Research And Application Of Automatic Optical Inspection System Based On Artificial Neural Networks

In recent years, with the development of Information Technology and Precision Manufacturing, more and more industrial products become sophisticated and miniature. The requirements of product quality is constantly improving,but the traditional manual testing methods have been unable to solve this problem. With the development of computer vision technology and image processing technology, Automatic Optical Inspection is widely applied in chip manufacturing,PCB manufacturing,intelligent detection and many other fields.But current research still weaked in intelligent defect detection,complex environment detection and defects classification problem.Based on the analysis of the advantages and disadvantages of nowadays technology,the core algorithm of Automatic Optical Inspection is studied and re-designed. Proposed an improved method of neural networks, which use the gray relational analysis method to determine the network nodes number of the hidden layer to optimize the network structure, use the PSO with genetic operator to train the BP network. This paper achieved the LED Chip Defect Detection System and the Train Lock Status Inspection System by using improved neural network algorithm, and proposes Feature Extraction methods and Locate methods in this two systems.Theoretical results of this study have been applied to the actual development of industrial equipment, theory is verified by experiments and comparative analysis by using prototype,the result shows that this improved neural network algorithm has a strong generalization abilitym, can be used to reduce the influence of environment light, and this method has high stability, high detection efficiency and low sensitivity. Solved the problem of current research in intelligent defect detection,complex environment detection and defects classification.