Anomalous Crowd Behavior Detection in Time Varying Motion Sequences
Imran Usman · 2019
Automated crowd behavior detection has become a prime research area in recent years. Due to inherent complexities in video sequences and foreground motion patterns, crowd motion analysis faces many challenges. This work uses a statistical model for representation and extraction of local motion patterns in order to generate the feature set. It then utilizes a Genetic Programming (GP) based classifier to classify normal and abnormal behavior patterns through a supervised learning mechanism. The developed classifier is generic in nature and can be easily implemented in hardware. Experimental results on public datasets validate that the proposed scheme outperforms contemporary techniques in terms of classification accuracy and effectiveness.