Ensemble Learning Approach for Anomaly Detection in Crowd Scene Classification

Anomaly detection in videos captured by surveillance cameras is a critical area of research that aims to automatically identify unusual or abnormal events captured in video footage. The process of manually analyzing video data can be difficult, time-consuming, and prone to human mistake. In surveillance camera footage, anomaly detection uses machine learning and deep learning to automatically identify typical patterns of behavior and detect deviations from these patterns. This paper provides an examination of anomaly detection through videos captured by surveillance cameras and videos included from YouTube and various sites and examines the different algorithms and techniques used for analyzing video footage. In our proposed model we have used an ensemble learning model which includes MobileNet and LSTM combination ensembled with one another. The model is properly able to detect and distinguish between four different types of scenes. We proposed an approach by executing experiments on several real-world datasets and show that our approach methods in terms of both accuracy and robustness. This research paper will provide a comprehensive understanding of anomaly detection in videos for crime scene activities happening in real life, including the latest advanced state-of-the-art algorithms and techniques, their limitations and challenges, and potential future directions of research.