Neural networks for the automated detection of chemical plumes and marine oilspills in airborne infrared multispectral remote sensing images

作     者：Chen, Zizi 

作者单位：University of Iowa 

学位级别：博士

导师姓名：Gary W. Small

授予年度：2020年

摘      要：The United States Environmental Protection Agency Airborne Spectral Photometric Environmental Collection Technology (ASPECT) program uses airborne remote sensing to support first responders in responding to incidents in which chemical or radiological materials are released into the environment. Two main tasks of the ASPECT program are the detection of chemical plumes in the atmosphere in accidental chemical releases and the detection of marine oil slicks on seawater in oil spill accidents. Detections are made in near real time by applying software-based mathematical classification models (classifiers) to the imagery data collected by a downward looking infrared (IR) multispectral sensor on the aircraft. The classifiers classify each image pixel as plume or non-plume or oil or non-oil, depending on the application. The research described in this dissertation focuses on the development of classifiers for future use in the above-mentioned emergency response applications.; Three classifiers have been developed for chemical plume detection. Methanol was used as the target compound to demonstrate the methodologies for the development of the classifiers. In the development of the first plume classifier, training data were collected from controlled methanol release field experiments that mimicked an accidental chemical release at an industrial facility. The classifier was built using multi-layer shallow neural networks (MSNN) on a set of optimized ratios of band intensities obtained from a series of feature extraction procedures applied to the IR radiance data.; The limitation of the first plume classifier was that it required time-consuming and expensive field experiments to obtain the analyte-active training data, which motivated the development of the second plume classifier. The second plume classifier utilized a simulation methodology to compute the plume radiances that comprised the analyte-active training data. This methodology used Planck s radiation law and