Abstract: Aiming at the problem of the image segmentation of farmland green crop plants under outdoor, we proposed an image segmentation method for farmland green crop plants which based on G-R color feature of green crop plants, combined with the Otsu method for the maximum variance between clusters and area threshold segmentation. The algorithm solved the difficult problem of segmentation of farmland green crop plants and background in the farmland images under the conditions of uneven illumination and complicated soil of outdoor. The experimental results show that the method is not only insensitive to crop, soil and light changes, but also can eliminate the effects of the shadow part of image, compared with the color indexing method EXGOtsu and RGB algorithm (G>R and G>B), the segmentation effect of the method is better.

Key words: green crop plant, image segmentation, uneven illumination, maximum variance between clusters, area threshold