Residual F-Lenet: an efficient hybrid deep learning model for abnormal event detection using surveillance videos

Abstract

Abstract Surveillance cameras are essential components for monitoring human activities and preventing annoying actions. In recent times, Security management professionals heavily depended on video surveillance to monitor corruption and prevent incidents that could harm society. There is a large amount of surveillance camera deployments that have been maximized intensely in remote and public regions to control communal actions. However, the conventional detection models required a large volume of labeled information for effective training, and the implementation of this system was time-consuming and more expensive. Therefore, this paper aims to design a ResidualF-Lenet for detecting abnormal events in surveillance videos. The process begins with extracting frames from input surveillance videos. Next, multi-object detection is done with YOLO v3. After the completion of multi-object detection, essential features including Local Optimal Oriented Pattern (LOOP), Pyramid Histogram of Oriented Gradients (PHOG), Median Binary Pattern (MBP), and Local Directional Pattern (LDP) are extracted. Lastly, abnormal event detection is performed using ResidualF-Lenet, which integrates Residual Network with Lenet. The evaluation outcome indicates that the proposed model achieved improved levels of accuracy, True Positive Rate (TPR), and True Negative Rate (TNR) as 91.681%, 90.740%, and 92.802%, respectively.