Automatic Visual Inspection Of Chip To Glass Deviation In Chip On Glass Process

With the development of display technology, high-density electrical connector technology has been evolved continued, in which COG(Chip on Glass) technology is becoming the mainstream of high-density electrical connection technology for its reliability, excellent performance, easy-to-production process. Almost all the phone screen and driver chips are connected by COG technology. But in COG detection process, the detection area is not only big but also haves a variety of shapes; the detection accuracy is high; and currently the research of detecting chip to glass deviation in chip on glass process has not been deep completeed enough. Becase of up reasons, the detection of deviation mainly relys on manual microscopic examination by optical microscope, resulting in inefficient detection, high error rate and high miss rate. At the same time, the repeatability and reproducibility of detection are insufficient to meet the increasing requirements for intelligent plant.In this dissertation, an automatic visual inspection system for detecting the chip to glass deviation in chip on glass process is researched. Through optical image acquisition, digital image processing technology and reasonable interactive design, gettingand processing good input image, finally show a quantitative measure of the deviation value.Firstly, the image acquisition module of the automatic visual inspection system for detecting the deviation after COG process is introduced, and the respective image acquisition sub-module is analysised detailed and designed based on the demand of the image acquisition module of automatic visual inspection system, and the hardware acquisition platform of the image acquisition module has been builded.Secondly, the algorithms using to automatic extract the region of interest is researched, including the processing of locating the coordinate of MARK point and the segmentating of line area to calculate.Thirdly, the algorithm of calculating deviation is studied. It focuses on how to use texture feature to extracte the region where the BUMP regional approach, and by comparing the difference between texture descriptors included the largest multi-local, local entropy, GLCM and local binary patterns about processing effectiveness and efficiency, a combination consider of two factors that effectiveness and efficiency, a method based on combined local variance calculating with local binary pattern is selected to extracte particle region.Finally, the two key modules of detection software, built standard module and real-time detection module are explained. A man-machine interface of detection software is achieved.The characteristic of this dissertation includes the following aspects: developed and implemented the automatic visual inspection systems of calculating deviation after COG process, and through by cooperating with the automation equipment suppliers, the automatic visual inspection systems is tested in several LCM module manufacturers to trial produce, the detect time of each is 3.5 seconds, erro rate is 0.1 percent, and the missing rate is 0.5 percent. It is proved that the systems could meet the requirements of real-time detecting. Verification, the way to implementating the algorithm is based on C ++, which greatly improves the efficiency of software development; furthermore, the automatic visual inspection systems of calculating deviation after COG bonding is proposed in this dissertation, to our knowledge, it is the first report in China.