摘要
全极化合成孔径雷达(Pol SAR)图像蕴含更丰富的散射信息,具有更多的可用特征。如何使用这些特征是极化SAR图像分类中非常重要的一步,但是目前尚未对此提出非常明确的准则。为了能够有效地解决上述问题,该文提出一种基于特征加权集成的极化SAR图像分类算法。该算法采用0-1矩阵分解集成方法对包括不同特征的数据集进行学习获得相应加权系数,并通过对每个特征集获得的预测结果进行加权集成来提高极化SAR图像分类性能。首先,输入极化SAR数据,获得极化特征作为原始特征集,并对其进行随机抽取获得不同的特征子集;然后,使用0-1矩阵集成算法得到每个特征值相对应的加权系数;最后,通过对各个特征子集的预测结果进行集成得到最终极化SAR图像分类结果。实测L波段和C波段极化数据的实验结果表明,该算法可以有效地提高极化SAR图像分类的准确度。
For Polarimetric SAR （PolSAR）, because it contains more scattering information, thus it can provide more available features. How to use the features is crucial for the PolSAR image classification, however, there are no existing specific rules. To solve the above problem, a supervised Polarimetric SAR image classification method via weighted ensemble based on 0-1 matrix decomposition is proposed. The proposed method adopts matrix decomposition ensemble to learn on different feature subsets to get coefficients, and weighting ensemble algorithm is employed via the predictive results to improve the final classification results. Firstly, some features are extracted from PolSAR data as initial feature group and are divided randomly into several feature subsets. Then, according to the ensemble algorithm to get the different weights based on the feature subsets, small coefficients are assigned to bad classification results to decrease the harmful impact of some features. The final classification result is achieved by combining the results together. The experimental results of L-band and C-band PolSAR data demonstrate that the proposed method can effectively improve the classification results.
出处
《电子与信息学报》
EI
CSCD
北大核心
2015年第6期1495-1501,共7页
Journal of Electronics & Information Technology
基金
国家973计划项目(2013CB329402)
国家自然科学基金(61271302
61272282
61202176
61271298)
国家教育部博士点基金(20100203120005)资助课题
关键词
极化合成孔径雷达
监督图像分类
集成学习
分类器集成
Polaximetric SAR （PolSAR）
Supervised image classification
Ensemble learning
Classifier ensemble
