Police and security teams guarding airports, docks and bordercrossings from terrorist attack or illegal entry need to knowimmediately when someone enters a prohibited area, and who theyare. A network of surveillance cameras is typically used to monitorthese at-risk locations 24 hours a day, but these can generate toomany images for human eyes to analyze. Now, a system being developed by Christopher Amato, a postdoc atMIT's Computer Science and Artificial Intelligence Laboratory(CSAIL), can perform this analysis more accurately and in afraction of the time it would take a human camera operator. "You can't have a person staring at every single screen, and evenif you did the person might not know exactly what to look for,"Amato says. "For example, a person is not going to be very good atsearching through pages and pages of faces to try to match [anintruder] with a known criminal or terrorist." Existing computer-vision systems designed to carry out this taskautomatically tend to be fairly slow, Amato says. |
"Sometimes it'simportant to come up with an alarm immediately, even if you are notyet positive exactly what it is happening," he says. "If somethingbad is going on, you want to know about it as soon as possible." So Amato and his University of Minnesota colleagues Komal Kapoor,Nisheeth Srivastava and Paul Schrater are developing a system thatuses mathematics to reach a compromise between accuracy - so thesystem does not trigger an alarm every time a cat walks in front ofthe camera, for example - with the speed needed to allow securitystaff to act on an intrusion as quickly as possible. For camera-based surveillance systems, operators typically have arange of computer-vision algorithms they could use to analyze thevideo feed. These include skin detection algorithms that canidentify a person in an image, or background detection systems thatdetect unusual objects, or when something is moving through thescene.
To decide which of these algorithms to use in a given situation,Amato's system first carries out a learning phase, in which itassesses how each piece of software works in the type of setting inwhich it is being applied, such as an airport. To do this, it runseach of the algorithms on the scene, to determine how long it takesto perform an analysis, and how certain it is of the answer itcomes up with. It then adds this information to its mathematicalframework, known as a partially observable Markov decision process(POMDP). Then, for any given situation - if it wants to know if an intruderhas entered the scene, for example - the system can decide which ofthe available algorithms to run on the image, and in whichsequence, to give it the most information in the least amount oftime.
"We plug all of the things we have learned into the POMDPframework, and it comes up with a policy that might tell you tostart out with a skin analysis, for example, and then dependingwhat you find out you might run an analysis to try to figure outwho the person is, or use a tracking system to figure out wherethey are [in each frame]," Amato says. "And you continue doing this until the framework tells you to stop,essentially, when it is confident enough in its analysis to saythere is a known terrorist here, for example, or that nothing isgoing on at all." Like a human detective, the system can also take context intoaccount when analyzing a set of images, Amato says. So forinstance, if the system is being used at an airport, it could beprogrammed to identify and track particular people of interest, andto recognize objects that are strange or in unusual locations, hesays. It could also be programmed to sound an alarm whenever thereare any objects or people in the scene, when there are too manyobjects, or if the objects are moving in ways that give cause forconcern.
In addition to port and airport security, the system could monitorvideo information obtained by a fleet of unmanned aircraft, Amatosays. It could also be used to analyze data from weather-monitoringsensors to determine where tornados are likely to appear, orinformation from water samples taken by autonomous underwatervehicles, he says. The system would determine how to obtain theinformation it needs in the least amount of time and with thefewest possible sensors. Amato and his colleagues will present their system in a paper atthe 24th IAAI Conference on Artificial Intelligence in Toronto inJuly.
The e-commerce company in China offers quality products such as Cotton Corduroy Fabric , China Cotton Yarn Dyed Fabric, and more. For more , please visit Cotton Nylon Fabric today!
Related Articles -
Cotton Corduroy Fabric, China Cotton Yarn Dyed Fabric,