基于高分辨率卫星影像的建筑物轮廓矢量化技术

高分辨率卫星影像包含着地表目标丰富的形状结构和纹理信息 ,使其成为城市研究的重要数据源。建筑物作为城市区域的主体单元 ,其轮廓的矢量信息在城市研究中有着重要的应用价值。二值点阵图的矢量化是图像处理的一个重要课题 ,现有的矢量化方法在提取建筑物轮廓信息时都存在局限性。本文针对矩形建筑物的轮廓矢量化提取技术 ,提出了一种新的基于角点识别和定位的矢量化方法。该方法自动提取建筑物二值图上的轮廓点 ,而后进行以基础点为划分点列组合依据的处理 ,再采用最小二乘法进行分段线性拟合 ,其中关键步骤就是如何准确识别参与拟合的轮廓点群。实验结果证明用该方法提取矩形建筑物轮廓具有运算速度快、效率高、精度满足实际应用需要等优点 ,有较好的实用价值。

A Study on the Vectorization of Building Profile based on High-Resolution Satellite Image

High-resolution satellite image provides abundant shape structure and texture information of landscape objects and it becomes important data source in field of city research. Building is principle part of city zone. Vector information of building profile holds vital application value in city research. The vectorization for the dot matrix image of binary is an important work in image processing. Existing vectorization methods are not good at extracting building profile information. This paper describes a new vectorization approach based on corners detection and positioning for the vectorization extraction of rectangle building. After automatically extracting binary image of building profile, the approach establishes combination way of boundary dots grouping by basal point, then adopts piecewise linear fitting by least-square algorithm, in which the key step is how to accurately detect group of boundary dots joining in fitting. The experimental results show that the approach is much fast in vectorization of rectangle building with robustness and higher precision.