Deep motion-aware representation learning for vehicle trajectory analysis
*B. Swarna JyothiCorresponding authorbadimela1508@gmail.comDepartment of Computer Science and EngineeringDayananda Sagar UniversityBangalore, Karnataka, 562112, IndiaView full profile → , S. K. Mouleeswaranmeetmoulee@gmail.comDepartment of Computer Science and EngineeringDayananda Sagar UniversityBangalore, Karnataka, 562112, IndiaView full profile → , M. Suleman Bashasuleman.ndl@gmail.comDepartment of Computer Science & Engineering (Data Science)Rajeev Gandhi Memorial College of Engineering & Technology (Autonomous)Nandyal, Andhra Pradesh, 518501, IndiaView full profile →
* Corresponding author · click or hover a name for details
- Received:
- 01 Mar 2026
- Published Online:
- 31 Oct 2026
- Article type:
- Research Article
- Language:
- EN
- Article no.:
- JIOS-2482
- Pages:
- 3545–3554
Abstract
Identifying vehicle trajectories from video surveillance is an essential task in modern traffic management systems to prevent accidents and reduce congestion. Detecting cluster structures in trajectory paths is crucial for distinguishing between normal and abnormal vehicle behaviours. Traditional visual clustering approaches, such as Visual Assessment of cluster Tendency (VAT) and improved VAT (iVAT), are effective for visualizing cluster patterns but are limited in capturing complex motion characteristics due to inadequate feature representation. To address this limitation, this paper proposes a novel Deep Motion-Aware Trajectory iVAT(DMAT-iVAT) framework that integrates motion-enhanced trajectory modelling with deep learning-based feature extraction. The proposed method augments trajectory data with velocity and acceleration information to capture rich spatio-temporal dynamics, followed by representation learning using a Bidirectional Long Short-Term Memory (BiLSTM) network with an attention mechanism. The learned embeddings are then analyzed using iVAT to reveal cluster tendencies more effectively. Experimental results on benchmark video trajectory datasets demonstrate that the proposed DMAT-iVAT framework significantly improves cluster structure visualization and discrimination compared to conventional VAT-based methods.
Keywords
Subject Classifications
References
[1] J. C. Bezdek and R. J. Hathaway, “VAT: A Tool for Visual Assessment of (Cluster) Tendency,” in Proc. 2002 Int. Joint Conf. Neural Netw. (IJCNN’02), vol. 3, Honolulu, HI, USA, pp. 2225–2230 (2002), doi: 10.1109/IJCNN.2002.1007487.
[2] P. Rathore, D. Kumar, J. C. Bezdek, S. Rajasegarar, and M. Palaniswami, “A Rapid Hybrid Clustering Algorithm for Large Volumes of High Dimensional Data,” IEEE Trans. Knowl. Data Eng., vol. 31, no. 4, pp. 641–654 (2019).
[3] T. C. Havens and J. C. Bezdek, “An Efficient Formulation of the Improved Visual Assessment of Cluster Tendency (iVAT) Algorithm,” IEEE Trans. Knowl. Data Eng., vol. 24, no. 5, pp. 813–822 (2012), doi: 10.1109/TKDE.2011.33.
[4] M. S. Basha, S. K. Mouleeswaran, and K. R. Prasad, “Detection of Pre-Cluster Nano-Tendency Through Multi-Viewpoints Cosine-Based Similarity Approach,” Nanotechnol. Environ. Eng., vol. 7, no. 1, pp. 259–268 (2022).
[5] Z. Xie, X. Luo, D. Zhao, and X. Chen, “Fuzzy C-Means Clustering Algorithm Based on Superpixel Merging and Multi-Feature Adaptive Fusion Measurement,” IET Image Process., vol. 18, no. 1, pp. 140–153 (2024), doi: 10.1049/ipr2.12939.
[6] B. Zhang, Y. Zhu, Y. Cao, S. Rajasegarar, G. Li, and G. Liu, “Kernel-Based iVAT with Adaptive Cluster Extraction,” Knowl. Inf. Syst., vol. 66, pp. 7057–7076 (2024), doi: 10.1007/s10115-024-02189-1.
[7] D. Kumar, H. Wu, S. Rajasegarar, C. Leckie, S. Krishnaswamy, and M. Palaniswami, “Fast and Scalable Big Data Trajectory Clustering for Understanding Urban Mobility,” IEEE Trans. Intell. Transp. Syst., vol. 19, no. 11, pp. 3709–3722 (2018), doi: 10.1109/TITS.2018.2854775.
[8] P. Gloaguen, L. Chapel, C. Friguet, and R. Tavenard, “Scalable Clustering of Segmented Trajectories Within a Continuous Time Framework: Application to Maritime Traffic Data,” Mach. Learn., vol. 112, no. 6, pp. 1975–2001 (2023).
[9] L. D. Nukala, T. A. S. Aruna Sri, and K. R. Prasad, “Improvised Deep Learning Techniques with the Spectral Concept for the Robust Skin Disease Image Classification,” J. Inf. Optim. Sci., vol. 47, no. 5-B, pp. 2095–2107 (2026).
[10] S. Datta, C. Karmakar, P. Rathore, and M. Palaniswami, “Scalable Cluster Tendency Assessment for Streaming Activity Data Using Recurring Shapelets,” in Proc. 44th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. (EMBC), Glasgow, U.K., pp. 1036–1040 (2022), doi: 10.1109/EMBC48229.2022.9871534.
[11] M. A. Mahdi, K. M. Hosny, and I. Elhenawy, “Scalable Clustering Algorithms for Big Data: A Review,” IEEE Access, vol. 9, pp. 80015–80027 (2021), doi: 10.1109/ACCESS.2021.3084057.
[12] G. Shkaberina, N. Rezova, E. Tovbis, and L. Kazakovtsev, “Visual Assessment of Cluster Tendency with Variations of Distance Measures,” Algorithms, vol. 16, no. 1, Art. no. 5 (2023), doi: 10.3390/a16010005.
[13] C. Nordahl, V. Boeva, H. Grahn, and M. Persson-Netz, “On Evaluation of Data Stream Clustering Algorithms: A Survey,” IEEE Access, vol. 13, pp. 139524–139546 (2025), doi: 10.1109/ACCESS.2025.3596435.
[14] Y. Hu, “VATdt: Visual Assessment of Cluster Tendency Using Diagonal Tracing,” Am. J. Comput. Math., vol. 2, no. 1, pp. 27–41 (2012), doi: 10.4236/ajcm.2012.21004.
[15] M. S. Basha, K. R. Prasad, S. K. Mouleeswaran, R. C. Poonia, and S. Sebastian, “Multi-Disease Detection System with X-Ray Images Using Deep Learning Techniques,” J. Inf. Optim. Sci., vol. 45, no. 5, pp. 1379–1388 (2024).
[16] Queen Mary University of London, “QMUL Junction Dataset,” [Online]. Available: http://personal.ie.cuhk.edu.hk/~ccloy/downloads_qmul_junction.html.
