Abnormal Events Detection in Surveillance Environments using Machine Learning

The ability of intelligent CCTV systems to detect unusual events, detect human behaviour and recognize objects is essential. This system enables the detection of anomalous events in the environment, anomalous human behaviour and the level of alertness in the environment. The system uses machine learning and machine vision capabilities to detect and identify specific anomalies in CCTV video feeds. This system often uses frames through frame processing, and supervised learning is often used as a training method. However, unsupervised and semi- supervised learning replace supervised learning in the system learning process due to the large variety of anomalies and the impossibility of predicting and preparing for all of them. This system can decrease or do away with the amount of labour-intensive human work needed to manually spot anomalies in the live CCTV feed and provide alerts. The initial monitoring of unusual events and low-quality archiving of typical scenarios also increases the efficiency of the system's storage. In addition, this system provides an extension for developing a distributed anomaly classification system so that only abnormal events are directed to different unique systems for categorization. Regular recordings, on the other hand, will be archived with low quality. Furthermore, this approach offers a distributed anomaly classification system development path.