| Multi-layer printed circuit board(ALIVH)is a printed circuit board formed by pressing several printed circuit boards(PCB)together after etching.It has the characteristics of high density,high precision and small space occupied.It is often used in computers,mobile phones,communications,instruments and other high-precision electronic products.As the carrier of electronic products,its production quality is more strict than the common single panel and double panel requirements,so it is very important to detect the defects of the inner plate before pressing.At present,most electronic factories detect PCB defects through manual visual inspection,electrical test and other methods,but the disadvantages of such methods are low efficiency,high cost and easy to detect by mistake.Therefore,an automatic detection system is needed to improve the efficiency and accuracy of defect detection of multi-layer circuit board.Inner circuit board as the main research object,this paper uses the linear array camera and transmission mechanism of combining the image acquisition device,based on HALCON deep learning algorithms developed a PCB defects online detection system,the system can quickly locate circuit board defect image coordinates,and through the neural network classifier accurately infer the type of defect.In this paper,according to the characteristics of the printed circuit board design process of online visual defect detection system and architecture,carried out in accordance with the system structure of form a complete set of hardware selection,mainly is the linear array camera and image acquisition card selection,comprehensive analysis of a variety of light source lighting effects,line array camera image quality and light source relations between position,Angle of the light source,in order to prevent the collection image distortion,The relation between the frequency of the linear array camera and the speed of the conveyor belt is analyzed,and the image acquisition scheme triggered by the external pulse signal combined with the infrared sensor and the encoder is determined.Then,the paper explores the way to transform the ODB++ file into the standard image of printed circuit board.According to the ODB++ file format,the paper designs the parsing process,reads the layer information and coordinate information of printed circuit board,and previews the parsing effect by writing software.Then,simple preprocessing and image correction were carried out on the PCB image.The corrected image and the standard image were subpixel difference operation to obtain the defect position,and the trained classifier was used to infer the PCB defect type.Finally,the upper computer software of defect detection based on Qt is designed to show the final implementation effect of the system scheme.The detection system designed in this paper can identify six common PCB defects.A total of 3000 defect images were collected in this paper.After repeated training,the classifier recognition accuracy reached 98.89%,and the actual test accuracy reached99.57%.The experiment proves that the HALCON-based PCB online visual defect detection system designed in this paper can accurately locate the defects of the circuit board and judge the defect type,which is a new scheme applied to the visual defect detection of PCB. |
