Research On Alignment Technology Of LED Dicing Saw Based On Machine Vision

At present,the automatic dicing saw is replacing the traditional dicing saw because of its high efficiency,high accuracy and high degree of automation in the process of LED wafer dicing.The visual positioning technology has replaced the manual and realized automatic alignment of LED wafer,which is very important to improve the speed and accuracy of dicing saw.However,the research of domestic automatic dicing saw is in the development stage,and there is still a gap between the domestic equipment and foreign advanced equipment in precision and speed.Most vendors develop image processing software of visual positioning system based on commercial software library,although shortens the research period,it raises the cost of equipment production,while hard maintenance and lack of pertinence.In this paper,the visual positioning system for the process of LED wafer dicing was developed,in view of the image characteristics of LED wafer,the accuracy and speed of the cutting.Firstly,the hardware platform of the visual positioning system was built according to the requirements of LED wafer dicing and the principle of visual positioning.Through the illumination experiment,the appropriate light source,camera and lens were selected for the system.Then,based on C#and Opencv,image-processing software was developed which realized functions including the image collection,display and processing,autofocus and automatic alignment.Finally,the position information of LED wafer was obtained to realize automatic dicing.An improved multi-template matching algorithm was proposed by observing LED wafer and comparing the existing image matching algorithm,LED wafer's angle and position can be detected.What's more,the algorithm can detect multiple targets at the same time.The experimental result was that the algorithm of position error was less than 1 pixel,the error of angle was less than 0.09°,and had less time-consuming than conventional template matching algorithm.The algorithm has the advantages of high precision,fast speed and stability by experimental verification.Meanwhile,the LED image processing was also studied,which contained preprocessing,segmentation and morphology transform.In addition,appropriate processing procedures and parameters were chosen,by comparing and analyzing the results of image processing,to improve visual positioning accuracy and stability.Finally,this paper combined the visual positioning system with the motion and feeding system of dicing saw to carry out the actual cutting experiment.The experimental results,that the position error of the gear mark was less than the actual production requirement,show that the visual positioning system designed in this paper can be applied in actual production.