TY  - GEN
A1  - Lenor, Stephan
UR  - https://archiv.ub.uni-heidelberg.de/volltextserver/20855/
N2  - Highly integrated and increasingly complex video-based driver assistance systems are rapidly developing nowadays. Following the trend towards autonomous driving, they have to operate not only under advantageous but also under adverse conditions. This includes sight impairments caused by atmospheric aerosols such as fog or smog. It is an important part of environmental understanding to thoroughly analyze the optical properties of these aerosols.

The aim of this thesis is to develop models and algorithms in order to estimate meteorological visibility in homogeneous daytime fog. The models for light transport through fog are carefully derived from the theory of radiative transfer. In addition to Koschmieder's well-established model for horizontal vision, a recursively-defined sequence of higher-order models is introduced which yields arbitrarily good approximations to the solutions of the radiative boundary problem.

Based on the radiative transfer models, visibility estimation algorithms are proposed which are applicable to data captured by a driver assistance front camera. For any one of these algorithms, the recording of luminances from objects observed at distinct distances is required. This data can be acquired from moving objects being tracked as well as from depth-extended homogeneous objects such as the road. The resulting algorithms supplement each other with respect to different road traffic scenarios and environmental conditions. All given algorithms are extensively discussed and optimized regarding their run-time performance in order to make them applicable for real-time purposes. The analysis shows that the proposed algorithms are a useful addition to modern driver assistance cameras.
AV  - public
TI  - Model-Based Estimation of Meteorological Visibility in the Context of Automotive Camera Systems
Y1  - 2016///
ID  - heidok20855
KW  - Fahrerassistenz
Frontkamera
Road Surface Luminance Curve
ER  -