Automatic Detection of Shoulder Bending Defects in Tire X-ray Images

The quality of tire determines the stability of vehicles. Among the various defects of tires, shoulder bending is a common and serious one. A novel method for shoulder bending detection in tire X-ray images is introduced in this paper. Firstly, a series of pre-processing operations including image inpainting, image blurring, adaptive binarization and image segmentation will be done on the images. Then a seed based cords searching algorithm is proposed to find the cords in pixel level. Finally, a feature extraction algorithm is designed to extract the feature vector of each cords which will be fed into the final judgement. In contrast with deep learning methods, the proposed solution does not require many samples to train the network and it is more adaptable when processing images produced from the factory assembly lines. Experimental results show that when processing different types of tire X-ray images, our method is not only more robust but also with even better precision and recall.