Abstract: It’s very important to monitor the spatial distribution of eucalyptus forests, especially for regional ecological environment protection and the government’s decision-making. Based on existing research, this paper uses multispectral GF-6 image as the data source, the first satellite in China to provide red-edge bands. In this paper, selecting Luzhai county in Guangxi as a typical research area, combines spectral characteristics, vegetation index characteristics and red-edge characteristics to design different classification schemes. A class hierarchy is established by object-oriented multi-scale segmentation method. According to the different scale levels, membership functions and CART decision tree models are used to classify eucalyptus plantation information. The experiment results shows that the red-edge feature had a significant influence on the construction of the CART decision tree model. The integration of the GF-6 red-edge features could effectively improve eucalyptus plantations’ classification accuracy, with an overall accuracy of 91.75%. Comparing with the classification scheme that only using the traditional bands and vegetation indexs, the classification accuracy has been improved by 11.25%. The research results have important theoretical significance and practical value in the identification of eucalyptus plantations using red-edge bands of Chinese satellite.

Key words: GF-6 satellite, red-edge bands, eucalyptus plantation forest, CART decision tree, accuracy evaluation