Research Article
Strip Steel Surface Defects Recognition Based on SOCP Optimized Multiple Kernel RVM
Table 4
Defect recognition performance comparison of SOCP-MKRVM, RVM and SVM.
| | Defect type | Scratches | Abrasions | Scar | Roller marks | Holes | Edge crack | Surface
upwarping | Pitting |
| | Testing sample number | 100 | 100 | 100 | 100 | 100 | 100 | 100 | 100 | | Error number (SOCP-MKRVM) | 0 | 0 | 1 | 2 | 0 | 0 | 1 | 2 | | Recognition accuracy (SOCP-MKRVM) | 100% | 100% | 99% | 98% | 100% | 100% | 99% | 97% | | Time consuming (ms) (SOCP-MKRVM) | 6.42 | 6.27 | 7.87 | 7.87 | 5.87 | 6.02 | 7.37 | 7.28 | | Error number (RVM) | 1 | 0 | 4 | 5 | 0 | 0 | 3 | 2 | | Recognition accuracy (RVM) | 99% | 100% | 96% | 95% | 100% | 100% | 97% | 96% | | Time consuming (ms) (RVM) | 6.81 | 6.34 | 7.89 | 7.88 | 6.03 | 6.95 | 7.89 | 7.31 | | Error number (SVM) | 4 | 1 | 6 | 5 | 0 | 1 | 5 | 7 | | Recognition accuracy (SVM) | 96% | 99% | 94% | 95% | 100% | 99% | 95% | 93% | | Time consuming (ms) (SVM) | 6.95 | 7.34 | 8.29 | 9.18 | 6.73 | 8.05 | 8.32 | 8.16 |
|
|
