In view of the problem that the eye can not detect the high voltage gauge and the standard high voltage meter, a digital display instrument automatic calibration system based on machine vision is proposed. The system using two cameras collects the character image of test high voltage measurement instrument and the standard high voltage measuring instrument. Then the system gets the binary image of the character using threshold value and makes the morphology processing .After that it segments the character image by horizontal-projection algorithm, and then the improved structural features are extracted ,finally using the K-Nearest Neighbor algorithm recognize the char. This method solves the problem of poor accuracy, low efficiency of visual identification. After testing, the system is successful in real-time, synchronous recognition of the high voltage measurement instruments and standard high voltage meter reading, and the recognition rate is above 99.6%.