Abstract: In order to reduce the application amount of pesticide, the intelligent weed recognition system was designed, which was composed of intelligent controller, nozzle assembly and installing bracket, and a new weed recognition method based on color and location characteristics was applied to the different environment in this paper. Experimental results showed that the intelligent weed recognition system was effective in the different environments, and could locate the weed rapidly, reliably and accurately. The recognition correct rate of the intelligent weed recognition system was 97.0% in the ordinary environment, 92.5% in strong illumination environment, and 89.2% in the shadow environment. The average recognition time was 160 ms. The research can provide a reference for design of the precise spraying weeding system based on machine vision．