Abstract: The gray level ratio between crop and weed in image has a great influence on the effect of distinguishing them. A method for utilizing multi-spectral image fusion to increase the gray level ratio was suggested in this paper. By utilizing multi-spectral images, a system for acquiring multi-spectral images was built up with black and white camera and four filters. Based on studying the gray level ratio with onion (crop), mustard (weed) and soil in multi-spectral images, many modes of multi-spectral image fusion were compared and studied through testing results, it was found that some modes of multi-spectral image fusion like b+ir-g-r give better results. Applying these modes to distinguish onion (crop) from mustard (weeds) in images, the results showed that the errors of identifying onion (crop) and mustard (weeds) were decreased by 22%. A method to evaluate the indices of gray level ratios of crop, weed and soil in images was also defined in this paper.