Studies On The Image Recognition Base-on The Artificial Neural Networks

In the modern industry automation, along with high accuracy, high velocity, the high grade request climbs unceasingly, the miniature bearing function increasingly is prominent. But each kind of electromechanical device intensifies to its dependence the miniature bearing to take day by day the important components, the miniature bearing quality to the equipment precision, the movement performance, the service life and so on all has the very important influence. Counts by the experiment indicated that, the bearing expiration as a result of the bearing surface flaw which in the bearing expiration form, which causes the crack, the crack creates to reach 65%. Therefore, must carry on the strict examination to the bearing quality, in particular to crack, crack examination.This article main research content is researches and develops one kind to apply face the enterprise in the miniature bearing surface defect examination system, is based on the neural network pattern recognition technology, the artificial neural networks are similar to the biology nervous system, is by the nerve cell one kind of non-linear auto-adapted dynamics system which (i.e. artificial neuron) composes for the fundamental operation unit. Carries on the training through the use reasonable study algorithm, the neural network has to the thing and the environment very strong from the study, auto-adapted and from the organization ability. Applies the merit the neural network which examines in the miniature bearing is may through to the system carry on the training to catch the bearing surface flaw complex class condition, but the shortcoming lies in the network to have to pass through the widespread structure adjustment and the sample training can obtain the good effect.This article introduced the neural network basic principle and the main characteristic, in view of the bearing examination domain particularity, have obtained a more effective neural network structure through the theoretical analysis and the massive experiments; The treatment mapping carries on the pretreatment likely, enhances the examination rate; The introduction study rate speeds up the BP neural network the training speed. Finally, proposed a process improvement based on the neural network bearing examination system.