| Summary: | 碩士 === 國立屏東科技大學 === 機械工程系所 === 95 === This research dedicated to advanced study, used Fourier transformation and Edge detection, on the inspection of surface defects of steel strip. The goal of this research was to reduce human error of visual inspection, or incapability of detecting surface defects in real-time, that render enormous loss of production cost.
The methodology of Fourier transformation, in this experiment, contributes to filtering out surface structure. A two dimensional Fourier transformation will transform images in two dimensional space domain into Fourier spectrum in frequency domain. After deriving the most suitable radius of the spectrum, images will be recovered to their original state by eliminating radius from the spectrum using Inverse Fourier transformation. The methodology of Edge detection, in this experiment, contributes to locating abrupt changes of image density. Among them, Sobel is one of the most fundamental and important algorithm.
Some common surface defects found on steel strip include holes, scratches, scales, slivers and roll marks. All images of such defects will firstly be categorized according to their properties and characteristics. After extraction of features from each image by image segmentation, image enhancement and denoising, a feature space module will be formed in order to reach the goal of automatic surface inspection.
|
|---|
