Thesis
A hypothesis-testing approach to low-overhead trajectory-based classification of aerial intruders
Washington State University
Master of Science (MS), Washington State University
05/2021
DOI:
https://doi.org/10.7273/000004280
Handle:
https://hdl.handle.net/2376/125168
Abstract
Motivated by security concerns in the growing use of unmanned aerial vehicles (UAVs), a method for classifying aerial intrusions based on intrinsic flight characteristics is presented. In contrast to existing approaches, the presented method utilizes a coarse physical model to aid in the classification of measured data. A hypothesis testing problem is posed wherein aerial intruders are modelled as seeking to regulate their speed on a linear trajectory in the presence of a noisy disturbance (e.g. wind), and a maximum a posteriori (MAP) detector is developed to classify intruders as a function of measured velocity samples. A reduced, computationally efficient form of the detector is determined, which fundamentally relies on two points in the sample autocorrelation of the input, and readily lends itself toward linear classification techniques. Further, a method for computing the a priori probability of error as a function of intruder models is presented, and this probability is shown to approach zero as the number of input samples is increased. Additionally, the probability of error conditioned on an input measurement is found, providing a confidence metric for detection results. The detector and the presented error metrics are confirmed via simulation using synthesized data, and considerations for the detector’s practical implementation are discussed.
Metrics
2 File views/ downloads
37 Record Views
Details
- Title
- A hypothesis-testing approach to low-overhead trajectory-based classification of aerial intruders
- Creators
- David Norman Petrizze III
- Contributors
- Sandip Roy (Advisor) - Washington State University, Electrical Engineering and Computer Science, School of
- Awarding Institution
- Washington State University
- Academic Unit
- Electrical Engineering and Computer Science, School of
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University
- Identifiers
- 99900896412801842
- Language
- English
- Resource Type
- Thesis