{"id":{"repo_id":"nps","oai_identifier":"oai:calhoun.nps.edu:10945/74687"},"canonical_url":"https://search.dev.ndltd.org/etd/nps/oai:calhoun.nps.edu:10945/74687","repository":{"repo_id":"nps","name":"Naval Postgraduate School","base_url":"https://calhoun.nps.edu/server/oai/request"},"display":{"title":"PASSIVE RADAR TRACK CLUSTERING: HIGHER FIDELITY OF TARGET IDENTIFICATION AND CLASSIFICATION OF UNLABELED TRACKS","abstract":"Effective radar signal classification is critical for naval electronic warfare systems like the SLQ-32, but data scarcity limits machine learning applications. This research evaluates semi-supervised learning (SSL) techniques to leverage unlabeled data for improved classification. Using a NIST dataset of simulated radar waveforms, we evaluated unsupervised learning (USL) and supervised learning (SL) approaches across three SLQ-32 preprocessed datasets: Raw Magnitude, Full Spectrogram, and Max Hold Spectrogram. We implemented SSL pipelines using K–Means and Gaussian Mixture Models (GMMs) to generate pseudo-labels from limited labeled data for classifier training. The Max Hold Spectrogram performs the best for both SL and USL approaches. GMMs outperform K–Means for all unsupervised clustering implementations. The SSL models are compared against the baseline results obtained by training each SSL’s corresponding SL classifier, using the same limited set of labeled data. We found that the SSL architecture improves accuracy from 72% to 79% compared to the Random Forests baseline, and from 78% to 79% compared to the XGBoost baseline. Additionally, SSL architectures that used GMMs in the SSL architecture outperformed their baseline SL counterpart. This research shows SSL's potential for radar signal classification in data-constrained environments, offering a viable solution for naval electronic warfare systems when obtaining large, labeled datasets is operationally challenging.","abstract_html":"Effective radar signal classification is critical for naval electronic warfare systems like the SLQ-32, but data scarcity limits machine learning applications. This research evaluates semi-supervised learning (SSL) techniques to leverage unlabeled data for improved classification. Using a NIST dataset of simulated radar waveforms, we evaluated unsupervised learning (USL) and supervised learning (SL) approaches across three SLQ-32 preprocessed datasets: Raw Magnitude, Full Spectrogram, and Max Hold Spectrogram. We implemented SSL pipelines using K–Means and Gaussian Mixture Models (GMMs) to generate pseudo-labels from limited labeled data for classifier training. The Max Hold Spectrogram performs the best for both SL and USL approaches. GMMs outperform K–Means for all unsupervised clustering implementations. The SSL models are compared against the baseline results obtained by training each SSL’s corresponding SL classifier, using the same limited set of labeled data. We found that the SSL architecture improves accuracy from 72% to 79% compared to the Random Forests baseline, and from 78% to 79% compared to the XGBoost baseline. Additionally, SSL architectures that used GMMs in the SSL architecture outperformed their baseline SL counterpart. This research shows SSL's potential for radar signal classification in data-constrained environments, offering a viable solution for naval electronic warfare systems when obtaining large, labeled datasets is operationally challenging.","abstract_has_math":false,"creators":["DelValle, Antolin D."],"institution":"Monterey, CA; Naval Postgraduate School","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":"Computer Science (CS)","school":null,"contributors":[],"advisors":["Barton, Armon C."],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-09","date_published":"2025-09","updated_at":"2026-07-27T20:26:50Z","subjects":[],"languages":[],"rights":["This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States."],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["NPS-25-N117-_"],"render_values":[{"text":"NPS-25-N117-_","href":null,"code":true}]}]},"links":{"outbound_url":"https://hdl.handle.net/10945/74687","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Barton, Armon C."]},{"key":"dc:contributor.department","label":"Department","values":["Computer Science (CS)"]},{"key":"dc:creator","label":"Author","values":["DelValle, Antolin D."]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-03-05T17:48:02Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2026-03-05T17:48:02Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-09"]},{"key":"dc:publisher","label":"Institution","values":["Monterey, CA; Naval Postgraduate School"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["NPS-25-N117-_"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/10945/74687"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Effective radar signal classification is critical for naval electronic warfare systems like the SLQ-32, but data scarcity limits machine learning applications. This research evaluates semi-supervised learning (SSL) techniques to leverage unlabeled data for improved classification. Using a NIST dataset of simulated radar waveforms, we evaluated unsupervised learning (USL) and supervised learning (SL) approaches across three SLQ-32 preprocessed datasets: Raw Magnitude, Full Spectrogram, and Max Hold Spectrogram. We implemented SSL pipelines using K–Means and Gaussian Mixture Models (GMMs) to generate pseudo-labels from limited labeled data for classifier training. The Max Hold Spectrogram performs the best for both SL and USL approaches. GMMs outperform K–Means for all unsupervised clustering implementations. The SSL models are compared against the baseline results obtained by training each SSL’s corresponding SL classifier, using the same limited set of labeled data. We found that the SSL architecture improves accuracy from 72% to 79% compared to the Random Forests baseline, and from 78% to 79% compared to the XGBoost baseline. Additionally, SSL architectures that used GMMs in the SSL architecture outperformed their baseline SL counterpart. This research shows SSL's potential for radar signal classification in data-constrained environments, offering a viable solution for naval electronic warfare systems when obtaining large, labeled datasets is operationally challenging."]},{"key":"dc:title","label":"Title","values":["PASSIVE RADAR TRACK CLUSTERING: HIGHER FIDELITY OF TARGET IDENTIFICATION AND CLASSIFICATION OF UNLABELED TRACKS"]}]}],"canonical_facts":{"dc:contributor.advisor":["Barton, Armon C."],"dc:contributor.department":["Computer Science (CS)"],"dc:creator":["DelValle, Antolin D."],"dc:date.accessioned":["2026-03-05T17:48:02Z"],"dc:date.available":["2026-03-05T17:48:02Z"],"dc:date.issued":["2025-09"],"dc:description.abstract":["Effective radar signal classification is critical for naval electronic warfare systems like the SLQ-32, but data scarcity limits machine learning applications. This research evaluates semi-supervised learning (SSL) techniques to leverage unlabeled data for improved classification. Using a NIST dataset of simulated radar waveforms, we evaluated unsupervised learning (USL) and supervised learning (SL) approaches across three SLQ-32 preprocessed datasets: Raw Magnitude, Full Spectrogram, and Max Hold Spectrogram. We implemented SSL pipelines using K–Means and Gaussian Mixture Models (GMMs) to generate pseudo-labels from limited labeled data for classifier training. The Max Hold Spectrogram performs the best for both SL and USL approaches. GMMs outperform K–Means for all unsupervised clustering implementations. The SSL models are compared against the baseline results obtained by training each SSL’s corresponding SL classifier, using the same limited set of labeled data. We found that the SSL architecture improves accuracy from 72% to 79% compared to the Random Forests baseline, and from 78% to 79% compared to the XGBoost baseline. Additionally, SSL architectures that used GMMs in the SSL architecture outperformed their baseline SL counterpart. This research shows SSL's potential for radar signal classification in data-constrained environments, offering a viable solution for naval electronic warfare systems when obtaining large, labeled datasets is operationally challenging."],"dc:identifier.other":["NPS-25-N117-_"],"dc:identifier.uri":["https://hdl.handle.net/10945/74687"],"dc:publisher":["Monterey, CA; Naval Postgraduate School"],"dc:rights":["This publication is a work of the U.S. Government as defined in Title 17, United States Code, Section 101. Copyright protection is not available for this work in the United States."],"dc:title":["PASSIVE RADAR TRACK CLUSTERING: HIGHER FIDELITY OF TARGET IDENTIFICATION AND CLASSIFICATION OF UNLABELED TRACKS"],"dc:type":["Thesis"]},"updated_at":"2026-07-27T20:26:50Z"}