A novel multispectral and 2.5D/3D image fusion camera system for enhanced face recognition

摘要

The fusion of images from the visible and long-wave infrared (thermal) portions of the spectrumudproduces images that have improved face recognition performance under varying lighting conditions.udThis is because long-wave infrared images are the result of emitted, rather than reflected,udlight and are therefore less sensitive to changes in ambient light. Similarly, 3D and 2.5D imagesudhave also improved face recognition under varying pose and lighting. The opacity of glass toudlong-wave infrared light, however, means that the presence of eyeglasses in a face image reducesudthe recognition performance.udThis thesis presents the design and performance evaluation of a novel camera system which isudcapable of capturing spatially registered visible, near-infrared, long-wave infrared and 2.5D depthudvideo images via a common optical path requiring no spatial registration between sensors beyondudscaling for differences in sensor sizes. Experiments using a range of established face recognitionudmethods and multi-class SVM classifiers show that the fused output from our camera system notudonly outperforms the single modality images for face recognition, but that the adaptive fusionudmethods used produce consistent increases in recognition accuracy under varying pose, lightingudand with the presence of eyeglasses.