Comparative Study On Land Cover Classification Of Multispectral Remotely Sensed Images Using Basic Color Spaces

Abstract

Land Cover Classification Of Remotely Sensed Images Plays An Important Role In Many Environmental And Socio Economic Applications. It Classifies The Multispectral Remotely Sensed Images Into Various Land Covers Such As Water Body, Urban, Vegetation, Forest Etc. Feature Extraction Is A Major Step In Multispectral Remote Sensing Image Classification Before Classifying The Image. In This Paper, Color Histogram Features Are Extracted Using HSV, LUV And LAB Color Spaces. The Extracted Features Are Trained And Tested By Random Forest Classifier. The Classification Performance Of The Color Spaces Are Compared Based On The Metrics Accuracy And Kappa Coefficient. An IRS LISS IV Multispectral Orthorectified Dataset Is Taken As The Input Image For This Experiment. It Is Observed That The HSV Color Space Outperformed When Compared To Other Color Spaces.