Study On Inspection Of SMT Chip Pins Based On Machine Vision

With the rapid development in microelectronics, surface mounting equipment is terribly needed. There are developed products available abroad, whereas few relating technique is open to public. On the other side, the research of surface mounting equipment in our country started late, and the demanded surface mounting equipment mainly depends on import. With the growing demands on the functioning index of equipment, machine vision inspection plays a more and more important role. This thesis is focused on vision inspection technologies in Surface Mounting Technology (SMT). The component and the work principle of a computer vision inspection system are discussed. Based on the demand of Surface Mounting chip pins inspection, the thesis develops an experiment system for their inspection, fulfills the inspection aim: the amount of the pins, the dimension of the pins, the pins distance and the pins height floating.Lighting and lenses play the most important role on the quality of image in a computer vision inspection application. The red flat LED illuminator and lenses with little distortion are selected as the main components of the inspection system, which can provide high quality images and this will make it easier for the image processing and the dimension calculation.A simple method of calibration is given according to the character of the experimental inspection system in the thesis. The method uses the grid template which is drawn by AutoCAD as a calibration standard reference, just calibrates the object plane resolution of the system. Compared with the calibration result of commercial machine vision software HexSight, the result of the method which this thesis adopts is credible.In this thesis, different edge detection operators are employed to process the chip pin image. It is testified that the Canny operator is the most suitable edge detection operator. Finally, the system uses Canny edge detection operator to detect pins'edge, then uses the contour track algorithm to get the pins' contour. The top-down poly-line splitting method is applied to express the contour. This method can retrieve the pins'contour very well by setting appropriate Normalize Max Tolerance.The thesis discusses the deficiency of template match based on gray scale correlation. So the contour 2-D moments are applied to identify the pins. The result shows that this method calculates fast and has a good effect, thus can content the demand of online detection.The thesis introduces how to realize the detection items in detail at last. The stability and repetition accuracy are tested and the qualitative analysis of various causes of system error are given. According to the test result, the whole detection system is reliable, fast and convenient to use, which can be up to the demand of the IC pins' detection very well.