Based on ecognition software platform, the method of combining object-oriented computer automatic classification with manual visual interpretation was used to extract the annual land use data of the study area. Finally, three methods are used to verify the data accuracy: field sample point survey, high resolution image recognition and Google Earth sample point recognition.

The remote sensing images are preprocessed by radiation correction, orthorectification, fusion and mosaic. The actual surface area corresponding to the minimum patch area is not less than 0.1 Hm2, the polygon has no overlap, no gap, and the patch attribute has no vacancy or redundancy. Before remote sensing image interpretation, remote sensing image, typical survey and field comparison were used to establish remote sensing interpretation marks of land use. Based on the remote sensing image, combined with the interpretation marks, the land use types are extracted. Reexamination of interpretation results: take no less than 5% of the total map spots for verification. The number and results of field verification samples meet the requirements of technical specification for remote sensing monitoring of soil and water conservation (sl592-2012). For the verification map spots, 10% of the verification samples are selected for field verification.