Face key point detecting method and face key point detecting device

Abstract

The invention discloses a face key point detecting method and a face key point detecting device. The face key point detecting method comprises the steps of: acquiring a human body depth image of a current user and extracting a face depth image in the human body depth image; generating an initial three-dimensional face point cloud according to the face depth image and preprocessing the initial three-dimensional face point cloud, eliminating noisy points and / or cavities in the initial three-dimensional face point cloud, and performing normalization processing to obtain a middle three-dimensionalface point cloud and detecting a face nose tip point according to the middle three-dimensional face point cloud; performing attitude correction on the middle three-dimensional face point cloud according to the face nose tip point, so as to obtain a target three-dimensional face point cloud; establishing an optimization model according to a preset Basel face model and the target three-dimensionalface point cloud, and generating a fitted three-dimensional face image according to the optimization model; and determining face key points according to the fitting three-dimensional face image. The detection method disclosed by the invention can improve the precision and obtain precise positions of the face key points.

Description

technical field [0001] The invention relates to the technical field of face detection, in particular to a method for detecting key points of a human face and a device for detecting key points of a human face. Background technique [0002] In recent years, the emergence of biometric technology can effectively protect people's property safety, personal safety, privacy security and social order stability. Biometric recognition includes voice recognition, palm recognition, fingerprint recognition, face recognition, etc., and the human Face recognition technology is a relatively new research direction of biometrics. Compared with other recognition technologies, face recognition has the advantages of non-contact, non-mandatory and parallel, and has been widely used in banks, enterprises, security, mobile phones , e-commerce and other fields. [0003] At present, two-dimensional face recognition is relatively mature in face recognition, which can meet people's needs under certain ...

AI Technical Summary

Problems solved by technology

[0005] At present, there are some three-dimensional face key point detection methods. For example, in related technologies, the nose point is obtained by judging the concavity and convexity of each point on the face, and some face features are constructed by using the nose point to match and recognize. This method only A key point is used, the extracted features are not rich enough, and only dozens of faces can be distinguished, which is not practical in the case of big data

For another example, in other related technologies, SI and some curvature features are combined to detect 14 key points on the human face (left inner and outer eye corners, nose heel point, right inner and outer corner of the eye, nasion point, left alar, right alar , nasal septum, left and right lips, upper and lower lips, front of the chin), this detection method has a relatively large error under the influence of noise and missing, which will directly affect the recognition rate of face recognition

In addition, in some related technologies, some average features around each key point are obtained by using some manually marked key points in the training set, and these average features are used to find the corresponding key points on the input face, because these average features do not have Unique characterization often leads to large deviations in key point detection, or even no correspondence at all, which is not conducive to subsequent face recognition

Images