Online Automatic Grading Method Of A Group Of Salted Eggs Based On Digital Image Technique

Fresh duck eggs, due to the difference of its origin, variety, rearing conditions, as well as their size, shape and shell thickness, are diverse from fresh eggs to become well-pickled salted eggs in the curing process. Coupled with conditions on marinated, pickled process, pickled technology, equipment and the quality of workers who operate procedure are uneven, making quality of salted eggs mixed. Salted egg products before the sale need to be detected and classified. However, nowadays salted eggs are grading by manually operated in China. It is slow, inefficient, and poor on accuracy. In order to solve the above problems, an online automatic grading salted eggs model is proposed in this study.The dynamic images of a group of salted eggs are acquired by industrial cameras, photoelectric senors and computer working together on the conveyor belt. Firstly remove the environmental image that is unrelated to salted egg grading, then preprocessing of color images must be done:To determine target area of the binary image by mathematical morphology arithmetic and removal of the object of a small area; according to the binary image is a convex figure, the target region leaked or not was determined. The effects of leaked region was eliminated by searching for mutation points, then fit salted egg boundary by the Least Square, and label the binary image and extract single target area.6description parameters were extracted with color space, and the linear,nonlinear regression equations based on RGB and HSV were established. Then the classification threshold was determined by Minimum error method. Finally the salted quality models were tested.Taking another group of salted eggs, they are graded and marked carefully by the artificial, and then all of salted eggs are mixed together and randomly placed online grading system for automatic classification. After grading completed, compared with previous marks,the number of those meet the model is counted. The experimental results says that the highest detection accuracy rate reached above93%and classification efficiency of5400/hour. It shows the model is feasible for salted eggs grading.