Abstract : Dense fog is a major road safety issue. Improving road users' safety involves prevention, forecast, detection, as well as signalling, lighting and driver assistance. However, in order to design suitable road or vehicle solutions compensating for the loss of visual information, accurate understanding of how fog reduces visibility is necessary. Evaluating such solutions in terms of driving behaviour requires the use of simulation, because tests in real foggy traffic are unpractical. The goal of this work is to describe how fog impairs the driver's visual environment, in order to model and simulate the visual effects of fog.
In the first bibliographic part, we notice the gap between the complex and varied microphysical and optical aspects of fog on the one hand, and the oversimplified notion of "visibility range" commonly used to describe its perceptive effects on the other hand. We show that frequency analysis is necessary to characterize the scattering-induced perturbations to the luminance distribution which forms the visual signal. In order to synthesise the images needed for this approach, we use a ray-tracing technique which allows physically-based rendering of scattering media.
In the second part, we first describe the Monte Carlo light tracing technique we implemented to simulate light propagation with multiple anisotropic scattering in a polydispersed medium such as fog. The resulting simulation code is then used to study the properties of the modulation transfer function of a "slab" of fog, considered equivalent to an optical filter. This leads to the definition of a frequency contrast operator which characterizes the halo effect of scattered light. Based on Koschmieder's law, we propose an extended model of fog visual effects - extinction, halo and veils (atmospheric and back-scattered) - which can be used to predict the modifications generated by fog in the visual environment of road users in any traffic situation. We also show that the proposed model is compatible with real-time driving simulation.
The photometric model of fog visual effects developed in this work has already been tested experimentally by driving psychologists. It has also been adapted by real-time graphic simulation specialists to enhance the fidelity of fog conditions on driving simulators. It will also be used to study the performance of road equipment in reduced visibility conditions.