Study On Key Techniques Of Hatching Egg Nondestructive Testing Based On Machine Vision

The hatching rate of eggs in our country is about 90%.Each year there are 480 million eggs failed to hatch.In order to reduce waste,it is necessary to pick up good ones before hatching.This paper uses the machine vision technology to study the hatching eggs to solve the problems caused by the manual identification.These issues include the labor intensity,low efficiency and being easy to be affected by the human factors.The main work of this paper includes:(1)According to the design criterion of the detection device,this paper discussed the structure and working principle of an egg yolk detection device and a surface defect detection device,then designed the mechanical structure of the detection device,finally verified the image acquisition effect of the detection device.(2)This paper presented a method for yolk image analysis based on the interval segmentation.At first,the image is divided into several intervals;and then the paper designed a mode of organization and management for the intervals to achieve a rapid operation,which includes: the search and insertion of lost points,the intervals merging operation,and the search and insertion of fault points etc.Secondly,the presented algorithm of yolk region extraction based on the interval segmentation refined the traditional algorithms which are based on the region segmentation.And the presented algorithm can reduce the error rate of extracting the target areas in the traditional methods.The target areas obtained by the proposed algorithm are more accurate than that obtained by the other traditional methods based on the region segmentation.Even if the average brightness of the image,or the size of the target area is very different,the proposed algorithm can still run stably and finally achieve the accurate extraction of the yolk area.(3)This paper proposed several characteristics reflecting the number of egg yolk,such as: roundness,the relative position of the centroid of the egg yolk in the egg etc.,based on which the paper proposed the AdaBoost algorithm suitable for the detection of the number of egg yolk.AdaBoost algorithm is used to train multiple weak classifiers to identify the number of egg yolk.The proposed AdaBoost algorithm can effectively detect the number of egg yolk.And compared with the existing algorithms,the proposed algorithm can achieve higher accuracy.(4)The detection of the surface damage of the eggs was studied,and an algorithm was applied to detect the surface damage.The defect detection algorithm for the egg surface,which is based on the statistical template matching principle,established the statistical models for each kind of defect in the training process;the target area shape and the shape of the statistical template image of a test object were compared in the detection process,so as to determine whether the image is defective.The experimental results show that the algorithm can effectively detect the surface defects,such as scratches,cracks,dirt,holes and so on.