The invention relates to the technical field of machine vision, and specifically discloses a face recognition method based on a central point tracking model, and the method comprises the steps: obtaining a plurality of frames of target images; performing face tracking on the target images of the front and back frames, and determining a corresponding tracking sequence ID; performing face alignmenton the face image of the tracking sequence ID, and extracting corresponding face features; after one tracking sequence ID is finished, constructing a pre-processing face ID sequence; performing quality evaluation on the face images in the pre-processed face ID sequence, and selecting face features in the first Num face images with the best quality as a cleaned face ID sequence feature group; and denoising face features in the face ID sequence feature group, and determining a final face recognition result through reordering of the face features. According to the method, the influence of factorssuch as illumination, pedestrian flow shielding and face angles on the face recognition effect in a dynamic video environment can be reduced, and the face recognition precision is improved.