Spatio-temporal based video anomaly detection using deep neural networks

Anomaly detection is the identification of unexpected events. The state–of-the-art algorithms try to reduce the reconstruction errors of training data, but there is still no guarantee that the reconstruction error will be smaller in the case of any event that is not usual. In this article, a framework is suggested to address the issues with anomaly detection. U-Network can capture spatial information, while long short-term memory (LSTM) can process temporal information well. The proposed method is a combination of U-Net and ConvLSTM to handle spatial information and temporal motion. To enhance the quality of the frames, bilateral filtering has been introduced, which will increase the range of the peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM), the two performance measures used. Using PSNR and SSIM, a weighted regular score function is derived to classify the frames based on the scores. Finally, results are compared with those of the state-of-the-art method to prove the superiority of the proposed method. © 2023, The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management.