Abstract: A new cotton identification method based on color differences was studied. It provided parameters for the motion of manipulator so that it could acquire precise location information of cotton, identify them from surroundings correctly, and accordingly pick up automatically. Two hundred cotton images were taken in natural environment and characteristics of different parts of cotton were analyzed in different color models. There exist color differences between cotton, cotton leaf and cotton stem, based on which a vision model was set up to identify ripe cotton from surroundings. Experimental results show that the proposed method has good performance in cotton identification and high efficiency without color model transition.