Face Authentication Using Image Signature Generated from Hyperspectral Inner Images

Abstract

Face recognition technologies are commonly used in access control systems. It is done by extracting selected features from the face image, taken by a 2D camera. This technique lacks the case of a picture placed in front of the camera. The system will mistakenly recognize it as a real live person and so, allow the access of the picture holder, which may be an unauthorized person. A new generation of security systems uses a three-dimensional face recognition. Although it is better than 2D, it lacks a similar case, where a 3D image is generated from many 2D images. The system will assume it is a picture taken from a live person, and mistakenly, allow the access. We propose an enhancement to the existing authentication process given 2D face image. It is based on inner images extracted from a hyperspectral camera. These images represent inner layers of the person tissue structure, which in general are different from person to person and so, may be used to differentiate between two persons. We use these generated features to generate an authentication signature. The authentication signature is a composition of processed inner layers features. To prove that this signature is universally unique and can substitute the current use of 2D image recognition system, there is a need to conduct a comprehensive testing and apply other technologies to prove it. We are not at this stage. Therefore, at this stage, we propose adding to each image a unique signature generated from the corresponding hyperspectral inner layers. When a person is trying to access, the access control system, using a hyperspectral camera, captures its standard image features, and in addition, calculates the inner images to generate a relatively unique signature, and compares both elements to the identification table. Experiments show that this combination generates a relatively unique identification key. From the beginning of our initial experiments, it kept its attentiveness and uniqueness for all we tried to challenge it. Further experiments prove the significant contribution of inner features for strengthening the person authentication.