Merging remotely sensed data and non-remotely sensed data is one of the important methods of improving the interpretation accuracy. The attributes of the RGB and IHS modes and color transformat ion theory are discussed.The TM images of the Yanhe basin are taken for instance, and the merging method of remotely sensed data and non-remotely sensed data has been studied with the support of ERDAS IMAGINE. Firstly, the TM image is transformed from the RGB mode to the IHS mode. Secondly, the H component of the IHS mode is replaced with the DEM, and then the components of I and Sare stretched. Thirdly , they are transformed to the RGB mode.The experiments indicate that the method replacing the H component of the IHS mode with DEM is better for supervised classificat ion, and the supervised classification accuracy of the merging data is 6% higher than that of the normal merging data.