论文部分内容阅读
提出了一种基于直角点统计特征的高分辨率遥感全色图像居民地提取方法。首先,分别检测全色图像中的直角点、直线段;然后,利用直角点邻边的长度约束进一步优化直角点;接着,通过构建直角点缓冲区投影的方式生成特征图像;最后,二值化特征图像提取居民地信息。实验结果表明,与角点、直线以及未限制邻边长度直角点的统计特征对比,邻边长度约束的直角点统计特征更能反映居民地信息,有效克服了同样具有较多角点与直线特征的道路、车辆、农田田埂等地物对居民地提取的负面影响。与已有的PanTex指数方法相比,所提方法不依赖于纹理的复杂程度,有效削弱了纹理复杂的非居民地区域对居民地提取的负面影响。
This paper presents a method for extracting high-resolution remote sensing full-color images from quarters based on rectangular point statistics. Firstly, the right-angle points and the straight-line segments in the panchromatic image are respectively detected. Then, the right-angle points are further optimized by using the length constraint of the right-angled points. Then, the feature image is generated by constructing the projection of the rectangular point buffer. Finally, Feature image extraction of residential information. The experimental results show that compared with the statistical features of the corner point, the straight line and the right-angled point of the unrestricted neighbor length, the right-angle statistical features of the constraint of the adjacent edge length can better reflect the information of residential areas and effectively overcome the features of the same corner and straight line Roads, vehicles, agricultural fields and other places on the extraction of residential negative impact. Compared with the existing PanTex index method, the proposed method does not depend on the complexity of the texture, effectively weakening the negative impact of the complex non-residential area on the residential extraction.