Camera calibration is one of the key technologies of Computer Vision. This paper presented a microscope camera calibration method based on grid sub-pixel corner detection for the particular application area of microscopic measurement. First of all, the actual corner coordinates information of the grid was obtained through the improved Harris corner detection method. Then, considering the distribution law of corner coordinates in the microscopic image, the paper obtained the sub-pixel corner coordinates by combining the quadratic surface and linear fitting. Finally, through the established non-linear camera model, the average error between fore-projection and re-projection grid corner coordinates was obtained by re-projecting the grid corner coordinates. Experimental results show that the sub-pixel corner detection algorithm is accurate and the final calibration error is 0.4637 pixels. Compared with improved Harris corner detection, the accuracy increased by about 52%, which is applicable to microscopic camera calibration.