论文部分内容阅读
该文提出了基于Fisher准则相对最优极化(Fisher-OPCE)的监督极化SAR图像的分类方法。首先,结合广义相对最优极化的思想,利用3个反映目标极化散射特性的参数对Fisher-OPCE进行了改进。以改进的模型为基础,提出了一种类似单边二叉树的分类方法,以保证功率差别较大的两类地物的错分现象尽量小;其次利用极化参数组合的系数对分类结果进行了优化。利用NASA/JPL的AIRSAR系统对美国旧金山地区的实际观测数据进行分类,结果表明用此方法可以清晰的显示出分类地物的纹理信息,每类目标的散射特性保持一致,实验结果验证了该方法的有效性。
This paper proposes a classification method of supervised Polarimetric SAR images based on the Fisher-OPCE. First, Fisher-OPCE is improved by using the three parameters that reflect the polarization characteristics of the target in combination with the idea of generalized relative optimal polarization. Based on the improved model, a classification method similar to unilateral binary tree is proposed to ensure that the misclassification phenomenon of two types of ground objects with large power difference is as small as possible. Secondly, the classification results are carried out by using the coefficients of polarization parameter combination optimization. Using the NASA / JPL AIRSAR system to classify the observed data in the San Francisco area of the United States, the results show that this method can clearly display the texture information of the classified objects, the scattering characteristics of each type of objects are consistent, and the experimental results verify the method Effectiveness.