Artificial Intelligence Tool for Chemical and Biological Plumes Detection

Accurate and timely detection and classification of chemical and/or biological threat agents (CBA) used by an adversary in a battlefield environment is a crucial task for reconnaissance and force protection systems. A key technology component in CBA plume sensing systems is the light detection and ranging (LIDAR) system. In Phase I a preliminary AI algorithm based on the autoencoding technique was developed. Field LIDAR data from CBA releases were compressed, filtered for artifacts and used to train the algorithm. The algorithm was tested on a separate set of data. The algorithm discriminated between different release mechanisms. In Phase II we will develop end-to-end software based on the Phase I algorithms. The algorithms will be extended to include a classifier to automatically discriminate different types of aerosols and releases. A database of synthetic data produced from high fidelity dispersion models will be produced to supplement field data. This will enable train the algorithms to discriminate within wider variety of meteorological and terrain conditions, and aerosol interferents. AI algorithm training will be performed on high-performance computers and also transitioned to a GPU system to reduce training time. Once trained the algorithms will operate in real time.