Auto-calibration of an SMT machine by machine vision

An SMT machine has many working coordinate frames—the fiducial mark camera frame, component camera frame, machine table frame, PCB frame, and reference frame. Because of many influences such as mechanical dimension errors, machine assembling errors, and camera lens distortions, all frames on the SMT machine must be calibrated to compensate for these machine errors. This paper applies machine vision techniques to auto-calibrate an SMT machine, including frame transformations and nozzle compensation, bringing the accuracy of this system to within ±0.1 mm. The coordinate mapping from the camera frame to machine table frame is a quadratic form, while the other frame mappings use linear forms. The merits for this machine’s vision-based, auto-calibration methods are: (1) It has little calibration time, (2) It does not need expensive calibration instruments, and (3) Its expense is very low.