Research On Automatic Defect Detection Algorithm Based On Tire X-Ray Images

With the development of the economy in China,automobiles have become one of the necessary transportations for people.Therefore,the quality of the tires determines the incidence of traffic accident,which is related to people's lives.The problem of automatic detection of tire defects has become one of the most concerned focuses of domestic and foreign tire factories.However,the current foreign tire defect detection software is not suitable for China's tire production environment.First of all,the software interface is mostly in English.It is difficult for domestic workers to operate,and the standards for tire defects at home and abroad are different.Therefore,the domestic tire defect detection still requires the workers to observe the X-ray images of the tires with eyes to obtain the result,it's not really realize the automatic detection.Moreover,the judgment result is greatly influenced by subjective factors,and the correctness and efficiency vary from person to person.When the worker has been working for a long time,the false detection rate caused by visual fatigue will rise linearly.It's very important to develop tire defect detection equipment suitable for China's actual tire production conditions,high performance and low cost,and simple operation.In this paper,by observing and analyzing the image features of each defect in the tire,the pretreatment method of tire images in different regions is proposed.For example,cord defects are generally classified into pitch problems and alignment problems.The characteristics of the cord can be highlighted by an image binarization operation,and then the cord defects can be detected by column scanning and line scanning.When dealing with bubble defects and debris defects,generally select the appropriate filter to filter the background of the image,then select the appropriate threshold to segment the defect from the image,and finally traverse the image for detection.Due to the complex shape of the tread in the belt,an improved template matching method is generally utilized to detect defects in the belt.In addition,this paper also combines the deep learning algorithm,based on the image processing of the tire X-ray image,using the neural network to learn the tire defect sample set,to achieve a true end-to-end automatic detection algorithm for tire defects.Based on the above automatic detection algorithm for tire defects,this paper independently developed an automatic tire defect detection system.The system includes many functions,such as real-time transmission of tire X-ray images,automatic detection of tire defects,establishment of defect databases,alarm functions,remote communication,and real-time control of conveyor switches.In the system development process,we analyze the basic requirements put forward by users,and design and develop various functional modulesof the system.After the system is tested,the defect detection report will be displayed on the touch screen,and the workers can confirm the test result again.The deep learning algorithm will continuously optimize its own parameters through the tire defect detection process,thus improving the accuracy of detection.The system will automatically store the detected tires in the background according to the model information,detection time,tire images,and defect reports,and establish a database of tires for the optimization of algorithms and the improvement of production equipments.Finally,a lot of tests were carried out on the software in the actual tire production process,and the ideal effect was obtained.Compared with the traditional detection method,the accuracy and speed were greatly improved.