Statistical abnormal crowd behavior detection and simulation for real-time applications

Abstract:

This paper proposes a low computational cost method for abnormal crowd behavior detection with surveillance applications in fixed cameras. Our proposal is based on statistical modelling of moved pixels density. For modelling we take as reference datasets available in the literature focused in crowd behavior. During anomalous events we capture data to replicate abnormal crowd behavior for computer graphics and virtual reality applications. Our algorithm performance is compared with other proposals in the literature applied in two datasets. In addition, we test the execution time to validate its usage in real-time. In the results we obtain fast execution time of the algorithm and robustness in its performance.

Año de publicación:

2017

Keywords:

• image analysis
• Real-time applications
• Abnormal crowd
• Surveillance

Fuente:

scopus

google