论文部分内容阅读
遥感图像分类对地面背景红外辐射特性仿真具有重要作用,提取的特征的性能直接影响分类精度。本文以高分辨率遥感图像为研究对象,提出了一种结合ReliefF和mRMR算法的特征降维算法,首先,通过ReliefF算法计算出各特征的权重系数,对特征集进行加权;然后利用mRMR算法选出与类别具有最大相关性且相互之间具有最小冗余性的特征。实验采用提出的算法对原特征空间进行优化,然后基于优化后的特征空间进行遥感图像自动分类,结果表明此方法能较好的提高分类精度。
Classification of remote sensing images plays an important role in the simulation of terrestrial infrared radiation characteristics. The performance of the extracted features has a direct impact on the classification accuracy. In this paper, a high-resolution remote sensing image is taken as the research object, a feature reduction algorithm combining ReliefF and mRMR algorithm is proposed. Firstly, the weight coefficients of each feature are calculated by ReliefF algorithm, and the feature set is weighted. Then mRMR Out features that have the highest correlation with categories and have minimal redundancy with each other. The proposed algorithm is used to optimize the original feature space, then the remote sensing image is automatically classified based on the optimized feature space. The results show that this method can improve the classification accuracy.