According to visual restaurant service robot application environment under the condition of uneven light and changeful, light and shade and occlusion, make the RGB image noise influence is very big which was collected by the binocular vision, difficult to use conventional image processing method for accurate identification of plate, the plate image is analyzed the characteristics and the traditional image processing technology, a adaptive edge of ellipse fitting method was proposed in this paper. Firstly, the image of dinner plate was preprocessed, and then the image edge information was extracted. Finally, according to the edge information, the dinner plate was identified by fitting ellipse with Hough transform. The experimental results show that under the circumstance of different light and shade in most restaurants, the ellipse feature of the edge of the plate can be accurately extracted by using this method, and the plate can be recognized, the validity of the method was verified.