| Entering the Post-Moore age,the minimum size of new chips has been reduced to less than 50 μm,and high density is the inevitable trend of chip technology development.Among them,the technology for mass accurate transfer of high-density chips is called mass transfer,but the chip carrier needs to be repaired by machine vision positioning detection technology before the chip can be transferred.Therefore,this thesis presents a new intelligent visual adjustable location algorithm for high-speed and high-precision repair of high-density chips,and carries out related research on high-density LED chips visual positioning repair system,the specific work schedule is as follows:(1)First,we need to deeply understand the difficulties of massive transfer technology and the current status of high-density chip growth which will inevitably lead to a large number of defects.In order to prevent a large number of defective chips from being transferred directly and to ensure the qualified rate after transfer and reduce the cost of later repair,we must repair the chip carrier in advance.Therefore,we design a system scheme to repair high-density chips based on machine vision positioning technology.Then carefully investigate the current status of machine vision detection algorithms,and make sure that these algorithms have shortcomings in accuracy and speed when facing high-density LED chip detection.Finally,the main research content of this thesis is determined based on the above research and analysis.(2)Based on the principle of laser peeling technology,we designed a repair scheme for the chip peeling to the empty position on the carrier by visual positioning detection and multi-axis motion platform adjustment.In the aspect of hardware,the selection and parameters of main equipment such as related light source,industrial camera,industrial lens,motion platform are preliminarily determined,and the camera parameters are determined through calibration experiments.In the software system,we designed the operation flow of detecting with LED chip.First,we detected large area missing in the low-power mode of the lens,then calibrated the angle of the chip and detected the precise positioning in the high-power mode of the lens,and designed the operation interface for the core positioning algorithm of the chip detection.(3)The core visual intelligent adjustable positioning detection algorithm in this system is based on the template matching algorithm and through four optimization strategies,the visual algorithm can be switched intelligently on three different resolution layers,high,low,and can achieve speed adaptive adjustment in both X-Y directions.The algorithm switches smart to the highest resolution layer when it is near the target and uses traversal mode detection.When it is far away from the target,the resolution is reduced and the skip scan mode is used,which allows the detection algorithm to greatly improve the detection efficiency while guaranteeing high-precision detection,so that both speed and precision can jump at the same time.Then,the reliability and robustness of the algorithm are analyzed through mathematical derivation.Finally,the experimental validation of the device system is carried out.The results show that the positioning error of the algorithm can reach 0.53 μm,and the detection time is l24.21 ms.The other four detection algorithms have the fastest measurement speed and the highest detection accuracy.Moreover,the algorithm is robust in different lighting simulation experiments and under the interference of pepper-salt noise with density up to(0.6/pixel)and Gaussian noise with maximum variance of 0.5. |
