The goal of a successful surveillance system to achieve persistence is to track everything that moves, all of the time, over the entire area of interest. The thrust of this thesis is to identify and improve upon the motion detection and object association aspect of this challenge by adding spectral information to the equation. Traditional motion detection and tracking systems rely primarily on single-band grayscale video, while more current research has focused on sensor fusion, specifically combining visible and IR data sources. A further challenge in covering an entire area of responsibility (AOR) is a limited sensor field of view, which can be overcome by either adding more sensors or multi-tasking a single sensor over multiple areas at a reduced frame rate. As an essential tool for sensor design and mission development, a trade study was conducted to measure the potential advantages of adding spectral bands of information in a single sensor with the intention of reducing sensor frame rates. Thus, traditional motion detection and object association algorithms were modified to evaluate system performance using five spectral bands (visible through thermal IR), while adjusting frame rate as a second variable. The goal of this research was to produce an evaluation of system performance as a function of the number of bands and frame rate. As such, performance surfaces were generated to assess relative performance as a function of the number of bands and frame rate.
Library of Congress Subject Headings
Electronic surveillance--Data processing; Signal processing--Digital techniques; Multisensor data fusion; Motion perception (Vision)--Computer simulation; Optical data processing; Optical pattern recognition
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Adams, Andrew J., "Multispectral persistent surveillance" (2008). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus