| Double-yolked eggs are common in poultry production.They are not suitable for hatching but have high nutritive value and commercial value.In recent years,the economic market for the production and processing of double-yolked duck eggs has become larger and larger.Double-yolked fresh duck eggs and their reconstituted eggs are exported to domestic and foreign markets.Therefore,it is very important to ensure the source of high-quality double-yolked duck eggs.At present,the poultry egg production and processing enterprises in our country still adopt the method of artificially illuminating eggs to identify double-yolked eggs,which has high labor cost and low detection efficiency.In order to solve this problem,the static and on-line identification of double yellow-yolked eggs were studied by using machine vision technology.The main research contents and conclusions were as follows:(1)The image acquisition device and on-line sorting device suitable for single-yolked and double-yolked duck eggs were designed and constructed.Aiming at the problem of poor light transmission caused by the thickness of fresh duck egg shell and mixed shell color,the models and working methods of light source,industrial cameras,industrial lens,sensor and controller suitable for double-yolked duck egg recognition were analyzed and determined.(2)A static identification method for double-yolked duck eggs was studied.To meet the needs of small batch sample detection,this method sequentially collected duck egg transmission images,cut out the region of interest of duck egg images,used open-based reconstruction to obtain the yolk region,smoothed the edge contour of the yolk region,used convex hull algorithm to obtain convex defects of the yolk region contour,and determined whether duck eggs were double-yolked eggs according to the size of the convex defects.The position of egg yolk in the whole egg was marked by watershed algorithm controlled by marking and ellipse fitting.The experimental result showed that the recognition accuracy of single-yolked and double-yolked duck eggs was 98.33% and 98.66% respectively.(3)A classification model of double-yolked duck eggs was established.According to the needs of large-scale sample detection,this method used convolutional neural network to classify duck eggs after determining the labels of duck egg images.The recognition effects of Alex Net,VGG16 and Inception-V2 were compared respectively.The accuracy rate of identifying double-yolked duck eggs by the three networks was as high as 99%,and the identification speed was very fast,which showed that convolution neural network had good classification effect on duck egg transmission images.This article choosed Alex Net network with the best effect.(4)The dynamic identification model of double-yolked duck eggs was established to realize on-line identification and sorting of double-yolked duck eggs.In order to meet the needs of on-line inspection of egg enterprises,accelerate the inspection speed and realize the automatic sorting function,an on-line identification and sorting method for double-yolked duck eggs was designed.PLC control program was programmed with ladder diagram and testing software was developed with MFC.Effective image processing was carried out on the dynamic image to remove light leakage interference in the vertical direction of duck eggs,and a convex hull algorithm was used to obtain duck egg recognition results.Each duck egg had three identification results,and a final result was comprehensively determined and stored in PLC.When duck eggs reached the sorting mechanism,the sensor was triggered,and the PLC read the stored duck egg identification result.If it was double-yolked egg,PLC controled solenoid valve to separate it,otherwise it would not be processed,thus achieving the sorting function.Through experiments,the online recognition accuracy of single-yolked and double-yolked duck eggs was 96% and 98% respectively. |
PDF Full Text Request