Research On 3D Face Recognition Based On The Projected Grating Phase Method

Face recognition is an identification technology based on facial features,it has been widely used in security monitoring and authentication.This paper applies face recognition to home service robots in order to improve their interactive capabilities and intelligence.At present,2D face recognition technology is widely used.However,the amount of information contained in 2D images is relatively small,so its performance is affected by factors such as ambient light,shooting angle and posture.The amount of information in the 3D image is more abundant,3D face recognition can overcome these influences and improve the recognition effect.A 3D face recognition method based on the projected grating phase method is proposed in this paper.The specific work is shown as follows:(1)A phase unwrapping method combining the least square method and the branch cut method is proposed in this paper.This method is used to unwrapping the wrapped map.Firstly,the idea of solving the residual point in the branch cut method is used to search the bad points,and fix the bad points according to the good points around them.Then,the least squares method is used for unwrapping the wrapped map in order to get a 3D face image.Finally,the method of curve fitting is used to repair the 3D face image.Through simulation and physical experiments,the effectiveness of the method is verified.(2)A 3D face feature location method based on stripes and shape index is proposed in this paper.Firstly,the basic information of the human face in the image is determined according to characteristics of the fringe image,and the ear position is located according to these information.Then,the search range of the nasal tip is determined in the 3D face image according to the position of the ear and the basic information of the face,and the position of nasal tip is located accurately according to height and shape index.Finally,the search range of the canthus is determined according to the position of the nose while the positions of canthus are found accurately according to the height and shape index.Experiments were conducted in dark environment and natural environment.The experimental results show that this method can effectively locate the nasal tip and canthus,and it is less affected by natural light.(3)This paper fuses the calculation results of side profile and horizontal contour near the nose at the decision level and uses the fusion result as the basis of recognition.Firstly,the contour from nasal root to nasal tip and the horizontal contour line,that is through the nasal tip and equal to the length between canthus,are extracted as characteristic curves according to the positions of nasal tip and the canthus.Next,the characteristic curve and the reference curve are registered using the ICP algorithm,and the average Euclidean distance of the corresponding point of the registered characteristic curve and the reference curve is calculated.Finally,the results of the two characteristic curves are weighted fused and the result of the recognition is obtained according to the result of the fusion.The experimental results show that,in ideal dark environment,the average matching time is 0.193 seconds and the recognition rate is 95.65%.When the face rotation angle is less than 20 degrees,the recognition rate is 97.67% in ideal dark environment and 94.83% in natural environment.