Image Processing for Surveillance and Security

Image Processing for Surveillance and Security

Deepayan Bhowmik (Sheffield Hallam University, UK) and Mehryar Emambakhsh (Heriot-Watt University, UK)
Copyright: © 2017 |Pages: 30
DOI: 10.4018/978-1-5225-2498-4.ch003
OnDemand PDF Download:
$37.50

Abstract

Security is a fundamental issue in today's world. In this chapter we discuss various aspects of security in daily life that can be solved using image processing techniques by grouping in three main categories: visual tracking, biometrics and digital media security. Visual tracking refers to computer vision techniques that analyses the scene to extract features representing objects (e.g., pedestrian) and track them to provide input to analyse any anomalous behaviour. Biometrics is the technology of detecting, extracting and analysing human's physical or behavioural features for identification purposes. Digital media security typically includes multimedia signal processing techniques that can protect copyright by embedding information within the media content using watermarking approaches. Individual topics are discussed referring recent literature.
Chapter Preview
Top

Image Processing For Visual Tracking

Automatic detection, tracking and anomaly identification occupy a considerable space in computer vision research and have many applications including intelligent surveillance, human-computer interaction (HCI), human-robot interaction (HRI), augmented reality (AR), medical applications and visual vehicle navigation etc. More importantly, recently there is a great deal of interest in robust visual tracking algorithms due to the increased need of automated video analysis relating safety in public places such as railway station, airport, shopping areas, religious festivals or governmental offices. This poses fundamental challenges in computer vision which involves input from image processing research along with input from machine learning community. In this section we discussed fundamental steps for visual tracking by dissecting the algorithms available in recent literature.

Image analysis for tracking consists of three key steps: 1) detection of interesting moving objects; 2) tracking moving objects in time lapsed frames and 3) analysis of tracked objects for behavioural study such as recognition, prediction and anomaly detection. One can broadly dissect the visual tracking algorithms in four different categories (Yilmaz et al., 2006; Haering et al., 2008; Yang et al., 2011): a) Object representation, b) Feature selection, c) Object detection and d) Object tracking. These categories can also be grouped and fitted into a classical image processing pyramid of low level, medium level, intermediate level and high level algorithms, based on the complexity and the type of data they process (Figure 1).

Figure 1.

Image processing pyramid of visual tracking algorithms

Complete Chapter List

Search this Book:
Reset