Critical Analysis of Machine Learning Technique for Discriminating the Similar Vegetation Pattern

Although the existing literature has classified the landcover by different techniques for change detection, the classified images obtained by different methods still deserve further discussion. For this reason, by analyzing Machine Learning technique, this study offers a comparison between the Tall vegetation (TV) change in Roorkee region of India due to Highway construction during the period of 2017–2018. In literature researchers attempted the optical and synthetic aperture radar (SAR) image classification with varying degrees of success. The challenge in this type of classifications is to identify urban and dense vegetation (tall vegetation) classes separately. Both these classes have very small amount of variations in the scatter power. A Random forest (RF) algorithm is trained on the combination for efficient classification for urban, dense vegetation and other classes. Random forest classification algorithm is belonging to the ensemble learning category, is the one of the best machine learning algorithms. It combines the predictions generated from multiple decision trees to make accurate predictions or classify data points. In this study the number of tall vegetation pixels decreased from 1977 to 31 due to construction of Highway and tall vegetation removal at other study regions taken at different stretches of Highway, results in significant decrease of tall vegetation area.