Dynamic background modeling for a safe road design
Anastasios D. Doulamis · 2010
In this paper we propose a new dynamic algorithm able to automatically identify the background from images under severe changes of the environmental conditions. The algorithm exploits motion estimation techniques and background subtraction methods. The proposed scheme is implemented in road construction application scenarios under outdoor environments in which the camera vision is dramatically varied with respect to the weather conditions. Despite such variations, the background can be automatically detected improving the tracking performance for the vehicles and just assisting survey engineers in defining possible errors in the design yielding as deviations between the actual and the ideal trajectory of the vehicle over a turn. Experimental results in real-life road surveillance systems reveal the efficiency of the proposed scheme regardless of significant background changes. In addition, the proposed algorithm requires low computational complexity.