Efficient Vehicle Counting and Classification using Robust Multi-Cue Consecutive Frame Subtraction

Abstract

The ability to count and classify vehicles provides valuable information to road network managers, highways agencies and traffic operators alike, enabling them to manage traffic and to plan future development of the network. Increased computational speed of processors has enabled application of vision technology in several fields such as: Industrial automation, Video security, transportation and automotive. The proposed method in this paper is a robust adaptive multi-cue frame subtraction method that detects foreground pixels corresponding to moving and stopped vehicles, even with noisy images due to compression. First the approach adaptively thresholds a combination of luminance and chromaticity disparity maps between the learned background and the current frame. The segmentation is further used by a two-step tracking approach, which combines the simplicity of a linear 2-D Kalman filter and the complexity of 3-D volume estimation using Markov chain Monte Carlo (MCMC) methods. The experimental results shows that the proposed method can count and classify vehicles in real time with a high level of performance under challenging situations, such as with moving casted shadows on sunny days, headlight reflections on the road using only a single standard camera.