3D Facial Data Acquisition Based On Binocular Stereo Views

3D face biometrics recognition as an important branch ofbiological recognition has been a challenging task, and recognition inrecent years as research in the field of biological hot spots. And 3Dfacial points for the reconstruction of 3D Face Recognition provide 3DFace model, three-dimensional face recognition is the front steps in theentire process of 3D face recognition plays an important role.Based on digital image processing technology and binocularstereo vision principle, a single-color active projection lines to obtainthree-dimensional facial points approach, The method used to Faceprojector projection monochrome lines, and then use two normal digitalcameras from two different angles by shooting two two-dimensionalpicture. According to binocular stereo vision and the principles of thecamera calibration parameters have been calculated Face 3D pointcloud coordinates. This paper expounds the theory and Zhang imagingcameras are Friends of the camera calibration method. Focuses on theprinciple of binocular stereo vision,the corresponding point matchingalgorithm and point reconstruction program of ideas and basicalgorithm.Based on the principle of binocular stereoscopic 3-Dreconstruction program is the key corresponding points in the twoimages of the stereo matching.This paper presents a projector with theinitiative projection technology to achieve color lines correspondingpoints Stereo Machining a variety of environments After repeated testswe use projection on the environmental impact of the smaller red linemethod to achieve this. These simple theories of technology,environmental adaptability, and low equipment requirements, have a certain applicationFirst, we collected by the camera image preprocessing. Thetraditional image processing techniques are directed at the gray-scaleimage processing. In this paper, in a variety of gray-scale imageprocessing technology based on the RGB color system characteristics,the gray image sharpening, contrast stretching to the use oftechnologies such as RGB color system, so as to realize the color of acolor image ( red) accurate extraction. Furthermore, in order to avoidthe obstruction, external factors such as noise and makes fault lines inthe situation, we relax the restrictions extract color lines to ensure thatthe extract is connected lines. Despite our projection of color lines onlysingle pixel width, but with digital cameras that collected the face of thecolor lines is often more than the width of a single pixel. In order toachieve our counterparts Matching program, we have adopted arefinement Rosenfeld lines on the extraction method for refining colorlines, which makes the image is a pixel line width, and that the detailsof the lines of the original lines of the skeleton, with connectivity.In order to adapt to the corresponding points of our matchprogram, we need to after pretreatment images numbering lines,making two images about the same number of lines is mutual match.That from the image up to down the middle search of the points wasencoded colors into 1... n. Then the code has been around theneighborhood of the point of difference on the same code in mind.Each of the lines on the part of all, we in the arbitrary lines on eachpoint to the midpoint of the initial origin, recorded as 0; left of theneighborhood were recorded as points 1,2, ... the right neighborhoodpoints were recorded as -1, -2,....This line coding schemes simple, we can meet the basic corresponding points matching the achievement ofprogram. However, due to the obstruction, light from the outside world,the impact of a variety of factors, we have made after pretreatmentdiscontinuous lines. How to ensure that the pretreatment of thecontinuity of color lines we now need is the will to resolve the problem.After pretreatment lines will be numbering, we have about two imagesof the match points, its matching principle is:1. Left image of the first lines of n points in the right image on thecorresponding points should be in the right image of the correspondinglines on the first n.2. Epipolar constraint based on the principle of the cameracalibration of the projection matrix, we can be obtained correspondingmatrix and on the basis of a line constraint equations. Image on the leftpoint in the right image on the corresponding point on should be acorresponding online.3. Order under the principle of consistency, if known about a pairimages should point (p1, p2), a left of p1 point in the right image of thecorresponding points on the left side of the p2; accordingly, p1 the rightpoint in the right image in the corresponding points on the right side ofp2.According to the above three principles, we are about imagesmatch, this simple principle of matching method. The camera angle andsurface inherent factors, will be making a one-to-many matching or newDual situation, we use the solution-Best Strategy First choose the bestmatch point.Finally, we use the intersection of the principle of linear space for the space coordinates, the two match points with the correspondingheart a two-line, two lines of intersection that is, we point the spacerequirements. But because of the error, we have made two straight-linemay be different, so we take a different line of the midpoint of thevertical as we need to calculate the space point. This paper presentsthe binocular stereo vision based on the initiative projection technologymonochrome lines to obtain three-dimensional point clouds Face.Although there are some technical problems to be solved, it issimple in principle, lower equipment requirements, and to give fullconsideration to the environmental impact of different colors and facethe adaptability of the equipment. So is the use by certain prospects...