The continuous development of computer application technology has accelerated the pace of artificial intelligence system in various fields. In agriculture, machine vision automatic weeding equipment has gradually become a hot research topic. Computer vision technology is based on the premise of image processing, in the soil background, accurately identify the distribution of crops and weeds, targeted weed control operations, contribute to realize precision agriculture. Corn visual knowledge grass, in this paper, combining color, shape, texture features of fusion technology, the use of principal component analysis method for feature dimension reduction, coupled with Bayes classification algorithm for recognition classification of maize and weeds. The simulation results show that this method can realize effective classification of weeds.