Study And Application Of The Recognition Technology For Cracked Egg Based On Computer Vision

During process and transportation,egg is easy to be cracked due to the unavoidable collision and compression.To ensure the food safety of egg and egg products,damaged or cracked eggs should be detected and removed before process or entering the market.The acoustic and vibration technology and the computer vision technology are two main automatic cracked egg detecting methods.Although the study on the acoustic and vibration based cracked egg detecting technology is mature,this technology has not widely applicated in China due to the high price.The manual detecting method is still used to remove the cracked eggs in most of the egg processing enterprises in China.The computer vision technology has the advantages of high detecting speed and low cost.However,the computer vision based cracked egg detecting technology is still not very reliable and in laboratory experiment stage.Cracked egg image recognition method with higher applicability and sensitivity,and the on-line cracked egg detection device based on computer vision technology need to be studied and developed.In order to make the computer vision based cracked egg detecting method suitable for on-line detection in egg industry,this work created cracked egg samples with different crack sizes by using an egg collision simulator.Then,a self-made egg image capture device was used to capture the image of the cracked and the intact egg samples,and the gray distribution of the crack area and the intact area in egg image was studied.After that,a new cracked egg image recognition method for hen egg and duck egg based on a sequenced wave signal extraction and classification algorithm was built and verified.Finally,an on-line cracked egg detection device was developed based on the new cracked egg image recognition method.The details were as following.(1)The study on the gray value distribution of cracked egg image indicated that:(ⅰ)the average gray value of the images of the same egg kind had a big dispersion,and the gray value distribution of single egg image was low in the middle and high on the edge;(ⅱ)the gray value of crack areas in most of the egg image could not be higher or lower than the gray value of the majority intact areas,and the threshold segmentation method was not suitable for crack area extraction;(iii)the gray value of the center line was higher or lower than that of the two side of the crack area.This gray value difference could be a basis to extract the crack signals in egg image;(iv)the translucent area of hen egg,and the shell texture and the air cell membrane of duck egg could disturb the crack recognition.(2)A cracked hen egg image recognition method based on a sequenced wave signal extraction and classification algorithm was built.Firstly,this method extracted the suspected sequenced wave signals from the hen egg images using a wave signal extraction algorithm and a wave signal connection algorithm.Then,the highest length-width ratio of segmented area(Rmax)and the gray value change index(D)of the suspected sequenced wave signals were calculated as two characteristic parameters.After that,a model of algebraic interface equation in discriminated fields was used to select the sequenced wave signals caused by cracks.Finally,the egg image with at least one crack caused sequenced wave signal was recognized as cracked egg image.The accuracy of this method for hen egg with the crack width lower than 20 μm,30 to 50μm and above 60 μm were 50.8%,76.3%and 98.6%,respectively,and for intact hen egg was 96%.The running time for calculating single egg image was 1.65 ±0.50 s.(3)A cracked duck egg image recognition method based on a sequenced wave signal extraction and classification algorithm was built.Firstly,this method extracted the suspected sequenced wave signals from duck egg image using a wave signal extraction algorithm and a wave signal connection algorithm.Then,the entire side slope(K),the side slope of the center region(Kc)and the distance between sequence wave signal and egg end(ym)of the suspected sequenced wave signals were calculated as three characteristic parameters.After that,a-model of algebraic interface equation-in discriminated fields was used to select the sequenced wave signals caused by cracks.Finally,the duck egg image with at least one crack caused sequenced wave signal was recognized as cracked egg image.The accuracy of this method for duck egg with the crack width lower than 20 μm,30 to 50 μm and above 60 μm were 60.0%,74.4%and 92.8%,respectively,and for intact duck egg was 93%.The running time for calculating single egg image was 0.98±0.06 s.(4)An on-line cracked egg detection device based on computer vision technology was developed.A triggering method of continuous image capture,a light source control method and a capture parameter adjust method were creatively applicated in this device to ensure the image capture stably,and the sequenced wave signal extraction and classification algorithm based cracked egg image recognition methods were used to recognition the cracked egg image.The cost of the device was about 20,000 CNY,and the detecting speed of this device was one egg per second.The accuracy for cracked and intact pink hen egg were 91%and 93%,respectively,for cracked and intact brown hen egg were 90%and 94%,respectively,and for cracked and intact duck egg were 84%and 87%,respectively.