To improve the accuracy of gun image stabilization calibration, a fitting calibration model was designed to improve the calibration property of a gun's collimator, which is expected to increase the shell's hitting probability. A threshold segmentation model fused via frequency-domain filtering was constructed to eliminate the influence of uneven imaging gray levels. A long line-fitting model and calibration-positioning axis model were constructed to calibrate the target center and realize accurate detection of the image's stabilization calibration, respectively. The results demonstrate that the proposed method exhibits good segmentation and denoising effect and can effectively eliminate the uneven phenomenon of imaging gray levels. The fitting correlation coefficient of the horizontal and vertical demarcated long lines is 0.995646 and 0.993913, respectively, and the calibration error of horizontal and vertical coordinates is less than or equal to 0.15% and 0.005%, respectively. Overall, the experimental results demonstrate that the designed calibration method can accurately achieve calibration detection of gun sight.