For the characteristics of dial board images, this paper proposes an accomplishment method of auto-calibration system of dial gauges. Uses computer vision technology into the calibration of dial gauges, substitutes the traditional calibration method which need human, and accomplishes automatic dial gauge calibration. The dial gauge calibration process consists of three parts: image pre-processing, scale marks recognition, and the pointer indication interpretation. This paper uses image subtraction method, dynamic threshold segmentation method to process in image pre-processing part. In the other two parts, this paper proposes a new method named as region-segmentation, to partition the dial image, only process the useful blocks of dial image, and not process the other areas. This method reduces the computation amount greatly, and improves the processing speed effectively. Then, uses least squares fitting lines method and central projection method to compute scale marks data and detected pointer value. Prove through experiments, this paper accomplishes an intelligent and rapidly dial gauge calibration system.