Golden Jackal Optimization with Deep Learning-Based Anomaly Detection in Pedestrian Walkways for Road Traffic Safety

Abstract

Road traffic safety discusses the procedures and measures utilized for preventing road users from being dead or critically injured. Archetypal road users contain horse riders, cyclists, pedestrians, vehicle passengers, motorists, and passengers of on-road public transport (mostly buses and trams). Anomaly detection in pedestrian pathways is a crucial investigation topic, generally utilized for improving pedestrian safety. Because of the varied consumption of video surveillance methods and the improved quantity of captured videos, the typical manual analysis of labeling abnormal proceedings is a tiresome task, thus, an automated surveillance method in which anomaly detection develops important betwixt computer vision researchers. At present, the progress of deep learning (DL) algorithms has obtained important interest in distinct computer vision procedures. Therefore, this article introduces a new Golden Jackal Optimization with Deep Learning-based Anomaly Detection in Pedestrian Walkways (GJODL-ADPW) for road traffic safety. The presented GJODL-ADPW technique aims to effectively recognize the presence of anomalies (such as vehicles, skaters) on pedestrian walkways. In the presented GJODL-ADPW technique, Xception methodology was exploited for effective extraction feature process. For optimal hyperparameter selection, the GJO algorithm is utilized in this study. Finally, bidirectional long short-term memory (BiLSTM) approach was employed for anomaly detection purposes. A widespread experimental analysis is performed to examine the enhanced performance of the GJODL-ADPW system. A detailed comparative analysis demonstrated the enhancements of the GJODL-ADPW technique over other recent approaches.