What if it were possible to scatter hundreds of wireless cameras in the wake of a natural calamity and have them create a single, three-dimensional, real time map of the entire area to assess all the damage?

“Camera nodes may be dropped out of helicopters onto a battlefield or they may be distributed throughout a hazardous environment by mobile search-and-rescue robots,” said Radke. “In search-and-rescue contexts, cameras could be deployed in destroyed buildings, radioactive environments or unstable structures that wouldn't bear the weight of several people, in order to locate trapped people, localize the sources of fire, etc.”

The networks could also prove invaluable in situations such as the aftermath of Hurricane Katrina. With the entire wired infrastructure of the city virtually destroyed, government and private contractors had to set up ad hoc networks of nodes to enable wireless communication in the area. With the aid of camera-equipped nodes, rescue personnel could, for example, locate people trapped on roofs, track boats navigating through the streets or monitor changing water levels.

“Having real time imagery of a disaster site is an integral part of having situational awareness,” said Gerard McEnerney, director of Emergency Preparedness at St. John's University in New York . “This is especially true when we deal with a natural disaster or a disaster that encompasses a significant area. Aerial imagery has played a major role in aiding emergency managers to assess the scope and extent of disasters. Having multiple synchronized cameras that could plot out the disaster area could aid in prioritizing response and identify, based on the area, where life safety issues were most critical.”

Where am I?

Presently, Radke is working on the challenge of having cameras in the network determine the presence of other cameras viewing the same scene from a different perspective and establish their positions relative to each other and their environment. This is done by creating a protocol in which each camera composes a short “digest” of distinctive features in its image and sending it around the network to see if any other cameras see some of the same features.

The unique part of his research is designing a totally decentralized system requiring no ordering on the set of cameras. Instead of all cameras directing their information back to a single powerful computer, each camera needs to send information only to its neighbors that see part of the same scene.

“In the wireless network scenario, there are many disadvantages to a centralized scheme,” said Radke. “If the ‘master node' is damaged or destroyed, then the whole network will be compromised. Furthermore, nodes that are close to the master node will have to relay a lot of messages back and forth and they, too, may burn out. On the other hand, in a distributed approach, all the nodes communicate with just their neighbors, and if one node burns out, the rest of the network is unaffected.”

The underlying premise of the scheme is that all of the information is distributed in pieces throughout the network until an outside entity asks for it. The network could undertake many tasks auto ­ nomously without reporting all the details back to a base. When all of the collected information is needed at one place—such as a map of all the terrain imaged by the cameras—all the data could be passed to one of many nodes connected to the base, ensuring that the different nodes did not duplicate the transmission of the same piece of information.

McEnerney believes that more should be done than just drop cameras in the disaster area. “If a camera rolls off a mountain and into flood waters, what would its value be?” he asked. “The system, as described, would need to be well integrated with existing capabilities of city, state and federal professional emergency response managers to determine the value added to our present readiness, response and recovery posture.”

Limitations

The limitations of the technology right now consist of the inability of the cameras to communicate beyond a short distance or maintain continuous contact due to power limitations and short-range antennas. Typically, wireless sensors are battery-operated and don't have a lot of power to send or receive a large number of messages or perform the same kinds of computations that a desktop computer can. Radke is working on systems that would make the cameras communicate efficiently within their power constraints.

He is currently experimenting with a 16-node network of normal cameras and is testing and developing the software required to run larger wireless networks within simulations. He eventually hopes to team up with interested parties to inexpensively build real wireless nodes that could perform in real time.

Other goals for the future include investigating higher-level applications on camera networks. “Once the cameras all know where they and their neighbors are,” he noted, “we can start to look at collaborative tasks like tracking multiple objects as they move through an environment, detecting changes in the world, creating maps of terrain or automatically detecting and relaying information about specific events. For example, the camera network could be ‘told' to detect smoke or fire and to prioritize communication about these important events. Such networks will be essential for 21st century military, environmental and homeland security applications.” HST