基于独立分量分析的高光谱图像目标检测算法

提出一种基于独立分量分析(ICA)的高光谱图像目标检测算法.首先利用无监督正交子空间投影进行端元提取,并将端元矢量构成矩阵作为快速定点独立分量分析的初始化混合矩阵,解决了独立分量在排序上的随机性;同时采用基于噪声调整的主分量分析(NAPCA)对原始图像数据降维,继而采用初始化后的快速独立分量分析从保留的主分量中依次提取出目标.利用AVIRIS高光谱数据进行实验研究,结果表明提出的算法能够有效地提取图像中的目标信息,其性能优于改进的CEM检测算法.

Target Detection Algorithm in Hyperspectral Imagery Based on ICA

The target detection algorithm in hyperspectral imagery based on Independent Component Analysis(ICA)was proposed.The unsupervised orthogonal subspace pnDjeetion operator was used to extract the target endmembers and the initialization mixing matrix of the Fast_ICA was made up of such endmember vectors.This method could solve the ordering randomicity of independent vectors.Meanwhile,the Noise-Adjusted Principal Component Analysis (NAPCA) was used to reduce the dimensionality of the original data.The ICA transformation of the reserved principal components was developed to detect the targets.The experimental results based on AVIRIS hyperspectral imagery have shown that it is more effective than the improved CEM method.