Discipline: Computer Sciences and Information Management
Subcategory: Computer Science & Information Systems
Malik Majette - North Carolina State University
Co-Author(s): Daquille Campbell, North Carolina Central University, Durham, NC
Face images acquired from thermal sensors and visible light cameras have large differences, and as such, recognition across these face modalities is very challenging. The handcrafted features designed in many existing algorithms (e.g., histogram of oriented gradients) often cannot uncover the underlying face structure. This project aims to develop a novel algorithm to match thermal face images to the data sets that contain visible light face images. In this work, convolutional neural networks (CNNs) are trained to extract and integrate face features shared by both modalities. The performance of the proposed algorithm is evaluated on three data sets that contain images acquired under different standoff distances and infrared (IR) spectra. We demonstrate that the proposed algorithm performs robustly for the data sets with different settings.
Funder Acknowledgement(s): This work was supported by the National Science Foundation Grant HRD-1238547
Faculty Advisor: Alade Tokuta,