A Novel Methodology for the Classification of Debris Scars using Discrete Wavelet Transform and Support Vector Machine

Abstract

Satellite images that capture surficial disasters can be used in estimating vulnerability of human settlements. High resolution satellite images have increased the ability to analyze and aid in hazard risk assessment. Current image processing techniques improved the accuracy in detecting surface morphology using rapidly developing techniques such as wavelet transforms. Coupled with new age machine learning algorithms these wavelets based image processing techniques are finding their way in hazard risk assessment. In this paper we would like to present a wavelet based image processing workflow based on Support Vector Machine (SVM) architecture which has been deployed to classify between debris scars and urban settlements. Around 100 images of debris scars and urban settlements from five Asian countries (India, Nepal, Japan, Taiwan and China) were obtained from Planet labs imagery with 3-5m resolution. Analysis for classification accuracy is done with various types of wavelets. Results obtained show that Daubechies4 (db4) wavelet and symlet give better accuracy of 96%. All simulations were computed using Matlab2019a.