The invention is suitable for the technical field of forensic
dentition science, and provides an automatic forensic
dentition science individual identification method based on a 3D
computer vision algorithm, and the method comprises the steps: carrying out the filtering, denoising and down-sampling
processing of an initial 3D
dentition model, so as to remove the
noise and reduce the data size; segmenting a dental crown region as a
region of interest through an automatic
algorithm, and removing gingiva and palatal plica soft tissues; and performing rigid registration on the segmented ROI model by adopting an iterative
nearest point algorithm, and calculating a root-mean-
square error and a registration fitting degree after registration. According to the method, the steps of filtering, denoising and down-sampling preprocessing are added before model segmentation, the 3D dental model segmentation, registration and comparison method is designed based on an
open source 3D vision algorithm, full-process
automatic processing from model preprocessing, model segmentation, model registration to model comparison is achieved, the registration and comparison precision can be guaranteed, and meanwhile the registration and comparison efficiency is improved. Registration efficiency is greatly improved, computing resources are saved, and subjective errors caused by
manual segmentation of a model are avoided.