Development and customization of in-house developed OCR and its evaluation

2018 ◽  
Vol 36 (5) ◽  
pp. 766-781
Author(s):  
Rajeswari S. ◽  
Sai Baba Magapu
Keyword(s):  
Character Recognition ◽  
Optical Character Recognition ◽  
Web Based ◽  
Content Type ◽  
Text Extraction ◽  
Manual Intervention ◽  
Key Phrases ◽  
Microsoft Office ◽  
Read Accuracy ◽  
Scanned Documents

Purpose The purpose of this paper is to develop a text extraction tool for scanned documents that would extract text and build the keywords corpus and key phrases corpus for the document without manual intervention. Design/methodology/approach For text extraction from scanned documents, a Web-based optical character recognition (OCR) tool was developed. OCR is a well-established technology, so to develop the OCR, Microsoft Office document imaging tools were used. To account for the commonly encountered problem of skew being introduced, a method to detect and correct the skew introduced in the scanned documents was developed and integrated with the tool. The OCR tool was customized to build keywords and key phrases corpus for every document. Findings The developed tool was evaluated using a 100 document corpus to test the various properties of OCR. The tool had above 99 per cent word read accuracy for text only image documents. The customization of the OCR was tested with samples of Microfiches, sample of Journal pages from back volumes and samples from newspaper clips and the results are discussed in the summary. The tool was found to be useful for text extraction and processing. Social implications The scanned documents are converted to keywords and key phrases corpus. The tool could be used to build metadata for scanned documents without manual intervention. Originality/value The tool is used to convert unstructured data (in the form of image documents) to structured data (the document is converted into keywords, and key phrases database). In addition, the image document is converted to editable and searchable document.

Multi-Oriented Text Extraction in Stylistic Documents

2015 ◽  
Vol 15 (01) ◽  
pp. 1550002
Author(s):  
Brij Mohan Singh ◽  
Rahul Sharma ◽  
Debashis Ghosh ◽  
Ankush Mittal
Keyword(s):  
Character Recognition ◽  
Optical Character Recognition ◽  
Morphological Operations ◽  
Document Images ◽  
Font Size ◽  
Text Extraction ◽  
Optical Character ◽  
Engineering Drawings ◽  
Flood Fill

In many documents such as maps, engineering drawings and artistic documents, etc. there exist many printed as well as handwritten materials where text regions and text-lines are not parallel to each other, curved in nature, and having various types of text such as different font size, text and non-text areas lying close to each other and non-straight, skewed and warped text-lines. Optical character recognition (OCR) systems available commercially such as ABYY fine reader and Free OCR, are not capable of handling different ranges of stylistic document images containing curved, multi-oriented, and stylish font text-lines. Extraction of individual text-lines and words from these documents is generally not straight forward. Most of the segmentation works reported is on simple documents but still it remains a highly challenging task to implement an OCR that works under all possible conditions and gives highly accurate results, especially in the case of stylistic documents. This paper presents dilation and flood fill morphological operations based approach that extracts multi-oriented text-lines and words from the complex layout or stylistic document images in the subsequent stages. The segmentation results obtained from our method proves to be superior over the standard profiling-based method.

scholarly journals Automated hierarchical classification of scanned documents using convolutional neural network and regular expression

2022 ◽  
Vol 12 (1) ◽  
pp. 1018
Author(s):  
Rifiana Arief ◽  
Achmad Benny Mutiara ◽  
Tubagus Maulana Kusuma ◽  
Hustinawaty Hustinawaty
Keyword(s):  
Neural Network ◽  
Convolutional Neural Network ◽  
Character Recognition ◽  
Optical Character Recognition ◽  
Regular Expression ◽  
Hierarchical Classification ◽  
Document Vector ◽  
Classification Prediction ◽  
Scanned Documents

<p>This research proposed automated hierarchical classification of scanned documents with characteristics content that have unstructured text and special patterns (specific and short strings) using convolutional neural network (CNN) and regular expression method (REM). The research data using digital correspondence documents with format PDF images from pusat data teknologi dan informasi (technology and information data center). The document hierarchy covers type of letter, type of manuscript letter, origin of letter and subject of letter. The research method consists of preprocessing, classification, and storage to database. Preprocessing covers extraction using Tesseract optical character recognition (OCR) and formation of word document vector with Word2Vec. Hierarchical classification uses CNN to classify 5 types of letters and regular expression to classify 4 types of manuscript letter, 15 origins of letter and 25 subjects of letter. The classified documents are stored in the Hive database in Hadoop big data architecture. The amount of data used is 5200 documents, consisting of 4000 for training, 1000 for testing and 200 for classification prediction documents. The trial result of 200 new documents is 188 documents correctly classified and 12 documents incorrectly classified. The accuracy of automated hierarchical classification is 94%. Next, the search of classified scanned documents based on content can be developed.</p>

scholarly journals Algorithm for auto annotation of scanned documents based on subregion tiling and shallow networks

2021 ◽  
Author(s):  
Komuravelli Prashanth ◽  
Kalidas Yeturu
Keyword(s):  
Neural Network ◽  
Character Recognition ◽  
Optical Character Recognition ◽  
High Performance ◽  
Performance Metrics ◽  
Training Data ◽  
Mean Average Precision ◽  
Average Precision ◽  
Squared Error Loss Function ◽  
Scanned Documents

<div>There are millions of scanned documents worldwide in around 4 thousand languages. Searching for information in a scanned document requires a text layer to be available and indexed. Preparation of a text layer requires recognition of character and sub-region patterns and associating with a human interpretation. Developing an optical character recognition (OCR) system for each and every language is a very difficult task if not impossible. There is a strong need for systems that add on top of the existing OCR technologies by learning from them and unifying disparate multitude of many a system. In this regard, we propose an algorithm that leverages the fact that we are dealing with scanned documents of handwritten text regions from across diverse domains and language settings. We observe that the text regions have consistent bounding box sizes and any large font or tiny font scenarios can be handled in preprocessing or postprocessing phases. The image subregions are smaller in size in scanned text documents compared to subregions formed by common objects in general purpose images. We propose and validate the hypothesis that a much simpler convolution neural network (CNN) having very few layers and less number of filters can be used for detecting individual subregion classes. For detection of several hundreds of classes, multiple such simpler models can be pooled to operate simultaneously on a document. The advantage of going by pools of subregion specific models is the ability to deal with incremental addition of hundreds of newer classes over time, without disturbing the previous models in the continual learning scenario. Such an approach has distinctive advantage over using a single monolithic model where subregions classes share and interfere via a bulky common neural network. We report here an efficient algorithm for building a subregion specific lightweight CNN models. The training data for the CNN proposed, requires engineering synthetic data points that consider both pattern of interest and non-patterns as well. We propose and validate the hypothesis that an image canvas in which optimal amount of pattern and non-pattern can be formulated using a means squared error loss function to influence filter for training from the data. The CNN hence trained has the capability to identify the character-object in presence of several other objects on a generalized test image of a scanned document. In this setting some of the key observations are in a CNN, learning a filter depends not only on the abundance of patterns of interest but also on the presence of a non-pattern context. Our experiments have led to some of the key observations - (i) a pattern cannot be over-expressed in isolation, (ii) a pattern cannot be under-xpressed as well, (iii) a non-pattern can be of salt and pepper type noise and finally (iv) it is sufficient to provide a non-pattern context to a modest representation of a pattern to result in strong individual sub-region class models. We have carried out studies and reported \textit{mean average precision} scores on various data sets including (1) MNIST digits(95.77), (2) E-MNIST capital alphabet(81.26), (3) EMNIST small alphabet(73.32) (4) Kannada digits(95.77), (5) Kannada letters(90.34), (6) Devanagari letters(100) (7) Telugu words(93.20) (8) Devanagari words(93.20) and also on medical prescriptions and observed high-performance metrics of mean average precision over 90%. The algorithm serves as a kernel in the automatic annotation of digital documents in diverse scenarios such as annotation of ancient manuscripts and hand-written health records.</div>

scholarly journals Comparison of text-image fusion models for high school diploma certificate classification

2020 ◽  
Vol 5 (1) ◽  
pp. 5-9
Author(s):  
Chandra Ramadhan Atmaja Perdana ◽  
Hanung Adi Nugroho ◽  
Igi Ardiyanto
Keyword(s):  
High School ◽  
Character Recognition ◽  
Optical Character Recognition ◽  
High School Diploma ◽  
Fusion Model ◽  
Text And Image ◽  
Digital Era ◽  
Adaptive Fusion ◽  
Scanned Documents ◽  
Better Than

File scanned documents are commonly used in this digital era. Text and image extraction of scanned documents play an important role in acquiring information. A document may contain both texts and images. A combination of text-image classification has been previously investigated. The dataset used for those research works the text were digitally provided. In this research, we used a dataset of high school diploma certificate, which the text must be acquired using optical character recognition (OCR) method. There were two categories for this high school diploma certificate, each category has three classes. We used convolutional neural network for both text and image classifications. We then combined those two models by using adaptive fusion model and weight fusion model to find the best fusion model. We come into conclusion that the performance of weight fusion model which is 0.927 is better than that of adaptive fusion model with 0.892.

scholarly journals Toward the optimized crowdsourcing strategy for OCR post-correction

2019 ◽  
Vol 72 (2) ◽  
pp. 179-197
Author(s):  
Omri Suissa ◽  
Avshalom Elmalech ◽  
Maayan Zhitomirsky-Geffet
Keyword(s):  
Character Recognition ◽  
Optical Character Recognition ◽  
Stage Structure ◽  
Historical Documents ◽  
Two Phase ◽  
Content Type ◽  
Historical Texts ◽  
Efficiency Measures ◽  
Historical Text ◽  
Amazon's Mechanical Turk

Purpose Digitization of historical documents is a challenging task in many digital humanities projects. A popular approach for digitization is to scan the documents into images, and then convert images into text using optical character recognition (OCR) algorithms. However, the outcome of OCR processing of historical documents is usually inaccurate and requires post-processing error correction. The purpose of this paper is to investigate how crowdsourcing can be utilized to correct OCR errors in historical text collections, and which crowdsourcing methodology is the most effective in different scenarios and for various research objectives. Design/methodology/approach A series of experiments with different micro-task’s structures and text lengths were conducted with 753 workers on the Amazon’s Mechanical Turk platform. The workers had to fix OCR errors in a selected historical text. To analyze the results, new accuracy and efficiency measures were devised. Findings The analysis suggests that in terms of accuracy, the optimal text length is medium (paragraph-size) and the optimal structure of the experiment is two phase with a scanned image. In terms of efficiency, the best results were obtained when using longer text in the single-stage structure with no image. Practical implications The study provides practical recommendations to researchers on how to build the optimal crowdsourcing task for OCR post-correction. The developed methodology can also be utilized to create golden standard historical texts for automatic OCR post-correction. Originality/value This is the first attempt to systematically investigate the influence of various factors on crowdsourcing-based OCR post-correction and propose an optimal strategy for this process.

Smart library book sorting application with intelligence computer vision technology

2020 ◽  
Vol ahead-of-print (ahead-of-print) ◽  
Author(s):  
Xiaohua Shi ◽  
Kaicheng Tang ◽  
Hongtao Lu
Keyword(s):  
Deep Learning ◽  
Radio Frequency Identification ◽  
Character Recognition ◽  
Optical Character Recognition ◽  
Conveyor Belt ◽  
Identification Algorithm ◽  
Content Type ◽  
Library Book ◽  
Blurred Image ◽  
Sorting System

PurposeBook sorting system is one of specific application in smart library scenarios, and it now has been widely used in most libraries based on RFID (radio-frequency identification devices) technology. Book identification processing is one of the core parts of a book sorting system, and the efficiency and accuracy of book identification are extremely critical to all libraries. In this paper, the authors propose a new image recognition method to identify books in libraries based on barcode decoding together with deep learning optical character recognition (OCR) and describe its application in library book identification processing.Design/methodology/approachThe identification process relies on recognition of the images or videos of the book cover moving on a conveyor belt. Barcode is printed on or attached to the surface of each book. Deep learning OCR program is applied to improve the accuracy of recognition, especially when the barcode is blurred or faded. The approach the authors proposed is robust with high accuracy and good performance, even though input pictures are not in high resolution and the book covers are not always vertical.FindingsThe proposed method with deep learning OCR achieves best accuracy in different vertical, skewed and blurred image conditions.Research limitations/implicationsMethods that the authors proposed need to cooperate and practice in different book sorting machine.Social implicationsThe authors collected more than 500 books from a library. These photos display the cover of more than 100 randomly picked books with backgrounds in different colors, each of which has about five different pictures captured from variety angles. The proposed method combines traditional barcode identification algorithm with the authors’ modification to locate and deskew the image. And deep learning OCR is involved to enhance the accuracy when the barcode is blurred or partly faded. Book sorting system design based on this method will also be introduced.Originality/valueExperiment demonstrates that the accuracy of the proposed method is high in real-time test and achieves good accuracy even when the barcode is blurred. Deep learning is very effective in analyzing image content, and a corresponding series of methods have been formed in video content understanding, which can be a greater advantage and play a role in the application scene of intelligent library.

scholarly journals AN ARCHITECTURE FOR AN INSCRIPTION RECOGNITION SYSTEM FOR SINHALA EPIGRAPHY

2016 ◽  
Vol 4 (12) ◽  
pp. 48-64
Author(s):  
Ruwanmini ◽  
Liyanage ◽  
Karunarathne ◽  
Dias ◽  
Nandasara
Keyword(s):  
Character Recognition ◽  
Optical Character Recognition ◽  
Geographical Location ◽  
Recognition System ◽  
Sri Lankan ◽  
Web Based ◽  
Optical Character ◽  
Scanned Images ◽  
Huge Challenge ◽  
Tracking Module

Sinhala Inscriptions are used as one of the major sources of getting information about ancient Sri Lanka. Revealing the Information from these inscriptions becomes a huge challenge for archeologists. This research paper focused on Sinhala character recognition in ancient Sri Lankan inscription. Our intention is to ease this process by developing a web based application that enable recognition of inscription characters through scanned images and store them in an inscription database.   Using this system people can track geographical location of inscriptions.  Epigraphist could be able to easily obtain Sinhala interpretation of Sri Lankan inscriptions via the optical character recognition feature in our system. Our work on this research project provides benefits to researchers in archaeology field, epigraphists and general public who are interested in this subject.   Inscription site tracking module will present a map that user can go around easily by tracking the locations of inscriptions. This paper presents the Architecture for this Sinhala Epigraphy system.

Sign in / Sign up

Export Citation Format

Share Document