Anomaly detection in surveillance videos

Abstract

The main task of this thesis is to review anomaly detection methods and then demonstrate the use of a state of the art approach to detect nudity as an anomaly in a surveillance video. The following contributions were made. One, the identified problem of anomaly detection based on nudity was solved by exploring methods in the reviewed literature. A model was developed and trained on a dataset. Furthermore, the trained model was used to develop a video-based anomaly detection application that could recognize normal and anomaly contained frames in surveillance video.