Egg Quality Detection Based On Computer Vision And Acoustic Technology

Chicken-egg is very popular with consumers for its high nutrition. Egg quality influences egg’s edible quality and safety. In sales and processing, if it can be graded and priced depends on the quality, it may not only protect consumers’ right but also help the managers and operators to adopt scientific management. Using computerized technologies in the processing and grading eggs system could release labor force and exclude subjective disturbance factors by human being, and give quick and precise results for egg quality assessment. Therefore, research and optimization on the egg quality detection system is an attractive and promising subject for improving marketing competition. This research is a part of the project entitled "The study on computer vision and dynamics for nondestructively detecting egg quality" (30371050) under National Natural Science Foundation of China. The central research is to detect the egg quality including crack (mainly about tiny cracked egg), dirty, freshness and improving detecting precision. The contents and results are as follows:1. Chicken Egg Crack Detection Based on Acoustic Resonance AnalysisAn experimental system was set up to measure acoustic resonance frequency of an egg excited with a light mechanical impact on different locations. The effects of physical properties (egg mass, egg shape index, shell thickness, shell stiffness and crack) and other factors (excitation point, detection point, impact intensity, and storage time) on the dynamic resonance frequency were analyzed. The results showed that detecting point, egg shape index, storage time and impact intensity have little effect on the dominant frequency. The dynamic response frequency has mighty difference with different impact points and if the intact egg was cracked. Egg mass, shell stiffness, shell thickness greatly affected the resonant frequency of an egg. The resonant frequency increased with increase of shell stiffness or shell thickness and decrease of the egg mass. The correlation coefficient between the dominant frequency and the egg mass was 0.699. The dynamic response frequency usually not affected by one factor, by the mechanism of the synthetic effects of the several factors.In order to find out eggshell crack detection nondestructively for on-line application, frequency responses of an egg excited with a light mechanical impact on different locations on the eggshell were studied. Through analyzing the response frequency by computer, it was found that the characteristic response frequency of intact eggs changed in a small range but the characteristic response frequency of cracked eggs were heterogeneous and varied in a larger frequency range. By analyzing coefficient of variation (CV) of the four characteristic response frequencies at the chicken egg equator zone, a sorting algorithm to detect crack eggs was developed. Based on this method and set CV to 1, the detecting rate was 91 % for cracked chicken eggs and 87 % for the total chicken eggs.2. Egg crack detection using computer vision and BP neural networkTo improve the accuracy of detection and classification of egg with cracks sequentially, an experimental system was set up based on computer vision for egg crack detection. Computer vision and BP neural network technology were applied to automatically identify and classify the eggs with cracks. Firstly, the picture images of egg with or without cracks were captured through computer vision system, then the images were processed, and 5 geometrical characteristic parameters of crack areas and noise areas were acquired. Secondly, with the 5 parameters as input, the best BP neural network (5 input nodes, 10 hidden nodes, 2 output nodes) by using MATLAB was employed to detect egg crack and classify eggs. The experimented results showed that the rate of detecting precision of cracked egg reached 92.9 % and the classification accuracy of total eggs can reach 96.8 % by the 5-10-2 BP neural network model.3. Egg crack detection based on computer vision and acoustic technologyFirstly, the picture images of egg shell were captured through computer visionsystem, then the images were processed, and 5 geometrical characteristic parameters of crack areas and noise areas were acquired. With the 5 parameters as input, the best BP neural network (5 input nodes, 10 hidden nodes, 2 output nodes) was employed to detect egg crack and classify eggs by using MATLAB. The experimented results showed that the rate of testing precision of cracked egg (mainly for tiny crack) reached only 68 % and the classification accuracy of total eggs can reach 98 % by the 5-10-2 BP neural network model. To analyze the data and results, the method which computer vision and BP neural network find egg with tiny crack and invisible crack from normal egg was difficult, and then show low accuracy. Secondly, the acoustic signals were captured and analyzed, then 6 parameters (F1 characteristic response frequency; F2 characteristic response frequency; F3 characteristic response frequency; F4 characteristic response frequency; CS mean value of coefficient of skewness; CE mean value of coefficient of excess) were pick up. With the 6 parameters as input, the best BP neural network (6 input nodes, 15 hidden nodes, 2 output nodes) by using MATLAB was employed to detect egg crack and classify eggs. The results showed that accuracy of using acoustic technology and BP neural network to identify egg shell crack (mainly for tiny crack) can exceed 90 % and reach 94 % for the total tested egg. Moreover the method to detect egg with large crack is relatively difficult.Thirdly, the egg with different degree crack or non-crack was detected by using acoustic technology and BP neural network, and then detected by using computer vision technology, and finanlly integrate the two results, the egg quality was showed. The method with fusion computer vision, acoustic technology and BP neural network was good for cracked egg quality detection, and the accuracy for the cracked egg can reach 98%.4. Research on dirt detection on brown eggs based on computer visionThe traditional artificial method of examining dirty spot of egg shell has unavoidable disadvantages, for example, the workers’ subjectivity, vision tiredness easily leading to low accuracy, so it can not satisfy the demand of the modern industry production. In this paper, an instrument of detecting egg dirty spot by using computer vision was built up, and the image of egg shell surface was captured, then the images were processed and analyzed. The algorithm and the way of classification were set up based on characteristic parameters obtained from the images. The results showed that the rate of detecting dirty egg could reach 92.7 %. The accuracy of classifying total eggs could exceed 90 %, and could possibly realize the automatic detection of egg dirt.5. Egg freshness detection based on computer vision and BP neural networkAn experimental system was build up based on computer vision. With the system,egg’s contents transmission images were acquired. After pre-processing H, S, I, a, b values of egg color were extracted. The egg shell color information (a~*, b~*) was measured. The weight of egg was measured using electronic balance and the height of egg’s albumen was measured using height vernier caliper. Egg freshness was calculated according to its weight and albumen height. The linear regression model for egg’s Hough unit and egg information (H, I, S, a, b, a~*, b~*, a-a~*, b-b~*) was established by SAS. Afterwards the 3 parameters (H, I, b) which is greatly correlated with egg freshness (HU, Egg’s Hough unit) was reserved.With 3 parameters (H, I, b) of input, the best BP neural network model (3 input nodes, 15 hidden nodes, 4 output nodes) was established by using MATLAB. On the BP neural network model of detecting the egg fresh degree, the automatic detection system was designed in this article, which can immediately show the results according to the egg’s color data after the network initialization. The results showed that the grading accuracy by using computer vision and BP neural network for egg freshness is exceed 90 %.6. Hardware and software of detection system for egg qualityThe hardware of nondestructively detecting egg quality (shell crack, dirty, freshness) system was set up based on multi-technologies including computer vision, acoustic technology and BP neural network. The primary function of software for this device was also showed.