Hybrid Motion Vector and Deep Learning Approach for Crime Event Detection and Face Identification in Surveillance Footage

The identification of unusual occurrences is one of the primary goals of the research and use of video surveillance. The objective of enhancing public safety through increased surveillance camera deployment has led to wider adoption in public areas such as roads, intersections and both banks and shopping malls. One of the most crucial tasks in video surveillance is spotting unusual activity, such as crimes, traffic accidents, or unlawful activity. Abnormal behaviour is frequently far less prevalent than normal behaviour. Finding the span of time during which the anomaly is occurring and promptly alerting users to behaviour that deviates from expected patterns are the two main objectives of a realistic anomaly detection system. Thus, by separating anomalies from typical patterns, anomaly detection might be regarded as a basic technique for understanding films. Once an anomaly has been identified, classification techniques can be used to assign it to one of the designated tasks. This article gives an overview of anomaly detection with an emphasis on banking operations applications. Employees, clients, debtors, and other parties are all involved in the routine, recurrent, and ad hoc activities and transactions that occur in banking operations. Early identification can greatly minimise any possible negative consequences and, in many cases, totally eliminate them, even though they may take some time to resolve. Anomaly detection based on time series is used to locate persons at unfavourable periods. This work uses an anomaly detection technique based on machine learning to identify both common and rare occurrences. The biometric distinctiveness of the face is noted and contrasted with the faces of infamous criminals. It is easy to identify and capture the criminal when a match is found.