Table 7.
Recognition accuracy (EER) of a recognition system that uses only visible-light or thermal images for the recognition problem using SVM and CNN features with PCA (unit: %).
| SVM Kernel | Accuracies of Recognition Systems Using Single Image Types | |||||
|---|---|---|---|---|---|---|
| Using only Visible-Light Images | Using only Thermal Images | |||||
| EER | FAR | GAR | EER | FAR | GAR | |
| Linear | 17.064 | 10.000 | 73.261 | 16.114 | 10.000 | 73.744 |
| 15.000 | 80.906 | 15.000 | 82.518 | |||
| 17.080 | 82.952 | 16.120 | 83.892 | |||
| 20.000 | 85.607 | 20.000 | 88.079 | |||
| 25.000 | 88.583 | 25.000 | 91.758 | |||
| RBF | 17.489 | 10.000 | 68.958 | 17.596 | 10.000 | 69.433 |
| 15.000 | 79.090 | 15.000 | 78.042 | |||
| 17.500 | 82.523 | 17.600 | 82.409 | |||
| 20.000 | 84.627 | 20.000 | 85.149 | |||
| 25.000 | 87.055 | 25.000 | 89.830 | |||