By: Cassidy Delamarter, University Communications and Marketing
What sounds like science fiction is inching closer to reality, thanks to the work of University of South Florida Assistant Professor John Murray-Bruce.
Inside the 国产福利资源 Bellini College of Artificial Intelligence, Cybersecurity and Computing, the computational imaging researcher and his students are using shadows and information not easily detected by the human eye to push the limits of what machines can see and how they see it. The work at its core is theoretical, but the real-world implications are vast: From mapping terrain under dense foliage in real-time for defense operations to detecting hidden threats around corners.
鈥淢y interest in pursuing this idea of seeing around corners grew after my family was t-boned in an intersection even though we had the right of way,鈥� Murray-Bruce said. 鈥淚f we had that superhuman vision of being able to peer into the intersection, perhaps we would have been able to react accordingly to avoid that crash.鈥�
With two major research projects underway, one funded by the Defense Advanced Research Projects Agency, the federal agency known for inventing the internet, and the a National Science Foundation鈥檚 , Murray-Bruce is developing powerful new ways to see the world, even when it can鈥檛 be directly seen.
The five-year NSF project, which began July 1, focuses on developing a rigorous mathematical framework to improve how computational systems interpret hard-to-read or incomplete data, which could transform how images are created.
鈥淭he research has created a foundation that could support a wide range of applications we hadn鈥檛 even anticipated, including long-range X-ray imaging and national security,鈥� Murray-Bruce said.
It鈥檚 seemingly science fiction, he added, which he says is what really piqued his interest. 鈥淓xtreme types of imaging 鈥� seeing around corners, seeing deep into human tissue or biological specimen at a very fine resolution. Think a million times finer than a human hair.鈥�
The research stems from a framework that Murray-Bruce and his students first explored through the study of shadows. What started as an effort to reconstruct 3D images of obstructed scenes from ordinary photographs of indirect light reflections evolved into an entirely new way of modeling imaging systems.
While the NSF grant focuses on the mathematical theory and its potential applications, the DARPA project is all about high-stakes testing to push X-ray imaging to extraordinary distances. A traditional X-ray system, such as those used in hospitals and airports, requires the object it is scanning to be within about 36 inches to produce an image. The challenge posed by DARPA: Build a system that works across several miles, while still producing usable images.
鈥淲e鈥檙e competing with some of the biggest names in defense technology on this project 鈥� teams from major government contractors,鈥� Murray-Bruce said. 鈥淎nd yet we鈥檙e standing firm. It鈥檚 a testament to the strength of our approach and relentlessness of my talented students.鈥�
Robinson Czajkowski, doctoral student, setting up a hidden scene
Generating images from the scene
Using shadows to see around the wall to see the Bulls logo
His students play a central role in the research and experiments. For Chibuike Ezeokoli,
a rising doctoral senior studying computer science and engineering, the hands-on environment in Murray-Bruce鈥檚 has helped him expand his research. 鈥淚鈥檝e had the opportunity to collaborate with
and learn from other students,鈥� Ezeokoli said. 鈥淚t is the ideal environment to nurture
the skills required for a future career trying to solve crucial real-world problems
through research.鈥�
Together, Murray-Bruce and his students developed novel algorithms and modeling techniques capable of interpreting data at unprecedented ranges. Their system currently achieves 97% accuracy just under one mile and 75% accuracy at nearly three miles.
Chibuike Ezeokoli in the lab running simulations of obstructed images
鈥淭hese results are promising, though we are not yet at the limit of what is theoretically possible. It鈥檚 an extreme version of shadow-based imaging,鈥� Murray-Bruce explained. 鈥淵ou send X-rays through an object, and depending on what gets absorbed, you capture shadows at the detector. The problem is, the longer the distance, the weaker and noisier the shadows become.鈥�
If selected by DARPA to continue to Phase 2 of the program, Murray-Bruce and his team will investigate scenarios with even greater distances and considerable motion blur.
Murray-Bruce is actively recruiting more graduate students and a postdoctoral researcher. For students interested, please email murraybruce@usf.edu.