Sensor-Based Vehicle Classification Using Machine Learning

Aug 3, 2026

Accurate vehicle classification is vital for civilian traffic management and national defense security. Traditional approaches often rely on frequency-domain signal processing to process sensor data for vehicle identification. This thesis evaluates machine learning and deep learning methods by training classifiers on seismic, acoustic, and magnetic measurements from Heavy and Light vehicles. Models using frequency-domain data and neural networks using time-series signals both achieve over 90% balanced accuracy. Average performance favors frequency-domain features, while results suggest further investigation into replacing traditional signal processing with deep learning.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages175

Reading Time7 h 21 min

AccessFree to download

Introduction

Research Objectives

Research Questions

Research Goals

Assumptions and Limitations

Document Overview

Background and Literature Review

Signals Processing

Transformations

Machine Learning Tasks

Related Works

Methodology

Data

Air Force Institute of Technology AFIT Scholar | Theses and Dissertations | Student Graduate Works | | --- | --- | | 3-2024
Sensor-Based Vehicle Classification Using Luke McFadden
Follow this and additional works at: [https://scholar.afit.edu/etd](https://scholar.afit.edu/etd)
Part of the Computer Sciences Commons | Machine Learning | Recommended Citation McFadden, Luke, "Sensor-Based Vehicle Classification Using Machine Learning" (2024) . Theses and Dissertations. 7783. [https://scholar.afit.edu/etd/7783](https://scholar.afit.edu/etd/7783) This Thesis is brought to you for free and open access by the Student Graduate Works at AFIT Scholar. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator of AFIT Scholar. For more information, [please contact AFIT.ENWL.Repository@us.af.mil](please contact AFIT.ENWL.Repository@us.af.mil). Sensor-Based Vehicle Classification Using Machine Learning THESIS Luke McFadden, Captain, USAF AFIT-ENG-MS-24-M-184 DEPARTMENT OF THE AIR FORCE AIR UNIVERSITY AIR FORCE INSTITUTE OF TECHNOLOGY Wright-Patterson Air Force Base, Ohio DISTRIBUTION STATEMENT A APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. The views expressed in this document are those of the author and do not reflect the official policy or position of the United States Air Force, the United States Department of Defense or the United States Government. This material is declared a work of the U.S. Government and is not subject to copyright protection in the United States. AFIT-ENG-MS-24-M-184 SENSOR-BASED VEHICLE CLASSIFICATION USING MACHINE LEARNING THESIS Presented to the Faculty Department of Electrical and Computer Engineering Graduate School of Engineering and Management Air Force Institute of Technology Air University Air Education and Training Command in Partial Fulfillment of the Requirements for the Degree of Master of Science in Computer Science Luke McFadden, B.S.C.E. Captain, USAF March 21, 2024 DISTRIBUTION STATEMENT A APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNLIMITED. AFIT-ENG-MS-24-M-184 SENSOR-BASED VEHICLE CLASSIFICATION USING MACHINE LEARNING THESIS Luke McFadden, B.S.C.E. Captain, USAF Committee Membership: Brett Borghetti, Ph.D Chair Anthony Franz, Ph.D Member Lt Col Christopher Rondeau, Ph.D Member AFIT-ENG-MS-24-M-184 Abstract Accurate vehicle classification is vital in fields ranging from civilian traffic management systems to national defense security measures. Many organizations have relied upon traditional frequency domain signal processing methods to perform sensor-based vehicle classification tasks. Recent advances in machine learning and deep learning have led many organizations to reassess their currently fielded systems of sensors. By integrating machine learning models and neural networks into currently fielded sensor systems, these organizations hope to improve the accuracy of their vehicle classifiers. Seismic, acoustic, and magnetic data were collected on ’Heavy’ and ’Light’ vehicle types. Machine learning models were trained to perform binary classification using frequency domain data, while neural networks were trained using time series data. The results found that both methods were highly accurate at distinguishing between the two vehicle types, with greater than 90% balanced accuracy being reported. While the frequency domain data yielded higher balanced accuracy on average, the classification ability of the neural networks using time series sensor signals demonstrated by this research calls for additional research into the ability of deep learning to replace traditional frequency domain signal processing. Table of Contents Page Abstract ............................................................... iv List of Figures ......................................................... vii List of Tables .....................................................

FAQs

What sensor data and vehicle types are used in the study?

The thesis uses seismic, acoustic, and magnetic data collected from Heavy and Light vehicle types.

How are the machine learning models trained for classification?

Binary classification is performed using frequency-domain data for traditional machine learning models, while neural networks are trained using time-series sensor signals.

What accuracy results are reported and how do the two approaches compare?

Both methods report greater than 90% balanced accuracy for distinguishing vehicle types. Frequency-domain features show higher average balanced accuracy, while time-series neural networks motivate additional research into deep learning replacing traditional signal processing.

Aug 3, 2026

Uploaded by Evangeline