Abstract
Researchers developed a portable computational imaging and deep-learning enhanced aerosol analysis device (c-Air) to identify and measure particulate emissions directly from traffic sources. Researchers found that significantly higher numbers of particles were collected per second when the car was in motion compared to the background particle levels measured when the vehicle was stationary. In addition, even more particles were generated during acceleration and braking. This mobile and cost-effective device is able to distinguish non-volatile as well as volatile and evaporating particles caused by brake and tire wear generated by a moving vehicle from background road dust, with a high degree of accuracy in the field. In addition to counting and sizing particles, this system can also classify particles based upon physical features, shape, color and volatility using computational imaging and deep learning.
