The Research On Rapid Template Matching Algorithm Applied To Led Wafer

As an important part of the LED production process, LED probe station is mainly to complete the LED chip detection. Prober station technology is widely applied in the complex, high-speed precision electrical measuring device. Image processing as a important link in automatic prober, play a very important role.Domestic prober is less automatic and less competitive, compared with the international first-class company’s products, largely because the vision algorithm and the core algorithm of image matching is the shortcoming.For specific application scenario for LED the probe station, this dissertation analyzes the match the actual LED chip industry requirements, and designs a kind of algorithm which can realize fast image matching of multiple LED chip with accurate positioning. This algorithm can achieve the image matching accuracy of subpixel level, and the matching speed can satisfy the actual industrial demand. Precise matching stage calculates rotation angle and precise match position by solving algorithm based on fast Fourier transform, and acquires the coordinates of the center of the target through getting mapping relationship. And this dissertation has improved the matching speed by pyramid algorithm and the matching precision by curved surface fitting method with a specific optimization, to enhance the algorithm performance. The dissertation also set up the software architecture of the proposed algorithm, the algorithm is encapsulated to make it suitable for LED automatic prober, which can finish the target recognition and LED chip positioning.Finally, testing algorithm with a series of images taken in LED automatic prober, teste experimental results show that the algorithm meets the project requirements. Compared with foreign commercial software Evision, the algorithm level offset maximum error is less than one pixel, rotation angle of the maximum error is less than 1.1°.And antinoise ability of the algorithm is strong,and this algorithm is not sensitive to illumination changes, which is an ideal matching algorithm.