Text Localization Research Articles

Vessel identification based on automatic hull inscriptions recognition.

The identification of ships plays a crucial role in security and managing vessel traffic for ports and onshore facilities. Existing video monitoring systems help visually identify a vessel where other systems are not present or sufficient. Readable vessel plates and hull inscriptions of detected ships in the video stream allow using text location and recognition methods to obtain ships’ identification names or numbers. The obtained information can be then matched with available ship registers. The automation of the process has met many challenges related to the often-low quality of available video streams, heterogeneous regulations on the marking of ships, and the specifics of natural scene text recognition, such as quickly alternating imaging conditions or the interference of the background. The main contribution of this research is a method that can identify any type of vessel in an image that has visible inscriptions (name, registration number) placed on the hull and must be registered in a public registry. The proposed method works with low-quality images with inscriptions placed under different angles and different, readable sizes. Our method recognised 91% of vessels from our test dataset. Obtained identification times have not exceeded 1s. The quality and efficiency of the proposed solution indicate that it is suitable for practical implementation in onshore monitoring systems.

Efficient and Reliable Geocoding of German Twitter Data to Enable Spatial Data Linkage to Official Statistics and Other Data Sources

More and more, social scientists are using (big) digital behavioral data for their research. In this context, the social network and microblogging platform Twitter is one of the most widely used data sources. In particular, geospatial analyses of Twitter data are proving to be fruitful for examining regional differences in user behavior and attitudes. However, ready-to-use spatial information in the form of GPS coordinates is only available for a tiny fraction of Twitter data, limiting research potential and making it difficult to link with data from other sources (e.g., official statistics and survey data) for regional analyses. We address this problem by using the free text locations provided by Twitter users in their profiles to determine the corresponding real-world locations. Since users can enter any text as a profile location, automated identification of geographic locations based on this information is highly complicated. With our method, we are able to assign over a quarter of the more than 866 million German tweets collected to real locations in Germany. This represents a vast improvement over the 0.18% of tweets in our corpus to which Twitter assigns geographic coordinates. Based on the geocoding results, we are not only able to determine a corresponding place for users with valid profile locations, but also the administrative level to which the place belongs. Enriching Twitter data with this information ensures that they can be directly linked to external data sources at different levels of aggregation. We show possible use cases for the fine-grained spatial data generated by our method and how it can be used to answer previously inaccessible research questions in the social sciences. We also provide a companion R package, nutscoder, to facilitate reuse of the geocoding method in this paper.

A New Deep Wavefront Based Model for Text Localization in 3D Video

With the evolution of electronic devices, such as 3D cameras, addressing the challenges of text localization in 3D video (e.g., for indexing) is increasingly drawing the attention of the multimedia and video processing community. Existing methods focus on 2D video and their performance in the presence of the challenges in 3D video, such as shadow areas associated with text and irregularly sized and shaped text, degrades. This paper proposes the first approach that successfully addresses the challenges of 3D video in addition to those of 2D. It employs a number of innovations, among which, the first is the Generalized Gradient Vector Flow (GGVF) for dominant points detection. The second is the Wavefront concept for text candidate point detection from those dominant points. In addition, an Adaptive B-Spline Polygon Curve Network (ABS-Net) is proposed for accurate text localization in 3D videos by constructing tight fitting bounding polygons using text candidate points. Extensive experiments on custom (3D video) and standard datasets (2D video and scene text) show that the proposed method is practical and useful, and overall outperforms existing state-of-the-art methods.

Reading signboards for the visually impaired using Pseudo-Zernike Moments

It is necessary to provide essential daily support for the visually impaired, allowing them to locate themselves in an underground station, to change trains, to read signboards like those in doctors offices, law firms, shops, etc. In this paper, we present a Reading Signboards System (called RSbS) based on Pseudo Zernike Moments (PZMs) and stroke features from a video captured by a camera. The system is made up of three main steps: (i) signboard detection, (ii) text localization and (iii) filtering of non-textual regions. In a video, signboard region contains all the necessary information. Therefore, the signboard region should be located and segmented from the other parts of a frame that will be considered as the background. Signboards localization and segmentation are performed using Pseudo Zernike Moments. Text localization and extraction is done using PZMs segmentation which is based on the HSV color space and specifically on the Value channel. Non-textual regions will be filtered out using the outline function. System evaluation shows that it is robust to changes in lighting, low resolution and camera tilt, and it gives very few false positives compared to other approaches. In addition, the resulting images are transmitted to an optical character recognition engine (the best known is OCR) to extract the text which will be converted to speech using one of the existing APIs like Google Text to Speech API (gTTS API).

Domain Adaptive Scene Text Localization in Night View Images

Scene text localisation in a night is a challenging problem and a less explored problem in the area of robust reading. The day and night images have a different distribution. Hence, a domain shift exists from day to night images. This paper introduces a domain adaptation network (DAN) to learn the domain shift from day images to night images. The proposed sub-network DAN can be attached to any existing state-of-theart text detector to learn the domain shift from day to night. For the training of the DAN, synthetic data generation has been done by utilising the generative adversarial network. The proposed method has been extensively validated on the public as well as on the synthetic datasets. The proposed DAN improves the performance of the underline text localisation model with a margin of 1.4-4.0%, 0.1-1.0%, and 0.1-3.1% on LP Night Dataset, ICDAR2015, and Total-Text benchmarking datasets.

A Novel Approach of Text to Audio Conversion for Blindly Disable Using Raspberry Pi

One of the most difficult tasks for visually impaired people is reading. Blind persons may now recognise boundaries, read labels or certain currencies, and even read written, typed, or printed information, thanks to a technology developed by a group of academics. In this work, an assistance system is designed for blind persons who are visually challenged by using Raspberry Pi as a 32-bit micro controller that utilizes the character recognition technology for image to audio conversion; this work proposes a more effective solution. A camera is the heart of the system, serving as the primary vision for identifying product reports, bills, bank account information, menu boards, school handouts, product packaging, and medication bottle directions. The proposed technique consists of four major steps: i) object identification, ii) text localization, iii) text extraction, iv) text to speech conversion. An optical character recognition (OCR) after recognizing printed text with a technique, the image is digitally processed to extract the label with the use of an open CV library. After the Region of Interest is extracted from the congested backdrop, the text localization technique is utilized to find and extract the text. The ROI's text is retrieved and converted into speech. For training the words independently, a convolutional recurrent neural network technique is proposed.

Constructing theoretically informed measures of pause duration in experimentally manipulated writing

This paper argues that traditional threshold-based approaches to the analysis of pauses in writing fail to capture the complexity of the cognitive processes involved in text production. It proposes that, to capture these processes, pause analysis should focus on the transition times between linearly produced units of text. Following a review of some of the problematic features of traditional pause analysis, the paper is divided into two sections. These are designed to demonstrate: (i) how to isolate relevant transitions within a text and calculate their durations; and (ii) the use of mixture modelling to identify structure within the distributions of pauses at different locations. The paper uses a set of keystroke logs collected from 32 university students writing argumentative texts about current affairs topics to demonstrate these methods. In the first section, it defines how pauses are calculated using a reproducible framework, explains the distinction between linear and non-linear text transitions, and explains how relevant sections of text are identified. It provides Excel scripts for automatically identifying relevant pauses and calculating their duration. The second section applies mixture modelling to linear transitions at sentence, sub sentence, between-word and within-word boundaries for each participant. It concludes that these transitions cannot be characterised by a single distribution of “cognitive” pauses. It proposes, further, that transitions between words should be characterised by a three-component distribution reflecting lexical, supra-lexical and reflective processes, while transitions at other text locations can be modelled by two-component distributions distinguishing between fluent and less fluent or more reflective processing. The paper concludes by recommending that, rather than imposing fixed thresholds to distinguish processes, researchers should instead impose a common set of theoretically informed distributions on the data and estimate how the parameters of these distributions vary for different individuals and under different conditions.

ASAYAR: A Dataset for Arabic-Latin Scene Text Localization in Highway Traffic Panels

The extraction of text information from traffic panels is one of the challenging problems in computer vision. Although the past decade has seen a promising shift and important progress in object detection, few works and a limited number of datasets focus specifically on extracting text from traffic signs. To address the lack of data for text detection in traffic panels, especially those with Arabic scripts, this paper introduces a new multilingual and multipurpose dataset named ASAYAR. It consists of three sub-datasets: Arabic-Latin scene text localization, traffic sign detection, and directional symbol detection. The dataset contains 1763 images collected on different Moroccan highways, and annotated manually, using 16 object categories. The fully annotated ASAYAR images contains more than 20000 bounding box objects. The paper also investigates the usability of the dataset, by evaluating the performance of the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">state-of-the-art</i> algorithms for object and text detection. Experimental results show good detection scores, demonstrating the potential contribution of ASAYAR in the development of methods for text extraction from traffic panels.

Spatial-Frequency Redistribution-based Saliency Region Detection for Thai Text Localisation

Saliency region detection plays an important role in computer vision applications and related areas, such as human fixation, figure-ground separation, face detection, and image compression. Nevertheless, state-of-the-art methods can moderately detect the saliency regions of a psychological pattern dataset. Therefore, this paper proposes a spatial-frequency redistribution (SFR) method to improve the efficiency of the detection of the saliency regions. The proposed SFR method consists of two main procedures: (i) adaptive cosine image construction-and-implantation and (ii) saliency region detection using complete multiple-object-implanted images. The former procedure constructs an adaptive cosine image by using a cosine function based on an object structure and then individually implants it into each object detectable. This stage provides the first significant property, spatial redistribution, to the object implanted. The adaptive cosine image redistributes the objects for controllability. Then, the latter procedure transforms a complete multiple object-implanted image into the frequency domain. At this point, the second significant property, frequency redistribution, provides the simple technique for identifying and separating the target object from the unwanted objects and background. In this paper, these properties are referred to as spatial-frequency redistribution. This method was realized as a computer program, and then such program was tested with a psychological pattern dataset and a Thai text dataset. The experimental results, when compared with state-of-the-art methods, show that the proposed SFR method can achieve the clear detection of the attention region in the psychological pattern dataset. Moreover, the proposed method can achieve a F-value of 80.28% with a recall and precision of 76.32% and 85.23%, respectively, for Thai text localisation in the Thai text dataset.

A Text Location Method for Sign Based on Pixel-level Gradient Ray Images and Edge Contour Mapping

Compared with deep learning methods, traditional image processing methods have lower computational costs and have advantages for typical problems such as sign text, this paper proposes a sign text location method based on pixel-level gradient ray map and edge contour mapping. First, based on image pre-processing, the pixel-level gradient ray map is obtained by using the Stroke Width Transform. Then, using the Maximally Stable Extremal Region method to locate the irregular text candidate area in the ray map. At the same time, add a mask layer to filter the irregular candidate area, and then regularize the candidate area and perform Non-Maximum Suppression processing. Finally, the background non-text candidate area is processed by extracting the mapping of the closed edge contour of the image, and the remaining text area positioning candidate frame coordinates are mapped back to the original picture. The experimental results show that the method is simple and clear, the effect is remarkable, and it has good robustness to the sign of different sizes.

Linguistic and Cultural Analysis of Translation Techniques in the Process of Localization of an English Fairy Tale

Introduction. This article is devoted to the consideration of linguistic and cultural study of translation techniques in the process of localization of an English fairy tale for a Russianspeaking readership. The relevance of the research is determined by the need for a comprehensive study of the localization process of various ethno-cultural phenomena displayed during translation, as well as the stylistic transformations used in this process for the purpose of adapting the original text to the linguistics and culture of the target language.Methodology and sources. The typology of translation transformations by L.K. Latyshev was used as a reference classification, within which the author distinguished lexical, morphological, syntactic, mixed and stylistic transformations. The research was conducted in the context of a linguistic-stylistic approach to the analysis of the text using continuous sampling method, the methods of contextual, linguistic and cultural analysis, as well as the descriptive and comparative method, which implies identifying and understanding of stylistic transformations performed by the translator, and semantic and cognitive analysis, consisting in the study of verbal representations of the concept in the language, were used. The research material is presented by an English-language fairy tale and its translation into Russian (the book by K. Lewis “The Lion, the Witch and the Wardrobe” and the translation made by G.A. Ostrovskaya).Results and discussion. Stylistic transformations implemented during the localization of a fairy tale significantly change the original text in accordance with the requests of the culture of the target language. The most common stylistic transformations during localization of an English fairy-tale texts are related to such stylistic devices as metaphors and idioms, due to discrepancies in the system of images between languages, as well as exotisms, that underlie the key concepts of a particular culture, and allusions, since, due to the distance between cultures, they may be unknown to the recipient of the translation.Conclusion. A fairy tale preserves the wisdom, traditions and morals of a particular society, and therefore is a way of reflecting national identity. Localization, as a process of adaptation of works of this genre, is directly related to stylistic transformations, since they play a key role in achieving an adequate translation.

A Pipeline Approach to Context-Aware Handwritten Text Recognition

Despite concerted efforts towards handwritten text recognition, the automatic location and transcription of handwritten text remain a challenging task. Text detection and segmentation methods are often prone to errors, affecting the accuracy of the subsequent recognition procedure. In this paper, a pipeline that locates texts on a page and recognizes the text types, as well as the context of the texts within the detected region, is proposed. Clinical receipts are used as the subject of study. The proposed model is comprised of an object detection neural network that extracts text sequences present on the page regardless of size, orientation, and type (handwritten text, printed text, or non-text). After that, the text sequences are fed to a Residual Network with a Transformer (ResNet-101T) model to perform transcription. Next, the transcribed text sequences are analyzed using a Named Entity Recognition (NER) model to classify the text sequences into their corresponding contexts (e.g., name, address, prescription, and bill amount). In the proposed pipeline, all the processes are implicitly learned from data. Experiments performed on 500 self-collected clinical receipts containing 15,297 text segments reported a character error rate (CER) and word error rate (WER) of 7.77% and 10.77%, respectively.

Multi-Script Video Caption Localization Based on Visual Rhythms

ABSTRACT Localization of video caption plays an important role in information retrieval in multimedia applications. In this work, we present and evaluate a novel method for localizing video captions using visual rhythms, which enable the representation and analysis of a specific feature throughout the time. We build visual rhythms from the text location maps produced by general text localization methods that are far more common in the literature than caption-oriented ones. Then, we process the maps properly to keep only the captions, generating caption localization masks. To meet the need for a standardized and large dataset, we constructed a new one, where captions with thirteen different scripts are added to the video frames, generating a total of 221 videos with ground truth. Experiments demonstrate that our method achieves competitive results when compared to other literature approaches.

Disguising Commercial Intentions: Sponsorship Disclosure Practices of Mexican Instamoms

Influencers have established themselves as key allies for brands by cultivating a powerful public image to promote them. In the case of Instamoms, these collaborations can offer moms a means of achieving economic stability. In a country like Mexico, where the gender gap in the labor market remains a contentious issue, digital work represents an opportunity for women. The similarity between the organic content and commercial content created by these profiles has strengthened the presence of hybrid advertising. This means of advertising has not spelled the end for the original content, and audiences may struggle to spot ads if sponsorship is not disclosed properly. It is important for consumers to be able to identify ads so their persuasion knowledge can be activated. This article examines the commercial messages and types of disclosure used by Mexican Instamoms to inform their followers of the commercial nature of their collaborations. The types of disclosure are analyzed based on language, location, and type of text. After a content analysis of 10,135 stories and more than 330 posts, 40% and 47% of the sample, respectively, was identified as advertising content. The analysis revealed that less than 5% of the Instamoms sponsored content was tagged as such and that sponsorship disclosure does not form part of the usual protocol for influencer-brand collaborations in a country where no legislation is yet in place and the sector is making little effort to control these practices.

Сопространственность, геокультуры и (не)локальные тексты: к транссемиотике провинциальных текстов

The semiotics of provincial texts is the content core of provincial research, ensuring their cognitive growth. Comparative semiotic analysis of provincial texts is faced with the impossibility of direct correlation of the compared texts due to their various phenomenological features. The processes of localization of provincial texts and the procedures for their semiotic identification are also directly related to the concepts of geo-culture and geo-culturality. Any local text assumes the presence or absence of a trans-semiotic ‘shifter’ that connects or disables the text itself to a particular provincial register. The provinciality of the text is formed as a situational mental distancing from both the particular signs of the geo-culture of the represented territory and the signs of its immediate locality. Provincial texts create mutually generating and reproducing spatial ontologies, which can be defined as co-spatialities. The constantly reproducing processes of ‘provincialization’ of texts in the course of increasing their co-spatialities are the ontological basis and subject of trans-semiotics of provincial texts. In ontological terms, provincial texts are spaces of many newly formed divergent meta-symbolic assemblages. The aim of trans-semiotic studies of provincial texts is to study the specifics of transformations and transfigurations of newly emerging meta-symbolic provincial assemblages.

ЛИНГВОКУЛЬТУРОЛОГИЧЕСКИЕ ОСОБЕННОСТИ ОБРАЗА ПЕТЕРБУРГА В БИОФИКЦИОНАЛЬНОМ ПОВЕСТОВАНИИ ДЖ. КУТЗЕЕ «THE MASTER OF PETERSBURG»

The article explores linguacultural characteristics of Petersburg in J. Coetzee’s English-language novel "The Master of Petersburg", the postmodern biography of the outstanding Russian writer F. M. Dostoevsky. For creating verisimilitude of his biofictional narration, the author skillfully constructs the external linguacultural context with the help of linguistic markers of Russian cultural text localization. However, the English author does not seek to create an actual representation of Petersburg, fulfilling the city space with fictional urban objects. This intentional play with the semantic correlation between reality and fiction provokes bilingual readers, acquainted with the toponymy of the Russian city, to suffer from cognitive dissonance

A Text Tracking Method using Maximally Stable Extremal Regions and Speeded up Robust Features

Content-based image retrieval is an active research area because of vast applications of image and video collections. Text embedded in video can provide significant contribution in identifying the contents of the multimedia data and facilitate the process of video indexing, retrieval and analysis. Text tracking is a vital part of text extraction process. It can speed up the text extraction process and also improves the text localization accuracy for videos. This paper proposes a novel approach for text tracking in videos. This method experimentally proposes Maximally Stable Extremal Regions (MSER) and Speeded Up Robust Features (SURF) descriptors for text tracking in videos. MSER is used for interest point detection as Extremal regions are affine invariant, so it is a suitable option for text tracking which can undergo scale, rotation and translation changes. SURF is chosen as a feature descriptor because of its scale and rotation invariance, speed and robustness. Proposed technique is testified on two datasets; the first is designed to test the tracking methodology as part of this research and second is publicly available Youtube Video Text(YVT) dataset. Succeeding analysis on diverse datasets endures verification to the fact that the projected technique demonstrates visible improvements to text tracking in videos

Metallic debossed characters industrial online non‐segmentation identification based on improved multi‐scale image fusion enhancement and deep neural network

This paper investigates accurate industry online scene text recognition techniques for metallic debossed characters (MDCs). As MDCs have low contrast with its background and easy to interfere with corrosion rust, oxidate skin or specular reflection and so forth, a multi-scale image fusion algorithm is first proposed to restore the MDCs’ shape and appearance information to enhance the sample's contrast by using the MDCs' depth characteristic with the help of four-directional illumination. With this method, the three most important reflection-related picture quality evaluation standards, namely contrast, saturation, and exposure, can be enhanced simultaneously. Next, based on the fusion result, a U-shaped network based on VGG-16 network architecture is used for text localization. In order to improve network receptive fields, an improved post-processing output module that combines a text sequence score map for coarsening localization with a single-character score map for fine localization is proposed, so that the network's adaptability and accuracy for long texts are improved. Finally, a convolutional recurrent neural network is adopted to realize the recognition of sequence characters. The advantages and effectiveness of the proposed method are statistically analyzed with the data from a liquefied petroleum gas cylinder annual periodic inspection production line.

Deep-CNNTL: Text Localization from Natural Scene Images Using Deep Convolution Neural Network with Transfer Learning

Deep-CNNTL: Text Localization from Natural Scene Images Using Deep Convolution Neural Network with Transfer Learning

”I don’t know if I can share this.”

In modern business organizations, digital practices are enacted daily, often when sharing texts, which is crucial for knowledge management. How professionals experience digital text sharing is an issue that is often overlooked. In this paper, we focus on a relatively new aspect of business digital literacy: the literacy practice of digital text sharing in workplaces. Our analysis was conducted on ethnographic data from business organizations. The results show that sociomaterial aspects are enacted by professionals by discussing 1) the protection of borders of their own and other organizations, and 2) the status and digital location of texts. The analysis highlights two means of expressing agency that indicate conflicting norms: joking and showing strong emotions. The study places the hitherto backgrounded literacy practice of digital text sharing in workplaces in the foreground, proposes methods for studying this phenomenon, and highlights issues concerning digital text sharing that should be addressed by organizations.