基于纹理方向和角点的居民地信息提取

高分辨率遥感影像中规则建筑聚集街区式居民地纹理方向性明显且含有丰富的角点,为了充分利用该特性,以简化滤波器参数设置,提高居民地提取精度及效率,提出基于纹理方向和角点的高分辨率遥感图像居民地信息提取方法。该方法首先对遥感图像进行特定方向角度的Gabor滤波变换;并将滤波器组输出的不同方向滤波影像进行重构。然后对滤波影像做Harris角点检测;并计算检测结果影像的角点密度似然度。最后将角点密度似然影像进行阈值分割以提取影像上的居民地信息。实验结果表明,该方法对乡村及山区居民地信息提取的Kappa系数分别达到0.954 2和0.894 4,保证了居民地提取的精度。同时,算法中各参数的鲁棒性较好,其中中心频率不需要反复的选取试验,从而提高了居民地信息提取的效率。

Information Extraction of Residential Area Based on Texture Directions and Corner Points

The aggregated residential area in the high resolution remote sensing image is quite directional, and its texture of the partial samples in principal direction have good consistency with the overall structure. On the ground of the above-mentioned views, the information extraction of residential area from high resolution remote sensing image based on texture directions and corner points will be proposed. Firstly, the experimental image should be transformed in particular directions through the Gabor filters, and then the likelihood function image is extracted from the Harris corner detection result of reconstructed image. Finally, through the threshold segmentation method, the residential area information can be extracted from the image. The analysis of the experimental results shows that the Kappa coefficient of residential area of village and mountainous areas can reach 0.9542 and 0.8944 respectively. The robustness of each parameter is good, and we do not need to repeatedly test and select the center frequency of the filter. The efficiency of the extraction of residential area information turns out to be higher.