Computer vision algorithms for PTZ camera based smart surveillance system

Abstract

Nowadays, the installation of surveillance cameras in unrestricted newlineplaces has been proliferated and subsequently a huge video data generation has newlinealso increased to a great extent. Smart surveillance has received a significant newlineattention of extremely active and widespread application-oriented research areas. newlineRecently, the advanced sensor technologies like PTZ (Pan Tilt Zoom) camerabased newlinecomputer vision techniques have remarkably increased. It can provide newlinedetailed information of data to cover a wider area. The objective of this thesis is newlineto present a brief survey and propose framework-based on motion segmentation, newlinefacial pose recognition, and facial expression analysis on the PTZ camera-based newlinesurveillance.Accordingly, the background modeling has an increasing significance newlinein the computer vision to segment the foreground objects for further analysis in newlinevideo surveillance applications. The survey attempts to address the challenges, newlinesolutions, key aspects of the PTZ camera-based foreground segmentation newlinemethods, categorization of different approaches as well as the available datasets, newlinethat are used for experimentation on this emerging area.The combination of Region-based Mixture of Gaussian (RMOG) and Extended Center Symmetric Local Binary Pattern (XCS-LBP) has been proposed for motion segmentation to cope with continuous pan, excess zoom, and sudden illumination conditions. Moreover, another key contribution of the work is the strong experimentation with different case studies on benchmark datasets, including Change Detection (CDnet 2014) dataset to show the newlinerobustness of the proposed work. newline newline