At the Microwave Exhibition 2010, Tokyo University’s Akira Hirose Lab exhibited mine visualization technology using subsurface radar. This mine detector utilizes a complex neural network, where computing is modeled on neurons in the brain. By observing the phase and amplitude characteristics of the reflected waves when EM waves are emitted from the antenna to the ground, this detector can distinguish mines from other ground features, which is difficult to do with the human eye.
“Previously, mine detectors worked as metal detectors, using characteristics such as specific wavelengths from metals. But by using a neural network, this detector can distinguish other objects that aren’t part of the ground. In other words, it can respond flexibly: rather than just detecting specific mines, it can detect improved versions of the mines, and it can make such judgments “at a glance”.”
This detector was successfully tested in a mine-detecting program in Cambodia. The Hirose Lab plans to continue its research, in order to make the detector even more accurate.
“Right now, we’re at the stage of understanding the fundamental theory of mine detection. Investigating the scope of its applications alone involves a lot of issues that take 10 or 20 hours to resolve. So there are various things that we want to improve before the detector becomes a commercial product. Also, neural network processing itself has many applications, so we’d like to work on that aspect as well.”