A novel hybrid approach using SVM and spectral indices for enhanced land use land cover mapping of coastal urban plains

Abstract

In present study, Landsat images of years 2001, 2009 and 2018 are used for LUC mapping of a coastal-urban-floodplain wherein built-up and coastal-barren classes have been identified to be the most confusing classes for interpretation. Otsu’s thresholding techniques have been used for mapping of waterbodies, built-up, and coastal-barren lands. The performance of most commonly used built-up indices have been assessed, among which BCI performed best for the study area. A new index, called Coastal-Barren-Index (CBI), has been developed using the spectral characteristics of SWIR1 and green spectral reflectance bands. A critical comparison of SVM and RFC classifiers are reported, and, finally, a hybrid approach is proposed as a combination MNDWI-CBI-SVM for mapping of the study area with Overall Accuracy 90.5% and Kappa value 0.87. The proposed approach is validated for an independent site, and, can be considered as generic in nature for LUC mapping of coastal urban plains.

Keywords:

• Spectral index
• land use
• Otsu’s thresholding
• Support Vector Machine classifier
• Landsat

Acknowledgements

The first author wishes to acknowledge Department of Science and Technology, Ministry of Science and Technology, Government of India for the financial support vide their letter no. DST/INSPIRE Fellowship/2018/[IF180589] dated 24 July 2019. The authors are grateful to the infrastructural support provided in Centre of Excellence (CoE) on ‘Water Resources and Flood Management’, TEQIP-II, Department of Higher Education (MHRD), Government of India, for conducting this research. The authors are thankful to anonymous reviewers and editor for their useful comments to improve the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The present research work was supported by Department of Science and Technology, Ministry of Science and Technology [DST/INSPIRE Fellowship/2018/IF180589 dated 24 July 2019].