Anomalous Trajectory Detection Based on Multi-Scale Spatiotemporal Dependency Modeling

View at publisher Summarise with VIRAT
Authors: Long Liu, Dawen Xia, Han Zhang, Wenlin He, Yang Hu, Huaqing Li
Cited by: 0
References: 20

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

With the rapid growth of smart devices and positioning technologies, mobile trajectory big data mining and analytics have become essential for understanding urban mobility patterns. Detecting abnormal travel routes is crucial for preventing passenger fraud and improving traffic management. However, most detection methods often face challenges in identifying time-dependent anomalies and lacking the capability of processing scale-dependent anomalies. To this end, we propose a Multi-Scale Spatiotemporal Dependency-based Anomalous Trajectory Detection model (MSSTD-ATD). Our model designs a multi-layer gated temporal convolutional network to capture dynamic temporal dependencies within trajectories, improving the identification of time-sensitive anomalies. Moreover, we construct a Multi-Scale Transformer Encoder (MSTE) module to extract trajectory features at various scales and introduce a learnable weight fusion mechanism that combines both local details and global patterns. Extensive experimental results on two real-world trajectory datasets demonstrate that MSSTD-ATD outperforms comparable methods in effectiveness and robustness across various anomalous scenarios.

Fields of Study & Subjects

Data Management and Algorithms Anomaly Detection Techniques and Applications Traffic Prediction and Management Techniques

Publication details

Journal / source
International Conference on Communication Networks and Machine Learning (CNML)
Pages
12–19
Publication date
30 January 2026
Language
EN