Combining the Decision Tree and Supervised Classify Techniques to identify the tobacco field in the Satellite Images: Luxi County of Yunnan Province in China as an example

摘要

Luxi County of Yunnan Province in China has the biggest areas of tobacco fields that belong to Chinese Red River Tobacco Company. And the areas of tobacco field in 2005 achieved more than 20,000 Chinese acres. So Luxi County is the ideal bed to identify the tobacco field by remotely sensed technology. The paper introduces SPOT5 imagery with the high spatial resolution of 5m and clear texture information and Landsat TM imagery with the medium spatial resolution of 30m and high spectrum resolution in study area. Firstly, we ortho-rectify the TM and SPOT imageries in study area, then uses the pansharp fusion method to fuse the above two Ortho-images with different spatial resolutions. Lastly, based on the spatial distribution patterns of the tobacco field with highly congregated in macro regions of continent & nation, and the small patch dispersible at the levels of the County & Town & Village, considered the tobacco spectrum characteristic and the terrain distribution characteristic, the paper introduces the altitude above sea level, the slope, the vegetation index (NDVI), the texture factor and so on to identify the tobacco fields in the fused imagery. The rate of accuracy of computer classifies by this method achieves 77.75%.