DocumentCode :
3095133
Title :
The research of classification method of arc welding pool image based on invariant moments
Author :
Fei, Gao ; Kehong, Wang
Author_Institution :
Sch. of Mater. Sci. & Eng., Nanjing Univ. of Sci. & Technol., Nanjing, China
Volume :
3
fYear :
2011
fDate :
8-9 Sept. 2011
Firstpage :
73
Lastpage :
76
Abstract :
In welding, the arc welding images are always caused uncertain deformation and deflection due to the deformation of the weldment. Therefore, the image can not be corrected by traditional correction methods. Corresponding to different qualities of arc welding images, the morphological characteristics are decided by different factors of welding. In this paper, HU moment invariant theory is used to reduce the translation, deflection and scale variations of welding images, which are caused by the welding deformation and so on. First, extracting the seven moment invariants of the welding pool image; second, constructing the similarity parameter according to the seven moment invariants; finally, classifying the images by the similarity parameter. Experiments show that: the worse the quality of the weld seam is, the larger the values of the moment invariants are; the similarity parameter of the image based on the same moment theory is closest to a homogeneous image. It is concluded that the type of the tested images can be determined accurately using the classification method of arc welding pool images based on moment invariants.
Keywords :
arc welding; feature extraction; image classification; method of moments; production engineering computing; HU moment invariant theory; arc welding pool image classification; deflection; deformation; image extraction; invariant moments; Classification algorithms; Educational institutions; Feature extraction; Image classification; Pattern recognition; Shape; Welding; HU moment invariants; Similarity; arc welding;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Power Engineering and Automation Conference (PEAM), 2011 IEEE
Conference_Location :
Wuhan
Print_ISBN :
978-1-4244-9691-4
Type :
conf
DOI :
10.1109/PEAM.2011.6135018
Filename :
6135018
Link To Document :
بازگشت