Utility Maximization of a Video Sensor Network
Video has many applications in the battlefield environment. Different than consumer networks, in tactical video networks, data must be delivered within a strict time interval because it is being used for decision making. Consequently, mobile ad hoc networks require capabilities to support the combined control and communication features of tactical video networks. These capabilities can be incorporated into the concept of a network of systems (NoS). A NoS is an abstraction that represents the interactions of multiple different systems and/or networks operating in a common framework. In a tactical video network, video sensor nodes capture data to be sent over a wireless system to be interpreted by a human for information extraction. A wireless network may not have enough capacity to support many high quality video streams. Consequently, quality variations are to be expected. The goal of a NoS would be to maximize the utility of the video frames delivered.
With that, WNCG graduate student Andrew Thornburg and professors Alan C. Bovik and Robert W. Heath Jr. developed a system to maximize the utility of a video sensor network. An algorithm to determine the feasible set of nodes and optimal packet schedule via a Markov decision process, given their perceptual constraints and channel qualities was proposed. The H.264/SVC codec is used for compression, while the visual information fidelity pixel-domain (VIFP) full-reference image quality assessment model is used to measure perceptual quality. The algorithm is implemented in custom C++ software with over-the-air transmissions using 802.11 WLAN hardware. Real-time encoding is done for over-the-air transmission of the video. The results indicate that gains of 10-30% can be realized for a three node system.
Research supported by Raytheon.
This work was presented at the 2013 Asilomar Conference on Signals, Systems, and Computers and is available on IEEEXplore.
Paper: Multi-User Real-Time Wireless Video with Perceptual Constraints