Glass Defects Based On Digital Image Recognition Technology, Online Testing Experimental Systems Research

Automated on-line detection technique of glass defects is a part of glass quality control, and it's also essential to stabilize glass quality. By this technique, glass defects can be distinguished and sorted. It is helpful to find the matters which exist in smelter, and guide technologists so that they can take some measures to improve glass quality. In addition, It is benefit to make division and sorting of glass more quickly, and that increase working efficiency and decrease working hardship. Presently, many corporations are doing research of this field in the world.In this paper, developing an on-line detection device of glass defects, the "real" glass production situation is simulated. With a CCD vidicon over glass and a lamp below glass, images with glass defects is gained, and then, by means of image processing methods, the objective images is processed, recognized and judged. The processing procedure is as follow: Firstly, glass image stored in memory is preprocessed, specifically, by means of gray processing and median filtering or neighborhood averaging, image noise is decreased, and the defects object edge is maintained. This prepare for next processing. Secondly, with image-processing methods of gray-level linear transformations, histogram equilibrium and sharpen, preprocessed image is photographically enhanced, contrast is increased, and appearance is improved. Thus, the processed image is more suitable for defects recognition and judgment. And then, after the image processed with proposed methods is treated with image segmentation, the object defects and glass background are separated. The defects features are abstracted. At last, some essential feature parameters are determined, and according to these parameters, defects is judged and recognized.In the study a vidicon image capture program is developed with Visual Basic system. It allows vidicon to capture with single frame or continuously, and then store the captured images in memory or the specified directory. Moreover, to process the captured images, the study work out the corresponding program based on Matlab language which is highly efficient and has the special image processing toolboxes and function sets.With similarity theory, feasibility and validity of the built model is analyzed. The effect of high or low resolution on glass defects recognition is discussed, then, optimal resolution is determined.