Research On Video Target Recognition Based On 3D Face

Face recognition is currently one of the most concerned biometric technologies and one of the hot topics in recent research.With the development of informatization,advances in the field of computer vision research,two-dimensional face recognition technology has been applied in various fields of society,such as the gradual adoption of two-dimensional face technology under the scene of customs,airports,finance,and security.For identity recognition,however,2D face recognition technology is still constrained in an unfavorable environment,and the gap between the recognition accuracy of actual applications and the experiment is also prominent,especially in the case of expressions,lighting,makeup,etc.The recognition rate of face recognition technology has dropped significantly.Business and academic circles have also increased research on 3D face recognition technology and achieved good results.The 3D face recognition technology is compared with 2D human face to obtain more comprehensive face information.In a non-ideal environment,such as uneven lighting,makeup,posture changes,etc.still have a higher recognition rate.For the purpose of accurately identifying pedestrian targets in video,this paper has carried out the following research work:(1)By constructing an acquisition environment,a multi-camera collaborative acquisition model of a 3D face is studied,and a 3D face modeling is implemented using a combination of SFS(Shadow Recovery Shape)and LMM(Local Deformation Model).This article integrates SFS and LMM methods to restore 3D information more effectively,and solves the problem that SFS has a good recovery effect on the forehead and cheek parts of the face,but it is not a good problem for the parts of the human face(eyes,nose,ears,etc.).High accuracy 3D face model.(2)Aiming at the problem that SIFT algorithm features high dimensionality and large computational complexity,an improved SIFT algorithm based on PCA is proposed.For a large number of feature points in the face image,128-dimensional means a huge amount of calculation.This algorithm improves the feature descriptors by reducing the dimension,reducing the dimensions of the descriptors to 48 dimensions.Compared with the original SIFT algorithm,the computational complexity of the improved algorithm has been greatly reduced,and it can more effectively improve the speed of face recognition.(3)Through the feature extraction algorithm improved in this paper,the video target detection and recognition based on the three-dimensional face model database,and compared with the video target recognition results based on two-dimensional face images.At the same time,the simulation of the human face is blocked and the video target recognition experiment based on the three-dimensional face model is performed and the result analysis is performed.(4)At present,the research and application of face recognition methods are mainly performed in non-complex environments.When faced with obstructions and non-positive faces,3D face recognition has a higher recognition rate and time than 2D face recognition.Degree of better results.