Research On Quality Inspection Of Eggs Based On Machine Vision

Machine vision was widely developed and used in quality inspection and automatic classification of eggs in china and abroad today, and was becoming one of more and more important inspection techniques. While machine vision and image processing technology were specializing, the application prospect of quality inspection of eggs using machine vision technology was becoming more and more attractive.Egg quality inspection included external quality inspection and internal quality inspection. Egg's size, shape coefficient and weight were important parts of its external quality and egg's contents and freshness were important parts of its internal quality respectively. Applying machine vision technique, egg size inspection algorithm based on edge, linear regression model, egg's size, shape coefficient and weight of external quality and its contents and freshness of internal quality were inspected under static condition, and egg's weight was classified using SOM neural network according to its four sizes extracted in this research. Main contents and results were shown as follows:1. The research advancements and achievements in the field of egg's external and internal quality inspection and classification using machine vision technique were reviewed. The existing problems and requirements of egg quality inspection method and equipments were put forward.2. Machine vision systems for egg's external and internal quality inspection were set up. The system for external quality inspection was composed of a lighting chamber, six fluorescent lamps (F40BX/840), a CCD camera, a frame grabber, and an industrial computer (ADVANTECH INDUSTRIAL COMPUTER 610). The system for internal quality inspection was composed of a lighting chamber, an incandescence lamp (PHILIPS), a CCD camera, a frame grabber, and an industrial computer.3. Egg's external quality images were acquired and pre-processed. The images were segmented with an indicator composed of R, G, B values. Egg edge was extracted with Laplacian Operator. The sizes of egg's vertical diameter and maximal horizontal diameter and the coefficient of its shape were detected and calculated using egg size inspection algorithm based on edge extracted and linear regression model. The correlation coefficients of egg's vertical diameter, maximal horizontal diameter and shape coefficient inspection models were 0.9923,0.9816 and 0.9579 respectively.4. Four sizes of egg's vertical diameter, maximal horizontal diameter, upper diameter and nether diameter were extracted using egg size inspection algorithm. The linear regression model for egg's weight and its four sizes was set up. The correlation coefficient of egg weight inspection model was0.9781 and the absolute error was no more than ±3g. Using SOM neural network classifier, egg weight was classified into three ranks: below 55g, between 55g and 65g (not 65g), 65g and above 65g. The accuracies of classification were 90.6%, 76.8% and 82.5% respectively.5. Egg's contents transmission images were acquired. After pre-processing, R, G, B, H, S, I mean values of egg contents' color feature were extracted. The weight of egg was measured using electronic balance and the height of egg's albumen was measured using height vernier caliper. Egg's haugh unit representing its freshness was calculated according to its weight and albumen height. The linear regression model for egg's haugh unit and its contents color feature mean values selected by SAS was set up. The correlation coefficients of egg freshness inspection models for eggs with brown eggshell and white eggshell were 0.8674 and 0.8929 respectively.