Visible light reconstruction method of thermal infrared face image based on kernel sparse canonical correlation analysis

Abstract

The present invention proposes a thermal infrared human face image visible light reconstruction method based on kernel sparse canonical correlation analysis, and its steps are as follows: use a thermal imager and an optical camera to collect several training samples at the same time, and establish a data set; for the global training of the data set, use The data obtained from the global training is globally reconstructed on the thermal infrared test image to obtain the reconstructed global visible spectrum image, and the local refinement training of the data set, the reconstruction of the local refinement of the thermal infrared test image to obtain the reconstructed visible spectrum residual image block, using the visible The average value of the overlapping pixels of the spectral residual image block is added to the reconstructed global visible spectrum image to obtain the visible spectrum face reconstruction image. The invention obtains the best projection direction of two sets of data of thermal infrared spectrum and visible spectrum, improves the separability of features, eliminates redundant information, obtains the best recognition effect, and solves the problem of human face under low-light environment Recognition problems, make full use of the existing visible light database for face matching.

Description

technical field [0001] The invention relates to the technical field of target face identity authentication in low-light environments, in particular to a visible light reconstruction method for thermal infrared face images based on kernel sparse canonical correlation analysis. Background technique [0002] At present, the face samples stored in the public security judicial system or other identity authentication systems are all visible light images. However, at night or in places with insufficient light, it is difficult for conventional monitoring systems to obtain effective visible light images, which leads to difficulties in obtaining key evidence in criminal cases. . Thermal imaging cameras operate in the thermal infrared spectrum, by capturing infrared radiation imaging, independent of illumination, dependent on temperature changes of objects. The distinguishing features of thermal infrared spectrum face images are determined by inherent factors such as human facial musc...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the technical problem that the existing visible light face recognition technology cannot cope with weak light environment, the present invention proposes a visible light reconstruction method of thermal infrared face image based on kernel sparse canonical correlation analysis, which solves the problem that the visible light face image acquired under low illumination cannot Problems Matching Existing Visible Light Databases

Images