Speed estimation of vehicle traffic

Abstract

The Metropolitan Manila Development Authority (MMDA) provides real-time traffic monitoring through strategically-located surveillance cameras all over Metro Manila. However, its current scheme reduces traffic analysis to 15-minute interval updates and three-state evaluation, not making use of the much available information offered by traffic videos. In this paper, a simple algorithm was introduced to address such issues by tracking vehicles in real-time and estimating their average speeds using linear regression and calibration. Speed distributions of two MMDA videos were analyzed. The algorithm was able to capture a wide speed range, from 7 km/h to 49 km/h. It was also able to contrast speed distribution between normal and bus lanes.