Symbolic Classification of Traffic Video Shots

Abstract

In this paper, we propose a symbolic approach for classification of traffic video shots into light, medium, and heavy classes based on their content (congestion). We propose to represent a traffic video shot by an interval valued features. Unlike the conventional methods, the interval valued feature representation is able to preserve the variations existing among the extracted features of a traffic video shot. Based on the proposed symbolic representation, we present a symbolic method of classifying traffic video shots. The symbolic classification method makes use of a symbolic similarity measure for classification. An experimentation is carried out on a benchmark traffic video database. Experimental results reveal the efficacy of the proposed symbolic classification model. Moreover, it achieves classification within negligible time as it is based on a simple matching scheme.