Decision hierarchical classification by FLD for vegetation application using PolSAR features

Abstract

Polarimetric synthetic aperture radar (PolSAR) features have great significance in application of vegetation classification, which can explain the scattering mechanism of the vegetation; in order to make full use of PolSAR features' scattering mechanism explanation, the decision tree classifier is chosen because of its simple and hierarchical classifier structure. Since all the classification methods are composed of two parts: feature selection and classifier selection, this method is established with PolSAR features as selected feature and decision tree as adopted classifier. As decision tree classifier is flexible in discriminant rules, the hierarchical classification process of multi-feature is built under the notion of Fisher Linear Discriminant (FLD); after the classification process, optimization of the branch sequence and boundary algorithms is made to improve the classification accuracy of the specific classes. The experiments of AIRSAR and AgriSAR data illustrate that this method can obtain good classification accuracy; at the same time, it can introduce expert knowledge into the whole framework to help improve the classification accuracy, and extract useful information of features and classifiers from the classification results as new expert knowledge.