Research On PCB Defects Automatic Optical Inspection

With the rapid development of the PCB (Printed Circuit Board), the quality of PCB production has become an important cornerstone of business survival, so the quality of PCB should be controlled strictly. The technology of defects automatic optical inspection can extract useful information from the collected images. And the useful information is extracted from optical images to process and understand. The actual detection results are ultimately finished. The technology is the main development direction of PCB defect detection because of its advantages, such as non-invasive, fast, and high precision.The Automatic Optic Inspection system consists of lighting system, image collection system, and image processing system. The hardware section includes lighting system and image collection system. The software section includes image processing system. This thesis has a further study with the image processing system. At first,1 circuit board image from the Gerber file is studied. The method takes into account the Gerber file, which is the standard data file of the circuit board, acquiring coordinate data of the circuit board lines and pads by deeply parsing the instruction in Gerber file, and generating reference circuit board image. Secondly, the filtering and segmentation algorithms of detecting circuit board image are studied. By comparing several filtering and segmentation algorithms, this thesis using the Mean Shift filtering algorithm effectively eliminates PCB noise in the image, and the lines and pads of circuit board image is separated from the background by the method of HSV color space threshold segmentation. Then the registration problem of reference circuit board image and detecting circuit board image is studied. A registration algorithm based on the center of positioning hole is used. We obtain the parameters of the positioning hole of detecting circuit board image by least squares estimate algorithm. The registration of two circuit board images accomplished by affine transformation based on parameters of the positioning hole. Finally, the automatic detection and recognition of defect circuit board image is studied. We get the defects of circuit board image by image contrast and mathematical morphology. And we achieve the automatic identification of defect types according to the characteristics, such as area, connected domain and Euler number. The experimental results show that the algorithms used in this paper can get more satisfactory results when it used in the field of PCB defect detection.