A Novel Principal Component Approach for Feature Extraction in Remotly Sensed Images

Lack of reliable training samples, especially in cases with broad dimensions, results in a decrease in supervised classifiers’ accuracy. One approach to solve such problem is applying feature extraction. In spite of some limitations and deficiencies when being applied, PCA and LDA are two methods, highly used in feature extraction. In this paper, a new method is proposed to rank principal components instead of arranging them in a descending order. First, using PCA method, principal components are found and instead of arranging them based on corresponding Eigen values, acquired weights used in LDA are applied. The method is not limited to a specific distribution function and it may be based on parametric and nonparametric methods. Results show that through using less number of features found by the method proposed, ML classifier’s accuracy and validity has increased

کلیدواژه ها:

principal component ، feature extraction ، remote sensing ، image processing