Ask a typical client what they want from a security camera, and they’ll tell you they want a system with a big, wide view that makes it easy for security to know what they are viewing and helps them identify situations before they become problems.
The Ohio State University (OSU) has a smart system that does all of that and more. The system scans views of thousands of people and identifies potential problems—people who may be intruders, intoxicated or just lost. Most people on campus do the same thing at the same time (e.g., enter a building at 8 a.m., or move purposefully toward a parking deck at 5 p.m.); so, it’s easy for the computer to detect unusual behavior.
Standard security cameras collect the data, and each camera has an established field of view.
“The system uses activity analysis and behavior analysis,” said James W. Davis, OSU professor of computer science and engineering, who developed the system and is expanding its capabilities.
Working with Davis in the lab are doctoral student Karthik Sankaranarayanan, who is funded by the National Science Foundation, and grad student, Matt Nedrich. The Air Force Research Lab also supports the work.
“If I am interested in a particular individual or aggregation, I can push that individual into the program and get a value of how typical their activity pattern is,” Davis said.
The longer the system runs, the more astute it becomes. It does not know what problem the person presents, only that they are acting abnormally.
“We care what you do, not who you are,” he said. “We are not trying to replace the human at the desk, but for a security guard trying to follow 100 cameras, this detects the onset of abnormal behavior and allows security to intervene.”
Security cameras pan, tilt and zoom (PTZ) and rotate 90 degrees up/down. But when security looks through these cameras, they get only a “soda straw” view. As they move the camera around, they can lose a sense of where they are looking within a larger context. In OSU’s system, cameras take snapshots from every direction, resulting in a 360-degree, high-resolution (1,000-by-1,000-pixel) image.
Once a view is displayed on-screen, operators can click anywhere within it, and the camera will PTZ to that spot for a live shot. Or, security could draw a line on the screen to track a particular route—a specific street, for instance.
Davis compares it to Google Earth. The operator uses the standard joystick to point and zoom in on an image. Using orthophotographs, the operator can get close to the action. In many cases, the operator might not know where the image is or may want to access a camera that gives a particular view. With the OSU system, there is no delay. Software drives the camera over the entire image, stitching together a series of images in a seamless panorama. The view is not the typical rectangular image but a fisheye view.
“Usually, a new security operator has to figure out the scene and stitch it together in his head,” Davis said. “This gives one, single image.”
Conversely, the operator might know which building has a problem but not recall which of the dozens of cameras in the array will give the best view. A simple click on, for example, a doorway on the map drives the program to the camera that covers that spot exactly.
“Security operators really like that feature,” Davis said. “They don’t have to find the best camera and then hunt up the area. The program pinpoints it for them.”
Currently, there are seven such cameras on OSU’s Columbus, Ohio, campus. They are standard Pelco Spectra III Series domes and Spectra IV PTZ units.
This year, OSU will incorporate the 500-plus cameras across campus into the network. A mix of Pelco and Bosch cameras will be used. The system also will work with Sony.
“When we developed this for the electrical contractor, we wanted something that was not tied to a single expensive camera but would work with any standard commercial camera,” Davis said.
The software is Windows XP-based and runs on Dell computers. Davis chose to go with a hard-wired, rather than wireless, interface for two reasons. First, on the busy campus, interference could be an issue. Second, all the building-based cameras in the buildings are networked already.
Cameras grab all video, convert to a USB video connection, and process it through the computer for analysis. While most new systems today are IP-based, OSU uses analog.
The next challenge Davis sees is tracking in a dense crowd.
“That’s not yet ready for prime time,” he said, noting the difficulty of tracking someone who goes behind a truck to reappear elsewhere in the scene or when the camera sees a person from the front at one moment and from behind—where the view is quite different—the next.
Contractors who would like to purchase the software should contact OSU. The university owns all the intellectual property.
HARLER, a frequent contributor to SECURITY & LIFE SAFETY SYSTEMS, is based in Strongsville, Ohio. He can be reached at 440.238.4556 and firstname.lastname@example.org.