Probabilistic Representation of Vehicular Trajectories as Input for Artificial Neural Networks

• Bianca Lahm Gomes Itaipu Technological Park Foundation (Itaipu Parquetec)
• Kame Haung Zhu Itaipu Technological Park Foundation (Itaipu Parquetec)

DOI: https://doi.org/10.5753/sbbd.2025.247718

Abstract

This work proposes a probabilistic approach to represent vehicular trajectories using data from license plate recognition (LPR) cameras, aiming at real-time processing. It is based on the Hierarchical Pattern Bayes (HPB) model to generate matrices of trajectory typicality and temporal density. These matrices were evaluated using the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm, resulting in a 75% increase in the Silhouette score and a 31.1% reduction in the Davies-Bouldin index compared to the use of raw coordinates. These results indicate a more structured representation, enabling the application of supervised models for real-time anomaly detection.

Keywords: Unsupervised learning, Vehicular trajectories, Probabilistic representation, Anomaly detection, Hierarchical Pattern Bayes (HPB), Real-time inference

References

Campello, R. J. G. B., Moulavi, D., and Sander, J. (2013). Density-based clustering based on hierarchical density estimates.

Cruz, A. B., Ferreira, J., Carvalho, D., Mendes, E., Pacitti, E., Coutinho, R., Porto, F., and Ogasawara, E. (2018). Detecção de anomalias frequentes no transporte rodoviário urbano. In Anais do XXXIII Simpósio Brasileiro de Banco de Dados, pages 271–276, Porto Alegre, RS, Brasil. SBC.

Filho, J. J. and Wainer, J. (2008). Hpb: A model for handling bn nodes with high cardinality parents. Journal of Machine Learning Research, 9(70):2141–2170.

Ghahramani, Z. (1998). Learning dynamic Bayesian networks, pages 168–197. Springer Berlin Heidelberg, Berlin, Heidelberg.

Liu, F. T., Ting, K. M., and Zhou, Z.-H. (2008). Isolation forest. In 2008 Eighth IEEE International Conference on Data Mining, pages 413–422.

Mao, Y., Shi, Y., and Lu, B. (2024). Detecting urban traffic anomalies using traffic-monitoring data. ISPRS International Journal of Geo-Information, 13(10).

Peralta, B., Soria, R., Nicolis, O., Ruggeri, F., Caro, L., and Bronfman, A. (2023). Outlier vehicle trajectory detection using deep autoencoders in santiago, chile. Sensors, 23(3).

Sun, L., Chen, X., He, Z., and Miranda-Moreno, L. (2020). Routine pattern discovery and anomaly detection in individual travel behavior. CoRR, abs/2004.03481.