Moving Target Detection Based on Improved Mixture Gauss Model - Multimedia and Signal Processing - page 261

Moving Target Detection
Based on Improved Mixture Gauss Model *

Gang Liu 1,** , Yugan You 1 , Siguo Zheng 2 , and Fanguang Li 1

1 School of Electrical Power and Automation Engineering,
Shanghai University of Electrical Power, Shanghai 200090, China
2 Shanghai Power Economic Research Institute,
Shanghai Municipal Electric Power Company, State Grid, Shanghai 200010, China
[e-mail]

Abstract. Based on mixture Gauss model to detect moving targets is easy to
produce "ghosting", smear and background update problem caused by light
mutation, so this paper proposes an improved mixture Gauss modeling method
background update problem. This method combines the frame difference
method to identify moving pixels and non-real motion pixels in the foreground
image. By giving non-real pixels a larger learning rate to make them fast
blending into the background, we solve the "ghost", smear and light mutations
problems. The experimental results show that this algorithm can effectively
detect moving targets.

Keywords: Moving object detection, Background subtraction, Gauss mixture
model.

1 Introduction

Intelligent visual surveillance system is an important research field in computer
vision, it has a wide range of applications in safety monitoring, communications,
medical and other fields. In intelligent monitoring system, moving target detection is
a fundamental and critical step. Moving target detection [1-2] is to real-time detect
moving targets in a series of video sequences. At present the commonly used methods
for target detection are background difference method, frame difference method and
optical flow method. Among them, optical flow method's computation is more
complex, its hardware requirement is very high, and it is not suitable for real-time

* Fund: This work was supported in part by National Natural Science Foundation of China
(No.61203224/F030307), the Leading Academic Discipline Project of Shanghai Municipal
Education Commission (No.J51303), and the Sic-Tech Innovation Foundation of Shanghai
Municipal Education Commission (No.13YZ101.).
** Gang, Liu (1977.2-) is a master instructor and a via-professor in Shanghai University of
Electrical Power, and mainly engages in the fields of the power quality analysis, the fault
diagnosis and automatic control of substation equipment, the scene identification and
planning of power plants, the multi-sensor image fusion and information fusion research.