Traffic Video Anomaly Detection for Static Vehicle Start-Time Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting abnormal traffic behavior in complex and unknown road scenarios require significant computational resources and time, making them inefficient for accurately estimating the start time of abnormal events such as static vehicles.
Innovation Solution
A method and system utilizing background modeling with MOG2, YOLOv3 target detection, and a network model combining twin cross-correlation with P3D-Attention for anomaly time estimation, which reduces computational costs and improves detection speed by retaining abnormal static vehicles in the background and using perspective view cropping for efficient anomaly detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If tracking-based methods are used for abnormal behavior detection, then measurement precision is improved, but loss of time increases significantly
Solution Approach 1:
The patent extracts and removes normal moving vehicles from the video frames through background modeling, retaining only the abnormal static vehicles in the background. This extraction approach eliminates the need for time-consuming tracking of normal vehicles while preserving the ability to detect abnormal static targets, thus reducing detection time while maintaining precision.
Solution Approach 2:
The patent performs background modeling and vehicle removal in advance, creating a pre-processed background that contains only potential abnormal static vehicles. This preliminary action eliminates the need for real-time tracking during anomaly detection, significantly reducing the detection time from 342 ms to 12.2 ms while maintaining high precision.
2Measurement precision
If complex tracking algorithms are used for anomaly detection, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts abnormal static vehicles directly from the background through background modeling, bypassing the need for complex multi-target tracking algorithms. This extraction-based approach simplifies the system architecture while maintaining detection precision by focusing computational resources only on potential abnormal targets.
Solution Approach 2:
Instead of tracking all vehicles and identifying anomalies through track analysis, the patent inverts the approach by removing normal vehicles and retaining only abnormal static vehicles in the background. This inversion simplifies the algorithm by eliminating the need for complex tracking and track estimation while maintaining anomaly detection capability.
3Measurement precision
If comprehensive vehicle tracking is performed for all targets, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent extracts only the essential information (abnormal static vehicles) from the video stream through background modeling, eliminating the need for comprehensive tracking of all vehicles. This selective extraction maintains precision for detecting abnormal targets while significantly improving detection speed and overall system productivity.
Solution Approach 2:
The patent performs background modeling and vehicle removal as a preliminary step, creating a simplified background representation that contains only potential anomalies. This preliminary processing enables faster real-time detection by eliminating the need for continuous tracking of normal vehicles, thus improving productivity while maintaining detection precision.
Data Source
AI summary
The present disclosure discloses a method and a system for detecting an abnormal traffic behavior. The method of the present disclosure includes: retaining an abnormal static target vehicle in a traffic surveillance video in a background through background modeling; performing abnormal target detection, and obtaining a cropped picture of an abnormal target vehicle and a cropped video clip through cropping; performing anomaly start time estimation, inputting the cropped picture and the cropped video clip to a network model combining twin cross-correlation with pseudo three-dimensional (P3D)-Attention, labeling a classification label on the cropped video clip, and determining a video frame when abnormal behavior occurs; and determining whether a to-be-matched vehicle is an abnormal target vehicle, and determining a start time and an end time of abnormal traffic behavior with reference to the video frame that is obtained when the abnormal behavior occurs.