Near-infrared and visible light cross-modal face recognition method based on modal augmentation

Abstract

The invention discloses a near-infrared and visible light cross-modal face recognition method based on modal augmentation. The method comprises the following steps: acquiring a near-infrared face image set and a visible light face image set; constructing a generative adversarial network for conversion from a near-infrared face to an auxiliary visible light face, training the generative adversarial network, and converting a near-infrared face image set into an auxiliary visible light face image; taking the auxiliary visible light face image as an auxiliary mode, and forming a three-mode image group together with the near-infrared face image set and the visible light face image set; constructing a convolutional neural network, and training by using the three-modal image group to obtain a cross-modal face recognizer; and inputting to-be-recognized data into the recognizer for cross-modal face recognition, and outputting a recognition result. Auxiliary modals can be added to make up the difference between the modals, and cross-modal invariant face feature information of near-infrared and visible light is fully extracted, so that the cross-modal face recognition accuracy of the near-infrared and visible light is improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a near-infrared and visible light cross-modal face recognition method based on modal augmentation. Background technique [0002] Face recognition is one of the most active biometric recognition technologies in the field of computer vision. By developing a deep convolutional neural network (CNN) model, the performance of face recognition in visible light has reached more than 99%. However, the face image under visible light has a disadvantage that the intensity of the face part of the image is largely dependent on the lighting environment, for example, in a dark street or indoor, the intensity of the face image under visible light will change significantly. [0003] Near-infrared imaging has been shown to be less sensitive to visible illumination changes, so near-infrared imaging technology provides an effective and straightforward solution for improving face ...

AI Technical Summary

Problems solved by technology

[0004] The different imaging principles of face images of different modalities lead to even the face images of near-infrared modalities and visible light modalities of the same person, but there are still huge visual differences between them, which is the difference between single-channel images and multi-channel images. The image information is seriously unequal caused by

The huge visual difference greatly reduces the recognition accuracy of traditional face recognition algorithms on cross-modal face datasets.

Images