Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


Thesis.pdf (5.29 MB)

ETD Abstract Container

Abstract Header

A Holistic Study on Electronic and Visual Signal Integration for Efficient Surveillance

Li, Gang
http://orcid.org/0000-0003-4113-8578
http://rave.ohiolink.edu/etdc/view?acc_num=osu1492639972377528

Abstract Details

2017, Doctor of Philosophy, Ohio State University, Computer Science and Engineering.
Surveillance systems are widely deployed with military, public security, urban management, and transportation applications. They are mainly used for monitoring people’s locations, behavior, and activities. In addition, they can collect information from other objects such as vehicles. Currently, visual (V ) signals play a major role in surveillance because they provide copious details about objects of interest. With an increasing number of surveillance cameras deployed, the volume of surveillance videos grows rapidly, which poses challenges for efficient surveillance. Besides visual signals, electronic (E) signals are very common in surveillance systems as wireless devices are pervasive. Electronic signals show great potential upon integration with visual signals for efficient surveillance. In this dissertation, we study efficient surveillance on both visual and electronic data. First, we explore how to process large surveillance video datasets efficiently with “big data” processing tools. This dissertation focuses on two types of objects of interest: vehicles and humans. We propose TaG, an augmented MapReduce framework for time-bounded analytics jobs on large traffic videos. By studying the characteristics of traffic videos, we propose a novel sampling algorithm based on motion information encoded in videos, which we embody in the MapReduce framework. Besides traffic analytics, we also study rapid retrieval in surveillance videos where humans are the objects of interest. We propose SurvSurf, a human retrieval system using large surveillance video data that exploits characteristics of these data and big data processing tools. We use the MapReduce framework to process video clips in parallel for human detection and appearance/motion-feature extraction. We design a distributed data store called V-BigTable to structuralize semantic information. Second, we explore how to integrate electronic and visual signals for efficient surveillance. We study two main problems in surveillance: localization and human identification. One main purpose of electronic surveillance is finding people’s locations. Channel impulse response (CIR) measurements can help extract Line-Of-Sight (LOS) received signal strength indicators (RSSIs), which can improve range estimation significantly. We propose EV-Sounding, a visual assisted electronic channel sounding system, which leverages cameras to help probe the wireless channel to find a high-resolution CIR rapidly. Such CIR measurements can help extract LOS RSSI to improve the localization accuracy. In visual surveillance, one main purpose is determining humans’ identities amidst different scenes. Traditional techniques process large V and E datasets separately, which does not serve our purposes well as each type of data alone is imperfect for information gathering and retrieval. Matching human objects in the two datasets merges these datasets’ advantages for efficient large-scale surveillance. In this light, we propose EV-Matching, a set of efficient parallel algorithms, to bridge big E and V data based on their spatio-temporal correlations. In this dissertation, we explore achieving efficient surveillance from two angles, using big data processing techniques and integrating electronic and visual signals. By addressing the challenges in current surveillance systems, our proposed solutions have important practical significance in advancing the field in both industry and academia.
Dong Xuan (Advisor)
128 p.
Computer Engineering; Computer Science
Surveillance; visual; electronic; MapReduce; traffic analysis; human retrieval; channel sounding; human identification

Recommended Citations

Citations

  • Li, G. (2017). A Holistic Study on Electronic and Visual Signal Integration for Efficient Surveillance [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1492639972377528

    APA Style (7th edition)

  • Li, Gang. A Holistic Study on Electronic and Visual Signal Integration for Efficient Surveillance. 2017. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1492639972377528.

    MLA Style (8th edition)

  • Li, Gang. "A Holistic Study on Electronic and Visual Signal Integration for Efficient Surveillance." Doctoral dissertation, Ohio State University, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=osu1492639972377528

    Chicago Manual of Style (17th edition)

osu1492639972377528
545
© 2017, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.