Target Detection of Multi-spectral Image Based on PCA and ICA

Abstract

Spectral features of various materials are different from each other,so the target can be detected among a group of ones with different materials based on the principle.Usually,the targets with the same features,such as color and shape, can not be detected in the panchromatic images;or the traditional methods can not give approving results because of the large scale of data and low precision of detection.So,this paper presents a new target detection method for multispectral images and hyperspectral images based on the Principal component analysis (PCA) and Independent component analysis(ICA).The data of multispectral images is firstly processed by PCA method in order to reduce the dimension of original multispectral images and take the redundancies out for getting the main components of images.The disposed data will be processed using the independent component analysis(ICA) algorithm,and then the target can be detected from the abstracted spectral features.This paper takes the advantages of PCA method and ICA method,and makes the target detection in multispectral and hyperspectral images quickly.At last,the validity of this method is verified by a test which detects two kinds of leaves(artificial and real),using the new method by combining the ICA and PCA.