Back to search

Institutional Repository of Vilnius University

Simbolių atpažinimas paveiksliukų apdorojime /

Abstract

dc:description

Because of high inflation in recent years and all prices increasing it has become even more important to track your expenses. To solve this issue optical character recognition is suggested. In this bachelor thesis is written image pre-processing steps which have been extended with perspective distortion, image flattening transformations. Another character segmentation method uses real character and receipt widths. A hypothesis is proposed that convolutional neural network which has been trained with images segmented using the new method should have better accuracy when classifying characters compared to segmentation method that only uses histograms. For experiments, two databases were created, both containing 8264 characters that were produced using different segmentation methods. To test the effect of different segmentation approaches to accuracy of convolutional neural network the network was ran 200 for both types when randomly creating training, testing and validation sets from gathered database. The average accuracy of all these runs was taken as a final result. The Segmentation method ,,histogram” 5 had 92,32 % accuracy, while ,,histogram + width” had 92,18 %. From these results we can conclude that the hypothesis was not correct.

Degree

thesis:*
Grantor dc:publisher
Institutional Repository of Vilnius University
Year dc:date
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Kripas, Gediminas,

Rights

dc:rights
Statement dc:rights
  • info:eu-repo/semantics/openAccess
Language dc:language
lit

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
Vilnius University
Base URL
epublications.vu.lt/oai
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Kripas, Gediminas,. Simbolių atpažinimas paveiksliukų apdorojime /. Institutional Repository of Vilnius University, 2024. https://repository.vu.lt/VU:ELABAETD210646438&prefLang=en_US