In the quest to enhance the whole body imaging technology that is being deployed to the nation’s airports the Department of Homeland Security (DHS) is taking two approaches in fostering the development of software that can automatically detect objects concealed on a person’s body, a department official says.
The Transportation Security Administration is currently working with the suppliers of the Advanced Imaging Technology (AIT) machines to develop the Automatic Target Recognition (ATR) algorithms, James Tuttle, head of the Explosives Division within the DHS Science and Technology Directorate, tells TR2. If successful, those spiral upgrades would be installed into existing equipment.
In addition, S&T recently issued a Broad Agency Announcement (BAA) aimed at funding third parties to develop ATR algorithms that could be installed on the AIT machines, Tuttle says (TR2, March 17).
Right now TSA doesn’t believe that the algorithms that the AIT vendors have developed to automatically detect anomalies concealed beneath a person’s clothing are ready for use, Tuttle says.
At Schiphol Airport in Amsterdam officials there have been testing ATR algorithms developed by L-3 Communications [LLL] for use on the company’s ProVision AIT machines that are deployed at some of the airport’s security and customs checkpoints (TR2, Jan. 20). Instead of an image that reveals a person’s private areas, Dutch screeners see a stylistic picture of a human figure with any objects on the person highlighted.
The use of the ATR algorithms permit more targeted searches of a person when they alarm. Using a picture of a stylistic human figure to represent any persons being imaged by the AIT machines reduces, of not eliminates, privacy concerns associated with the personnel screeners.
Sen. Collins on Schiphol ATR Testing
Maine Sen. Susan Collins (R), the ranking member on the Senate Homeland Security and Governmental Affairs Committee, says the AIT machines being used at Schiphol with the ATR algorithms are superior to what the TSA is deploying in the U.S. Following a recent trip to Schiphol, Collins told Maine’s Capitol News Service this month that the AIT machines being used in Amsterdam are superior because “No radiation is involved, it uses radio saves. Second, it avoids the privacy issues that have arisen with the style of full- body scanners that the Department of Homeland Security is deploying in this country.”
However, Tuttle says that S&T’s Transportation Security Laboratory has tested L-3’s ATR algorithms but that “TSA feels that they are not where they need to be. There needs to be further development.”
L-3’s ProVision system is based on millimeter wave technology. So far TSA has purchased and deployed 40 of these systems. TSA has also contracted with OSI Systems‘ [OSIS] Rapiscan division to acquire 150 backscatter X-Ray based Secure 1000 AIT systems. The agency is in the process of deploying the Rapiscan systems, which is slated to be finished this summer.
TSA plans to purchase and deploy another 300 AIT machines this year. So far only L-3 and Rapiscan have AIT machines that are qualified by TSA.
TSA officials have told Congress they are hopeful that they’ll be able to install improved privacy or ATR-type algorithms on AIT machines in the near-term. If so, that won’t obviate the need to continue to improve on the algorithms, Tuttle says. Overall, it’s a “tough challenge” to automatically discover anomalies on people who come in different shapes and sizes, he says.
S&T is planning a two year ATR effort. Phase one is for one year and entails government evaluation of technical reports, proposals and design reviews. In the second phase further development of selected algorithms will ensue. S&T has budgeted $2 million for the algorithm development although Tuttle says the amount of spending will depend in part on how many proposals are selected and what they call for.
The BAA that S&T has released is open to any third party developers, whether they are private firms, universities or academia. Tuttle says that S&T has already provided AIT related data sets to the academic community so they are aware of the development needs and challenges.
The academic community is a “good example” of people who “have never even applied their math to this problem,” Tuttle says. S&T is also reaching out to other agencies such as the Defense Advanced Research Projects Agency for help, he says.
Having third parties that could contribute software to existing AIT systems, or any screening systems purchased by DHS, would represent a strong step by the department toward open architecture type solutions. For example, the Transportation Security Administration currently relies on the suppliers of its screening equipment, whether it is for checked bags, carry-on bags or passengers, to provide any software upgrades to installed equipment.
In the BAA the S&T Directorate is looking for developers to supply ATR algorithms that have the flexibility to work with more than one type of machine.
“Additionally, the algorithm design shall include the capability to perform as a modular component, integrated with multiple screening systems,” the BAA says.
“Obviously we don’t want to get into developing something that only one company can use,” Tuttle says. “I don’t want to say that the algorithms can work for any technology. There might be a set of algorithms that work great for millimeter wave, another set that works for backscatter, and another set for terahertz. We’ll see where that goes.”
By allowing other vendors to compete for these software upgrades the TSA would reap the benefits of increased competition in the form of more advanced products and or lower costs.
There is no clear cut path moving forward. Tuttle says that TSA is currently defining the “trade space,” working with the AIT vendors to understand their interest in working with third party software developers and their willingness to release screening images for use by others to develop the targeting algorithms.
“There’s a whole bunch of different things that could be done so they’re working with the vendors to see what the vendors would be applicable to,” he says. “Obviously they (TSA) want to be able to deploy them on the current machines so they don’t have to start with a whole new inventory of equipment. That’s one of the things we pushed for. It’s got to work with existing equipment and can be incorporated into future stuff.”
The TSA hasn’t given S&T specific requirements for the ATR algorithms, Tuttle says. But in the development the goal is to get the false alarm rate to 10 percent or better, he says.
“Meaning if one out of 10 persons alarmed, assuming no one had threats, [TSA] would have to deal with one out of 10 persons,” Tuttle says.
As for any future concepts of operations that would include the use of ATR capabilities with AIT machines, Tuttle says that’s a TSA matter.
More than likely though, once an ATR capability is introduced, “that would directly reduce the manpower needed to support this and it would address the privacy issues,” Tuttle says.