Open AccessOptical inspection of appearance faults for auto mirrors using Fourier filtering and convex hull arithmetic
Author(s) -
YuanShyi Peter Chiu,
Hong-Dar Lin,
Hsu-Hung Cheng
Publication year - 2021
Publication title -
journal of applied research and technology
Language(s) - English
Resource type - Journals
ISSN - 2448-6736
DOI - 10.22201/icat.24486736e.2021.19.4.958
Subject(s) - reflection (computer programming) , fault (geology) , convex hull , constant false alarm rate , computer science , process (computing) , artificial intelligence , curved mirror , regular polygon , hazard , hull , computer vision , false alarm , fourier transform , optics , mathematics , engineering , geometry , geology , physics , chemistry , organic chemistry , seismology , marine engineering , programming language , operating system , mathematical analysis
Auto mirrors are indispensable essential in reflection of objects behind the car and act a crucial part in driving security. In manufacturing stages of auto mirrors, certain tasks operated unusually will cause producing scratches, chips, pinholes, bubbles, damaged edges, the general surface and profile faults on auto mirrors. Those appearance faults sometimes will severely have an impact on standard of the mirror reflection and grow the driving hazard. At traditional examination of auto mirrors in manufacturing process, almost all works are performed by human examiners. Manual examination is simple to be disturbed by foreign objects reflected on the mirror surfaces and arouse causing mistaken determinations of fault examination. Thence, this study works toward investigating the automatic appearance fault detection of auto mirrors. We propose a fault enhancement technique based on Fourier high-pass filtering and the convex hull arithmetic to inspect appearance faults on auto mirrors. This approach only utilizes their own information of testing images to judge whether there are any irregular appearance changes without the need of standard patterns for matching. Experimental outcomes illustrate that the appearance fault detection rate reaches to 95.13%, and the false alarm rate decreases to 1.88%, and the correct classification rate attains to 98.11%.