Detection of Anomalies in Surveillance Video using Improved Cheetah Optimizer Algorithm along with Variational Auto Encoder Classifier
Ankit Bansal, Praneeth Reddy Amudala, Swathi Suddala, Ruchi Bansal, Aayush Singhal · 2025
Anomaly detection in surveillance videos involves identifying abnormal and unusual patterns within video data, like accidents or theft. Traditional techniques including Machine Learning (ML) algorithms have been used for detecting such anomalies; however, these methods often fail to accurately detect anomalies. Hence, this research proposes the Improved Cheetah Optimizer (ICO) Algorithm for the feature selection process for anomalies detection in Surveillance Video. The objective of this research is to enhancing detection rate through effective feature selection and classification. The University of California, San Diego (UCSD) dataset is used as input to train the model. The data is first pre-processed using normalization and temporal smoothing techniques. The Variational Autoencoder (VAE) approach is introduced for the classification of various video surveillance. The proposed ICO-VAE method achieves the better Area Under Curve (AUC) of 93.56% and an Equal Error Rate (EER) of 94.59% as compared to the traditional Support Vector Machine (SVM) approach.