"Shape activity": a continuous-state HMM for moving/deforming shapes with application to abnormal activity detection.

Area of Science:

• Computer Vision
• Machine Learning
• Data Science

Background:

• Traditional group activity modeling methods like co-occurrence statistics and dynamic Bayesian networks struggle with large numbers of interacting objects.
• There is a need for robust methods to detect abnormal activities in complex, multi-object scenarios.

Purpose of the Study:

• To develop a novel approach for modeling group activity using shape dynamics of point objects (landmarks).
• To utilize these models for effective abnormal activity detection in scenarios with many moving objects.

Main Methods:

• Objects are treated as point landmarks, and their configurations are modeled as moving and deforming shapes using Kendall's shape theory.
• A continuous-state hidden Markov model is employed to capture landmark shape dynamics.
• Abnormal activity is defined as a deviation from the learned shape activity model.

Main Results:

• The proposed shape dynamics model successfully captures the configuration of multiple moving objects.
• The method demonstrates effectiveness in detecting abnormal activities in real-world multi-object scenarios.
• The approach is applicable to scenarios where previous methods fail due to a large number of objects.

Conclusions:

• Modeling group activity as landmark shape dynamics provides a powerful framework for abnormal activity detection.
• This method offers a scalable and effective solution for complex multi-object interactions.
• The approach advances the field of abnormal activity recognition in dynamic environments.