Road Surface Guard: AI Paved Safety
P. Srinivas Rao 1, Syed Ismat Maalikah 2, Telapolu Poojitha 3, S. Anjani Kaustub4, T. Lokesh5
1 Assistant Professor (AIML) 1,2,3,4,5 Sreyas Institute of Engineering and Technology
Abstract
Pothole detection is a critical aspect of road maintenance and safety, with the potential to prevent accidents and reduce infrastructure repair costs. Early detection and timely repair of potholes can help prevent accidents and reduce maintenance costs. Deep learning techniques have shown success in several computer vision tasks, including object detection and segmentation. The proposed system leverages Convolutional Neural Networks (CNNs), You Only Look Once (YOLO) object detection algorithm and Light Detection and Ranging (LiDAR) technology to identify and locate potholes in real-time. The system's architecture comprises data collection from vehicle-mounted cameras, image preprocessing, and a deep learning model for pot-hole detection. A labeled dataset of road images with annotated potholes is used to train the model, allowing it to learn the distinctive features of potholes, such as shape, depth, and texture. These images are then utilized to train a CNN-based model using deep learning techniques. The trained CNN model is then employed to detect pot-holes in real-time road images captured by vehicle-mounted cameras. Evaluation of the proposed system on diverse road surfaces and lighting conditions demonstrates its robustness and accuracy. It achieves a pothole detection rate of over 90%, outperforming traditional methods. The system's ability to provide instant alerts to drivers and municipal authorities enhances road safety and expedites pothole repair efforts. The proposed approach can be integrated into existing road monitoring systems, aiding in the timely identification and remediation of road hazards, ultimately improving road safety, and reducing maintenance costs.
Keywords: Deep Learning, Convolutional Neural Networks (CNNs), Light Detection and Ranging (LiDAR), Real-time Monitoring, Vehicle-mounted Cameras.