Abstract | ||
---|---|---|
Impurity detection on tire X-ray image is an indispensable phase in tire quality control and the widely adopted manual inspection could not attain satisfactory performance. In this work we propose an idMKL method to automatically detect impurities by leveraging multiple kernel learning (MKL). idMKL first applies image processing techniques to separate different regions of a tire image and suppress their normal texture characteristics. As a result, candidate blobs containing both true impurities and false alarms are obtained. We extract different features from the blobs and evaluate their effectiveness in impurity detection. MKL is then employed to adaptively combine the features to maximize the detection performance. Experiments on thousands of images show that idMKL can well separate the blobs and achieves promising results in tire impurity detection. Moreover, idMKL has been adopted as a mean complementary to the manual inspection by tire factories and shown to be effective. |
Year | DOI | Venue |
---|---|---|
2017 | 10.1007/978-3-319-77380-3_33 | ADVANCES IN MULTIMEDIA INFORMATION PROCESSING - PCM 2017, PT I |
Keywords | Field | DocType |
Tire X-ray image,Multiple kernel learning,Defect detection | Computer vision,Impurity,Pattern recognition,Computer science,Multiple kernel learning,Image processing,Artificial intelligence | Conference |
Volume | ISSN | Citations |
10735 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 9 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Shuai Zhao | 1 | 0 | 0.34 |
Zhineng Chen | 2 | 192 | 25.29 |
Baokui Li | 3 | 0 | 0.34 |
Bin Zhang | 4 | 60 | 40.23 |