An Efficient Dynamic Background Subtraction Algorithm for Vehicle Detection Tracking System

Abstract

Background subtraction is an important role in video surveillance system in ITS, yet in complex scenes, it is still a challenging problem; hence, it is required to model the background before subtraction. Various illumination changes and dynamic backgrounds form the major key aspects for background modeling. In this paper, an algorithm (TCO-DBS) is proposed to develop an efficient background subtraction framework to solve the above problems. Here, texture and color features are considered for background modeling, thereby separating the foreground and background video frames. The texture features mainly depend on scale values used, i.e., number of neighboring pixels used for describing local texture description. Among this, local binary pattern (LBP) is mostly used in computer vision applications. LBP texture features along with color feature give a promising result when compared to other methods.