The paper studies and makes an automatic temperature verification system based on computer vision in high accuracy temperature measuring to control technique, computer vision measuring technique and digital image processing. In order to acquire the higher temperature accuracy and long-term stability relative to the normal electricity measures, first-class standard mercury thermometers are adopted directly as a temperature measurement standard. The mercury thermometer indication was detected by digital camera imaging, and image processing methods are used to determine the calibration line and the exact location of the mercury column vertices, including image enhancement, image smoothing, image thresholding, edge detection, edge linear data fitting etc. After edge extraction, the system obtains the pixel location in the image of the standard mercury thermometer scale and mercury column vertex, and obtains accurate temperature values by calculation. Finally, the paper acquires the data by testing of image processing to first-class standard mercury thermometer images and the results of person's eye, contrasted and analysed with ideal temperature values, which proves the accuracy of computer vision measuring is an order of magnitude higher than of person's observation.