The 3DID face alignment system for verifying identity

摘要

The 3DID system verifies the identity of a cooperative person by matching a sensed 3D surface of the face to a face model stored during a prior enrollment. First, anchor point detection is performed based on a shape index; then, a rigid alignment is determined between the observed and model face anchor points. A best alignment is determined using an improved Iterative Closest Point (ICP) algorithm that aligns the surfaces allowing for trimming of 10% noise points. Trimmed Root Mean Squared (RMS) error for the same person is almost always smaller than 1.3 mm; whereas for different persons, it is almost always larger than this threshold. Performance analysis shows that the 3DID system is fast enough (<2 s on a 3.2 MHz P4), reliable enough (1% equal error rate with 1.5% reject rate), and flexible enough (handles 30° of yaw and 15° of roll and pitch) to be practical in several applications. 3DID is also user friendly, providing several displays of intuitive value to human agents either in online or delayed analysis mode. An inexpensive scanner is needed for widespread use.