An Overview of Various Techniques Involved in Detection of Anomalies from Surveillance Cameras

Abstract

In recent years, the use of surveillance cameras is rapidly increasing in both public and private areas to enhance the security measures. Many companies are recruiting people to monitor the activities captured by surveillance cameras and due to human error they may failed to monitor the abnormal events. So, an automated system to detect the anomalous events acts as a significant approach in surveillance applications. Due to sparse occurrence of anomalous activities, the detection of anomalies is remaining as a challenging task. To overcome these drawbacks, many researchers have worked to develop an effective anomaly detection methods using different approaches. This study prioritized some existing approaches to detect anomalies takes place in surveillance videos. The existing researches utilized University of Central Florida (UCF) Crime video dataset to collect the data about the anomalous activities, UCF crime video dataset consist of 13 categories of anomalies which consist of 1900 surveillance videos. The key parameters such as accuracy, recall, F1 score and Area Under Curve (AUC) are evaluated to analyse the efficiency of the existing anomaly detection methods. This survey acts as a tool for future researchers to overcome the drawbacks in the existing methods and create a novel anomaly detection approach.