Advanced framework for illumination invariant traffic density estimation

CCTV cameras are becoming a common fixture at the roadside. Their use varies from traffic monitoring to security surveillance. In this paper an advanced two-stage framework for estimating vehicular traffic density on a road segment is presented. The proposed approach is computationally efficient and robust to varying illumination. The method is novel because it combines state-of-the-art image processing techniques with a simple traffic model in order to increase robustness. Experimental results have shown that the proposed framework can achieve higher performance than existing state-of-the-art techniques under conditions of varying illumination.