The use of image processing for the detection and classification of defects has been a reality for some time in science and industry. New methods are continually being presented to improve every aspect of this process. However, these new approaches are applied to a small, private collection of images, which makes a real comparative study of these methods very difficult. The objective of this paper was to compile a public annotated benchmark, that is, an extensive set of images with and without defects, and make these public, to enable the direct comparison of detection and classification methods. Moreover, different methods are reviewed and one of these is applied to the set of images; the results of which are also presented in this paper.
2
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Automatic quality inspection of ferrite products is difficult as thier surfaces are dark and in many cases coverred with traces of grinding. A two-stage vision system for detection and measurement of crack regions was devised. In the first stage the regions with strong evidence for cracks are found using a morphological detector of irregular grightness changes with subsequent morphological reconstruction. In the second stage the feature-based k-Nearest Neighbors classifier analyzes the pixel indicated in the first stage. The classifier is optimized by using procedures of reclassification and replacement carried out on the reference set of pattern pixels to achieve a low error rate and a maximum speed of computation.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.