In order to solve the disadvantage of image segmentation by K-means clustering to extra-green character used to be adopted in agricultural images, an image segmentation method based on the particle swarm optimization and the K means clustering was proposed. Firstly, image pixels value was fast clustered with the K-means clustering. Regarding the results as the position of a particle, PSO can be used and the new class centers also can be re-calculated with the K means clustering. Subsequently,the position of all particles got updated and the optimal threshold was obtained. Experimental results proved that the improved algorithm was an effective method for segmenting the object accurately from images, and applicable for various kinds of agricultural images with extra-green character.