Projects

• Home
• Research
  • Human Gesture Analysis
  • Smart Environments
  • Smart Camera Architecture
  • Vision-based Localization
  • Surveillance
• People
• Publications
• Download
• Courses
• Contact

Surveillance

• Hybrid-Resolution Mote for Surveillance

Click here for a description of Stanford MeshEye mote

• Face Profile Collection and Tracking

  

• Distributed Intelligent Surveillance

  In most today's implementation of surveillance systems, pan-tilt-zoom cameras are distributed across the deployment area and their raw video output is streamed to a surveillance center, in which a panel of monitors displays the video streams. Obviously, this implementation requires sufficient bandwidth for video streaming, has high installation cost, and most of all is hardly scalable. We consider any surveillance solution that performs processing of the video stream right at the camera and hence reduces bandwidth requirements as an intelligent system.
  As a first level of intelligence, the camera nodes use a motion detection scheme such that only moving scenes are streamed to the surveillance center. At a second level, the camera nodes could perform object detection and classification such that only moving scenes containing persons or more general objects of interest are forwarded. Going even further, the smart camera nodes could collaborate to identify objects and only transmit their textual description along with a snapshot. Continuing this train of thought of adding intelligence to surveillance, a network of smart cameras could possibly just notify the surveillance center in case of events of interest by providing a hybrid textual/visual or fully textual description of the event. As the level of intelligence increases, bandwidth requirements on the underlying data transmission network decrease accordingly.
  

• Attribute-based Tracking

  This project presents an implementation of a color-based multiple agent tracking algorithm targeted for wireless image sensor networks. It uses multiple image sensors to track and graphically display the path of autonomous agents moving across the overlapping fields of view (FOV) of the sensors. A color histogram is constructed for each agent to identify the dominant hue of each agent, and this hue value is used as a means of identification. The algorithm is able to reliably track the agents when collisions occur and to locate the relative position of each agent during collisions. This work also studies the possible use of color histograms in image sensor color balance self-calibration, which would be a possible future extension of the project. This algorithm has low computational requirements and its complexity scales linearly with the size of the network, so it is feasible in low-power, large-scale wireless sensor networks.
  

• Event-driven Routing

  Security monitoring systems are deployed specifically to detect events that occur rarely but require immediate notification. On the other hand, the design of sensor network-based systems that comprise energy constrained nodes is typically dictated by bandwidth efficiency and longevity concerns. Therefore, the design of such systems must not only strive to provide packet delivery guarantees over potentially multiple hops, but also consider dynamically changing parameters at network nodes such as queue sizes.
  In security monitoring systems, the observer may not be interested in all images but only the images of certain events. Detection of situations that may need the observer's attention or intervention, monitoring the rate at which moving objects flow through the observed environment, or registering the types and quantities of certain objects or events are among such applications. These applications may only require occasional transmission of images to the observer. For example, in an application to monitor an environment, the nodes of the network periodically transmit packets to acknowledge that it is alive to base station. Occasionally, when a node detects a suspicious target, it may buffer and transmit a few image frames, which can be used for further analysis.
  In this project a distributed routing scheme with adjustable priority support for event-driven wireless surveillance networks is developed. The proposed algorithm employs a cost function based on the network topology, current queue lengths, and remaining energies at neighboring nodes as a basis for next hop selection, and can provide improved end-to-end delay and lifetime performance. The network nodes are assumed to generate periodic data packets that are reported to the destination via multihop routing. Nodes may also infrequently detect an event from which a set of image frames need to be reported. In order to report the observed events on time, we assign different priorities to the image packets based on the importance of the packets such that the observer can still obtain high quality images when we have delay constraints in the base station. Simulation results indicate lower end-to-end average and maximum delays, significantly reduced buffer size requirements for the network nodes, and improved network lifetime. In addition, the proposed algorithm can provide better quality images in terms of the average peak signal-to- noise ratio of the transmitted images.