| Project/Area Number |
18K13955
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 25010:Social systems engineering-related
|
|---|
| Research Institution | Kanagawa University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2018-04-01 – 2020-03-31
|
| Project Status |
Completed (Fiscal Year 2019)
|
| Budget Amount *help |
¥4,030,000 (Direct Cost: ¥3,100,000、Indirect Cost: ¥930,000)
Fiscal Year 2019: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2018: ¥2,860,000 (Direct Cost: ¥2,200,000、Indirect Cost: ¥660,000)
|
|---|
| Keywords | 深層学習 / 外観検査 / 機械学習 |
|---|
| Outline of Final Research Achievements |
Photographing of a product is conducted with various light source in a visual inspection process. An inspection operator check an appearance of the product using the multiple images. A reduction in accuracy of defect detection is caused by accumulation of fatigue, stress, reduction in concentration etc. To efficiently perform an inspection work at an inspection place, a total inspection system using multiple images for supporting an inspection operator is required. This study developed an inspection system using simple image superposition for supporting operator. Demonstration experiment reveals that the proposed system is effective to improve the inspection accuracy for the defect whose images to be used as teacher data are adequately prepared.
For the detection of defects not prepared, measuring of similarity of feature vectors extracted from deep layer in CNN which learns from the above‐mentioned training data is effective.
|
| Academic Significance and Societal Importance of the Research Achievements |
サンプル数が十分にある欠点に対して、その検出精度を明らかにした点は、産業界における外観検査において、深層学習の利用を促進させる意味があり、社会的意義が認められると考えた。また、サンプル数の少ない欠点を検出するひとつのアプローチの有効性を示した点について、学術的意義があると考えた。
|
