复杂城区建筑物和植被分类研究

Study on classification of building and vegetation in complex urban area

针对城市复杂环境点云数据中建筑物和植被较难分离的问题,提出利用二者高度、面积、坡度等信息先将建筑物分离,再利用建筑物结果作为约束,较为准确地将二者进行分离的方法。首先将倾斜摄影密集匹配点云数据进行插值生成DSM,针对非地面点集,依据建筑物规则特征,将其从地物中首先分割出来。其次,利用最小二乘拟合平面法获得近似建筑物屋顶的拟合平面,经过边界优化得到准确建筑物边界。最终将建筑物作为约束条件,提取植被信息。实验结果表明本文提出的方法较面向对象法有更好的分类结果,且能得到地物三维信息,具有较强实用性。

To solve the problem that buildings and plants were difficult to separate from each other in complicated urban area,this research came up with a method making use of the information of height,area and slope and so on to separate them.Secondly,using results of buildings as constraints,at the end of the experiments the two were separated more accurately.First of all,DSM was built through interpolation of point clouds gained from oblique aerial images.Buildings were split out from mess surface features,on basis of the regular character of them against the non-ground point set.Then,fitting plane of building roofs were got by least-square plane fitting method,acquiring accurate boundary of buildings at the end by means of optimization of edges.Finally,information of plant was extracted with the building boundary as constraint condition.Experiments show that the way raised by this paper has better classification result than object-oriented method,even can get three dimensions information of features,hence it is relatively practical.