Agricultural image segmentation is the basic work of computer vision technology in agriculture. This paper presents an image process method based on color characteristic. The algorithm separates the crop area from soil background according to the color feature, which is got by analyzing the color difference between crop and background. In order to translate the color images to gray ones effectively, fourth different transformation methods in RGB color space and three components of HSI were compared. Test results show that the excess green 2G-R-B is the best index for segmentation. This index has a characteristic of consuming little time, good quality of segmentation and not being influenced by sunshine intensity and can meet the real-time navigation in the target extraction.