一种MBR约束下的高分光学影像直角建筑物提取与标绘方法

针对高空间分辨率遥感影像中建筑物信息提取与标绘问题，提出了一种MBR约束下的高分光学影像中直角建筑物信息提取与标绘方法。首先采用多尺度影像对象分割与CART决策树分类技术，提取影像中的建筑物区域；其次用Candy算子提取出建筑物的粗轮廓，并将其转化为点集形式表示；然后通过轮廓点集计算建筑物最小外包矩形（MBR），对建筑物的轮廓进行分段拟合与优化；最后通过交点方向决策器确定建筑物的角点，依次连接各角点实现建筑物的标绘。通过计算建筑物的面积与周长，确定周长相对精度为93.3%，面积相对精度为96.1%，本文方法可以有效提高建筑物的标绘精度。

A Method for Right Angle Building Extracting and Mapping Based on MBR Constraints in High Resolution Optical Image

Aiming at the problem of building extraction and mapping in high spatial resolution remote sensing images,a method based on MBR constraints is proposed.Firstly,this paper extracts the building from the image by using the multi-scale segmentation method and CART decision tree classification technology. Secondly,this paper extracts the rough contour of the building by using Candy operator,and expresses as point set for calculating building MBR,piecewise fitting and optimization the contour of buildings.Finally,this paper through the point direction decision maker seeks building's angular points and realizes building's mapping.After calculating building's area and perimeter,we can ensure that perimeter relative accuracy is 93.3%,area relative accuracy is 96.1%.The presented method can effectively improve the precision of buildings mapping.