Local Template-based Segmentation Algorithm for Strip Steel Images 论文.pdfVIP

Local Template-based Segmentation Algorithm for Strip Steel Images ·1 · Local Template-based Segmentation Algorithm for Strip Steel Images Peng Tiegen , He Yonghui (Baoshan Iron Steel Co., Ltd., Research Institute, Shanghai, China) Abstract Inspired by the background subtraction method in motion detection algorithm, we proposed a new local template-based segmentation algorithm to detect the defects in steel strip images. The differences between the horizontal and the vertical templates were used to determine the location of the defect pixels. First, a block in original image is taken as template and moved it along the horizontal and vertical directions respectively to generate a relative motion between the template and the defects. Secondly, the defect pixel location is determined by the gravity center of the template differences. Experiments results show effectiveness of this segmentation algorithm. Key words local template, image segmentation, strip steel 1 Introduction Surface inspection system is very important in steel strip manufacture because the surface defects are important elements in determining the quality of steel roll products [1]. Commonly, defect segmentation algorithms employed in surface inspection system are gradient operator [2], such as Sobel, Kirsch etc. Inspired by the background subtraction algorithm used in motion detection, we proposed a new method to segment defects in steel strip images. The template was moved along the horizontal and vertical directions in an image to get the pixels gray differ