Skip to main content
Publication

The Mertens Unrolled Network (MU-Net): A High Dynamic Range Fusion Neural Network for Through the Windshield Driver Recognition

Publication Type
Conference Paper
Journal Name
Proceedings of SPIE
Publication Date
Page Number
18
Volume
11415
Conference Name
Autonomous Systems: Sensors, Processing and Security for Vehicles & Infrastructure 2020
Conference Location
Anaheim, California, United States of America
Conference Sponsor
SPIE
Conference Date
-

Face recognition of vehicle occupants through windshields in unconstrained environments poses a number of unique challenges ranging from glare, poor illumination, driver pose and motion blur. In this paper, we continue the development of a camera system to overcome these challenges. Using this system we collect a small dataset of through-windshield image captures of known drivers. We then formulate the classical Mertens-Kautz-Van Reeth HDR imaging method as a neural network, called the Mertens Unrolled Network (MU-Net), for the purpose of High Dynamic (HDR) fusion of these images. This novel HDR method is then evaluated as the input to a state-of-the-art facial recognition network and compared against the performance of traditional HDR methods and a previously developed GAN method for HDR image correction.