基于对数极坐标变换的遥感影像居民地提取

Extraction of residential areas from remote sensing images based on log-polar transformation

本文研究基于对数极坐标变换和傅里叶变换实现全色影像居民地提取。首先对影像进行对数极坐标变换,将影像的旋转信息转换为极坐标系的行平移,然后对变换后的极坐标影像进行傅里叶变换,以傅里叶变换功率为旋转不变特征实现居民地的提取;针对提取结果的不足,采用数学形态学进行后续处理。实验表明:采用该方法可以有效提取遥感影像中的居民地。

The residential area panchromatic remote sensing image extraction method of based on fourier transformation and log- polar transformation was presented in the paper. First, the panchromatic image was translated by log-polar transform to convert the rota- tion to translation in different directions, then the fourier transform was used for the texture image of log-polar transform, and at last, the power of fourier transform was used to segment the texture as rotation invariant feature. Moreover, a mathematical morphology meth- od was used for subsequent processing. Experimental results showed residential areas could be acquired from the remote sensing images by the method.