Spectrum Analysis And Information Obtaining On TM Image Of Xianghuang Banner Grassland In Inner Mongolia

The spectral characteristics and distribution of different types of grassland were analyzed by using TM image, the results indicate that there were significant differences among spectral characteristics in different places where grazing pressure and soil characteristics were different, and grassland spectrums in different years were significantly different too; There were great variations on spectral characteristics among different types of grassland, but several grassland types were easily being confused.The way and process for grassland classification were studied by using TM image, and analyzed by using the data of spot survey and past years. The results show that unsupervised classification methods were better compared with the supervised ones in speed and precision of grassland classification; The images composed by means of bands 2, 3, 4, 5, 7 and TNDVI (a kind of vegetation index) for unsupervised classification were more precise than those obtained through bands ratio, principal components analysis and original bands of TM image composite for supervised classification. It was difficult to identify the type of the grassland using remote sensing technique with 5 indexes (coverage, height, density, frequency and weight), because the eatable herbages were almost exhaustion by grazing, the data obtained can't reflect precisely the real grassland condition. Spectral reflectance of the grassland type distributes at the different levels, the reflectance distribution of different grassland types were probably overlapped, however, variations among different formations (a unit of grassland type) could be found in small sections of spectral response. In grassland classification, incorporation and adjusting of formations was done by combined the similar and not easily identified formations reflectance features. Adjusted 9 kinds of formations represented different combinations of vegetation structure, landform and soil, and they could truthfully reflect actual productivity and utilization of different grassland types.The method mentioned above was preferable for increasing the speed and accuracy of remote sensing (RS) classification in grassland. The classification accuracy of formations was above 80%.