Text detection and text extraction on images with Tesseract OCR

SAMOSIR, WILLIAM KAMDESU (2021) Text detection and text extraction on images with Tesseract OCR. Other thesis, Universitas Katholik Soegijapranata Semarang.

	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-COVER_a.pdf Download (2MB)
	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-BAB I_a.pdf Download (2MB)
	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-BAB II_a.pdf Restricted to Registered users only Download (2MB)
	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-BAB III_a.pdf Download (2MB)
	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-BAB IV_a.pdf Download (2MB)
	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-BAB V_a.pdf Download (2MB)
	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-BAB VI_a.pdf Download (2MB)
	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-DAPUS_a.pdf Download (2MB)
	Text 16.K1.0032-WILLIAM KAMDESU SAMOSIR-LAMP_a.pdf Download (2MB)

Abstract

In this day, documents is very important. Documents can be in the form of archives in notes or in the form of typed files, for documents in the form of notes, its is usually in print out or handwriting. For documents in printed or handwritten form, they usually have difficulty in the storage process because documents documents records can be damage, for example such as faded print ind and easily torn print paper. In modern time this can be overcome by using OCR(Optical Character Recognition), which is image processing that can detect text and text exctraction into a documents file format that can be edited and stored on a computer device for easier storage of documents text. OCR (optical character recogniton) is an image processing that can detect text and text extraction. through OCR (optical character recognition), the text of the document will be processed using the LSTM (long short term memory) algorithm to perform text detection and text extraction. LSTM (long short term memory) image will be pre-processed using tresholding which will help the process of detecting text. then the image will be processed in convolutional which will turn the image into a matrix, then the batch normalization process is carried out to add stability to the neural network (CNN). After that using Leaky Relu (Leaky Rectified Liniear Unit) is a type of activation function based on a ReLU, but it has a small slope for negative values instead of a flat slope as layer function , max pooling layer as the output or the final result of the detection. The image detected by the text character will be extracted into a document format in the form of a .txt file which is ready to be processed and stored. Based on the final results of OCR (optical character recognition) using the LSTM (long short term memory) algorithm, it has a satisfactory level of accuracy for text detection, while the process speed in recognizing character letters is good enough. The detected language recognition still has limitations due to the written character of the language

Item Type:	Thesis (Other)
Subjects:	000 Computer Science, Information and General Works > 004 Data processing & computer science
Divisions:	Faculty of Computer Science > Department of Informatics Engineering
Depositing User:	mr AM. Pudja Adjie Sudoso
Date Deposited:	18 May 2021 02:30
Last Modified:	18 May 2021 02:30
URI:	http://repository.unika.ac.id/id/eprint/25048

Actions (login required)

View Item