Vision Based For Classification Of MIG Butt Welding Joint Defect Using Occurrence Matrices And Gray Absolute Histogram
This paper will introduced a new approach of vision which is enable to overcome the problems in the vision inspection systems. This system uses 2D gray pixels coocurrence matrix and gray absolute histogram of edge amplitude as the input features extract from the MIG butt welding joints. Images of th...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
International Journals Of Engineering And Sciences (IJENS)
2017
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/20878/2/IJMME2.pdf http://eprints.utem.edu.my/id/eprint/20878/ http://ijens.org/Vol_17_I_05/170105-9898-IJMME-IJENS.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This paper will introduced a new approach of vision which is enable to overcome the problems in the vision inspection systems. This system uses 2D gray pixels coocurrence matrix and gray absolute histogram of edge amplitude as the input features extract from the MIG butt welding joints. Images of the welding surfaces are captured using one CCD camera that is mounted on the top which is parallel with the work benches. The images are segmented and the 2D gray value coocurrence matrix consists of energy, correlation, homogeneity and contract, and absolute histogram of the characteristic feature in these images will be calculated. The same process will be applied in zooming image factor by 0.5 to calculated the next characteristic feature values. Finally both feature value is used as the input value in GMM and MLP classifier to classify the welds defect into three categories which are good weld, excess weld and
insufficient weld. Results are taken from the 18 MIG butt
welding joints samples were tested in overall accuracy
recognition rate for MLP is 94.4 % while for GMM is 83.3%. In terms of total computation time, the overall time for MLP is 1.96 m/s and GMM is 1.175 m/s. |
|---|