一种遥感图像中建筑物的自动提取方法

目前，传统建筑物识别方法通常采用基于建筑物边缘线性特征的识别算法，这类方法具有简单高效的优点，但识别率较低。该文提出了一种综合利用建筑物的若干特征进行自动识别的方法。首先用Canny算子提取边缘。然后根据建筑物的空间分布特点和Hough变换特性。在Hough变换域进行建筑物边缘方向统计来筛选边缘线段，提取出潜在的目标边缘线段；接着该文提出了对建筑物的几何特征（例如矩形特征、角点特征和阴影特征等）和灰度特征进行识别的算法，将其识别结果做为判定建筑物目标的依据，最终准确地提取出建筑物。大量实验证明该方法相比较单一的线性特征检测方法，速度快、准确率高．具有较强的实际应用价值。

An Automatic Building Detection Method from Remote Sensing Images

Utilizing linear feature is now widely used in building detection.These linear feature-based methods are simple but with low accuracy.This paper proposes a new method of automatical building detection from remote sensing images based on buildings' multi-characteristics.The method first adopts Canny Algorithm to detect edge lines from image.Then based on features of building distribution and Hough transform,it employs directional statistics algorithm to filter edge lines.Then we use a series of geometrical characteristics(such as rectangular characteristic,corner characteristic,shadow characteristic etc) and gray characteristic of buildings as criteria for building judgment to eliminate disturbs,reduce false detection and at last successfully detect the buildings.Large amounts of experiment results show that the method,compared with common linear feature building detection methods,is of high speed,accuracy and good robustness,fit for practical applications.