一种新的高分辨率遥感影像城区提取方法

Automatic Urban Area Extraction Using a Gabor Filter and High-Resolution Remote Sensing Imagery

利用城区特有的局部特征,提出了一种新的高分辨率遥感影像城区提取方法。该方法首先对原图像做多角度和多频率组合的Gabor变换,然后利用Ostu方法阈值分割所有变换结果,并在每个中心频率上对各个方向的阈值分割图做逻辑与运算,其次根据运算结果图上Gabor特征的分布情况,确定适合影像的最优中心频率,最后利用滤波器在最优中心频率上的特征提取结果,结合高斯函数构建空间投票矩阵,最终提取城市区域。实验表明,该方法可以成功地提取高分辨率影像城市区域,且具有较高的准确度。

A new automatic urban area extraction method from high-resolution remote-sens- ing imagery that exploits the unique local features of urban area is presented in this paper. The proposed algorithm contains the following steps. First, it obtains the filtering response images with Gabor filters grouped at various central frequencies and orientations; Secondly, we use the Ostu^s method to implement threshold segmentation,and then realize the logical and operation in the various orientations of every central frequency~ Thirdly, we determine the optimal central frequency with the Gabor features distribution information~ Finally, with the information above, we extract the urban area by forming the spatial voting matrix with the Gaussian function. Experimental results show that the approach is able to detect the ur- ban areas in the high-resolution remote-sensing imagery. The results of a performance evalu- ation also support the high precision of this approach.