Comparative Analysis of Generalized Intersection over Union and Error Matrix for Vegetation Cover Classification Assessment

Abstract

The result of vegetation cover classification greatly depends on the classification methods. Accuracy analysis is mostly performed using the error matrix in remote sensing. In recent remote sensing, image classification has been carried out on the basis of deep learning. In the field of image processing in computer science, Intersection over Union (IoU) is mainly used for accuracy analysis. In this study, the error matrix, which is frequently used in remote sensing, and IoU, which is mainly used for deep learning images, were compared and reviewed to analyze their accuracy levels for the results of vegetation index calculation. The results of vegetation index calculation were applied to the comparison of the accuracy levels of IoU and the error matrix. According to the results of accuracy analysis using the error matrix, which is based on random points, the accuracy of the normalized difference vegetation index (NDVI) was shown to be 82.4% and that of deep learning was shown to be 93.7%, with a difference of about 11.3%.