Anomaly Detection from CCTV Camera Feed
Aakash Aakash, Lakshay Chauhan, Shubham Sharma, Sujay Deb · 2024
The paper presents a novel approach for detecting anomalies like arson, explosion, vandalism, etc in public spaces using CCTV camera feeds. We propose a deep learning-based system that analyzes video streams to identify unusual events or behaviours that deviate from normal patterns. Our method combines Convolutional Neural Networks (CNNs) and Spiking Neural Networks (SNNs) for feature extraction with Long Short-Term Memory (LSTM) networks to model temporal dependencies in the video sequences. The system is trained on the UCF Crime Dataset to learn normal activity patterns and can flag potential anomalies in emulated real-time situations. To enhance accuracy, we incorporate additional computer vision technique, like optical flow, with CNN/SNN and LSTM to get various combinations of models (OF-CNN, OF-CNNLSTM, OF-SNN, OF-SNNLSTM, CNNLSTM) for improved anomaly detection. The proposed system has potential applications in enhancing public safety and security monitoring in urban environments.