OCR Engine Research Articles

Optical character recognition system using artificial intelligence

Abstract A technique termed optical character recognition, or OCR, is used to extract text from images. An OCR the system's primary goal is to transform already present paper-based paperwork or picture data into usable papers. Character as well as word detection are the two main phases of an OCR, which is designed using many algorithms. An OCR also maintains a document's structure by focusing on sentence identification, which is a more sophisticated approach. Research has demonstrated that despite the efforts of numerous scholars, no error-free Bengali OCR has been produced. This issue is addressed by developing an OCR for the Bengali language using the latest 3.03 version of the Tesseract OCR engine for Windows.

Efficient Vehicle Registration Recognition System: Enhancing Accuracy and Power Efficiency through Digital Image Processing

Number-plate recognition technology uses optical character recognition on images to read vehicle registration plates using OpenCV, PyTesseract OCR Engine, and YOLOv4 model. Images of the license plates are provided to the YOLOv4 model using a web app. The model is trained with the help of images from the Kaggle dataset. Then the number plate is extracted from the vehicle using the model. The string from the image is extracted using PyTesseract OCR Engine. Tested on several datasets this method gave us a success rate of 94%. This method can also be used in real-time incidents.

Neither Corpus Nor Edition: Building a Pipeline to Make Data Analysis Possible on Medieval Arabic Commentary Traditions

We have built a suite of tools in Python to proficiently analyze text reuse and intertextuality for a specific kind of set of medieval Arabic texts (commentaries) available in print. We take these printed editions, scan them, pre-process the images, give it to an OCR engine, clean the results, and store it in a data structure that mimics the explicit intertextual relation the texts have, and continue to perform data analysis on it. Digital approaches to medieval Arabic texts have either been at the micro-level in what has become known as a ‘digital edition’, i.e. the digital representation of one text, densely annotated, most commonly in TEI-XML, or it has been done at the macro-level in what is called a ‘digital corpus’, consisting of thousands of loosely encoded and sparsely annotated plain text files, accompanied by an entire infrastructure and high-performing software to perform broadly scoped queries. The micro-level generally is at the level of tens of thousands of words while the macro-level can be at the level of over a billion words. The micro-level is explicitly designed to be human readable first, while the macro-level is built to be machine readable first. At the micro-level, every little detail needs to be correct and in order, while at the macro-level a fairly large margin of error is still negligible as a mere rounding error. Amidst these levels we have been seeking a meso-level of digital analysis: neither edition nor corpus, but rather a group of texts at the level of hundreds of thousands to millions of words, with a small but perceptible margin of error, and a light but noticeable level of annotations, principally geared towards machine readability, but with ample opportunity for visual inspection and manual correction. In this paper we explain the rationale for our approach, the technical achievements it has led us to, and the results we so far obtained.

Build a Trained Data of Tesseract OCR engine for Tifinagh Script Recognition

This article introduces a methodology for constructing a trained dataset to facilitate Tifinagh script recognition using the Tesseract OCR engine. The Tifinagh script, widely used in North Africa, poses a challenge due to the lack of built-in recognition capabilities in Tesseract. To overcome this limitation, our approach focuses on image generation, box generation, manual editing, charset extraction, and dataset compilation. By leveraging Python scripting, specialized software tools, and Tesseract's training utilities, we systematically create a comprehensive dataset for Tifinagh script recognition. The dataset enables the training and evaluation of machine learning models, leading to accurate character recognition. Experimental results demonstrate high accuracy, precision, recall, and F1 score, affirming the effectiveness of the dataset and its potential for practical applications. The results highlight the robustness of the OCR system, achieving an outstanding accuracy rate of 99,97 %. The discussion underscores its superior performance in Tifinagh character recognition, exceeding the findings in the field. This methodology contributes significantly to enhancing OCR technology capabilities and encourages further research in Tifinagh script recognition, unlocking the wealth of information contained in Tifinagh documents

Applicability of OCR Engines for Text Recognition in Vehicle Number Plates, Receipts and Handwriting

Optical character recognition (OCR) is a computer vision technique that enables computers to recognize text from images. Text detection and computer vision have made significant advancements, leading to the development of various OCR technologies. However, selecting the most suitable OCR system for a specific purpose has become a challenging task. This research paper aims to explain the theoretical concepts and mathematical formulas underlying OCR engines, providing a better understanding of their functioning and performance. The analysis covers various aspects, including the theories, algorithms, and techniques employed by OCR engines. This paper presents experiments conducted on five different image categories: vehicle number plates, receipts, handwriting, symbols and plain text images. Evaluation metrics such as Character Error Rate (CER), Word Error Rate (WER), Insertion Error Rate (IER), Deletion Error Rate (DER), End-to-end recognition rate (EEER), Word Error Rate (WER), Recall, Precision, and F1-score were utilized. The findings reveal that OCR systems perform well on plain documents, with recall and F1-score values exceeding 0.85 and 0.8, respectively. However, there are still areas for improvement in OCR, which are discussed in detail in this paper. This research provides valuable insights for researchers, developers, and practitioners interested in employing OCR technology for their commercial projects.

Design and Implementation for BIC Code Recognition System of Containers using OCR and CRAFT in Smart Logistics

The BIC (Bureau International des Containers et du Transport Intermodal) Code is the identification code for ocean shipping containers and is crucial for logistics, transportation, and security. Accurate recognition of container BIC Code is essential for efficient import and export processes, authorities' ability to intercept illegal goods and safe transportation. Nevertheless, the current practice of employees recognizing and manually entering container BIC codes is inefficient and prone to error. Although automated recognition efforts have been made, challenges remain due to the aging of containers, manufacturing differences between companies, and the mixing of letters and numbers in the 11-digit combination. In this paper, we propose the design and implementation of a BIC Code recognition system using an open source-based OCR engine, deep learning object detection algorithm, and text detector model. In the logistics industry, various attempts are being made to seamlessly link the data required at each stage of transportation between these systems. If we can secure the stability and consistency of BIC Code recognition that can be used in the field through our research, it will contribute to overcoming the instability caused by false positives.

Automatic Car Number Plate Detection using Morphological Image Processing

One of the most common uses of computer vision, automatic number plate recognition (ANPR) is also a pretty well-explored subject with numerous effective solutions. Due to regional differences in license plate design, however, these solutions are often optimized for a specific setting. Number plate recognition algorithms are often dependent on these aspects, making a universal solution unlikely due to the fact that the image analysis methods used to develop these algorithms cannot guarantee a perfect success rate. In this research, we offer an algorithm tailor-made for use with brand-new license plates in Iraq. The method employs edge detection, Feature Detection, and mathematical morphology to find the plate; it was developed in C++ using the OpenCV library. When characters were found on the plate, they were entered into the Easy OCR engine for analysis.

Are Searches in OCR-generated Archives Trustworthy?

Abstract Digitised archives are revolutionary tools for research that, in a few seconds, generate results that earlier often took years to obtain. But do they provide all results for the terms searched for? The accuracy of searches was tested by performing sample searches of leading newspaper databases. The test revealed several weaknesses in the search process, including an average 18 percent error rate for single words in body text, and a far higher error rates for advertisements. Such high error rates encourage a critical look at the 20-year-old sector. Although these errors can be reduced by a re-digitation and with new improved OCR engines and new search algorithms, searches will nevertheless return manipulated results. In response, and to identify infringed bias and skewed representation, database owners need to provide thorough metadata to ensure source criticism.

Vehicle Number Plate Recognition Using Raspberry Pi

Vehicle number plate recognition (VNPR) systems have gained significant attention due to their wide range of applications in traffic management, parking enforcement, and security systems. This abstract presents an overview of a VNPR system that utilizes computer vision techniques and machine learning algorithms for license plate detection and recognition. The system is implemented using a Raspberry Pi platform, which provides a cost-effective and portable solution. The system follows a systematic approach to perform license plate recognition. Firstly, the camera captures images or video frames, which are then processed using the OpenCV library to enhance the visibility and quality of the license plates. License plate detection algorithms, employing edge detection or machine learning-based object detection techniques, are applied to locate and extract the license plate regions in the processed images. Following license plate detection, character segmentation techniques are utilized to isolate individual characters on the plate. Finally, integrated into the system, the Tesseract OCR engine recognizes the segmented characters and extracts the text information, effectively providing the license plate data.

Adaptive dewarping of severely warped camera-captured document images based on document map generation.

Automated dewarping of camera-captured handwritten documents is a challenging research problem in Computer Vision and Pattern Recognition. Most available systems assume the shape of the camera-captured image boundaries to be anywhere between trapezoidal and octahedral, with linear distortion in areas between the boundaries for dewarping. The majority of the state-of-the-art applications successfully dewarp the simple-to-medium range geometrical distortions with partial selection of control points by a user. The proposed work implements a fully automated technique for control point detection from simple-to-complex geometrical distortions in camera-captured document images. The input image is subject to preprocessing, corner point detection, document map generation, and rendering of the de-warped document image. The proposed algorithm has been tested on five different camera-captured document datasets (one internal and four external publicly available) consisting of 958 images. Both quantitative and qualitative evaluations have been performed to test the efficacy of the proposed system. On the quantitative front, an Intersection Over Union (IoU) score of 0.92, 0.88, and 0.80 for document map generation for low-, medium-, and high-complexity datasets, respectively. Additionally, accuracies of the recognized texts, obtained from a market leading OCR engine, are utilized for quantitative comparative analysis on document images before and after the proposed enhancement. Finally, the qualitative analysis visually establishes the system's reliability by demonstrating improved readability even for severely distorted image samples.

Enhancing Optical Character Recognition on Images with Mixed Text Using Semantic Segmentation

Optical Character Recognition has made large strides in the field of recognizing printed and properly formatted text. However, the effort attributed to developing systems that are able to reliably apply OCR to both printed as well as handwritten text simultaneously, such as hand-filled forms, is lackadaisical. As Machine printed/typed text follows specific formats and fonts while handwritten texts are variable and non-uniform, it is very hard to classify and recognize using traditional OCR only. A pre-processing methodology employing semantic segmentation to identify, segment and crop boxes containing relevant text on a given image in order to improve the results of conventional online-available OCR engines is proposed here. In this paper, the authors have also provided a comparison of popular OCR engines like Microsoft Cognitive Services, Google Cloud Vision and AWS recognitions. We have proposed a pixel-wise classification technique to accurately identify the area of an image containing relevant text, to feed them to a conventional OCR engine in the hopes of improving the quality of the output. The proposed methodology also supports the digitization of mixed typed text documents with amended performance. The experimental study shows that the proposed pipeline architecture provides reliable and quality inputs through complex image preprocessing to Conventional OCR, which results in better accuracy and improved performance.

Handwritten Documents Validation using Pattern Recognition and Transfer Learning

Handwritten documents in an Enterprise Resource Planning (ERP) system can come from different sources and usually have different designs, sizes, and subjects (i.e. bills, checks, invoices, etc.). Given these documents were filled manually, they have to be inspected to detect various kinds of issues (missing signature or stamp, missing name, etc.) before being saved in the ERP system or processed by an OCR engine. In this paper, the authors present a transfer learning approach to detect issues in scanned handwritten documents, using an award-winning deep convolutional neural network (InceptionV3) and different machine learning algorithms such as Logistic Regression (LR), Support Vector Machine (SVM) and Naive Bayes (NB). The experiment shows that the combination of InceptionV3 and LR got an accuracy of 91.8% for missing stamp detection. This can allow using this approach in an ERP system as an automatic verification procedure in a document processing flow.

BLPnet: A new DNN model and Bengali OCR engine for Automatic Licence Plate Recognition

The development of the Automatic License Plate Recognition (ALPR) system has received much attention for the English license plate. However, despite being the sixth-largest population around the world, no significant progress can be tracked in the Bengali language countries or states for the ALPR system addressing their more alarming traffic management with inadequate road-safety measures. This paper reports a computationally efficient and reasonably accurate Automatic License Plate Recognition (ALPR) system for Bengali characters with a new end-to-end DNN model that we call Bengali License Plate Network (BLPnet). The cascaded architecture for detecting vehicle regions before vehicle license plate (VLP) is proposed to eliminate false positives, resulting in higher detection accuracy of VLP. Besides, a lower set of trainable parameters is considered for reducing the computational cost, making the system faster and more compatible for a real-time application. With a Convolutional Neural Network (CNN) based new Bengali OCR engine and word-mapping process, the model is characters-rotation invariant, and can readily extract, detect and output the complete license plate number of a vehicle. The model feeding with 17 frames per second (fps) of real-time video footage can detect a vehicle with the Mean Squared Error (MSE) of 0.0152, and the mean license-plate-character recognition accuracy of 95%. While compared to the other models, an improvement of 5% and 20% were recorded for the BLPnet over the prominent YOLO-based ALPR model and the Tesseract model for the number-plate detection accuracy and time requirement, respectively. • We propose an end-to-end Deep Neural Network (DNN) model called BLPnet. • BLPnet operates in two separate detection phases minimizing false positive detection of number plate. • With a lower set of trainable parameters, BLPnet offers impressive computational efficiency. • BLPnet’s new CNN based OCR engine offers rotation-invariant character recognition with notably higher accuracy.

A Novel Memory and Time-Efficient ALPR System Based on YOLOv5.

With the rapid development of deep learning techniques, new innovative license plate recognition systems have gained considerable attention from researchers all over the world. These systems have numerous applications, such as law enforcement, parking lot management, toll terminals, traffic regulation, etc. At present, most of these systems rely heavily on high-end computing resources. This paper proposes a novel memory and time-efficient automatic license plate recognition (ALPR) system developed using YOLOv5. This approach is ideal for IoT devices that usually have less memory and processing power. Our approach incorporates two stages, i.e., using a custom transfer learned model for license plate detection and an LSTM-based OCR engine for recognition. The dataset that we used for this research was our dataset consisting of images from the Google open images dataset and the Indian License plate dataset. Along with training YOLOv5 models, we also trained YOLOv4 models on the same dataset to illustrate the size and performance-wise comparison. Our proposed ALPR system results in a 14 megabytes model with a mean average precision of 87.2% and 4.8 ms testing time on still images using Nvidia T4 GPU. The complete system with detection and recognition on the other hand takes about 85 milliseconds.

Resonate: Website on Text to Speech

Abstract: Synthesizing speech is quite complex as it is heavily reliant on language. Meaning, the language processing section in a TTS system inherently has the largest chunk of linguistic knowledge for a particular language. The technical as well as theoretical challenges faced while building such a high-quality system can be quite daunting and hard to navigate. To ensure that the system has relevant and updated linguistic information, one must make sure it has access to the most natural and unrestricted text to ensure quality and authenticity. We will also need extensive studies to achieve the same. At the heart of this software engine lies an OCR Engine (Optical Character Recognizer) which inherits crucial morphological operations required for image conditioning & transformation, accompanied with python libraries used for character classification. Further the processed textual data is transformed into speech signals using various Text-to-Speech synthesis techniques

A Machine Learning and NLP Approach for Analyzing Eligibility Based on Resume and CV

Today Jobs are curved to consider the wishes of the organization. Ideally, job design is destined to be a tired response to maintain the operational model and hence the functions needed. Every job should serve a purpose and contribute to the interests of business outcomes. All changes within the business should consider the impact on jobs and managers should re-design those jobs to suit. Good jobs may even contribute to motivating and inspiring generations to come. Dangerous job design can fail to make that motivation. Job descriptions are in all probability the foremost necessary employment-related documents for a firm. employment description underpins the utilization contract by commencing what management wants from an employee. They are thus a kind of the premise of all employer-employee interactions leading on to appraisal, development, pay and succession discussions. The application that will be presented will try to solve the above concern by acquiring the CV/Resume and related examination scores for the job description the candidate has applied to and by using Text Extraction Techniques the information extracted will be sent in the backend of the application and use the necessary data extracted and run it on the ML model and compare it with some of the top Resumes, Personality and Technical Knowledge evaluation and provide the candidate with good insights on various aspects. The proposed system will first extract the useful terms by using OCR engines from the scanned PDF of CV/Resume, analyze and split the data into categories like Education Qualifications, Experience, Skills, Projects, Candidate Description etc. And based on the data gathered, an evaluation for a particular job description will be done.

DETECTION AND RECOGNITION OF HINDI TEXT FROM NATURAL SCENES AND ITS TRANSLITERATION TO ENGLISH

India is a country with many cultures and if you travel from one place to another, you might find yourself in totally different culture. This also means the languages change from place to place in India and it gets very difficult to read signboards, shop names and even many other common things written in local languages. This can create problems for not only the travelers travelling from other countries but also the people who move withing the country from different regions. But most of the signboards, shop names or other landmarks mostly use English or Hindi in most of the regions. Here we propose a complete text detection & recognition as well as transliteration system that will help travelers read text written in Hindi on any signboards or shops and then transliterate that detected text into English. The proposed system is capable of detecting text written in Hindi language in natural environment using Progressive Scale Expansion algorithm and then transliterating the detected text into English language. Our proposed system can detect text in tough scenarios, and it can even detect curved text from natural images. Our system after detecting text region, extracts the text from the detected area using PyTesseract OCR engine and then the extracted text is further transliterated into English text with the help of seq2seq MultiRNN LSTM model which gives us accurate transliterations without losing the actual pronunciation of the original Hindi words. We use a synthetic dataset for Hindi Text images containing approx. 100000 for Text Detection and FIRE2013 dataset for transliteration. The overall system is evaluated using BLEU score.

Traffic Violation Data Security System

Since the road resources are limited and the number of vehicles on road are increasing at a tremendous rate, traffic monitoring has become a huge challenge and higher management costs have resulted in implementing a manual traffic monitoring system. Many traffic violations are difficult to capture with human observation. It is incontrovertible that a more intelligent and less cost scheme to solve the traffic management problem is necessary. In this paper, we aim to design a smart traffic management system using image processing, which can identify a vehicle’s license plate and add the violations in a blockchain. The computerised traffic monitoring system relies heavily on licence plate detection. The algorithm uses Canny Edge Detection to detect license plate. Tesseract is used as an OCR engine to recognize the characters from the detected license plate. OpenCV module is used with python for image processing and cloud storage is used to store the details of registered license plates. IoT cameras will be placed at every junction and the license plates of traffic violators will be captured. Then the license plate number along with the violation will be added to blockchain and a notification will be sent to violator. If the license plate is not registered, then the anomaly will be reported to Motor Vehicle Department.

In-depth analysis of the impact of OCR errors on named entity recognition and linking

AbstractNamed entities (NEs) are among the most relevant type of information that can be used to properly index digital documents and thus easily retrieve them. It has long been observed that NEs are key to accessing the contents of digital library portals as they are contained in most user queries. However, most digitized documents are indexed through their optical character recognition (OCRed) version which include numerous errors. Although OCR engines have considerably improved over the last few years, OCR errors still considerably impact document access. Previous works were conducted to evaluate the impact of OCR errors on named entity recognition (NER) and named entity linking (NEL) techniques separately. In this article, we experimented with a variety of OCRed documents with different levels and types of OCR noise to assess in depth the impact of OCR on named entity processing. We provide a deep analysis of OCR errors that impact the performance of NER and NEL. We then present the resulting exhaustive study and subsequent recommendations on the adequate documents, the OCR quality levels, and the post-OCR correction strategies required to perform reliable NER and NEL.

Blpnet: A New Dnn Model and Bengali Ocr Engine for Automatic License Plate Recognition

The development of the Automatic License Plate Recognition (ALPR) system has received much attention for the English license plate. However, despite being the sixth largest population around the world, no significant progress can be tracked in the Bengali language countries or states for the ALPR system addressing their more alarming traffic management with inadequate road-safety measures. This paper reports a computationally efficient and reasonably accurate Automatic License Plate Recognition (ALPR) system for Bengali characters with a new end-to-end DNN model that we call Bengali License Plate Network(BLPnet). The cascaded architecture for detecting vehicle regions prior to vehicle license plate (VLP) in the model is proposed to eliminate false positives resulting in higher detection accuracy of VLP. Besides, a lower set of trainable parameters is considered for reducing the computational cost making the system faster and more compatible for a real-time application. With a Computational Neural Network (CNN)based new Bengali OCR engine and word-mapping process, the model is characters rotation invariant, and can readily extract, detect and output the complete license plate number of a vehicle. The model feeding with17 frames per second (fps) on real-time video footage can detect a vehicle with the Mean Squared Error (MSE) of 0.0152, and the mean license plate character recognition accuracy of 95%. While compared to the other models, an improvement of 5% and 20% were recorded for the BLPnetover the prominent YOLO-based ALPR model and the Tesseract model for the number-plate detection accuracy and time requirement, respectively.