Objective To analyze the problems of very slow generation of feature point descriptors and incapacity of self adaption and adjustment of the distance ration threshold during the matching process in original SIFT algorithm, and to propose a fast adaptive image matching algorithm based on improved SIFT. Methods The efficiency of the algorithm was improved through simplifying the feature point descriptor generation process, and the robustness of the algorithm was improved through adaptive adjustment of the distance ration threshold parameters. Results The test results showed that the algorithm matched more accurately with smaller amount of calculation, though the matched points became fewer. Conclusion With the advantages of efficiency and accuracy, the proposed algorithm was suitable for computer vision in industrial nondestructive detection, printing image detection, printing dot image detection and packaging image detection,with good promotional value.