Error measurement and analysis for a 3D face surface matching system

Purpose – To report on the evaluation of error of a face matching system consisting of a 3D sensor for obtaining the surface of the face, and a two‐stage matching algorithm that matches the sensed surface to a model surface.Design/methodology/approach – Rigid mannikin face that was, otherwise, fairly realistic was obtained, and several sensing and matching experiments were performed. Pose position, lighting and face color were controlled.Findings – The combined sensor‐matching system typically reported correct face surface matches with trimmed RMS error of 0.5 mm or less for a generous volume of parameters, including roll, pitch, yaw, position, lighting, and facecolor. Error accelerated beyond this “approximately frontal” set of parameters. Mannikin results are compared to results with thousands of cases of real faces. The sensor accuracy is not a limiting component of the system, but supports the application well.Practical implications – The sensor supports the application well (except for the current co...