论文部分内容阅读
针对传统全极化合成孔径雷达(SAR)图像信息提取方法存在的问题,结合林地、居民地的散射机理,文章提出了一种综合多特征(多种极化特征、几何形状和尺度特征)的全极化SAR林地和居民地信息提取方法。该方法采用分形网络演化算法实现综合多特征的多尺度分割;基于对象选择极化特征并制定分类规则来提取林地和居民地。实验结果表明,该方法能有效提取研究区的林地和居民地,结果明显优于H-α-A-Wishart分类方法。
Aiming at the problems existing in the traditional method of image information extraction for fully-polarized synthetic aperture radar (SAR), combining with the scattering mechanism of woodland and residential area, this paper proposes a comprehensive multi-feature (multi-polar features, geometric shapes and scale features) Method of extracting information from all - polarized SAR forest land and residential area. This method uses the fractal network evolutionary algorithm to realize the multi-scale segmentation of multi-feature synthesis. The forest land and residential area are extracted based on the polarization characteristics of the object and the classification rules. The experimental results show that this method can effectively extract the forest land and residential area in the study area, and the result is obviously better than the H-α-A-Wishart classification method.