9th International Conference on ICT for Sustainable Development

Authors - Rashmi S Bhumbare, Aarti Machhindra Chavan, Mayuri Sanjay Chavan, Pratiksha Vijay Chavan, Shejal Sanjay Pawar
Abstract - In contemporary times, video surveillance assumes a crucial role, especially given the widespread adoption of cutting-edge technologies such as artificial intelligence, machine learning, and deep learning. These advancements have fostered the creation of intricate systems capable of discerning various forms of questionable conduct from live video streams. Identifying suspicious behavior, often subtle and challenging to discern from typical actions, proves to be one of the most demanding tasks. To tackle this hurdle, a blend of methodologies, including deep learning, is utilized. Deep learning algorithms are trained to identify anomalous patterns of behavior across diverse settings, notably within educational institutions. The surveillance process typically encompasses the examination of frames extracted from video recordings. These frames undergo a dual-stage processing: initial feature extraction followed by classification using these features to ascertain the presence of suspicious behavior. In essence, these dynamic systems harness state-of-the-art technology to bolster surveillance capabilities, facilitating prompt identification and response to potential security risks. Upon detecting suspicious activities, timely alert notifications are dispatched to relevant authorities, ensuring a proactive stance towards upholding safety and security.