Investigation of intelligent adaptive image enhancement to aid night time driving

• Main Author: Rio, A.
• Other Authors: Thompson, C. P.
• Language: en
• Published: Cranfield University 2016
• Online Access: http://dspace.lib.cranfield.ac.uk/handle/1826/11091

Driving at night is a difficult task. In an attempt to ease this task, most automotive companies are developing systems that aim to increase the safety of the driver and his/her passengers at night. Jaguar Cars Ltd have been involved in such project for several years and have developed a Night Vision System (NVS) based upon the Near Infrared (NIR) and Head-Up Display (HUD) technologies. This thesis is concerned with the application of digital image enhancement algorithms to further increase the driver's visual range at night. The purpose of this research work is to provide the driver with a safe and non-disturbing, enhanced view of the road scene ahead, which is presented on a head-up display. In this automotive environment, specific requirements such as real-time processing, robustness and reliability must be kept in mind to design algorithms that will not compromise the safety of the driver, his/her passengers and other road users. To fulfill these requirements, we have developed a novel intelligent image enhancement scheme for night time driving that actively adapts to the road scene. This scheme results in the enhancement of the contrast in a portion of the projected HUD road scene as if extra headlamps were directed to the region of the image that represents where the road is going. Human Factors studies have shown that this region is where the driver is concentrating his attention when driving. The position of the region of interest is defined by the computation of an approximation of the vanishing point of the road, updated for every frame using a novel, reliable and optimised road edge detection algorithm. The enhancement of the contrast within the region of interest is obtained by applying several novel low-level algorithms based upon the grey level segmentation of the image into regions and the use of the global histogram equalisation and quantised bi-histogram equalisation algorithms. These novel algorithms have all been implemented on the Matrox Genesis board based upon the multitasking, multiprocessor and parallel DSP TMS320C80 chip from Texas Instruments. All algorithms described in this thesis are able to sustain real-time processing at the NTSC frame rate of 30 frames per second. This new concept for a night time driving aid is an attractive solution that meets the numerous requirements driven by Human Factors research in an automotive environment, in particular safety requirements.