Abstract: Abstract: 【Objective】To reduce noise and improve land use classification accuracy, we need to explore a classification method that combines texture and spatial information. 【Method】Using remote sensing images obtained from hyperspectral imaging spectrometers in Henan Zhenping and learning from the classification of a decision tree, a hyperspectral remote sensing classification method combining spectrum, texture, and spatial information was used. The land use type was initially determined by extracting the spectral features, and then hierarchically determined in detail by extracting the texture features and vegetation, which was based on spatial information. Finally, the accuracy of the classification of various land use types was verified by ground test points, and the extraction efficiency of different methods was compared. 【Result】The land-type layered-extraction system combined many characteristics, adopted various band spectral reflectances to classify large ground classes, used textural features to classify ground classes with smaller spectral differences, and then classified vegetation based on spatial information. The total classification accuracy of the method combining texture and spatial information was 86.7%, which was 13.3% higher than the total accuracy of maximum likelihood classification. 【Conclusion】The remote sensing classification method that combined hyperspectral and texture features, and spatial information can effectively reduce noise and improve the classification accuracy. This provides a reference for the study of land use classification.