文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > 自然科学 > 物理 > Moving Target Detection and Tracking
侵权投诉

Moving Target Detection and Tracking

运动小目标检测

Research Journal of Applied Sciences, Engineering and Technology 6(24): 4642-4645, 2013 ISSN: 2040-7459; e-ISSN: 2040-7467 © Maxwell Scientific Organization, 2013

Submitted: February 27, 2013 Accepted: April 02, 2013 Published: December 25, 2013

Moving Target Detection and Tracking

Zhan Xu, Du Lingyan, Lei Yuerong, 1Zeng Huiming and 2Chen jianling

1

Department of Electronic Engineering, Sichuan University of Science and Engineering, China

2

Dong Fang Boiler Group CO., Ltd., Zigong, Sichuan 643000, P.R. China

Abstract: This study proposes a background update algorithm based on the correlation coefficient, overcome the shortcomings of fuzzy image of the average background updating. Then, it uses subtraction background method to detect moving targets and record the moving target region and use projection histogram to resize the center of the target and calculate the confidence coefficient of the current frame and the subsequent frame to research the center of the target in the subsequent frames. Then, the moving target could be tracked in the subsequent frames. The experimental results show that the algorithm can accurately detect and automatically track moving targets.

Keywords: Background update, projection histogram, track

INTRODUCTION the moving target region and adopts projection

histogram method to track moving targets.

The problem of tracking non-rigid targets and recognizing them in video sequences is of increasing METHODOLOGY importance to many applications. Examples include a

motion detector in video surveillance systems, action Background method: We randomly save the current analysis for animation medical imaging and Human scene in which no targets are moving as the Computer Interaction (HCI). Tracking a deformable background. But the scene will be change according to target in consecutive frames is of particular concern in time, light and weather condition and so on. It is video surveillance systems. There have been various necessary to update background in real time. Therefore, researches for video-based target extraction and the paper proposes a method to update background. The tracking. One of the simplest methods is to track main steps are shown in the following:

regions of difference between a pair of consecutive

Saving the current scene image and divided into frames (Haritaoglu et al., 2000) and its performance can

the scene image matrixes of size 16*16 be improved by using adaptive background generation

Calculating the scene image matrixes’ variance and subtraction. Although the simple difference-based

tracking method is efficient in tracking an target under Dividing the background image into the matrixes

of size 16*16, and calculating their variance noise-free circumstances, it often fails under noisy,

complicated background. The tracking performance is Calculating covariance and correlation coefficient

ρ between the scene image matrix of size 16*16 further degraded if a camera moves either intentionally

and its corresponding background matrix or unintentionally. For tracking targets in the presence

of shadows, noise and occlusion, a non-linear target Comparing ρ and threshold, if ρ is bigger than

threshold, using the scene image in place of its feature voting scheme has been proposed in Amer

corresponding background image, otherwise, (2003). As an alternative method to frame difference

saving the original background image based tracking, a blob or region tracking, can be used to

locate the centroid of an target. Based on the

Correlation coefficient:

Moving Target Detection and Tracking

assumption of stationary background, Wren et al.

(1997) proposed a real-time blob tracking algorithm.

(1) Another blob tracking methods that use targets ρ=

histogram and mean shift approach have been proposed

in Comaniciu et al. (2003).

This study proposes a background update where,

: The ground image and the scene image of algorithm based on the correlation coefficient, B, I

size M*N, their two-dimensional matrixes overcome the shortcomings of fuzzy image of the

are expressed as f (x, y), g (x, y), (1≤x≤M, average background updating. Then, it uses background

0≤y≤N) subtraction method to detect moving targets and record

Corresponding Author: Zhan Xu, Department of Electronic Engineering, Sichuan University of Science and Engineering,

1

1

1

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email