Road safety in India presents a multifaceted challenge due to the dynamic nature of road environments. Diverse el-ements such as vehicles, pedestrians, and livestock, considered as road friction elements, frequently disrupt traffic flow and demand efficient management. Additionally, the roads often lack clear markings or divisions, making navigation difficult for Advanced Driver Assistance Systems (ADAS). To address these challenges, we propose a computer vision-based solution for detecting and classifying critical road friction elements on Indian roads. From the Indian Driving Dataset (IDD), which includes 34 annotated classes, we focus on seven main classes: animals, motorcycles, bicycles, trucks, cars, persons, and riders. Additionally, two new classes-carts and garbage-are introduced, making a total of nine classes considered as road friction elements. To locate these friction elements and enhance understanding of road environ-ments' the road label is utilized for road segmentation. The YOLOv5 model handles the detection of friction elements, while road segmentation is managed by the U-Net model. Experimental results indicate a high mAP of 0.92 for the YOLOv5 model trained on common labels and 0.89 for the model focused on less common labels. We establish ground truth by annotating images for binary classification of detected objects into friction or nonfriction elements. Using an AdaBoost classifier, we achieve 94.8 % accuracy in classifying these elements as either friction or nonfriction. This method demonstrates significant promise for advancing road safety, traffic management, and autonomous driving in India.
Enhancing Road Safety: Road Friction Object Detection
2024-12-07
8211268 byte
Conference paper
Electronic Resource
English
VIRTUALIZED ROAD TRAFFIC SIGN GENERATION FOR ENHANCING ROAD SAFETY
European Patent Office | 2022
|VIRTUALIZED ROAD TRAFFIC SIGN GENERATION FOR ENHANCING ROAD SAFETY
European Patent Office | 2022
|ON-ROAD OBJECT DETECTION SYSTEM AND ON-ROAD OBJECT DETECTION METHOD
European Patent Office | 2024
|Enhancing Road Safety with AdaBoost-Based Drowsiness Detection for Drivers
Springer Verlag | 2024
|