Abstract: The aim of the project is to develop an effective method for anomaly detection and recognition in live streams. Anomaly detection refers to the task of identifying instances or events that deviate from the expected or normal patterns. In the context of live streams, anomalies can include activities such as fighting, accidents, robbery, and other unusual events. Rather than requiring continuous human monitoring of surveillance feeds, automated anomaly detection can relieve the burden on security staff. The system can alert operators only when potentially significant anomalies are detected. Effective anomaly detection can help reduce false alarms in live streams. By accurately distinguishing between the normal and anomalous activities, security personnel can avoid unnecessary responses to non-threatening situations. The proposed method works better by recognizing the abnormal behaviors in long-distance captured images and reducing the false alarm rate by effectively distinguishing between the normal and abnormal activities.
Keywords: Anomaly detection, live streams, false alarms, unusual events, abnormal behaviors, normal activities, effective recognition
|
DOI:
10.17148/IJARCCE.2024.134115
