Texture—Invariant Detection Method For Tire Crack

The tire industry plays an important role in our economy, in this process many different kinds of defects exist due to some factors such as rural materials and the manufacturing technique. These defects have great influences on the safety and wear resistance of tires and they low the quality and performance of the product.For example, cord cracking can easily lead to the internal tire suffering from uneven force. So it may result in significant security risks. Therefore, there is a need for imaging the interior of the tire cord structure by means of X-ray technology and by analyzing the tire image, to achieve the detection of defects, and improve the quality of the tire. The tires are mainly composed of two parts:the sidewalls and crown. The texture of Sidewall portion is relatively simple, so the defects there can be easily detected. The crown portion is composed of t multilayer texture and its texture is very complex, the main defect of this section is the cords cracking.Now there have been some methods which can be used to texture defect detection: For texture defects detection, these prior methods must satisfy a prerequisite:they must use normal images of the same kind or similar texture as a reference. In fact, tires of different textures in the detecting process come randomly, so the above preconditions cannot be satisfied, thus it limits the application of the above method in the detection of cracks tread. Existing methods are influenced by the texture, they can only detect images within few kinds of textures, so now it is very necessary to find a detection method which is applicable to a variety of textures.The existing methods start from the point of view of extracting the texture feature of normal images. However, it is very difficult to give a general description of the model of the different types of textures. So this paper starts from the new view of extracting the characteristics of crack-defects and proposes a new method based on the projection of density of lines, thus it convert the problem of images into the problem of curve shapes. This new method use curve shapes shared by the crack defect images as the determination criteria, regardless of image texture interference, it do not need to use the normal image within the same or similar texture as a reference and is a texture-invariant detection method. Experimental results show that it has high detection efficiency on crack defects; what is more, it has low complexity, can be easily achieved, and can be used for online real-time detection of the crown crack.