Visual Surveillance of Human Activities: Background Subtraction Challenges and Methods

2019 ◽  
pp. 43-66
Author(s):  
Thierry Bouwmans ◽  
Belmar García-García
Keyword(s):  
Background Subtraction ◽  
Human Activities ◽  
Visual Surveillance

A Novel Background Subtraction Method Using Multiclass Support Vector Machine

2014 ◽  
Vol 701-702 ◽  
pp. 265-269
Author(s):  
Shao Na Zhou ◽  
Shao Rui Xu ◽  
Hua Xiao
Keyword(s):  
Support Vector Machine ◽  
Background Subtraction ◽  
Visual Surveillance ◽  
Support Vector ◽  
Subtraction Method ◽  
New Approach ◽  
Vector Machines ◽  
Multiclass Support Vector Machine ◽  
Multiclass Support Vector Machines

Background subtraction, where the foreground is segmented from the background, is the first step of data analysis and processing in automated visual surveillance. Aiming to solve the problems associated with dynamic, multi-modal background, we explore a new approach which can handle the unconstrained environment. Based on multiclass support vector machines, a new MSVM is proposed for the classification of the background and the foreground. The simulation indicates our proposed algorithm is feasible.

Automated Alarm Based on Intelligent Visual Analysis

2013 ◽  
Vol 340 ◽  
pp. 701-705
Author(s):  
Zhi Hua Li ◽  
Qiu Luan Li
Keyword(s):  
Object Tracking ◽  
Event Detection ◽  
Background Subtraction ◽  
Visual Analysis ◽  
Visual Surveillance ◽  
Appearance Model ◽  
Connected Component ◽  
Abnormal Event Detection ◽  
Multi Target Tracking ◽  
Surveillance Applications

Abnormal event detection and automated alarm are the important tasks in visual surveillance applications. In this paper, a novel automated alarm method based on intelligent visual analysis is proposed for alarm of abandoned objects and virtual cordon protection. Firstly the monitoring regions and cordons position are set artificially in the surveillance background scenes. The forground motion regions are segmented based on background subtraction model, and then are clustered by connected component analysis. After motion region segmentation and cluster, object tracking based on discriminative appearance model for monocular multi-target tracking is utilized. According to the motion segmentation and tracking results, alarm is triggered in comparison with the monitoring regions and cordons position. Experimental results show that the proposed automated alarm algorithms are sufficient to detect the abnormal events for alarm of abandoned objects and virtual cordon protection.

Sign in / Sign up

Export Citation Format

Share Document