Abnormal Activity Recognition in Private Places Using Deep Learning: A Survey

Abstract

Abstract: The method of analysing human motion using computer and machine vision technologies is known as "human activity recognition," or HAR. One of the applications of human activity recognition in security systems is anomaly detection. Surveillance cameras have been widely placed as the foundation for video analysis as the demand for security has grown. Identifying aberrant behaviour necessitates considerable human effort, which is one of the major challenges in surveillance video analysis. It is important to set up video recording in order to detect unusual activity automatically. Our intelligent video surveillance system can detect an abnormality in a video using deep learning technologies. Real-time detection of activities is also conceivable, and these video frames will be saved in the system as images for the user to study. The proposed Abnormal Activity Recognition system was designed with the purpose of finding and detecting anomalies in the financial industry, especially in an ATM context, using a live stream. The first part of the research focuses on the use of image deep learning algorithms to recognise different products and detect anomalous behaviour utilizing ATM monitoring systems