周期性分析下的城市高分遥感影像同形态建筑物群提取

Extraction of Homomorphic Buildings from Urban High-resolution Remote Sensing Imagery Based on Periodic Analysis

针对遥感影像数据建筑物提取算法目标单一、运算复杂等问题,提出了一种利用周期性实现城市一定范围内建筑群提取的方法。通过图像分割、轮廓检测、矩形检测和辅助特征分析,得到建筑群初提取结果,再结合对称性分析、间距统计、灰度震荡规律、属性统计4种方法,分析建筑群周期性排列规律,剔除干扰地物和零散建筑,实现城市某区域同形态建筑群目标提取。在进行建筑物提取时以此作为结束,可以提高检测效率。实验表明,通过探索建筑群周期性排列模式,可准确有效地从遥感图像中识别出规则、集群分布的矩形类建筑物,同形态建筑物群提取精度可达到80%以上。

Aiming at the problems of single extraction target and complex operation of building extraction algorithm in remote sensing image data,a method of extracting buildings within a certain range of a city by periodicity is proposed.The results of initial extraction of buildings are obtained through image segmentation,contour detection,rectangle detection and auxiliary feature analysis,and then the periodic arrangement of buildings is analyzed by combining symmetry analysis,interval statistics,gray scale oscillation rule and attribute statistics to eliminate disturbing objects and scattered buildings,so as to realize the target extraction of homomorphic buildings in a certain area of a city.Taking this as the end of building extraction can improve the detection efficiency.Experimental results show that the regular and clustered rectangular buildings can be accurately and effectively identified from remote sensing images by exploring the periodic arrangement pattern of buildings,and the extraction accuracy of homomorphic buildings can reach more than 80%.