Publication

Abstract : COVID-19 has greatly affected social mobility habits and choices in the past three years. The pandemic has caused a modal shift from the use of public transportation to private vehicles. This global shift has caused congestion and is one of the reasons for increased potholes. Potholes are road surface defects with distinctive features such as depression or cavity, irregular shape, varying size, depth, rough edges, loose material, associated cracks, water accumulation. The potholes have a significant impact on the ride quality, cause not only discomforts to the passengers, but also deaths due to vehicle accidents. The cost of repairs caused by potholes has always been a concern for vehicle owners. Proactive communication about the pothole presence and location provides safe and comfortable travel. This paper proposes an artificial intelligence-based application that can identify, update, and render potholes. The proposed application includes a front end, pothole detection module, rendering module and a back end. The front end serves as a friendly platform for the users to upload / view images. The back end stores the captured images. The pothole detection module employs YOLO v5 model to detect the presence and location of pothole(s). The rendering module allows graphical visualization of pothole. The proposed system is tested with unseen real time images of potholes located at Tiruvannamalai highway, Tamil Nadu, India. The performance is analyzed for different real time images under poor lighting and heavy rain.