Design And Implementation Of A Three-dimensional Face Binocular Measurement System Based On Time-phase Unwrapping

Compared with other biometrics,face recognition has the advantages of non-contact and convenient acquisition.Most of the existing face recognition methods are two-dimensional face recognition,but two-dimensional face recognition technology has many defects which are hard to overcome,such as illumination influence,posture influence and expression influence.3D face recognition method can effectively avoid many defects in two-dimensional face recognition.But for now,its development is facing many difficulties,among which collecting difficulties is the first difficulty of 3D face data acquisition.Therefore,this thesis studies a best fringe coding method based on temporal phase unwrapping.It can not only improve the measurement accuracy of the system,but also improve the measurement efficiency and stability of the system,and build a set of 3D face binocular measurement system based on this method.The main work of this article includes the following:(1)Based on all kinds of phase unwrapping algorithms,a perfect mathematical model for the anti-noise ability is established.And according to the mathematical model,an optimal temporal phase unwrapping algorithm is selected to improve the accuracy of the phase unwrapping of the face.(2)Based on the optimal temporal phase unwrapping algorithm that has been selected,through the rigorous theoretical analysis,multi-directional simulation comparison and actual experimental verification of the established noise model,a set of effective coding and decoding schemes for sinusoidal fringe coding are determined to improve the efficiency,accuracy and stability of 3D data acquisition.(3)This thesis takes the basic principle based on the monocular reconstruction.The three dimensional data obtained from both the left and right angles are preprocessed by filtering,then the processed 3D face data are spliced to get a complete 3D face.Based on the above work,the system successfully achieves the 3D information of the face in less than 0.1 seconds,and achieves the measurement accuracy of the root mean square error about 0.04mm,providing a powerful guarantee for 3D face recognition.