The work aims to propose an image matching algorithm based on improved FAST operator coupled projection constraint rule for the current problems that the projection correlation between feature points is ignored which results in reduction of algorithm robustness and more matching errors since the current image matching algorithms mainly optimizes the matching feature points by measuring the similarity of feature points. Firstly, FAST operator was used to extract feature points of image, and Harris operator was used to remove pseudo feature points from feature points to obtain stable feature points. Then, the Gaussian curvature value of pixels in the circle domain was adopted to describe the feature points. Finally, the normalized cross correlation coefficient was used to match the feature points. The projection value of the feature points was calculated by the projection relation function between the feature points, and the projection constraint rule was established according to the projection value to remove the mismatch points and obtain the matching results. From the experiment data, the proposed algorithm had better robustness and matching effect than the current image matching algorithm. Under multiple geometric attacks, the correct matching rate of proposed algorithm was kept above 90%. The proposed algorithm still has good matching accuracy under various geometric algorithms and has good reference value in image processing, trademark retrieval and other fields.