Three-Dimensional Face Imaging and Recognition: A Sensor Design and Comparative Study

Many commercially available 3D sensors suitable for face image capture employ passive or texture-assisted stereo imaging or structured illumination with a moving light stripe. These techniques require a stationary subject. We describe a design and evaluation of a fixed-stripe, moving object 3D scanner designed for human faces, called the Static Light Screen (SLS) Scanner (Patent Pending). Our method of acquisition requires the subject to walk through a light screen generated by two laser line projectors. Triangulation and tracking applied to the video sequences captured during subject motion yield a 3D image of the subject's face from multiple images. To demonstrate the accuracy of our initial design, a small-scale facial recognition experiment was executed. In an experiment with 476 images of 161 unique subjects, we achieved 46.3% rank-one recognition using an Iterative Closest Point (ICP) based matching method, demonstrating the feasibility of the technique. This is the subset of our full SLS dataset with 2270 images of 379 unique subjects, which has a subject velocity less than 0.33 m/s. We also developed techniques to improve face recognition based on ICP using fusion techniques and score normalization techniques. We improved rank-one recognition on a data set (FRGC v2) of 4007 faces of 466 unique subjects from 96.4% rank-one recognition to 98.6%.