基于空频域纹理特征的高分辨率遥感图像居民地提取

为了解决高分辨率遥感图像中居民地信息因光谱和结构复杂度高造成的提取精度低，速度慢等问题。提出一种基于空频域纹理特征的高分辨率遥感图像居民地提取算法，该算法首先对高分辨率居民地图像分别进行特定方向Gabor滤波和分形维数的计算，然后依据得到的空频域纹理图像的局部纹理灰度特征对居民地信息进行提取，最后对提取初步结果进行形态学优化得到最终的提取结果。实验结果表明，该算法对乡村地区和山区居民地信息提取的总体精度达到97%以上，与传统的分形维数方法和Gabor滤波方法相比，误提率降低了45%以上。实现了全自动、有效的提取平原、山区两种地貌的居民地信息。

Extraction of Residential Areas from High-Resolution Remote Sensing Images Based on Spatial and Frequency Texture Features

In order to solve the problem of low accuracy and speed in extracting residential area information from high-resolution remote sensing images, an extraction method based on spatial and frequency texture features is proposed. Firstly, images were subjected separately to fractal dimension computation and specific direction Gabor filtering to obtain both spatial and frequency texture features. Then, residential area information was extracted according to the local gray-scale features of the two texture features. Finally, the extraction results were morphologically optimized. The experimental results show that the Kappa Coefficient of the proposed algorithm reach up to 0.97. Compared to the traditional fractal dimension and Gabor filtering methods, the Commission Error of the proposed algorithm is reduced by more than 45%. The proposed method achieves automatic and effective extraction of residential area information in both rural region and mountainous region.