自动识别航空CCD图像上建筑物的方法

New method to recognize buildings from airborne CCD image

针对高分辨率航空CCD图像上建筑物的特点,提出了一种自动识别规则建筑物的新方法。通过系统级几何校正和平滑处理实现图像的预处理,组合拉普拉斯图像锐化和阈值分割方法准确检测CCD图像上的边缘信息。应用二值图像像元的邻接连通量提取线段信息,提出了根据图像双向投影图确定建筑物角点位置的方法,并对角点位置序列进行搜索和匹配识别,从而提取建筑物轮廓信息。通过实际的航空CCD图像处理说明方法是可行的,识别的有效率达到85%左右。

A new method to effectively recognize regular man-made buildings from high resolution airborne CCD image is proposed. The airborne CCD image is rectified systemically and noise in the image is also smoothed in order to provide good quality image. A method in combining laplacian sharpening algorithm with threshold segmentation is adopted to detect edges and connectedness of pixels is used to extract possible building lines. The possible vertexes of building are determined based on cross direction projection and these possible building vertexes are matched with possible building lines to find the best suitable vertexes that belong to the same building. Real airborne CCD image are processed and the result is satisfactory with rate of correct recognition up to 85%.