Modified Genetic Algorithm (MGA) based feature selection with Mean Weighted Least Squares Twin Support Vector Machine (MW-LSTSVM) approach for vegetation classification

Abstract

Vegetation classification using remotely sensed images is an advancing approach predominantly in area developmental schemes. It is very common that the same vegetation type on ground may have different spectral features in remotely sensed images. Also, different vegetation types may possess similar spectra, which makes very hard to obtain accurate classification results. In the recent work, there are number of classifiers that are proposed by different researchers to solve this problem. Though many solutions are available, high dimensionality of samples become a major issue. The prime objective of this work is to increase the classification efficiency of agricultural area. This research presents a novel object identification and feature selection algorithm. At the initial stage of the work, Modified Fuzzy Possibilistic C-Means clustering is applied for the proficient segmentation of objects. In addition texture and the spectral features of the segmented image are extracted for efficient vegetation classification and these features are selected based on the Modified Genetic Algorithm based wrapper feature selection algorithm. Finally, vegetation classification is performed by using Mean Weight-Least Squares Twin Support Vector Machine (MW-LSTSVM). Thus the vegetation classification is achieved accurately. The experimentation results prove that the MW-LSTSVM provides higher values in regard to accuracy, recall, precision and F-measure justifying its efficiency. MW-LSTSVM efficiently improves the classification of remotely sensed images in an agricultural area when compared to existing classifiers.