Anti-spoofing 3D face reconstruction using infrared structure light
Aspects of the present disclosure describe systems, methods, and structures for anti-spoofing 3D face reconstruction using infrared structured light that advantageously reconstructs 3D face structures for facial recognition and detect face surface material(s) such that human skin may be effectively distinguished from artifacts thereby providing additional security for facial recognition including immunity from 2D/3D print attacks including face masks and special make-ups.
1. An anti-spoofing 3D face reconstruction system comprising:
a structured light generator including an infrared laser light source configured to generate infrared light at 980 nm and a light pattern generator configured to generate different structured patterns of the generated infrared light, the light pattern generator including a digital multimirror device (DMD) configured to generate a plurality of structured light patterns from received light;
a light beam expander interposed between the infrared laser light source and light pattern generator, the light beam expander configured to expand a beam of infrared light generated by the light generator;
a camera configured to detect and digitally image the different structured patterns after they are reflected from an object comprising a face;
a second light beam expander consisting of a double concave lens interposed between the structured light generator and the face, the second light beam expander configured to further expand a structured IR light beam to a proper light spot area for the face; and
an image processor configured to generate a 3D surface reconstruction of the face from the digital images of the reflected structured patterns and simultaneously detect a surface material of the face;
wherein the image processor reconstructs the 3D surface of the face including any distortion of the structured light patterns by triangulation, determines the surface material by comparing a strength of the reflected structured patterns of infrared (IR) light with a baseline strength of human skin, detects abnormal areas by a convolutional neural network (CNN) as part of the surface material detection, and marks or highlights the detected abnormal areas as non-skin segments/areas on the reconstructed 3D surface.