Detection of Defects on Highways Followed by Logistics Transport Trucks with Deep Learning and Providing Safe Driving Support for Drivers

Highway infrastructure is of vital importance in providing safe and efficient transportation. However, highway imperfections such as cracks, potholes and uneven surfaces can pose significant driver risks. In this study, we developed a method based on deep learning using machine vision to detect highway defects on the highway. The images used in the study are taken with a camera from the highway and these images are processed with image processing methods. The processed images are given as input to MobileNet, where MobileNet is a deep learning model that classifies images. Transfer learning was used during the training of MobileNet. The pre-processed image supplied to MobileNet is divided by the model into three groups as good, poor, and very poor. The route is recommended to the driver according to these classes. In order to train the MobileNet model, datasets in the literature dataset collected by our trucks., and datasets created by data augmentation methods were used. The MobileNet, which was customized in this study with the tests performed, detects highway defects in 0.054 seconds and classifies the results. Thanks to these classifications, truck drivers are warned and decision support is provided.