Open AccessAn Investigation on Land Cover Mapping Capability of Classical and Fuzzy based Maximum Likelihood Classifiers
Author(s) -
B. R. Shivakumar,
S. V. Rajashekararadhya
Publication year - 2018
Publication title -
international journal of engineering and technology
Language(s) - English
Resource type - Journals
ISSN - 2227-524X
DOI - 10.14419/ijet.v7i2.10743
Subject(s) - multispectral image , pixel , land cover , computer science , fuzzy logic , pattern recognition (psychology) , artificial intelligence , data mining , contextual image classification , margin (machine learning) , remote sensing , machine learning , image (mathematics) , geography , land use , civil engineering , engineering
In the past two decades, a significant amount of research has been conducted in the area of information extraction from heterogeneous remotely sensed (RS) datasets. However, it is arduous to exactly predict the behaviour of the classification technique employed due to issues such as the type of the dataset, resolution of the imagery, the presence of mixed pixels, and spectrally overlapping of classes. In this paper, land cover classification of the heterogeneous dataset using classical and Fuzzy based Maximum Likelihood Classifiers (MLC) is presented and compared. Three decision parameters and their significance in pixel assignment is illustrated. The presented Fuzzy based MLC uses a weighted inverse distance measure for defuzzification process. 10 pixels were randomly selected from the study area to illustrate pixel assignment for both the classifiers. The study aims at enhancing the classification accuracy of heterogeneous multispectral remote sensor data characterized by spectrally overlapping classes and mixed pixels. The study additionally aims at obtaining classification results with a confidence level of 95% with ±4% error margin. Classification success rate was analysed using accuracy assessment. Fuzzy based MLC produced significantly higher classification accuracy as compared to classical MLC. The conducted research achieves the expected classification accuracy and proves to be a valuable technique for classification of heterogeneous RS multispectral imagery.