A system was developed to monitor and map grassland vegetation using hyperspectral imaging at the field level. The system is essentially divided into two main techniques : a pushbroom image scanning system for image acquisition and an image processing system for mapping grassland vegetation based on hyperspectral images. In addition, a data mining engine was designed for model calibration and validation. The model validation result showed that the coefficient of determination were 0.38-0.71 (grass chemical compositions) and 0.48 (weight of dead matter). The vegetation map generated using this system reflected the spatial distribution of the actual vegetation characteristics, demonstrating the potential of this system for creating high spatial resolution maps of grasslands.