Study on the forest vegetation restoration monitoring using HJ-1A hyperspectral data

Abstract

In this paper, Xunke County was studied using HJ-1A hyperspectral data for monitoring vegetation restoration after forest fires. The pre-processing procedure including data format conversion, image mosaicing and atmospheric correction. Support vector machine classification was used to perform surface feature identification based on the extracted spectral end-members. On that basis, the image area was divided into seven categories and statistical analysis of classification types was performed. The results showed that HJ-1A hyperspectral data had great potential in fine classification of surface features and the accuracy of classification was 91.8%. The mild and severe fire-affected area extraction provided useful reference for disaster recovery monitoring. Furthermore, the distinction between coniferous forest and broadleaved forest can offer useful information for forest fire prevention and early warning to some extent.