MODELING REFLECTION PROPERTIES OF ROAD SURFACES BY DATABASE METHOD
Wenyi Li, Zhiyong Zhang, Yong Min Yang, Muqing Liu, Haiping Shen · PROCEEDINGS OF the 29th Quadrennial Session of the CIE · 2019
This paper proposes a new method to model r-tables without an analytical expression, but by database method.A so-called 'knowledge base' for road surfaces is setup, which is constructed by verified r-tables.The data model is trained and built up by some selected r-tables, then predictions are made for the other r-tables in the knowledge base, and prediction errors are then analysed to verify the model.In this paper, 287 r-tables of dry road surfaces is tested.The model is built up by its r-tables with even number indices.The r-tables with odd number indices are then predicted.Results show that 90% of the prediction errors are below 15%, and a great number of examples are distributed around 5%.Only two worst cases are of error around 50%.These results are much better than those we obtained by analytical expression methods.By this database model, if we have a large enough data base with reliable r-tables, covering road surfaces of different materials, service and weather conditions, we can make predictions for most road surfaces with a few measured value, then make a more precise road/street lighting design.