Broersma, A. (2001) Modular Neural Network Classifier for Optical Occluded Character Recognition. Master's Thesis / Essay, Computing Science.
|
Text
Infor_Ma_2001_ABroersma.CV.pdf - Published Version Download (2MB) | Preview |
Abstract
Optical Character Recognition (OCR) or the automatic recognition of characters is being done for quite some years now. In general, very good results have been achieved here. In practice however, various forms of distortion can cause the recognition rate to decrease Automatic License Plate Recognition. For example dirt,drastically. One of the areas this applies to is ALPR or screws or tow bars can cause the license plate and its characters to become partially obscured. Another common problem are fenders that, because of the angle in which some images are taken, cause the top of the license plate to become obscured. Because of this, the characters become occluded, which means part of their top is cut off. In most cases this results in not recognizing the characters. By building a modular neural network, which includes networks that are especially trained with occluded characters, we can increase the recognition rate drastically. It proves that by using a classifier after the first ttempt to recognize these characters with the traditional network, we can improve our results on recognizing occluded characters without a loss in normal recognition.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Degree programme: | Computing Science |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 15 Feb 2018 07:29 |
| Last Modified: | 15 Feb 2018 07:29 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/8848 |
Actions (login required)
 |
View Item |