论文部分内容阅读
本文提出了一种基于多尺度图分解的极化SAR图像分割算法。该算法首先构造亲和矩阵,得到无向连接图,然后通过对数据进行多尺度采样,构造多尺度连接图,最后结合规范切准则,引入并行处理思想,实现数据的快速有效分割。其中在构造亲和矩阵时除了利用强度和轮廓参数,还兼顾极化SAR数据的特性,引入了反Wishart距离。最后分别利用星载和机载极化SAR数据进行对比实验,验证了算法的可行性和有效性。
This paper presents a polarimetric SAR image segmentation algorithm based on multi-scale graph decomposition. First of all, the algorithm constructs the affinity matrix and obtains the undirected join graph, and then constructs the multiscale concatenated graph by multiscale sampling the data. Finally, with the help of canonical slicing criterion, the parallel processing idea is introduced to realize the fast and efficient segmentation of data. The anti-Wishart distance is introduced in the construction of the affinity matrix in addition to the use of intensity and profile parameters and the characteristics of polarized SAR data. At last, the contrast experiments of on-board and on-board polarimetric SAR data are carried out to verify the feasibility and effectiveness of the algorithm.