基于向量角度差和拟合曲线融合的建筑物点云提取方法

为了提升城市建筑物提取完整度并消除植物的影响,提出一种向量角度差和拟合曲线融合的建筑物提取方法。该方法以扫描线为研究对象,通过计算相邻方向向量角度差提取平面屋顶;利用曲线拟合计算同一扫描线内数据点的残差值,基于多数抽样样本建立残差阈值进而提取曲面屋顶;最后,通过细化后处理完成建筑物提取。选取典型城市实验区数据集,并与成熟的不规则三角网法(triangulated irregular network,TIN)进行对照实验。结果表明:所提方法能够更有效地提取曲面建筑以及抑制植物的影响,测区总体精度和kappa系数分别为98.96%和0.9148,优于TIN算法。该方法具有运算简单,稳定性高等特点,为有效解决曲率复杂和具有高差突变性的建筑物屋顶提取问题提供了思路。

Building Point Cloud Extraction Method Based on Fusion of Vector Angle Difference and Fitting Curve

In order to improve the completeness of the extraction of urban buildings and eliminate the influence of plants, a building extraction method combining vector angle difference and fitting curve is proposed. This method takes the scan line as the research object, extracts the flat roof by calculating the angle difference of adjacent direction vectors; uses curve fitting to calculate the residual value of the data points in the same scan line, and establishes the residual threshold based on most sampling samples to extract the curved roof; Finally, the building is extracted through refined post-processing. Select a typical urban experimental area data set, and conduct a control experiment with the mature Irregular Triangulation Method (TIN). The results show that the method proposed in this paper can more effectively extract curved buildings and restrain the influence of plants, the overall accuracy of the survey area and the kappa coefficient They are 98.96% and 0.9148 respectively, which are better than the TIN algorithm. This method has the characteristics of simple operation and high stability, and provides an idea for effectively solving the problem of building roof extraction with complex curvature and sudden height difference.