The objectives of the study were to analyze spectral characteristics of red soils different in SOM (soil organic matter) content in Ji’an, Jiangxi, for establishment of county-scale SOM quantitative estimation models based on SOM reflectance bands, it was found that red soil and paddy soil differed sharply in soil spectral characteristic. The wave range of 560～710 nm was the spectral absorption band specific of SOM in the soils of Ji’an county. Correlation coefficient of the log of the absorption area in the range of characteristic absorption bands with soil SOM was 0.86, fitting equation was y=-20.91 ln(s)-27.26, and determination coefficient was 0.74. After calibration with the data of SOM in various types of soils, the determination coefficients for prediction(R²), the root mean square error (RMSE) and the relative error analysis(RPD) was found to be 0.75, 0.61 and 1.88, respectively. The partial least squares regression (PLSR) model, established on continuum removal and logarithm processing of reciprocal of the reflectance, was the highest in estimation accuracy, with prediction determination coefficient R2 being over 0.83, RMSE less than 5.00 and RPD more than 2.4. These findings indicate the model based on specific absorption bands 560~710 nm can be used to quantitatively estimate soil SOM in red soil regions.