SUSPICIOUS ACTION AND BEHAVIOR DETECTION USING CNN

Surveillance is the monitoring of behaviour, activities to manage or direct. This includes observation from a distance with the help of electronic equipment, such as closed-circuit television (CCTV). Most of the surveillance systems are still under human surveillance. Violence within the premises could degrade the decorum of the institution and need to be addressed with at most sincerity. Luckily, the recently emerging AI technique can automatically detect the anomalies. Such anomaly detection is fast and can be further used as the pre-processing mechanism to filter out the surveillance videos, and then forward the anomaly videos to perform the examinations by other highly accurate algorithms. One such algorithm is a convolutional neural network (CNN) that takes the input video frames and outputs the features to the Long Short-Term Memory (LSTM) to learn global temporal features and eventually classify the features by fully-connected layers. This network can not only be implemented by the pre-trained models in ImageNet but also have the adaptability to accept variable-length videos.