Automatic Smoke and Obscurant Cloud PR from Visible and Thermal Imagery

The US Army has a requirement to develop and deploy systems to automatically detect smoke and obscurant clouds using visible and infrared imagery. The proposed program will implement a method to determine the spatial extent of clouds using a combination of techniques such as advanced image processing, pattern recognition, knowledge based procedures (KBP), and neural networks (NN). The complexity of detection systems requires highly sophisticated algorithms for operations such as classification, recognition identification and characterization. The capabilities of systems to perform on-board, real-time processing has increased enormously over the past several years. The objectives of this work are to perform the research and implement the software to develop a system for the development and evaluation of algorithms for cloud detection and classification. Phase I will encompass the implementation of a simple system for algorithm evaluation, and make recommendations for a hardware platform and software development. Extensive development and validation will be completed in Phase II.