<![CDATA[Car tires play a crucial role in ensuring the safety of drivers and passengers. Tire wear that is not properly identified can reduce vehicle stability and traction, thereby increasing accident risk. Conventional tire wear assessment is generally carried out through manual visual inspection, which is subjective and prone to inconsistency, leading to potential misjudgment of actual wear conditions. This study proposes an automatic tire wear detection system using a Convolutional Neural Network (CNN) combined with an edge detection approach to provide a more objective and accurate assessment. The research stages include data preprocessing, dataset construction, and system evaluation. Model performance was tested using two data partitioning strategies, namely random and stratified classification, while maintaining the same architectural configuration consisting of HiddenL64, 10 epochs, maxpooling2, and iterationmax 20. The evaluation aimed to compare detection performance across both classification schemes. The experimental results indicate variations in training and testing accuracy between the two methods. The CNN model achieved its best performance under the random classification approach, with training accuracy reaching 99.15% and testing accuracy of 86.62% using a 90%–10% data split. Despite these results, the observed performance gap between training and testing data suggests that model generalization is still influenced by dataset size and distribution, indicating potential overfitting. Therefore, the findings should be interpreted within the scope of the experimental setup and dataset limitations. Overall, the proposed CNN-based system demonstrates promising capability for automatic tire wear classification and has potential to support preventive maintenance and timely tire replacement in automotive applications to improve road safety.]]>
Copyrights © 2025
