Soil moisture is very important to crop yield. Based on spot data of the soil moisture in the second branch irrigation canal of Baoji Xia, the data obtained from October, 2003 to March, 2005. Normalization processing was made for bands TM5 and TM7 of Landsat-TM, and then monitoring model for remote sensing and soil moisture by using difference value between normalized difference soil moisture index and reference normalized soil moisture index. Quantitative inversion of soil moisture based on established model and remote sensing image in June 28, 2005. The results show that inversion accuracy reach over 80%, and the best depth of soil layer is 0～40 cm layer. The monitoring model is able to fulfill the soil moisture monitor requirements in large scale irrigation area.