Share Email Print
cover

Proceedings Paper

Weak scratch enhancement algorithm based on frequency domain characteristics
Author(s): Xiaobo Hu; Huanyu Sun; Shiling Wang; Menghui Huang; Jin Huang; Xiaoyan Zhou; Fengrui Wang; Hongjie Liu; Yuhao Zhou; Chong Liu; Dong Liu
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

Surface defect detection of optical components is commonly used in the fabrication of high quality optical components. The presence of defects reduces the local damage threshold of the component. When the laser energy is higher than the threshold, it will cause damage due to absorption of too much energy. Somethings it even causes cracking of the entire component. Therefore, the quality inspection of optical components and the formation of accurate quality assessment are technologies in the production and processing of high quality optical components. The data discussed in this paper all come from dark-field scatter imaging, which is a commonly used detection method. In dark field imaging, surface defects of the optical component will produce scattered light under the illumination of the light source, which is captured by the detector and reflected on the captured image. However, some scratches with a narrow width and insufficient depth are difficult to generate enough scattered light, which is usually caused by uneven force during polishing and grinding. It is an extremely weak straight line when reflected on the image. Because of the limited brightness, it is difficult to capture by the human eye or ordinary method of machine vision. The primary cause of weak scratches is that the signal intensity exhibited in the image is extremely low. Weak scratches’ gray value is usually only slightly higher than normal noise, and some are even submerged in noise, which leads to difficulties in identification. The enhancement algorithm proposed in this paper avoids the low gray value of weak scratches in the spatial domain, and successfully finds its characteristic with obvious discrimination, that is, its frequency domain characteristics. The pixels of weak scratches are usually connected stably and smoothly in a specific direction to form a straight line, which is their extremely continuous spatial continuity, although they are low in intensity. Through Fourier transform, the spatial continuity of weak scratches in a certain direction can be integrated, and the signal appearing in the vertical direction of the frequency domain is similar to strong scratches. Both of them show a very clear distinction. In contrast, other elements in the image, such as dust and pitting, are basically in the form of a circular shape, which is uniform in all directions and is submerged in the frequency domain by Fourier transform. In addition to scratches, dust, and pitting, there is another element that cannot be ignored in the detected image, which is noise. It usually includes the noise of the sensor itself, the noise caused by the light source and so on. Due to the uncertainty of the noise, the frequency distribution is also unpredictable and will also interfere with the final test results. In this paper, the inverse filtering method is adopted. The scratch information with discrimination is removed in the frequency domain, and the image after inverse transformation contains almost all noise, dust and pitting. After the difference with the original image and result, a pure scratch image was obtained at the expense of the vast majority of image. Since there is basically no noise interference, the detection threshold of the Hough line detection can be directly reduced, and the weak scratch can be easily captured. Without above work, a large amount of noise will lead to failure of scratch detection.

Paper Details

Date Published: 18 December 2019
PDF: 6 pages
Proc. SPIE 11338, AOPC 2019: Optical Sensing and Imaging Technology, 1133812 (18 December 2019); doi: 10.1117/12.2543103
Show Author Affiliations
Xiaobo Hu, Zhejiang Univ. (China)
Huanyu Sun, Zhejiang Univ. (China)
Shiling Wang, Zhejiang Univ. (China)
Menghui Huang, Zhejiang Univ. (China)
Jin Huang, China Academy of Engineering Physics (China)
Xiaoyan Zhou, China Academy of Engineering Physics (China)
Fengrui Wang, China Academy of Engineering Physics (China)
Hongjie Liu, China Academy of Engineering Physics (China)
Yuhao Zhou, Zhejiang Univ. (China)
Chong Liu, Zhejiang Univ. (China)
Dong Liu, Zhejiang Univ. (China)


Published in SPIE Proceedings Vol. 11338:
AOPC 2019: Optical Sensing and Imaging Technology
John E. Greivenkamp; Jun Tanida; Yadong Jiang; HaiMei Gong; Jin Lu; Dong Liu, Editor(s)

© SPIE. Terms of Use
Back to Top
PREMIUM CONTENT
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?
close_icon_gray