Očitavanje rukom pisanih korekcija odgovora u ispitnim obrascima

undergraduate thesis

• Title (english): Reading Handwritten Answer Corrections in Exam Forms
• Author: Mislav Gillinger
• Mentor: Tomislav Hrkać (mentor)
• Committee member: Tomislav Hrkać (predsjednik povjerenstva)
• Committee member: Karla Brkić (član povjerenstva)
• Committee member: Miroslav Popović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2017-07-13, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Cilj ovog rada bio je razviti sustav za detekciju i ekstrakciju ćelija u tablicama na slikama skeniranih ispitnih obrazaca, a na slova upisana u ćelije primijeniti postupak za raspoznavanje znakova. Metoda za prepoznavanje tablice radi na principu da traži uzorak piksela koji odgovara određenom vrhu. Za prepoznavanje slova brine se neuronska mreža koja na izlazu daje informaciju da je ulaz bio jedno od slova A-F ili prazna ćelija. Metoda zadužena za detekciju tablice i ćelija testirana je na 31 primjeru skeniranih tablica, te je uspješno detektirala vanjske linije 97% tablica. Konačni program testiran je na 28 tablica koje su sadržavale po nekoliko slova, od kojih su neka bila jasno, a neka deformirano napisana. Program je uspješno prepoznao sva slova u 18 tablica, dok je na ostalima prepoznao 0 do nekoliko slova, ali ne sve.

Abstract (english)

The goal of this paper was to implement a system for detection and extraction of table cells in scanned images of exam forms. After detecting a table, the system needs to recognize the letter written in each cell. The method used for table detection searches the pattern of pixels which represent one of the table edges. Artificial neural network takes care of letter recognition. On its output it gives information that the input was one of the letters A-F or an empty cell. Tested on a set of 31 table images, the table recognizing method successfully worked on 30 images, which gives 97% accuracy. The final program was tested on 28 tables, each containing a few correctly or deformed letters, and it successfully recognized all letters in 18 tables. In other tables only some of the letters were recognized.

• Keywords: detekcija tablice; umjetna neuronska mreža; ispitni obrazac; Keras
• Keywords (english): table detection; artificial neural network; exam form; Keras
• Language: croatian
• URN:NBN: urn:nbn:hr:168:168107
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2019-03-13 20:27:00