University of Cincinnati
Networked video surveillance systems are indispensable components in today’s public safety
infrastructure. The situational awareness of public safety professionals could be greatly improved if they receive high quality video observations along with their analytics results in real time. Overall, the proposed research targets at evaluating the quality of information (QoI) for public safety videos and designing QoI-based sensing, compression, and communication solutions. More specifically, the outcomes from the first two years of this project could impact the design of video sensing and compression solutions, offering new ways for improving the quality of videos for analytics tasks and fostering future research initiatives in this area. -July 2019
Principle Investigator: Rui Dai
University of Cincinnati
Video surveillance systems have been used heavily in public safety for various applications aiming at improving the situational awareness of first responders. In recent years, more and more wireless cameras are deployed for video surveillance. To achieve satisfactory coverage and observation quality for public safety use cases, there is an increasing demand of bandwidth for delivering surveillance videos, which puts pressure on the wireless networking infrastructure. It is essential to make good use of the available wireless bandwidth to extract meaningful information from raw surveillance videos. Traditionally, multimedia networking protocols aim to leverage network resources to satisfy QoS (qual-ity of service) or QoE (quality of experience) requirements. However, QoS or QoE parameters do not directly reflect the amount of information that could be obtained from a networked system. It is a largely unexplored task to quantify the quality of information in wireless surveillance systems. Moreover, little is known about how video networking schemes should be designed to boost the quality of information for surveillance tasks.
This project will design an information-driven video communication framework for wireless video surveillance systems with the objective to maximize the information gain for public safety applications. With the advances in computer vision and artificial intelligence, automatic video analysis tools, which attempt to detect, recognize, track objects, and understand their behaviors, have been applied in surveillance systems. In contrast to centralized video analysis, applying automatic video analysis at network edges not only enables better real-time response to users but also could potentially alleviate the bandwidth pressure through selective transmission of meaningful video clips.
This project will design and develop communication solutions for maximizing the information that human operators could gain with the help of distributed automatic video analysis.
The major challenges for this project are: i) to accurately quantify the quality of information for surveillance applications; ii) to enable efficient automatic video analysis in dynamic wireless networking environments; and iii) to deliver bandwidth-consuming visual information to first responders in real-time. Our research plan will consist of three major tasks. First, we will build analytical models for the quality of visual information obtained from automatic video analysis tools in networked surveillance systems. We will establish relationships between information quality and parameters that are observable in the net-work. Second, we will design communication protocols to support efficient and real-time distributed in-network video analysis. Third, we will design an information-driven video streaming application to effectively disseminate the information extracted from surveillance cameras to human operators.
Success of the proposed research will greatly enhance the situational awareness of first responders and increase the speed and precision of decision making for various incidents in public safety. The communication prototype from this research will be demonstrated to the public safety stakeholder community. The software and simulation platforms developed in this project will be made available to the research community for further studies on networked video surveillance. Results from the proposed research will be disseminated to audience with broad background through publications and presentations in research seminars, workshops, conferences, and journals.