Neural Network-based Egg Breakage Detection

AbstractIn this paper, the author researches the automatization of the diagnosing of egg's dilapidation situation in eggs production. The computer is substituted for human power to accomplish building the egg's module and its classifying by combining the relation of systems' vibration mode and its construction with application of the artificial neural networks in mode recognition.In this paper, the current situation of the diagnosing of egg's dilapidation during the egg production in our country is briefly related, then the research content, ways and aim are decided. To acquire sound signal the prompting device is made for prompting eggs. In addition, the author puts forward that the eggs could be recognized based on the difference in the frequencies of their prompted sounds. Because the eggs' prompted sounds are lied on low frequency bands, the microphone and sound amplifier that have excellent duration response should be used. The output signal from the sound amplifier is transferred to data signal by A/D card before it is put into computer. The data signal is in the form of discrete. We will do Fast Fourier Transferring to it. Then we could get power-frequency graph and the graphics of power distributing in every frequency band. After that, the data that describes the power distributing situations in every frequency will be taken as the input of the neural networks. Then these data form eggs' two modes.The author compiles the neural networks program and design interface by VB6.0.The BP arithmetic of the neural networks is used in this paper. At the same time the author discusses how to speed the training and avoid the process trapping into local minimal as well as expounds the virtues and the shortcomings of this arithmetic .In the end the training is done with two modes to get the synaptic weights. Then these synaptic weights are used to classify eggs.