Taxonomy of Anomaly Detection Techniques in Crowd Scenes

Area of Science:

• Computer Vision
• Artificial Intelligence
• Surveillance Systems

Background:

• Crowd anomaly detection is crucial for intelligent video surveillance.
• Manual monitoring of crowds is labor-intensive and prone to errors.
• Existing literature offers various approaches to detect abnormal crowd behavior as outlier detection.

Purpose of the Study:

• To provide a comprehensive review of recent anomaly detection methods in crowd analysis.
• To introduce a novel taxonomy for organizing existing research in crowd analysis and anomaly detection.
• To summarize current reviews, datasets, and challenges in the field.

Main Methods:

• Reviewing recent developments in anomaly detection from a computer vision perspective.
• Analyzing various datasets used for crowd anomaly detection.
• Classifying and organizing existing works into a new taxonomic structure.

Main Results:

• A detailed overview of computer vision-based anomaly detection methods for crowds.
• A new taxonomic organization of current research in crowd analysis and anomaly detection.
• A summary of relevant datasets, existing reviews, and identified research challenges.

Conclusions:

• The paper offers a structured overview of crowd anomaly detection research.
• It highlights research trends and future directions in intelligent video surveillance.
• This work serves as a valuable resource for researchers in computer vision and surveillance.