Automatsko prepoznavanje japanskog slikovnog pisma

undergraduate thesis

• Title (english): Automatic recognition of Japanese ideograms
• Author: Tessa Slovenc
• Mentor: Marko Subašić (mentor)
• Committee member: Marko Subašić (predsjednik povjerenstva)
• Committee member: Roko Franetović (član povjerenstva)
• Committee member: Franko Šikić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-06-27, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Automatsko prepoznavanje ručno pisanog ili tiskanog pisma je tehnika koja se razvija od početka postojanja računala, te se i danas otkrivaju razna poboljšanja u brzini i kvaliteti detekcije. Prepoznavanje može biti napravljeno za opću upotrebu, ili se može specijalizirati za određenu vrstu pisma ili količinu danih podataka.

Kod specijaliziranih tehnika, važno je razmatrati tip pisma koji se prepoznaje, te ostale informacije koje možemo dobiti iz danih znakova. Kod vrsta pisma koja imaju veliki skup potencijalnih znakova, potrebno je što više umanjiti broj usporedbi radi optimizacije brzine. To se u slučaju detekcije veće količine teksta može raditi prediktivnim metodama, a u slučaju manje se mogu koristiti metode dekompozicije.

U ovom radu je implementirano detektiranje ručno pisanog teksta japanskog kanji pisma, a optimizacija se vrši pomoću analize i usporedbe broja poteza nacrtanog znaka, te rastavom znaka na radikal i pod-znak kojim se istovremeno mogu odrediti djelomične vjerojatnosti više znakova koji sadrže identični radikal. Znakovima se dalje rekurzivno određuje vjerojatnost po potrebi. Također je napravljena analiza kvalitete programa ovisno o podesivim postavkama programa, koristeći ispitni slučaj koji se sadrži od tiskanog skupa svih znakova koje program potencijalno može prepoznati.

Abstract (english)

Automatic recognition of handwritten or printed text is a method that has been developing since the inception of computers, and various improvements in speed and quality of detection are still being discovered to this day. It can be made for general use, or specialized for a certain script or amount of information given.

With specialized techniques, it is important to consider the type of script that is recognized, and other data we can infer from the given characters. For writing systems that have a large set of potential characters, it is necessary to reduce the number of graphic comparisons in order to optimize speed. This can be done with predictive methods in the case of detection of a larger amount of text, and in the case of a smaller set, decomposition methods can be used.

In this paper, the detection of handwritten text of the Japanese kanji alphabet is implemented. The optimization is performed by analyzing and comparing the number of strokes of the drawn character, and by dividing the character into a radical and a sub-character which can simultaneously determine the partial probabilities of several characters containing an identical radical. The results determine if further recursive calculation of probability is needed. An analysis of the program quality was also made depending on the adjustable program settings, using a test case consisting of a printed set of all characters that the program can potentially recognize.

• Keywords: prepoznavanje ručno pisanog teksta; kanji; prepoznavanje japanskog pisma; dekompozicija; rastav pisma; modeliranje pisma; vjerojatnosno prepoznavanje
• Keywords (english): handwritten text recognition; kanji; Japanese character recognition; decomposition; character separation; character modeling; probabilistic recognition
• Language: croatian
• URN:NBN: urn:nbn:hr:168:348830
• 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: 2023-10-11 10:27:57