Mizzou team focused on anti-IED surveillance system
A team of Mizzou Engineering researchers is seeking to develop complex surveillance software that would alert American troops to deadly explosives before they can be deployed.
Most U.S. combat casualties in Iraq have been caused by explosive devices, according to Department of Defense records. Military officials have sought to thwart these improvised explosive devices-commonly known as IEDs-with such measures as radio jamming and manned video surveillance systems.
Tony Han, Guilherme DeSouza and Zhihai “Henry” He, all Mizzou electrical and computer engineering assistant professors, are working on an automated video surveillance system that would detect and alert the military to behaviors considered red flags for terrorist activities. The surveillance software itself would be able to identify people in a scene and determine whether their activities are suspicious, dramatically reducing the need for human supervision, Han said.
“The hardware is already in place,” DeSouza said. “This software on top of that hardware can relieve the burden of watching the video-and make it much more effective.”
The team’s research is funded by a $417,704 grant from the Leonard Wood Institute (LWI), a non-profit Missouri research organization financed by the U.S. Army Research Laboratory. The yearlong project calls for the Mizzou team to deliver its software programs to LWI by next September.
The MU team’s project envisions an integrated video system that would prevent IED attacks by warning military authorities if it detects the “signature steps” that typically precede them. Those steps often include reconnaissance, digging, dropping a container with explosives and attaching a detonator to it, Han said.
While existing surveillance techniques might not coordinate scenes of those steps in time to detect an IED before it explodes, Mizzou’s proposed video system aims at automatically detecting and integrating those signature activities and sounding an alarm, Han said.
“We want to pick up a series of suspicious activities,” he said.
Three programming tasks are at the heart of the project: designing software that can effectively determine which objects in the camera’s view are people; software that can accurately follow their movements; and software that can recognize suspicious activities.
First among these is programming that will allow the computer to differentiate people from their surroundings, a challenging task given natural variations in appearances and postures, Han said. By integrating color, texture and shape as features, Han said he has been able to devise software that can better distinguish people in a scene from background clutter.
Since tracking people once the computer discerns them also is made difficult by cluttered backgrounds and human variations, team members are working on new mathematical models describing human posture that are less computationally intense than existing methods.
That means finding new ways to describe the angles of a human body-that is, such joints as the shoulders, elbows and knees-based on accurate estimations rather than painstaking three-dimensional calculations, DeSouza said.
The Mizzou team is tackling the final software development task with the aid of a video database showing people performing IED-related activities. Through increasingly selective screenings, the software will search the video scenes for actions which when linked seem suspicious, team members said.
Taken together, the team’s anti-IED software should prove substantially more accurate-as well as faster and more efficient-than current surveillance techniques, Han said. Small errors are not automatically corrected and accumulate in existing monitoring systems, leading to system failures, he said.
Moreover, Han said the proposed video surveillance system also could be used to track down suspects when terrorists succeed in setting off an IED. Analysts could quickly and easily search the video for human activity in the bomb area, pulling up relevant information that could help authorities find those involved and prevent future attacks, he said.
“Fully applied, the success of our project may decrease casualties caused by IEDs,” Han said.