Design and development of intelligent 3d model recognition and inspection system in die stamping industry
Discussion on development of the vision system which focused on 3D model recognition and inspection is centered in this research. This research embarks on two main issues that pose as main problems that need to be resolved. The first issue is based on the fact that many industries produce different...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Technical Report |
| Language: | en en |
| Published: |
UTeM
2012
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/27072/1/Design%20and%20development%20of%20intelligent%203d%20model%20recognition%20and%20inspection%20system%20in%20die%20stamping%20industry.pdf http://eprints.utem.edu.my/id/eprint/27072/2/Design%20and%20development%20of%20intelligent%203d%20model%20recognition%20and%20inspection%20system%20in%20die%20stamping%20industry.pdf http://eprints.utem.edu.my/id/eprint/27072/ https://plh.utem.edu.my/cgi-bin/koha/opac-detail.pl?biblionumber=122604 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Discussion on development of the vision system which focused on 3D model recognition and inspection is centered in this research. This research embarks on two main issues that pose as main problems that need to be resolved. The first issue is
based on the fact that many industries produce different type of product in the same production line. To overcome this problem, cost-effective appearance-based method is proposed to replace traditional method which based on model of object. This
method is based on neural network that characterize each type of stamped part by using image moments. The second issue is based on the fact that many existing defect inspection algorithms are not feasible for Small and Medium Industries (SMis) due to complexity of techniques and equipments. To overcome this problem, a new defect inspection
algorithm for stamped part is developed and tested. The method is based on Canny edge detector to extract the surface defects based on optimal thresholding value. The data used for the experiments are five stamped parts which are collected using CCD camera. In 3D recognition results, it is observed that the first three image moments were able to increase neural network accuracy more than 95%. This results shows that
the appearance-based method is very promising for 3D recognition purpose in automated visual inspection. On the other hand, the inspection results show that pose of stamped parts is significantly reduced the algorithm performance. For each pose, only partial surface defects that is able to be detected. This shows that the specular reflectance that responsible for the shiny surface of stamped parts should be consider in future research. |
|---|