Intelligent Crime Scene Recognition: Advancing Public Safety through Deep Learning Architectures and Event Sequence Analysis
Anil M. Hingmire, Aman Sheikh, Saurabh Shukla, Rahul Shah, Sunayana Jadhav, Tatwadarshi P. Nagarhalli, Amrita Ruperee · International Journal of Electrical and Electronics Engineering · 2024
In contemporary society, ensuring public safety is a paramount concern, and one of the significant challenges faced by law enforcement agencies is the swift detection and classification of criminal activities from surveillance footage. Current crime scene detection systems often lack real-time analysis and struggle with the prompt identification of criminal acts, hindering the timely response required for effective law enforcement. Consequently, there is a critical need for an advanced Crime Scene Detection System (CSDS) capable of classifying the type of crime occurring in real time, triggering immediate alarms, and aiding in the rapid identification of criminals captured within the surveillance footage. In this paper, we have highlighted various types’ of techniques that can be used to detect crime scenes using Artificial Intelligence (AI). Crime detection using Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) involves the application of deep learning techniques to analyse and interpret complex patterns in crime-related data. The primary objective is to accurately identify and classify criminal activities, thereby assisting law enforcement agencies in taking proactive measures to ensure public safety. This multidimensional approach is essential for addressing the dynamic nature of criminal behavior, the diversity of criminal activities, and the need for real-time data processing. The analysis reveals that the LRCN model excels in accurately identifying crime events, achieving an impressive 94% accuracy. In contrast, the CNN-LRCN model lags behind with an 84% accuracy rate.