| LED wafer sorter is one of the necessary equipment of LED packaging production line,and vision system is the key part of LED wafer sorter.However,most of the wafer sorters used in China are developed and produced in the United States,the Netherlands and Taiwan,and the imported equipment is generally expensive.Although there is also research and development of related equipment in China,there are still some problems such as slow sorting speed and insufficient precision.Therefore,it is of great significance to study the vision system of wafer sorter with independent property rights.In this paper,a lot of key technologies such as hardware,software and vision algorithm are explored.The main research works are as follows:Firstly,on the basis of the existing hardware,the whole structure of the visual system software is designed,the interface development of the software of the LED chip automatic sorting machine is completed,and the key functions of the vision system are realized,such as automatic alignment and calibration,automatic pre scanning of the whole disk chip,and the simulation diagram of the whole disk chip.In the whole disk chip pre scanning process,the relative coordinates and offset information are used to accurately allocate the LED chip data obtained from the positioning,and realize automatic pre scanning.In view of the long timeconsuming problem in the separation process of LED chip sorter,the "S-shaped" sorting strategy is designed to make the separation speed of LED chips faster and the separation efficiency higher.Secondly,in order to solve the problem that the existing LED chip positioning methods are slow and unable to recognize the rotation angle of the chip,a variety of image recognition and positioning methods are deeply understood and explored,and the recognition algorithm based on edge gradient is selected to complete the chip positioning.Canny operator is used to extract the edge points of wafer image,and the gradient corresponding to the edge points is saved.The correlation between the template and the gradient of the image to be searched is calculated by the normalized cross-correlation matching method,and the rotation angle and position center of the wafer are obtained according to the correlation.Finally,according to the characteristics of chip image multi-target,a LED chip multi-target recognition method is proposed,which makes the algorithm can locate all complete chips in the image quickly and accurately.Through the experimental test,the realization of the key module function of the vision system software makes the operation of the vision system software of LED chip sorter more convenient and improves its automation;The chip positioning algorithm based on edge gradient can complete the accurate positioning of a chip within 3.4ms,the speed is better than Hex Sight,the mainstream visual software library of industrial vision,the accuracy and recognition rate are equivalent to Hex Sight,and the rotation angle of LED chip can be recognized,which meets the requirements of speed and accuracy in the actual work of LED chip sorter.The average sorting speed of the chip sorter using this vision system is about 125 ms,which reaches the leading level in the field of chip sorting in China. |
