从机载激光雷达点云提取建筑物屋顶边界的活动轮廓模型改进方法

Improved Active Contour Model for Building Roof Boundary Extraction from LiDAR Point Cloud

基于边缘与局部信息提出一种处理多波段图像的活动轮廓模型,并将其应用于LiDAR数据的建筑物边界提取。首先将分类得到的屋顶点云数据转换为栅格数据,并作为模型的输入图像,进而采用变分水平集方法解求模型能量函数的最小解,得到建筑物的边界。该模型消除了其他活动轮廓模型对初始曲线和所处理图像类型的限制,适于任意形状的建筑物边界的自动提取;水平集规则项的添加,减小了模型的计算时间。试验结果表明:与IAC模型、GACcolor模型相比,本文模型在建筑物边界提取的应用中可以达到更高的匹配度、形状相似度以及位置精度。

Based on the edge and the local region information,a new active contour model is proposed in this paper,which can process multi-spectral image.And it is used to extract the building roof boundary from LiDAR data.The input image of this model is processed by microstation software,the LiDAR point cloud is classified firstly and then the classified results are converted to raster format.This model is solved by variational level set method,and the minimal solution is the exact building roof boundary.It can eliminate the restrictions on the initialization and the image types of ACM,and it is suitable for the automatic extraction of any shape of building roof boundaries.In addition,the computational time of the new model is reduced by adding the level set rules.Building roof boundary extraction experiment result indicates that this model can obtain higher accuracy in matched rate,shape similarity and positional accuracy than that of the IAC model and the GACcolormodel.