Robustness Classification by Machine Learning from Vehicle Tire Surface Abrasions

The safety and durability of vehicle tires is an important variable in terms of driving safety and cost effectiveness. Different methods such as visual inspection, tire air pressure control, pattern depth measurements, rotation and balancing can be used to evaluate these factors. In this study, different machine learning algorithms such as ResNET50, DenseNET121, AlexNET, CNN, which are image-based, are used to analyse the images of the tire surface to determine the surface wear of the vehicle tires and to perform robustness classification. For the training of the models, 1447 vehicle tire surface images of different categories (very good, good, bad, very bad) were used. The dataset containing the images belongs to the authors of this study and is unique. In the future, it is aimed to make the dataset available for copyrighted use on an open platform. The results obtained from the trained models are compared. The CNN algorithm, which showed the most successful results, was selected as the final algorithm. In conclusion, this paper represents an important step towards solving safety and efficiency issues in the automotive industry by introducing a machine learning approach to detect surface wear and robustness classification of vehicle tires. This technology has the potential to optimize tire management and maintenance.