Jeremy Dawson, associate professor of the Lane Department of Computer Science and Electrical Engineering, is the recipient of a four-year subcontract consisting of one base period and two option periods valued at $750,000 to collect biometric data that will be used to develop software algorithm-based systems capable of performing whole-body biometric identification at distances as far as 300 meters or more.
"The face and whole-body imagery we collect will help improve human recognition performance, making biometric systems more equitable for different age, gender and ethnicity groups and allowing better recognition in challenging conditions," Dawson said.
The large-scale biometric dataset that will be collected will consist of images and videos of a person's face, gait, body shape and type. The data will be used to improve the accuracy of the algorithm's ability to identify a person from low-quality images, like those captured from security cameras or drones.
"The controlled data will be collected at a relatively close distance with high-quality cameras; then, we will go out into the field and collect imagery from a longer distance and more extreme angles while asking our participants to walk, interact with their phone or other specific actions," Dawson said. "Gathering enough data in situations that would make it harder to recognize people is our primary focus."
This multi-year research effort from the Intelligence Advanced Research Projects Activity (IARPA) is called the Biometric Recognition and Identification at Altitude and Range (BRIAR) program. Dawson's research group aims to provide the biometrics research community with whole-body imaging data from hundreds of participants on to address the challenges of identifying people at extreme distances and angles. The data will be collected from individuals who volunteer to participate, requiring individuals to offer full consent for their data to be collected and used in research.
“Face recognition performance is severely impacted by distance, the individual’s pose, the lack of adequate resolution; all those things weigh heavily on how well facial recognition systems have been able to correctly identify people,” Dawson said. “The data that we collect will enable the development of new face recognition algorithms that are robust enough to handle those challenging conditions that are present in in real world operational scenarios.”
WVU is a subcontractor to Systems and Technology Research (STR) and will provide the data to STR and IARPA. Then the data will become a resource for the general research community in biometrics and researchers will be able to request the data set to use for their systems.
"The algorithms that are currently used for facial recognition are all based on artificial intelligence and machine learning," Dawson said. "These types of systems need to be developed using a lot of data captured under many different conditions to perform really, really well. The data collected in all three phases of this project, even though we're subcontracted to one company, will eventually get dispersed out into all the performers funded through the BRIAR program."
Undergraduate and graduate students at the Benjamin M. Statler College of Engineering and Mineral Resources will gain real-world experience through this research project by developing the systems and testing the procedures used to collect the data and creating novel techniques for biometric data manipulation.
"Our graduate students will oversee the development of the systems. They will test the systems that will collect the data, and even preprocess the data to get it ready for delivery," Dawson said. "There's also going to be the opportunity for undergraduates to help with data collection by being our sensor operators. After the grad students develop a lot of the systems, the undergraduate operators gain invaluable experience with new camera technologies, system implementation and troubleshooting."
Dawson said this research will lead to the creation of new algorithms, but it will also improve existing systems.
"This opens the door for developing new algorithms that can take advantage of the richness of the variety of data that we're collecting," Dawson said. "We're not just collecting face, gait or the whole body, but we're collecting all of these things that will be fused in the recognition pipeline to enable better performance across all systems."
Republished courtesy of WVU.