Abstract:Traffic anomaly detection (TAD) in driving videos is critical for ensuring the safety of autonomous driving and advanced driver assistance systems. Previous single-stage TAD methods primarily rely on frame prediction, making them vulnerable to interference from dynamic backgrounds induced by the rapid movement of the dashboard camera. While two-stage TAD methods appear to be a natural solution to mitigate such interference by pre-extracting background-independent features (such as bounding boxes and optical flow) using perceptual algorithms, they are susceptible to the performance of first-stage perceptual algorithms and may result in error propagation. In this paper, we introduce TTHF, a novel single-stage method aligning video clips with text prompts, offering a new perspective on traffic anomaly detection. Unlike previous approaches, the supervised signal of our method is derived from languages rather than orthogonal one-hot vectors, providing a more comprehensive representation. Further, concerning visual representation, we propose to model the high frequency of driving videos in the temporal domain. This modeling captures the dynamic changes of driving scenes, enhances the perception of driving behavior, and significantly improves the detection of traffic anomalies. In addition, to better perceive various types of traffic anomalies, we carefully design an attentive anomaly focusing mechanism that visually and linguistically guides the model to adaptively focus on the visual context of interest, thereby facilitating the detection of traffic anomalies. It is shown that our proposed TTHF achieves promising performance, outperforming state-of-the-art competitors by +5.4% AUC on the DoTA dataset and achieving high generalization on the DADA dataset.

A Feature-Based Method for Traffic Anomaly Detection

Anomaly Candidate Identification and Starting Time Estimation of Vehicles from Traffic Videos.

Real-Time Detection of Anomalous Taxi Trajectories from GPS Traces.

Detection of Anomalous Traffic Patterns and Insight Analysis from Bus Trajectory Data

On Detection of Emerging Anomalous Traffic Patterns Using GPS Data

Deep learning detection of anomalous patterns from bus trajectories for traffic insight analysis

A Fused Method of Machine Learning and Dynamic Time Warping for Road Anomalies Detection

Text-Driven Traffic Anomaly Detection with Temporal High-Frequency Modeling in Driving Videos

Real Time Anomalous Trajectory Detection and Analysis

Traffic Anomaly Detection Based on Spatio-Temporal Hypergraph Convolution Neural Networks

Global Spatio-Temporal Fusion-based Traffic Prediction Algorithm with Anomaly Aware

Quantifying Urban Traffic Anomalies

Detecting Traffic Anomalies in Urban Areas Using Taxi GPS Data

A Framework of Abnormal Behavior Detection and Classification Based on Big Trajectory Data for Mobile Networks

Traffic System Anomaly Detection Using Spatiotemporal Pattern Networks

A Three-Stage Anomaly Detection Framework for Traffic Videos

Good Practices and A Strong Baseline for Traffic Anomaly Detection

Road traffic anomaly monitoring and warning based on DeepWalk algorithm

Traffic incident detection based on a global trajectory spatiotemporal map

A Novel Method for Road Anomaly Objects Detection in the Traffic Environment With Multi-Mechanism Fusion

A Time-Frequency Detecting Method for Network Traffic Anomalies