Urban Area Classification Using High Resolution Remote Sensing Data: A Hybrid Classification Approach

Increasing urbanization has resulted in reduction of agricultural land. There is immense requirement of keeping track urban development's to avoid depletion of agricultural land. This paper explains urban area classification using remote sensing and GIS Mapping of classification is done with help of top sheet E43D/5, with scale 1:50, 0000 and IRS-P6 LISS-IV image of Feb. 2014 having resolution of 5.8 m and GPS receiver My GPS coordinates is used. Classification methods like supervised and unsupervised both are used and detailed results analysis is done. For doing this the acquired image went through the series of process namely Preprocessing, Classification and Result analysis. Some image enhancement techniques were also performed to improve the satellite imagery for better visual interpretation. ENVI 4.4 image analysis tool were used for data processing and analysis. Maximum Likelihood, Mahalanobis distance, and Isodata classifiers were performed for urban classification in this study. Seven urban classes have been identified from the satellite image classification processes. It is observed that area is covered by barren land, agricultural area (with crops and without crops), hilly area, buildings and water bodies and play ground etc. The experimental results were compared with different supervised and unsupervised classifiers, such as Mahalanobis distance, and Maximum Likelihood and Isodata classifiers respectively. The performances of the classifiers were evaluated in terms of the classification accuracy with respect to the ground truth. The result of the classifications suggests that, each used and covered type were best classified by the Hybrid classification technique. Hybrid classification is done by first applying unsupervised classification like ISODATA classification, followed by various supervised classification techniques like Maximum likelihood, Mahalanobis classification. Hybrid classifier is best classifier of all. Combination of Supervised and Unsupervised classifier results in better accuracy as compared to supervised classification alone.