Abstract: A novel method for weed detection using the color feature of corn seedling was developed. The leaves of corn seedling were dark green, but its centre zone was peak green. The unique feature could be reflected by the saturation index, which depended upon the relative dominance of pure hue in a color sample. The saturation of centre zone had a maximum saturation value for corn seedling. That was used to extract the centre zone of corn seedling after soil background was segmented with the green-red index. For the segmented foreground of green seedling, the connected region with the extracted centre zone was classified as corn seedling. On the contrary, the unconnected region with them was recognized as weed. The results showed that the correct classification rate of corn plant and weed was 88% and 84%, respectively. For a frame image with 720×576 pixels was processed, the mean processing time was only 120ms.The correct classification rate of corn plant was mainly influenced by the integrity of centre zone, whereas the correct classification rate of weed was mainly affected by the occluding degree of corn and weed leaves. Therefore, the future work will be on the control of field view and the segmentation of occluding leaves of corn and weed.