论文部分内容阅读
散射熵能较好地反映目标散射的随机性,但忽略了相干矩阵特征分解后3个相干散射成分之间的关系。为了更充分地利用极化信息提取更有效的特征,该文提出一种描述目标散射成分一致性的新参数,并利用该参数进行图像分类。新参数融合了相干矩阵的特征值分布信息与各正交散射成分之间的相似性信息,反映了目标的整体散射机制接近于某种单一相干散射的程度。利用该新特征替代散射熵,先对AIRSAR的旧金山L波段数据进行初始分割,然后进行基于Wishart分类器的迭代调整。实验结果表明:利用该特征能够更准确地实现图像分类,展现地物细节,从而证实了该特征的有效性。
The scattering entropy can well reflect the randomness of the target scattering, but ignores the relationship between the three coherent scattering components after the eigen decomposition of the coherent matrix. In order to make more efficient use of polarization information to extract more efficient features, a new parameter describing the consistency of the target scattering components is proposed and used to classify the images. The new parameter combines the information of the eigenvalue distribution of the coherence matrix and the similarity information of the orthogonally scattered components, reflecting the degree to which the global scattering mechanism of the target is close to some single coherent scattering. Using this new feature instead of scattering entropy, the AIRSAR San Francisco L-band data is initially segmented and then iteratively adjusted based on the Wishart classifier. The experimental results show that the feature can be used to classify the images more accurately and display the details of the features, thus confirming the validity of the feature.