A Detection Method of Weed in Wheat Field on Machine Vision

Abstract A method of weed detection based on position and edge feature was studied. Firstly, the plant pixels are separated from soil background using the color difference of green plant and soil. Secondly, according to drilling crops are arranged in rows, this paper use pixel histogram method to select centerline of crop rows and set the centerline as starting point and crop rows edge as ending point, then fill crop area and eliminate crop pixels. Finally, weed detection is completed via the feature that weeds tend to grow in small associations and distribute closely. Experiments show that the algorithm obtains good weed recognition rate.