Hybrid YOLOv5-CNN Framework with Grey Wolf Optimization for Enhanced Accident Prevention in Traffic and Industrial Environments

Abstract

Accidents on roads and in industrial areas pose a serious threat to life and property. This study introduces a novel approach that combines advanced computer vision techniques with optimization techniques to improve the effectiveness of accident prevention strategies. The study uses the highly efficient YOLOv5 object search algorithm to quickly identify potential threats. In addition, Convolutional Neural Networks (CNNs) are used to analyze image data to provide context that enhances detection accuracy. The frameworks dataset, sourced from Kaggle, consists of meticulously curated and annotated traffic camera images from various countries, featuring diverse weather and lighting conditions, with bounding box labels for multiple objects such as cars, people, and traffic signs, making it ideal for real-world object detection applications. Deep learning models with powerful CNNs are particularly well suited for visualization when they self-learn reference sequences Model performance is further optimized by Gray Wolf Optimization (GWO) an implemented over, an algorithm inspired by Gray Wolf hunting techniques implemented on Python, and its effectiveness is evaluated through tests on detailed data sets of real-world images. The hybrid YOLOv5-CNN-GWO model exhibits this accuracy compared to widely used object detection methods such as DSTA, SSD, and embedded computer vision. The results show that the YOLOv5-CNN-GWO model performs significantly better than traditional methods, making it a promising tool for crash prevention programs. With an accuracy of 98%, the proposed method outperforms RCNN, FAYOLO, FESSD, and YOLOV4+NSF by 9.82%. The YOLOv5-CNN-GWO framework contains the potential for use in a variety of applications including robotics, autonomous vehicles, and analytics, where accurate and efficient object detection and classification is required.