DEEP NEURAL NETWORKS FOR SUSPICIOUS ACTIVITY RECOGNITION IN VIDEO SURVEILLANCE

Abstract

The objective of this investigation is to enhance automated security monitoring systems by investigating the utilization of deep neural networks (DNNs) to identify suspicious activity in video surveillance. To accurately identify abnormal or potentially hazardous behaviors, the research employs advanced architectures, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to extract temporal and spatial features from video streams. The proposed methodology effectively distinguishes between normal and suspicious behavior by incorporating pattern recognition, motion analysis, and feature learning. In order to address obstacles such as occlusion, variable lighting, and complex crowd dynamics, the investigation implements rigorous training methodologies and extensive, annotated datasets. Experimental results suggest that deep learning models outperform conventional machine learning methods in terms of adaptability, scalability, and accuracy. Subsequently, they are optimal for contemporary surveillance applications in public safety, transportation systems, and smart cities.