(G.H.Raisoni College of Engineering and Management, India.)
Mr. Harshal G. Mundle is an Assistant Professor in the Department of Civil Engineering at G. H. Raisoni College of Engineering and Management, Nagpur, Maharashtra, India. He is engaged in civil engineering education and academic activities. His contribution focuses on the integration of multimodal AI technologies into smart city monitoring and urban anomaly detection.Dr. P. Sumalatha is an Assistant Professor in the Department of Basic Sciences and Humanities at Seshadri Rao Gudlavalleru Engineering College, Gudlavalleru, Andhra Pradesh, India. She is involved in teaching and academic activities in basic sciences and humanities. Her contribution focuses on multimodal data fusion for air quality monitoring and pollution prediction.
38. Anmol W. Dongre
S. Malathi, Vaibhao V. Maind, Harshal G. Mundle. In: Multimodal Artificial Intelligence: Foundations, Intelligent Applications, and Emerging Trends — ISBN: 978-81-17543-28-3. Pages: 250 - 269
Intelligent surveillance and monitoring systems are essential for smart cities to enhance public safety, regulate traffic, safeguard infrastructure, and respond to emergencies. Conventional surveillance systems predominantly rely on video cameras and human oversight, which may restrict their capacity to identify intricate events, comprehend contextual data, and adapt to swiftly evolving urban environments. This section introduces a multimodal artificial intelligence system for surveillance and anomaly detection in smart cities, encompassing data collection, preprocessing, feature extraction, cross-modal integration, anomaly scoring, event categorisation, and alert creation. Advanced methodologies in deep learning, including convolutional neural networks, visual Transformers, recurrent neural networks, audio encoders, and multimodal attention mechanisms, are examined for urban surveillance applications. The section additionally explores privacy safeguarding, algorithmic prejudice, erroneous alerts, data protection, clarity, and ethical implementation. A proposed evaluation methodology incorporates event-detection metrics, false-alarm rates, detection delay, and cross-location generalisation.
