Detection And Identification Of Tire X-ray Image Defects Based On Machine Vision

The development of the automatic detection of tire defect is of great significance to national tire industry,especially in the context of current national manufacturing 2025,Industry4.0 and other development strategies.Due to the differences of tire materials,tire types and production environments between domestic and foreign,imported tire automatic detection equipment can not be completely suitable for domestic detection needs.Currently,domestic tire X-ray image defects detection is mainly rely on visual inspection method.The method has the problems of subjectivity classification,visual fatigue and the detection rate affected by the worker status,which seriously affects the development of China's tire manufacturing industry.On the basis of this,we propose the tire X-ray image defect detection methods as followed:1.In order to improve the detection rate and accuracy of defect detection,a method of tire image segmentation based on texture structure is adopted.This method is through the selection of the optimal Gabor filter parameters and then the original structure in accordance with the tire structure is divided into tire beads,sidewall and crown three parts.In addition,in order to meet the real-time requirements of tire enterprises,this paper has proposed a method of salient region detection based on integral image.By using the integral map to calculate the amount of small,high efficiency features,the method not only meets the real-time requirements of tire segmentation,but also provides candidates for tire impurity detection.2.By analyzing steel texture of tire sidewall cord,a cord defect detection method has been proposed.Aiming at the enhance the structure of tire steel cord,gamma correction is adopted to weaken the background noise and highlight the structure of the tire sidewall cord.Then we extract contour of the steel wires and filling them by using Canny edge operator.The cord defects are classified by designing several strategies.Experimental results show that the proposed scheme can effectively detect defects in the tire sidewall.3.A tire impurity detection method is proposed.Features including texture,geometric and gray of tire impurity defect have been extracted.Based on the rule of experience classifier,the candidate regions of impurity defects are classified and judged.4.A method of tire impurity classification based on machine learning is proposed.In face of high false alarm rate of tire impurity detection,this paper presents an approach by extracting higher dimensional feature vectors of impurity defects and using more robust support vector machine classifier to improve accuracy of impurity classification.Finally,we also come up with a idea of effective feature selection based on multiple kernel learning.Comparisons with conventional image edge detection methods our scheme presents more promising during test dataset and application.