Repository logo

Detection of a weak radiological source in ambient background using spectral analysis




Meengs, Matthew Richard, author
Brandl, Alex, advisor
Johnson, Thomas, advisor
Kokoszka, Piotr, committee member

Journal Title

Journal ISSN

Volume Title


The detection of radiation requires the use of statistical tools due to the probabilistic nature of the emission and interaction properties of radiation, an analysis that includes the testing of a hypothesis regarding the presence or absence of a source against background. Traditionally, a false positive rate of 5% is used to calculate a y*, the decision threshold, above which a source is determined to be present. However, in radiological conditions where a source is both improbable and weak, and where counting time is limited, detection of a source becomes increasingly challenging using this traditional method. The detection of clandestine fissile materials presents such a challenge, and with the increasing risk of nuclear proliferation, there exists a growing desire to research more optimal methods in detecting these sources, especially where a missed detection is of such high consequence. Previous research has shown that using a string of measurements, and calculating a detection limit based on a certain number of false positives within that string, consistently outperforms the traditional method of basing the detection limit on just one measurement. Such research to date has only been applied to counts of all energies (gross counts). The purpose of this research is to apply the success of this new detection algorithm to certain energies within the spectrum, and to discover whether further optimization is possible by this process. Optimization was evaluated using receiver operator characteristic (ROC) curves, where special emphasis was placed at the lower false positive values. Over the course of this research, two hypothesis were tested. The first hypothesis conjectures that it is indeed possible to further optimize source detection when using an energy bin other than gross counts. The second hypothesis postulates that if the first hypothesis is true, than there exists a mathematical criterion that predicts this behavior. Both hypothesis were verified to be correct.


2018 Summer.
Includes bibliographical references.

Rights Access



Associated Publications