Defect Detection Of Tire Forming Based On 3D Vision

In the last few years,the economic level has been highly improved in our country.People's travel mainly depends on cars nowadays and the safety of people's travel has a great relationship with the quality of tires.Qualified tires are closely related to the safety of personal and fortune.So detecting the defects of tires in the tire forming process is crucial.It can not only find defects during the process of tire in time to adjust the production process,minimize the wastage of materials,but also provide protection for people to travel.Most of the tire manufacturers use X-ray images to detect the formed tires.Nowadays,the defect detection in the tire forming process is not mature.The main contents of this paper are as follows:(1)In this paper,c4-1280 camera and 3D laser are used to build the data acquisition system.The 3D camera is synchronized with the belt drum servo motor,and the sampling frame rate can be up to 75000 frames / s.The system is flexible in installation and fast in scanning speed,and can complete the collection of 3D data without affecting the tire forming process.Then the height of the collected data is represented by the gray value of the image,which makes the analysis and processing of the data more simple and convenient.(2)Because the data in this paper are collected by multiple cameras from different angles,the collected data are many small pictures with the size of 1280 * 640.In view of the problem that the collected data are small pictures with high gray value,this paper firstly transforms the gray value of the collected data from 0 to 255,so as to display and analyze the data.After that,they are merged into one long image.Because the long image contains a lot of information which is not related to the target area,it may affect the segmentation result.Therefore,it is necessary to cut these regions and cut the remaining parts into head,material and tail images.(3)Aiming at the problem that the similarity between the drum surface of the first belt and the target area is too high to be segmented.Firstly,RANSAC algorithm is used to correct the drum surface and remove the noise.Then use it and the defined rectangular box as the input of the Grab Cut algorithm.Because the material tail image of the first belt is also attached with the material head part,the material head part will cause great interference and obstruction to the segmentation of the material tail image.Therefore,this paper first differentiates it with the segmented material head image,so that the material head part can be removed.Then the image segmentation operation can be completed successfully.Because the second belt is attached to the first belt,this paper first finds its corresponding position on the first belt and carries out the difference,then completes the segmentation processing.Then,by comparing with grabcut algorithm and u-net algorithm directly,and using dice parameter evaluation index,it shows that this algorithm is better than the other two algorithms.(4)This paper independently designed and completed the tire molding process defect detection system.For completing the function of image processing,the system was developed on the QT platform and used the Opencv library.The system is composed of four interfaces: login interface,main interface,teaching board interface and continuous operation interface.Before entering the main interface,the user is required to login and verify correctly.The teaching version interface mainly has a series of image processing functions such as gray scale conversion,cutting,merging,RANSAC algorithm,Grab Cut algorithm,etc.,which can realize batch and step-by-step processing of images.The continuous operation version of the interface realizes the continuous scrolling display of the tire image and the function of outputting the test results.Finally,a rigorous and meticulous test was done on the entire system.