Land utilization classification based on CBERS-04 multi-spectral images---A case study in northwest Hainan, China

ObjectiveChina-Brazil Earth Resources Satellite 4(CBERS-04) multispectral remote sensing data ac-quired in 2015 at three stages were used to explore their potential for land utilization classification in tropical areas.MethodBy combining spectral and phenological information, a case study in northwest region of Hainan was conducted using maximum likelihood method and decision tree classification approach based on the CBERS-04 images obtained at three stages. ResultThe maximum likelihood method based on single image can achieve relatively ideal classification ac-curacy, reaching an overall classification accuracy of 85.8%-88.8%and kappa coefficient of 0.80-0.84. However, the clas-sification accuracy has improved significantly by inputting multi-temporal images obtained at three stages and using maxi-mum likelihood method and decision tree classification. The overall accuracy reached 91.61%-92.61% and kappa coeffi-cient 0.88-0.89. ConclusionThe study demonstrates the broad application prospects of multiple stages of CBERS-04 multispectral data for land utilization classification in tropical regions.