Periodical calibration is the important method for making instruments in good condition. In this paper, a new measurement instruments calibration scheme is introduced, which adapt the techniques of the computer vision discrimination, incorporate human’s ability of visual and apprehensive models, various meter calibration can be optimized automatically. The central idea of the proposed method is to capture the measuring instrument images in real-time when the standard adjustable source outputting analog, the location of needle and dial can be ascertained using computer image recognition. Based on computer vision techniques, which allows the implementation of completely automatic calibration procedures even in the case of the instruments are not equipped with computer interfaces.