POTHOLE DETECTION

Abstract

Roads are the most important form of nation's transportation system. It is extremely crucial to maintain them in good situation. Potholes are a type of road problem that can harm vehicles and have a detrimental impact on drivers' ability to drive safely, which can result in traffic accidents. Potholes that develop on the road must be filled to keep the roadways in excellent condition. It is essential that you keep them in good shape. It can be difficult to locate potholes in the road, particularly in India where there are millions of km of roadways. In a complicated road environment, effective and proactive management of potholes is crucial for ensuring driver safety. Driver safety is significantly improved by effective and proactive treatment of potholes in a complex road environment. Additionally, it is anticipated to help maintain traffic flow and assist to the reduction of traffic accidents. To get around this problem, a number of strategies have been developed, including manual reporting and government initiatives to help auto-detect pothole zone. There have been several strategies created to get around this problem, from human reporting to authorities to take action for automatic detection of pothole zones. Automated methods for spotting potholes have recently been developed, and these systems include a number of fundamental technologies, including sensors and signal processing. Considering the technology used in the process of identifying potholes, three different types of automated pothole detection systems can be categorized- methods based on vision, vibration, and 3D reconstruction. Building an autonomous model of pothole detection is the major goal of this endeavour, which aims to find potholes as soon as possible. Therefore, it is necessary to automate pothole detection with high speed and real-time accuracy. Our major objective is to train and analyze the YOLOv5, YOLOv7, and YOLOv8 model for pothole identification. These models are trained using a collection of data for potholes images, and the results are examined by assessing the model's accuracy, computational speed, recall, and size, which are then contrasted with those of previous YOLO algorithms. The methodology used in this paper will greatly aid in road maintenance by decreasing expenses and speeding up the detection of potholes. In this article, 84.6%, 87.1% and 85.4% accuracy has been achieved for yolov5, yolov7 and yolov8 model respectively.