Anomalous Event Detection in Crowd Scenes using Histogram of Optical Flow and Entropy
Divya Dileep, K. G. Sreeni · 2021
This work proposes an algorithm that makes advances in anomaly detection in a crowded video. The proposed model uses optical flow magnitude, orientation, and entropy to capture anomalies in a given crowded video. Assuming that the video is recorded from a stationary camera, a feature descriptor, Histogram of Optical Flow and Entropy (HOFE) that uses spatial and temporal information, is introduced. An SVM classifier is used to classify anomalous frames from the extracted features. From the conducted experiments, it is found that the performance of the model is superior to other existing methods.