A RESEARCH APPROACH TO THE DETECTION AND TRACKING OF MOVING OBJECTS IN VIDEO SEQUENCES
DOI:
https://doi.org/10.62405/osiconf.2026.02Keywords:
machine learning; video sequence; object detection; object tracking; trajectory; computer vision.Abstract
The aim of the study is to investigate the influence of moving object detection parameters on the accuracy and stability of their subsequent tracking in video sequences. An experimental approach based on sequential machine-learning detection, inter-frame association, and trajectory formation with variation of the detection confidence threshold is proposed. The evaluation is performed using detection precision and recall, average detection accuracy, tracking quality and stability, the number of identity changes, and video processing speed. The results demonstrate that an excessively low confidence threshold increases false detections and incorrect associations, whereas an excessively high threshold causes the loss of temporarily occluded and low-contrast objects. The most stable trajectories are obtained at intermediate parameter values. The findings confirm the need for joint adjustment of detection and tracking parameters while considering the trade-off between detection accuracy, identity stability, and computational efficiency.
References
Gragnaniello, D., Greco, A., Saggese, A., Vento, M., & Vicinanza, A. (2023). Benchmarking 2D multi-object detection and tracking algorithms in autonomous vehicle driving scenarios. Sensors, 23(8), 4024. https://doi.org/10.3390/s23084024
Medaglini, A., Peccerillo, B., & Bartolini, S. (2025). High-performance multi-object tracking for autonomous driving in urban scenarios with heterogeneous embedded boards. IEEE Access, 13, 8649–8663. https://doi.org/10.1109/ACCESS.2024.3484129
Stadler, D., & Beyerer, J. (2023). An improved association pipeline for multi-person tracking. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) (pp. 3170–3179). IEEE. https://doi.org/10.1109/CVPRW59228.2023.00319
Stanojevic, V. D. & Todorovic, B. T. (2024). BoostTrack: Boosting the similarity measure and detection confidence for improved multiple object tracking. Machine Vision and Applications, 35, Article 53. https://doi.org/10.1007/s00138-024-01531-5
