Harnessing optimization with deep learning approach on intelligent transportation system for anomaly detection in pedestrian walkways

Anomaly Detection (AD) in pedestrian walkways is significant in urban safety and security methods. It is generally employed for perceiving unusual or abnormal situations, behaviours, or actions in regions devoted to pedestrian traffic, like pedestrian bridges, sidewalks, or crosswalks. The main aim is to enhance security, efficiency, and safety in the urban environment by classifying changes and observing the activities of pedestrians from recognized models. This type of AD usually has sensors, surveillance cameras, and advanced software methods. By applying innovative computer vision (CV) and machine learning (ML) methods, this approach always examines the pedestrian area to detect potential anomalies and attacks. Deep learning (DL) aided AD in pedestrian walkways, which displays a new and very effective method to progress safety and security in urban environments. Therefore, this study designs a novel Harnessing Optimization with a DL Approach on an Intelligent Transportation System with Anomaly Detection in Pedestrian Walkways (HODLAITS-ADPW) method. The presented HODLAITS-ADPW method utilizes the CV and DL process to detect pedestrians accurately. To attain this, the HODLAITS-ADPW method performs image pre-processing using Median Filtering to enrich the quality of input frames. Next, the HODLAITS-ADPW method utilizes an enhanced YOLOv7 object detector for the object's detection in the frame. Besides, the Chameleon Swarm Algorithm is employed for the optimum hyperparameter tuning method. The HODLAITS-ADPW method uses an Attention Pyramid Convolutional Neural Network model for anomaly recognition. The experimental values of the HODLAITS-ADPW technique were certified on the pedestrian dataset, and the outcomes can be examined in numerous aspects. The empirical findings reported better outcomes for the HODLAITS-ADPW technique when compared to other models.