The Research On Non-destructive Method Of Egg's Properties Based On VFW

With rich nutrition, The egg is important animal-nutrition food in daily life, but it is easy to metamorphose that is difficult to be detected by manual work, thus, it is of real value to do some research about non-destructive detection of egg's properties and some designs about mini-detection system.This task is to firstly find the method of obtaining images by VFW(Video for Windows) and then build math models about egg greenness and weight by image manipulation. The testing equipment is made up of the image-detected box(including USB vidicon, lighting and darkroom) and the computer. USB vidicon takes digital images and transmits to the computer and images are captured by VFW which is a pure software capturing and has no relation with equipment and is more universal compared to the traditional video capturing by SDK(Soft Develop Kits).Image manipulation is including pretreatment and pick-up eigenvalue of the image: the pretreatment uses the method of selective average in the space to deal with the spot in the captured image; the egg's edge detection take the method of Robert grads; then greenness image manipulation module searches and obtains the color information of the core such as H, S, I; measuring the value of haugh and the average is regarded as greenness index; building the greenness model is by step-regression and checking out it is by F and correlation and passed approach analysis. The result indicates the regression of the white egg greenness model is most prominent due to (Pr>F_0.05)<0.0001 and the correlative coefficient is 0.8024 and the model is as follows:Haugh=-212.69905+1.04298*H+210.05943*I;the regression of the brown egg greenness model is prominent due to (Pr>F_0.05)=0.0019 and the correlative coefficient is 0.562228 and the model is as follows:Haugh=-212.69905+1.04298*H+210.05943*I;The two colors of egg's haugh are both related to H and I and less related to S; eggshell color is recognized according to the I value with fuzzy math because the eggshell color is high related to the I value of the egg core.In addition, egg weight image manipulation module firstly searches the egg edge by the same way above mentioned and then obtains pels sum(Count) , the pels of the long axes(Length) and the pels of the short axes(width) of the image. The several distinct egg weight models are built by step-regression as follows:Model1 Weigh=2.11771+0.00205*Count-0.12967*LengthModel2 Weigh=-115.46211+0.57072*Width+42.99928*BiModel3 Weigh=-12.22456+0.00159*CountCompared with each other, the best model is:Weigh=2.11771+0.00205*Count-0.12967*LengthIt shows that Weigh is in direct ratio to Count and in inverse ratio to Length; the model is most prominent due to (Pr>Fo.os)<0.0001 and the correlative coefficient is 0.9702.Through verifying the two models, the result shows that the correct discerning rate of the white egg is 86.3% and of the brown egg is 83.5% and the error of the weight detection is between -0.9g and 0.9g. It indicates the models can be accepted.