Anomalous Detection System in Crowded Environment using Deep Learning

In recent years, surveillance systems have become very important due to security concerns. These systems are widely used in many applications such as airports, railway stations, shopping malls, crowded sports arenas, military etc., [1]. The wide deployment of surveillance systems has made the detection of anomalous behavioral patterns in video streams to become increasingly important. An anomalous event can be considered as a deviation from the regular scene; however, the distribution of normal and anomalous events is severely imbalanced, since the anomalous behavior events do not frequently occur, hence it is imperative to accurately detect anomalous behavioral pattern from a normal pattern in a surveillance system. This paper proposes a Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM) technique. The CNN is used to extract the features from the image frames and the LSTM is used as a mechanism for remembrance to make quick and accurate detection. Experiments are done on the University of California San Diego dataset using the proposed anomalous behavioral pattern detection system. Compared with other existing methods, experimental analysis demonstrates that CNN-LSTM technique has high accuracy with better parameters tuning. Different analyses were conducted using the publicly available dataset repository that has been used by many researchers in the field of computer vision in the detection of anomalous behavior. The results obtained show that CNN-LSTM outperforms the others with overall F1-score of 0.94; AUC of 0.891 and accuracy of 89%. This result shows that the deployment of the proposed technique in a surveillance detection system can assist the security personnel to detect an anomalous behavioral pattern in a crowded environment.