Traffic analysis using discrete wavelet transform and Bayesian regression

Abstract

Traffic analysis using Discrete Wavelet Transform and Bayesian Regression is used to estimating the size of inhomogeneous traffic, composed of vehicles that travel in different directions without using explicit object segmentation or tracking is proposed. Using the dynamic texture motion model, here the traffic is segmented into components of homogeneous motion. From each segmented region, a set of holistic low-level features are extracted using 4-level discrete wavelet transform. Using the 4 level discrete wavelet transform, I calculate the energy of wavelet coefficients and a function that map features into estimates of the number of vehicle per segment is learned with Bayesian regression. Here two Bayesian regression models are examined. The first is a Gaussian Process Regression with a compound kernel, which accounts for both the global and local trends of the count mapping but is limited by the real-valued outputs that do not match the discrete counts. I addressed this limitation with a second model which is based on a Bayesian treatment of poisson regression that introduces a prior distribution on the linear weights of the model. Experimental results show that regression-based counts are accurate regardless of the traffic size. Velocity of each car can be calculated.