Nondestructive Detection Research On Salted Egg Based On Computer Vision

Duck's egg was rich in protein, vitamin, calcium, potassium, iron and phosphorus nutrients.Salted egg that made by it had a special taste,with rich nutrition and was eaten conveniently.What was more,it could be storaged for a long time.At present,it was detected mainly by manual work,which was affected by subjective factors with low detection efficiency and quality.That could not meet the scale production requirements. Therefore, exploring new technology for nondestructive detection on salty egg had very important significance.Based on computer vision system,using the knowledge of image processing and analysis and multivariate statistical to process data to realize the nondestructive detection on salty egg and to establish correlation model. The computer vision system was composed of the computer, digital camera, image acquisition card, light source, the light box and the detection work bench etc.Image manipulation of salted egg by computer:including pretreatment such as color image gray processing,binary, image segmentation,image enhancement, image denoising, edge detection etc. The optimal threshold method was used to achieve image segmentation. The media filtering was used to enhance image. With wavelet to march image denoising. After comparing Roberts edge detection operators, Sobel edge detection operators, Prewitt edge detection operators,Laplace operators,Gaussian-Laplace operators and Krisch edge detection operators. Sobel operators were chosen to detect the edge.Characteristic parameters extraction of salted egg by computer:extracting the R, G, B, H, S. I color value of salted egg. the energy, nertia, local entropy, local stability values of texture features as the characteristic parameters. By using statistical principle, the categories models of RGB, HSI, texture features were established by Bayes step-regression discriminance and checking out them were by F analysis.Through verifying the four models,the results showed as follows:the accuracy of salted egg based on RGB color models was 96.6%, the accuracy of salted egg based on HSI color models was 85.9%, the accuracy of salted egg based on texture characteristics models was 82.5%, the accuracy of salted egg based on color and texture characteristics models was 85.3%.From the results of the validation, they were showed that based on RGB color models'accuracy was highest, using Typel, Type2 and Type3 respectively represented good egg, black and yellow egg and retreat dissolve egg, the models as follows: Typel=-170.61367+1.77659R-0.43885G Type2=-12.96038+0.43004R-0.06405G Type3=-166.48462+1.92103R-0.60913GThe models were prominent due to (Pr>Fo.o5)<0.0001.