Land cover/use information is a key parameter for estimating biophysical process and energy exchange between land and atmosphere, and meanwhile it is also an important parameter for regional and global climate models and ecosystem process models. Remote Sensing is an important and effective tool for mapping land cover /use on a large scale. But the quality of remote sensing data and weather condition have great impacts on the accuracy of land cover / use information from remote sensing data. This research proposes a new methodology for mapping land cover /use of Guang Dong province using object-oriented classification with remote sensing data and field survey information as inputs.Object-oriented classification has been increasingly employed for mapping land cover/use because of its advantages of using multi information of remote sensed objects such as spatial distribution, shape, size, spectral and texture over some traditional classification methods which often only use spectral information of objects. The remote sensing data used in this research include multitemporal and multi spectral China HJ-1A/B and TM data. Firstly multi temporal remote sensing data were processed and geo-rectified. The classification rules were then established using object-oriented classification method with processed remote sensing data, DEM, history land cover/use data and field observation data as inputs. Finally a new classification method is developed by integrating multi information from remote sensing data and field observations into classification rules. In situ observation data is used to validate classification result. The result shows that the average accuracy of mapping land cover /use of Guangdong province is 85%, much higher than that of traditional classification method. The accuracy can be further improved by using more effective classification rules and more field survey in future.