SBIR-STTR Award

LIDAR is a powerful tool for the measurement and characterization of aerosols in the atmosphere. In battlefield environments, a wide variety of aerosol constituents and plume types are present. Traditional LIDAR analysis methods aim to characterize plumes in well controlled environments with a minimum of atmospheric clutter. To extend the abilities of plume characterization and identification in cluttered environments, Michigan Aerospace Corporation proposes to combine its long history of experience in atmospheric LIDAR with its machine learning algorithm development capabilities to generate a robust and powerful real-time tool for use in the battlefield. This tool will provide real time analysis and classification of plume types as measured with atmospheric LIDAR units being operated in the field. Initial results using an in-house LIDAR demonstration system are encouraging; Phase I work will use government-furnished data as indicated in the topic description. The proposed work will allow us to validate measurement performance in more complex and relevant environments.

In battlefield environments, the atmosphere is littered with aerosols from a variety of sources: dust from vehicles and foot traffic; smoke from firearms and explosives; and, potentially, plumes of weaponized aerosols. In order to improve situational awareness, it is critical to develop a real-time knowledge of the presence, type, and structure of aerosol plumes on the battlefield. Elastic LIDAR systems are powerful remote sensing tools capable of detecting, quantifying, and classifying clouds of aerosols.LIDAR instruments have a long history of use in atmospheric remote sensing for making measurements of atmospheric aerosol types and concentrations. Beyond atmospheric science, LIDAR has use in the measurement and quantification of man-made aerosol releases ranging from smoke grenades to chemical explosives to anthrax.To extend the ability of LIDAR systems tocharacterize and classify plumes in cluttered environments, Michigan Aerospace Corporation proposes to further develop itsplumesight algorithm. Operating on raw LIDAR backscatter returns, the plumesightalgorithmreconstructs high fidelity images of the plumeusing machine learning algorithms, and then usesdeep convolutional recurrent networksto learn how plumes of differentsubstances evolve over time. By exploiting these complex spatiotemporal patterns, the plumesight algorithm is able to classify plumes using elastic LIDAR returns.