Classification of weld bead defects based on image segmentation method


Anuar, Nurul Izah and Nordin, Mohd Hisham and Mohamed Noor, Ahamad Zaki and Akmal, Suriati and Idris, Syahril Anuar and Abdullah, Zulkeflee and M. Fauadi, Muhammad Hafidz Fazli and Awang, Nurfadzylah (2018) Classification of weld bead defects based on image segmentation method. Computer Science, 12 (4). pp. 51-60. ISSN 1985-3157

[img] Text
75e5e9058e014c1b8c80bcd4223bb34ce001.pdf - Published Version
Restricted to Repository staff only

Download (1MB)


Defect is an imperfection that could impair the worth and utility of a finished good. The defects show some disorder of the product and it is opposite the standard or criteria that have been stated. In defining and detecting the defects occur, many ways have been discussed and observed. However, the techniques or ways are not appropriate or not suitable for some condition and situation. In addition, welding process is one of the critical processes in detecting and defining defect to ensure the quality of the weld bead. To overcome the problem in detecting and defining defects, image processing is one of the methods in improving the process of detecting defects. The defects are classified based on automatic thresholding method that automates detecting and defining the defects. This study proposes a Decision Tree-based classification of weld bead defects through segmentation of image. The result obtained shows that the classification is effective in identifying the weld bead defects with 89% accuracy. For future work, the focus will be made to improve the detection accuracy by integrating suitable filters.

Item Type: Article
Uncontrolled Keywords: Welding, Defect; Weld Bead Defect; Image Processing; Otsu Method
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK4601-4661 Electric heating
Divisions: Faculty of Engineering and Technology (FET)
Depositing User: Ms Suzilawati Abu Samah
Date Deposited: 13 Mar 2021 19:28
Last Modified: 13 Mar 2021 19:28


Downloads per month over past year

View ItemEdit (login required)