hh.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
An Experimental Study on ObjectTracking
Halmstad University, School of Information Technology.
2025 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
Abstract [en]

This thesis investigated the robustness of 3D object-tracking algorithms undersnowy weather conditions, focusing particularly on snowy scenarios affecting au-tonomous vehicle perception systems. The principal objective was to evaluateand compare the performance of four tracking methods: Kalman Filter, ExtendedKalman Filter, Particle Filter, and ByteTrack. Each method was assessed using Li-DAR data obtained from the Canadian Adverse Driving Conditions (CADC) dataset,representing harsh winter conditions, and the nuScenes dataset, representing clear,optimal weather conditions.

The methodology involved processing sequential frames of LiDAR data, detected3D bounding boxes, and tracking objects through association and state estima-tion. Standard metrics such as HOTA, IDF1, AMOTA, and AMOTP were used tomeasure tracking accuracy and consistency across both datasets. Results indicatedsignificant performance degradation for all algorithms under snowy weather con-ditions compared to clear weather. The Kalman methods suffered from the linearbehaviorand the noise, while the Particle Filter provided a more robust estimationdue to its ability to cope with the uncertainty via its multiple hypotheses.

Deep learning-based solution ByteTrack demonstrated better performance, withbetter accuracy and fewer identity switches inchallenging scenarios. It wasfoundthat deep learning based tracking can provide more solid guarantee on point tra-jectoryThe study concluded that deep learning-based tracking methods offer enhancedreliability for autonomous vehicles in challenging environments.

Abstract [sv]

Denna avhandling undersökte robustheten hos 3D-objektspårningsalgoritmer un-der snöiga väderförhållanden, med särskilt fokus på snöiga scenarier som påverkarperceptionssystem i autonoma fordon. Det huvudsakliga målet var att utvärderaoch jämföra prestandan hos fyra spårningsmetoder: Kalmanfilter, Utökat Kalman-filter, Partikelfilter och ByteTrack. Varje metod bedömdes med hjälp av LiDAR-data från CADC-datamängden, som representerar hårda vinterförhållanden, samtnuScenes-datamängden, som representerar klart och optimalt väder.Metodiken innebar bearbetning av sekventiella LiDAR-ramar, detekterade av 3D-bounding boxes och spårning av objekt genom association och tillståndsestimer-ing. Standardmått som HOTA, IDF1, AMOTA och AMOTP användes för att mätaspårningsnoggrannhet och konsistens i båda datamängderna.Resultaten visade en tydlig försämring i prestanda för alla algoritmer under snöigaväderförhållanden jämfört med klart väder. Kalmanbaserade metoder hade svårtatt hantera detta på grund av sina linjära antaganden och känslighet för brus,medan Partikelfiltret visade ökad robusthet genom att hantera osäkerhet effek-tivt via flera tillståndshypoteser. ByteTrack, som bygger på djupinlärning, prester-ade konsekvent bäst, med högre noggrannhet och färre identitetsfel under svåraförhållanden.Studien drog slutsatsen att djupinlärningsbaserade spårningsmetoder erbjuderhögre tillförlitlighet för autonoma fordon i utmanande miljöeer.

Place, publisher, year, edition, pages
2025. , p. 35
Keywords [en]
Autonomous Vehicles, LiDAR, 3D Object Tracking, snowy Weather Conditions, Kalman Filter (KF), Extended Kalman Filter (EKF), Particle Filter (PF), ByteTrack, LSTM (Long Short-Term Memory), Canadian Adverse Driv- ing Conditions (CADC) Dataset, Multi-Object Tracking (MOT), Tracking Algo- rithms, Machine Learning, Bird’s-Eye-View (BEV)
Keywords [sv]
Autonomous Vehicles, LiDAR, 3D Object Tracking, Adverse Weather Conditions, Kalman Filter (KF), Extended Kalman Filter (EKF), Particle Filter (PF), ByteTrack, LSTM (Long Short-Term Memory), Canadian Adverse Driv- ing Conditions (CADC) Dataset, Multi-Object Tracking (MOT), Tracking Algo- rithms, Machine Learning, Bird’s-Eye-View (BEV)
National Category
Engineering and Technology Artificial Intelligence
Identifiers
URN: urn:nbn:se:hh:diva-56272OAI: oai:DiVA.org:hh-56272DiVA, id: diva2:1966144
Presentation
2025-05-23, D415, Högskolan i Halmstad, Halmstad, 13:00 (English)
Supervisors
Examiners
Available from: 2025-06-09 Created: 2025-06-09 Last updated: 2025-10-01Bibliographically approved

Open Access in DiVA

fulltext(453 kB)81 downloads
File information
File name FULLTEXT02.pdfFile size 453 kBChecksum SHA-512
8aa87d8690c76ebeadab9814d3ba9f005ebc07b234c35dd96357d0f7eb7980505c7d71a2dbcf62369cf6a017fda41173e53147b4aef152f3f162796f12fec97f
Type fulltextMimetype application/pdf

By organisation
School of Information Technology
Engineering and TechnologyArtificial Intelligence

Search outside of DiVA

GoogleGoogle Scholar
Total: 81 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 291 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf