95.91%。Appearance quality grading for fresh corn ear was implemented by computer vision based on HSI color model. Bare tip position was detected and removed using projection method. Defects of fresh corn ear were identified by the first order differential operation on H. Characteristic parameters of appearance quality, such as defect proportion, ear length, ear maximum diameter, aspect ratio and rectangle factor were obtained. General regression neural network with five characteristic parameters as input was developed for grading. Experiment showed that average errors of bare tip position, ear length and ear maximum diameter were 2.27mm, 1.96mm and 0.54mm, respectively. Mistake rate of defect proportion was 3.00%, and grading average ratio was up to 95.91%.