Tire Defect Detection Based On Integration Of Spatial And Frequency Domains

With the general improvement of people’s living standard, the Family Sedan has become the principle means of transport in recent years. With the corresponding, it makes the industry of tire continue to flourish in China. Along with the quality of life, people’s safety consciousness is generally improved, tend to choose a more secure product. In the situation of this consumption, it is necessary to improve the quality of tire. Restricted by production process, the tires may have all kinds of defects in the process of production, which directly affects the overall quality of tire. These quality problems will directly affect the experience of consumers and the personal safety, even threaten people’s life, so tires must be strictly inspected before delivery. At present, the automated digital processing of the tires is mainly rely on computer on the X-ray image of the tire for various defects detection. However, most tire defect detection algorithms are in the stage of theory, existing the problems that high time complexity, not up to standard detection accuracy, low algorithm robustness and so on, which cannot meet the practical requirements of tire enterprises.According to the actual characteristics of tire X-ray image, analyzing tire defect problem, this paper put forward a set of image processing method combined with spatial domain and frequency domain. According to the characteristics of the tire image, this method improves the existing image segmentation algorithms, using longitudinal dynamic threshold binary conversion method. On the basis of the segmentation, this paper proposes a new image preprocessing method based on gray level co-occurrence matrix theory. The method specifically satisfy tire image texture of horizontal and longitudinal, which has regular features. This paper proposes a new local smoothing filter algorithm based on the GLCM and greatly improves the time efficiency of the algorithm. The tire crown part of the complex image texture can be greatly simplified after processing, and boosted its normal area of sparse. On this basis, this paper further puts forward a tire detection algorithm based on the characteristics of tire image sparse. Tire image is transformed to frequency domain by frequency domain transformation, then we can get tire defect detection image using inverse transform.The main work of the thesis is as following: The experimental data from the Tire manufacturer of Linglong tire factory, adopted by the image processing library which have full kinds of defects and the type of defection. Compared to many algorithm produced in the environment of lab, our algorithm has a strong practical value. Our algorithm combines the individual merits of the surface defect detection in the field of spatial and frequency domain, through targeted improvement, adopting algorithm of bicycle images binary in the preprocessing. The tire is divided into several image segmentation with their respective characteristics, which will increase the detection accuracy of the algorithm. Filter algorithm is adopted to the simple texture, which will weaken the tire crown part of the normal regional complex texture characteristics. A new gray level co-occurrence matrix entropy is put forward and frequency detection is adopted, which will improve the detection efficiency and accuracy. Our algorithm in this paper has applied in the defect inspection of tire enterprises, a large number of testing data of X-ray image tires show that this algorithm has obvious advantages in efficiency and accuracy of tire defects detection, compared with other algorithm.