Citation

Chougule S, Barhatte A. Int. J. Intell. Transp. Syst. Res. 2023; 21(3): 483-492.

Copyright

(Copyright © 2023, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s13177-023-00363-3

PMID

unavailable

Abstract

Potholes are a threat on roads, and their presence compromises driver, vehicle, and pedestrian safety. In developing countries, the primary reason for road accidents is bad road conditions, resulting in human life and property loss. In countries like India, Road maintenance is a challenging activity. Accidents rates are increasing year by year due to the up-surging potholes count. This paper presents the system as "Forward View Guidance for pothole detection for Indian passenger Car." The camera captures video images, and a deep learning algorithm is used to classify the images as potholes and regular roads. The camera will also provide a view of the vehicle's front, highlighting the pothole. Deep learning YOLOv3 and YOLOv5 algorithms are used to train the model and tested with the Kaggle pothole detection datasets to predict the model's accuracy for detection. The proposed system will help monitor the road's condition, count the number of potholes on the road, and generate an alert signal. The performance of the proposed system is evaluated using precision, recall, and average precision (AP). The experimentation results show that the YOLOv5 algorithm performs best than the YOLOv3 algorithm. The precision, recall, and average precision (AP) for YOLOv5 are obtained as 0.763, 0.548, and 0.635, respectively. The system algorithm is implemented on the Raspberry Pi 4B model, which can be easily fitted as an addon system in the vehicle.

Language: en

Keywords

ADAS; CNN; Potholes; Raspberry Pi; Yolo