Detecting tire cracks using cascading deep features of transfer learning models and ensemble classifiers

This study proposes an artificial intelligence-supported worn tire detection system to detect cracks in vehicle drivers' tires. In general, drivers are aware of the significance of tire tread depth and air pressure, but are not aware of the risks associated with tire oxidation. However, tire oxidation and cracks can lead to significant problems affecting driving safety. In this study, a new hybrid architecture called CTLDF+EnC is proposed for crack detection in tires using deep features obtained using pre-trained transfer learning methods and ensemble learning methods. The proposed hybrid models include nine transfer learning methods to extract features and Stacking, Soft and Hard voting ensemble learning methods as classifiers. Unlike most studies in the literature based on X-Ray images for industrial use, the study works with images that can be obtained with any digital imaging device. Within the scope of the study, the highest test accuracy value of 76.92% was obtained with the CTLDF+EnC (Stacking) hybrid model. For CTLDF+EnC (Soft) and CTLDF+EnC (Hard) architectures, accuracy values were obtained as 74.15% and 72.92%, respectively. The results of the study show that the proposed hybrid models are effective in detecting tire problems. In addition, a low-cost and feasible structure is presented.