Baron W.J. (2002) Real-Time Camera-Based OCR. Master's Thesis / Essay, Artificial Intelligence.
|
Text
AI_Ma_2002_WJBaron.CV.pdf - Published Version Download (1MB) | Preview |
Abstract
This paper describes a system for performing real-time text detection and optical character recognition in fairly complex indoor scenery. It is initially intended to provide navigational cues to robot agents and it operates by coupling text detection utilizing localized measures (Mirmehdi) with neural network based optical character recognition techniques through a region growing process. Text detection using statistical measures has been shown to be strongly invariant to the colour, orientation and size (including illegibly small) of text regions while convolutional neural networks display the same relative invariance with respect to the position and scale of individual characters. The two techniques complement each other and ease the real-time requirement; text detection provides information on where text may be, region growing on its orientation and size, and convolutional networks identify the characters that are actually present. The camera-based reading process works as follows; various local properties of a (video) image such as the variance, edge count, density, and orientation are employed by a feed forward network to identify regions as text or non-text. Likely candidates are then selected using a fast constrained region growing technique and these are normalized and fed to a convolution feed forward neural network which performs optical character recognition. Depending on the application, the resulting characters may be processed further or simply be projected back onto the original location to restore the spatial orientation. While the current project targets robot vision, a robust system would have many applications such as providing an aid to the visually impaired or searching digital media, as well as offering some fascinating possibilities in the fields of augmented reality and wearable computing.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 15 Feb 2018 07:30 |
| Last Modified: | 15 Feb 2018 07:30 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/9056 |
Actions (login required)
 |
View Item |