBR)algorithm,we introduced the boundary point category division algorithm,and used the least square method to fit the oblique line contour and B-Spline limiting endpoint position to fit the curve contour.The results show that ACMBR algorithm can maintain the orthogonality of building contour,and correctly extract oblique line contour and curve contour,which extends the application scope of MBR algorithm.For oblique line contour and curve contour,the distance mean value and root mean square error of ACMBR algorithm are smaller than those of IMBR algorithm,and the accuracy of building contours extracted by ACMBR algorithm is better than that of IMBR algorithm.