Research Article
Modeling and Implementing Two-Stage AdaBoost for Real-Time Vehicle License Plate Detection
Table 3
Performance comparison of some typical ALPR systems for LPD.
| Methods | Main procedures for license plate detection | Database size | Image conditions | LPD Rate | Processing time | Real time | Plate format |
| [10] | Sliding concentric windows, histogram | 40 images | 640 × 480 pixels (Different distances and weather, road) | 82.5% | — | — | Korean plates |
| [11] | Vertical edge, edge filtering, and morphological operation | 350 images | Different distances and weather and road | 95.2% | — | — | Iranian plates |
| [12] | Vertical edge detection, unwanted line elimination | 664 images | 640 × 480 pixels (various weather conditions, road) | 91.65% | 47.7 ms | Yes | Malaysian plates |
| [13] | Scan line, texture properties, color, and Hough transform | 332 images | 867 × 623 pixels (various illumination and different distances and road) | 97.1% | 0.53 s | No | Taiwanese plates |
| Our proposed method | Cascade AdaBoost algorithm and adaptive thresholding | 1800 images | 1280 × 720 pixels, various weather conditions and different illumination | 98.38% | 49 ms | Yes | Korean Plates |
|
|