Parallelepiped and Mahalanobis Distance based Classification for forestry identification in Pakistan

Abstract

Rapid deforestation has been witnessed in Pakistan over the past few years. It is taking its toll on Pakistan economy, infrastructure, and environment in the form of frequent floods. In order to keep the numbers steady frequent surveys need to be conducted. Identifying lush green forests through remote sensing is quite effective when it comes to collecting ground truth reality through extensive ground surveys. In the following study two pixels based supervised classification algorithms i.e. Parallelepiped and Mahalanobis Distance Classification Algorithms are compared for classifying forests in Pakistan. For that purpose High Geometric Resolution Imagery of SPOT-5 (2.5m) is used as the base image. According to our results Parallelepiped Classification is proved to be the better one of the two with overall accuracy of 95.4% and kappa coefficient value of 0.937, with reference to the Mahalanobis Distance classifier with overall accuracy of 85.97% and kappa coefficient value equal to 0.8115. On the basis of these findings Parallelepiped Classifier is preferred to be used for the remote sensing of forestry in Pakistan.