Abnormal Activity Recognition in Private Places Using Deep Learning

Abstract

Using computer and machine vision technology, the process of analysing human motion is known as "human activity recognition," or HAR. Anomaly detection in security systems is one of the situations in which human activity recognition is useful. As the demand for security growing, surveillance cameras have been widely installed as the foundation for video analysis. Identifying anomalous behaviour demands strenuous human effort, which is one of the main obstacles in surveillance video analysis. It is necessary to establish video recording in order to automatically catch anomalous activities. Using deep learning methods, our intelligent video surveillance system can identify an anomaly in a video. Real-time detection of the actions is also possible, and these video frames will be afterwards preserved as photographs in the system for the user to examine.The suggested Abnormal Activity Recognition system was created with the goal of identifying and detecting irregularities through a live feed in the banking sector, more specifically in an ATM setting.The initial phase of the study focuses on the application of image deep learning techniques to recognise various items and spot unusual behaviour using ATM monitoring systems.