System For Duck-egg Shell Crack Detection Based On Machine Vision

Eggs food is very important for people’s daily life. During the egg processing, sorting is an indispensable step. Broken eggs need to be picked out from non-crack ones. Sorting can not only promise the quality of eggs, but also improve economic efficiency. At present, egg processing industry in China is still using manual testing methods, such as candling and knocking, respectively identified through the eyes and ears of people. Manual inspection requires skilled workers to do long production operations, which are prone to visual or auditory fatigue and resulting in bad inspection. And the medium-sized and small enterprises are facing the shortage of workers, so automated egg sorting technology becomes urgent need.At present, researchers have done a lot of research in the quality testing of eggs, based on image recognition or voice recognition, or two both. Since image recognition still have difficulties in inspecting the duck eggs with unclean surface, so many researchers use voice recognition. But image can contain more information than voice, and it is a non-contact method, which will not cause secondary damage. So I chose image method in this paper.This paper is based on machine vision. We use the improved algorithm of local contrast enhancement, with the new image acquisition method, which shot two picture respectively under the back-light and the front-light, aiming at removing interference of unclean things. Finally, we utilize the probabilistic neural network to make recognition based on the characteristic parameters, which are selected through experimental experience. In the experimental environment, with a total of357sampling duck egg images, the crack identification rate is100%, the false detection rate for non-crack eggs is12.7%, and the speed of system is an average of0.4seconds each, which basically meet the industry requirements.