基于小波与遗传算法的特征提取与特征选择

高维遥感数据的分类与识别与传统的多光谱遥感分类技术具有明显的区别。本文提出了一种基于遗传算法和小波／小波包分析相结合的特征提取方法用于高维遥感数据降维与分类。该方法综合了遗传算法的全局优化和小波／小波包分析的多尺度、多分辨率的特点。首先,通过离散的小波变换(DWT)或小波包变换(WP)将高光谱信号变换到特征域进行光谱分解。由于DWT变换是一种线性变换,不同尺度的DWT系数可作为线性光谱特征。然后,对这些线性光谱特征利用遗传算法结合训练样本计算类内／类间距离搜索最优分类子集,其具体染色体编码取可能的特征号,适应度函数基于样本平均Jeffries-Matusita距离计算。所用的分类器采用最大似然分类器。试验结果表明该方法与常规特征提取算法如主成分变换(PCA)、判别分析特征提取(DAFE)、决策边界特征提取(DBFE)相比,能提高分类精度约1.1％-6.5％。

Feature Extraction and\n\rFeature Selection Based on Wavelet and Genetic Algorithm

Classification and pattern recognition of high dimensional remote sensing data are distinctly different from traditional multi-channel remote sensing classification techniques. In this paper, a newly integrated feature extraction algorithm based on GA and wavelet/wavelet packet (WP) transform is proposed for high dimensional data reduction and classification. The proposed algorithm combines the advantages of GA's global optimization and wavelet's multiresolution and multi-scale analysis. Hyperspectral signals are firstly transformed to feature domain by using a discrete wavelet or wavelet packet decomposition strategy. Since the discrete wavelet transform (DWT) is a linear transform, the DWT coefficients at specific scales could be directly used as linear features. Followed by the decomposition phase is optimal feature subset selection, in which the optimal feature subset acquired the best divergence is obtained according to interclass/intraclass distance of the training samples. This procedure is implemented by a Genetic Algorithm, with each possible feature subset encoded as chromosome. Fitness scores in GA are calculated and evaluated based on Jeffries-Matusita distance of the selected training samples. Hyperspectral data are classified with maximum likelihood classifier ( MLC). Experimental results show that the use of DWT/WP and GA-based feature extraction technique improves the overall classification accuracy by 1. 1%-6. 5% , as compared to the use of conventional feature extraction techniques, such as principal component analysis ( PCA) , Discriminant Analysis Feature Extraction ( DAFE) and Decision Boundary Feature Extraction ( DBFE).