Research On Grading Of Apple's Color Based On Computer Vision

Apple is a sort of fruit that is liked by all the people. Our country has the highest total quantity of apple yield in the world, and the profits and the exportation quantity have being notable improved in recent years, but is still far from the advanced countries. There are many reasons towards this result; the most dominant one is that the post-process is despised. The post-process includes: selection, washing, waxing, grading and packaging, and the most important one is grading. In our country the grading is still mainly depends on manual grading, and it has many defects such as: low efficiency, slow speed. The grading result is mostly depends on the level of workers, so the apple quality is unable to be ensured. Manual grading has been the bottleneck factor that restricts the efficiency of the processes; so an automatic grading system is required imminently. The color and the red area are important appearance quantities of apple, and they are also one of the important standards of grading. In this paper this grading system is studied through color.The whole process system includes: preprocess, character pick-up, grading. Preprocess is a very important step in image manipulation. Only the proper preprocess method can ensure the success of character pick-up. After considering the actual status and the process effect, the following methods are determined at last: grey-degree linear transformation, smoothness, minimum-error image segmenting and the profile extraction. The final results show that these methods are proper for the preprocess.Character pick-up is the main part of the process system. Only if it is carried out properly, the grading, could be done successfully. By using a method of three-dimensional counterchanges, the true area of an apple's surface can be calculated, based on the fact that an apple is an approximate sphere (this fact is approved in the paper). This method is more reasonable in getting the proportion of the red-area on an apple, than that by taking count of the number of pixels on the projection. After comparing the two methods we can see that the areas calculated with the two methods are not quite different while the red-area is concentrated in the center of the image, but if the red-area is close to the border of the apple in image, the result will be quite different. According to the analysis towards the hue-distribution curves of the apples, hue-threshold is chosen to divide the cardinal red and the thicken red.The grading method is the last step in the system, and in this paper a rule of direct membership function is selected to deal with the final grading. This rule is similar to the thinking pattern of the human brain. Membership function is determined according to theactual aspect, and the final grade is operated after the value of membership function is calculated. Experimental results show that this method has a good effect with a veracity of 89.2%.