Research Of PCB Defect Inspection Technology Based On Machine Vision

With the development of modern electronic technology,printed circuit board(PCB)has been widely used in a variety of common areas.Today,China has become the largest circuit board manufacturing power.The export volume of the PCBs are also among the highest in the world.In the production process,the defective PCBs can be manufactured inevitably.If these PCBs put into the market,the risks or the result will be great.Therefore,in order to avoid this,it is an important part of the production to detect the defective PCBs.This paper uses the automatic optical detection method to detect the defects of the PCBs.According to the changes of the parameters in the PCB,the defects of the circuit board are detected by using local detection methods.We use the neural network to detect the defects of the solder joint in PCBs.The main work of this paper is as follows:1.In order to solve the problem that the segmentation is not accurate and the efficiency is low,an improved method for circuit board segmentation is proposed in this paper.This method,firstly,pre-segment the PCB image by using the improved genetic algorithm,calculate the initial segmentation threshold.Then the initial threshold is optimized by using the fast algorithm of two-dimensional Otsu method to calculate the final segmentation threshold.Through the experiment,it is proved that the object is extracted more accurately,the algorithm is faster and the computer storage space is smaller by using this method.2.In order to solve the problem that defect classification can not be detected by comparing the relative parameters of the PCBs.We use a local detection method to detect defects in the circuit board.When a defect is detected,the system first extracts the location where the defect exists,then comparing the local reference changes of the defect area to identify the type of the defect.Experiments show that the system can be well identifyed nine major defects,such as short-circuit,broken circuit,bulge,circuit sag,more line,less line,Cavity,Pad missing and Pad blockages.3.In order to solve the current situation of low defect detection efficiency of PCBs,the optimized image preprocessing algorithm,image matching algorithm and recognition algorithm are used to detect the circuit board defects in the detection system.The results show that detection efficiency and the detection accuracy of the proposed algorithm are improved.4.Aiming at the problem of low convergence speed when BP neural network is used to detect solder joints,We use the LMBP neural network which is of high convergence speed to detect the PCBs' solder joints.Training the samples of defective solder joints and normal solder joints by utilizing LMBP neural network.If the network training was successful,the network will be saved.When the solder joints need to be detected,the network which is trained successfully will be invoked to identify the defects of solder joints.The results show that the LMBP neural network is more accurate and faster than the traditional BP neural network.