Traffic Detection at Nighttime Using Entropy Measurement

A vision-based traffic detection method at nighttime based on entropy measurement is proposed. Due to limited precision and complicate impact factors exist at night time, the impacted factors beams and reflected lighting should be firstly eliminated. After then, the vehicular tail/heading lighting will be the only underlying features to represent the detected vehicle. The exponential entropy of the measured area is then computed. Finally, the Fourier analysis is applied to extract the fundamental entropy value and the ripple factor as the basis to detect traffic. Experimental results demonstrate the feasibility and validity of our proposed method.