OIF Based Indeces Oriented Ecological Classification Using LANDSAT TM Digital Data – A Case Study on Beluchary and Dhulibasan Island Groups, Sunderban, West Bengal, India

Abstract

The classification of vegetation from remotely sensed data has long attracted the attention of remote sensing community as the results are fundamental sources for many environmental applications. There are different approaches and techniquesto improve the classification accuracy. However, different uncertainty or errors may be introduced into classification due to many factors like complexity in the landscapes under investigation, selected remotely sensed data, image processing approaches, the availability of reference data etc. So much efforts should be devoted to identify these major factors in the image classification processes and then to improve them. In the present study, different vegetation indices (VIs) have been adopted for the betterment of vegetation classification accuracy. The analysis of correlation and standard deviation of each VI was used to identify the best combination for the separability analysis. The selection of the best combination was done using Optimum Index Factor technique based on the total variance within bands and correlation coefficient between bands. The OIF technique was applied to all the calculated seven VIs. A number of twenty one colour combinations were produced and analyzed using OIF. The combination having the highest OIF value has been selected for the classification in which a distinct spectral dissimilarity has been observed, which is very helpful for information extraction. Finally overcoming the spectral self similarity, after classification five ecological classes has been got from the Beluchari and Dhulibasan islands. Finally the technique of OIF has been successful in conclusively deriving the five ecological classes in Beluchari and Dhulibasan Islands by overcoming the spectral self similarly.