Abstract: The study chooses 0.8 m panchromatic/3.2 m multi spectral resolution GF2 remote sensing image as the basic data. Information extraction processes and methods based upon GF2 resolution satellite images are compared and analyzed. In order to obtain spectral characteristics of different types of information, the spectrum of GF2 satellite image is analyzed. Then, the spectral and spatial information of remote sensing images are fused. Furthermore, the study compared quality evaluation of different fusion methods. Then, the study makes the agricultural information extraction experiments by using the threshold, the relationship between the spectral method, unsupervised classification and object-oriented method. The paper also verifies the accuracy of information extraction results and takes results analysis. Tests showed that Pansharp fusion algorithm was the best method among these methods, it has the advantage of normal color image, no ghost, high definition, class contrast to normal, clear texture, high entropy, high correlation coefficient of fusion images and the original multispectral images. Threshold and spectral relations method applies to extract a single element of agricultural information. Unsupervised classification method is suitable for obtaining land use in the study area quickly and preliminarily. The method of object-oriented has the high precision for whole information extracted. Overall, the different information extracted method has its own advantages. In practical, the people can select the best method according to the spectral characteristics of the class.

Key words: agricultural information, GF2 satellite, spectrum analysis, image fusion, information extraction, object-oriented