Design And Algorithm Research Of Camera Calibration Module In MV Vision Algorithm Library

Innovation and industrialization of key technologies in IC(Integrated Circuit)encapsulation equipment is the strategic direction advanced by the12thFive-Year Plan. Asone of the key technologies in IC encapsulation equipments, high speed machine visionalgorithm was monopolized by foreign manufacturers. National Science and TechnologyMajor Project proposed a target of developing a vision algorithm library called MV VisionAlgorithm Library and industrializing it on typical IC encapsulation equipments. Cameracalibration is one of the kernel modules in the MV Vision Algorithm Library, and widely usedin vision orientation and vision measurement.Acoording to the features of IC encapsulationequipments, software designment and algorithm designment of cameral calibration module inthe MV Vision Algorithm Library were done.The major work is as follows:Function partition of camera calibration module was finished based on the functionalrequirement analysis of typical IC encapsulation equipments and the function partition of MVVision Algorithm Library. Performance index designment of camera calibration module wasfinished based on the accuracy requirement analysis of typical IC encapsulation equipments.Software overall designment of camera calibration module including architecture, data flow,data structure and program interface was accomplished, according to the result of functionpartition and performance index designment.An optimized algorithm based on Delaunay triangulation was proposed to correspond theimage coordinates and grid coordinates of every feature point automatically. Judgmentcodition of area and relative roundness were used to selecte the feature point region; Contourextraction and tracking was used to get the edge of feature point region and least square fittingof elliptic equation was used to calculate the image coordinate of feature point; Afour-adjoining lookup table was set up according to the law of effective Delaunaytriangles.Then, a recursive search started from the selected seed point was done to finished thecorrespondence of feature points’ image coordinates and grid coordinates. Experiments usingsimulation image and real image show that the algorithm can help the equipment accomplishcamera calibration without being suspended or manual intervention.An optimized algorithm based on least squares fitting in divided image region and asimplified algorithm based on interpolation of Delaunay triangles was proposed, in view of the research on the system error of traditional algorithm. The former algorithm was proved tobe more accurate than traditional algorithm in industry application circumstance and in worstdistortion circumstance. The latter algorithm was proved to have same precision level withtraditional algorithm in industry application circumstance. Application experiment shows thatboth the two algorithms could confine the image correction fluctuation less than1%andinsure that the error of coordinate transfoamation is less than5um.