Research Of Image Processing And Defects Recognition Based On X-ray Real-time-image System

With the development of industry, the industry ray real-time imagery examination technology is the new technology and the development direction of nondestructive. At present, Ray testing occupy forty percent in industry manufacture, therefore it share a very important position.Traditional ray testing is bases on film imaging but has some disadvantages, for example, it can't satisfy the requirements of real time imaging, testing and evaluation. Besides it has an expensive cost and inconvenience in image management. Aiming at this problem, the author study and design a real-time-image-processing image detecting and recognition system based on X-ray photography. After getting the X-ray image by collecting, much research about the digital image has been done in this paper, these research are image enhancement, preprocess of image, analysis of image, extracting defect's characteristic information and classified defect using BP neural network.Preprocess of image is the base of following image processing. Aiming at the problem of using traditional and single filtering method to remove mixed impulse noise and Gaussian noise, an adaptive hybrid filter algorithm based on the adaptive weighted median filter and the Gauss weighted mean filter is proposed. The algorithm first detects the type of the noise and then applies the adaptive weighted median filter and the Gauss weighted mean filter to removing impulse noise and Gaussian noise, respectively.The method of linear gray and nonlinear gray and histogram intensification aims at X-ray image's character of low gray contrast and faint edge are applied in this thesis, they intensified object's edge in image, to a certain extent it widen gray distributing, the X-ray image becomed clear and bright.Edge detector is an effective method to line out defect in character extracting. The paper analyzed the theory of some classical edge detecting algorithms, and presented a new edge detecting method that is based on morphological gradient. To get the better thresholding segmentation results, this paper presents a fuzzy set of multi-attribute thresholding segmentation algorithm which in view of the overall and local image attributes divided by the threshold. The emulation results indicate that the methods of edge detection on X-ray image are effective.Selection and extraction of defect's characteristic parameter are precondition for defect classification, it directly affect the result. Studied character of defect, author presented some eigenvectors that factually reflected defect's essential properties and their respective computation. They mainly included geometry characteristic, gray characteristic and image invariant moments. The self-organized and self-adaptive back propagation (BP) neural network algorithm is used to intelligently recognize the X-ray image of defect's classification. After analysising of the truss of the system and designing the system, the author design and implement an image-processing software system based on X-ray with visual c plus plus. The above-mentioned solution is applied to intelligent defect recognition of solar battery panels and has a perfect effect.