Study on the precision evaluation index in building seismic damage information extraction from remote sensing image

Abstract

Precision evaluation is an important step in remote sensing classification. The Kappa is widely used to evaluate the overall accuracy of classification result, but it cannot reflect the classification effect of a certain class. A new precision evaluation index R can apply to evaluate the accuracy of a certain category in classification image. In this paper, we extracted building damage information based on objectoriented method, and discussed the effect and robust of index R and Kappa when different categories of the classification result were involved in precision evaluation. The study shows that the standardized index R can reflect the efficiency of the classification result more stably and it has certain advantages in classification evaluation.