DocumentCode
2690136
Title
Illumination invariant measurement of mechanical dimensions using a multiple segmentation method
Author
Ohliger, Kevin ; Edeler, Torsten ; Hussmann, Stephan ; Mertins, Alfred
Author_Institution
Dept. of Electr. Eng. & Inf. Technol., Westcoast Univ. of Appl. Sci., Heide, Germany
fYear
2010
fDate
3-6 May 2010
Firstpage
1008
Lastpage
1013
Abstract
In many industrial processes the need for better quality assurance leads to a higher degree of process automation which often includes an optical inspection system. In our case the mechanical dimensions of printed micro structures on ceramic substrates have to be evaluated. Therefore an optical inspection system providing high resolution images with 12288 pixel × 14000 pixel was developed. The segmentation task in this setup is challenging caused by the huge amount of data and the different appearances of the micro structures depending on size, material, and humidity. For this purpose we present a multiple segmentation method and evaluate the results against common used segmentation methods on a number of different real images. Experimental results show that the proposed optical inspection system and algorithms can achieve an illumination invariant high resolution measurement of the micro structures.
Keywords
image segmentation; optical resolving power; high resolution images; illumination invariant measurement; mechanical dimensions; multiple segmentation method; optical inspection system; printed micro structures; process automation; Automatic optical inspection; Automation; Ceramics; Image resolution; Image segmentation; Lighting; Mechanical variables measurement; Optical materials; Pixel; Quality assurance; AMSER; MSER; ROI; adapted maximally stable extremal regions; high resolution; illumination invariant; maximally stable extremal regions; optical inspection system; region of interest; segmentation;
fLanguage
English
Publisher
ieee
Conference_Titel
Instrumentation and Measurement Technology Conference (I2MTC), 2010 IEEE
Conference_Location
Austin, TX
ISSN
1091-5281
Print_ISBN
978-1-4244-2832-8
Electronic_ISBN
1091-5281
Type
conf
DOI
10.1109/IMTC.2010.5488256
Filename
5488256
Link To Document