Today, most of the urban and extra-urban transportation services are provided by highways. Potholes on the highway significantly affect driver safety by reducing the road quality. The potholes on the highway cause serious traffic accidents as well as financially damaging the vehicles on the highway. In order to increase driver safety and reduce traffic accidents, a driver assistance system that detects potholes on the highway surface should be integrated into vehicles. The realization of a system that detects potholes on the highway and gives a pre-warning for driver assistance systems will prevent material damage to highway vehicles and reduce traffic accidents. For this reason, the development and application of deep learning algorithms, which have been used in different object detections and have shown great success in real life problems in recent years, make it possible to detect highway potholes. In this study, a system that detects potholes on the highway is proposed using Yolo-v4, one of the deep learning algorithms. In order to realize the proposed system, a dataset consisting of a total of 681 highway images, including 329 highway potholes images and 352 normal highway images, was used. Using this data set, the model training and testing of the proposed system and the performance test used in real-life problems were carried out. According to the experimental results, it was seen that potholes on the highway were detected quickly, reliably and successfully using the Yolo-v4 deep learning algorithm.
Anahtar Kelimeler: Deep Learning, Convolutional Neural Network, Yolo-V4, Pothole Detection
|
