Abstract: For meeting the demand for large scale and operational cotton monitoring system using the remote sensing technology, the identification of cotton information from all kinds of crops on the remote sensing image must be rapidly, precisely and reliabley conducted. In cotton growth season, the spectral information of cotton and main crops was collected in field measurements using a spectroradiometer and also from Landsat TM images in the experimental area. Through analyzing the spectral features, the best time for cotton identification on TM images was confirmed and a spectral information based cotton identification model on TM images was developed. The model was validated by mathematical analysis and practical application. The wodel is simple, efficient, and more accurate, and is suitable to be used in the Cotton Operational Monitoring System using Remote Sensing in Xinjiang.