Image Character Recognition Research Articles

Incomplete handwritten Dongba character image recognition by multiscale feature restoration

Incomplete handwritten Dongba character often appears in heritage documents and its recognition is significant for heritage and philology. However, all previous methods always suppose that a complete Dongba character is used as input, and thus fail to achieve satisfactory performance when applied to incomplete Dongba character recognition. In this paper, an end-to-end network (DB2RNet) is proposed for incomplete handwritten Dongba character image recognition by multiscale feature restoration. Specifically, we first develop datasets that contain different levels of incomplete Dongba characters. A restoration module is proposed to restore the input incomplete Dongba character, and then a recognition module is employed to recognize Dongba character. By introducing an inter-module residual connection between the restoration module and recognition module, the DB2RNet can strengthen feature information transmission and boost the recognition performance. In addition, novel multiscale feature blocks are introduced, which can provide more effective texture and contextual information transmission for Dongba character image restoration, and thus yielding better restoration effects and better recognition results. Extensive experiments are conducted on Dongba character, Chinese character and Oracle character datasets and validate the effectiveness, superiority and robustness of our methods. Experiments results demonstrate that our proposed DB2RNet achieves competitive Dongba character restoration and recognition performance and outperforms the current state-of-the-art methods.

The meta-learning method for the ensemble model based on situational meta-task.

The meta-learning methods have been widely used to solve the problem of few-shot learning. Generally, meta-learners are trained on a variety of tasks and then generalized to novel tasks. However, existing meta-learning methods do not consider the relationship between meta-tasks and novel tasks during the meta-training period, so that initial models of the meta-learner provide less useful meta-knowledge for the novel tasks. This leads to a weak generalization ability on novel tasks. Meanwhile, different initial models contain different meta-knowledge, which leads to certain differences in the learning effect of novel tasks during the meta-testing period. Therefore, this article puts forward a meta-optimization method based on situational meta-task construction and cooperation of multiple initial models. First, during the meta-training period, a method of constructing situational meta-task is proposed, and the selected candidate task sets provide more effective meta-knowledge for novel tasks. Then, during the meta-testing period, an ensemble model method based on meta-optimization is proposed to minimize the loss of inter-model cooperation in prediction, so that multiple models cooperation can realize the learning of novel tasks. The above-mentioned methods are applied to popular few-shot character datasets and image recognition datasets. Furthermore, the experiment results indicate that the proposed method achieves good effects in few-shot classification tasks. In future work, we will extend our methods to provide more generalized and useful meta-knowledge to the model during the meta-training period when the novel few-shot tasks are completely invisible.

Aplikasi Pengenalan Pola Huruf Aksara Kaganga Menggunakan Jaringan Saraf Tiruan Dengan Metode Ekstraksi Fitur Geometri

This research aims to enable the Rejang Lebong community to relearn the Kaganga script from elementary school children to the elderly. The Kaganga script is only learned through books, teenagers or adults no longer want to learn the Kaganga script, so it is important for the Rejang Lebong people to maintain their very valuable culture. The method used is a qualitative method which is a method that focuses on in-depth observation. Data collection was carried out using an observational study conducted in the form of experimental research using a case study approach per individual for 7 (seven) image inputs consisting of 28 (two eight) Kaganga script letters to 7 (seven) different people to obtain the number of executions. The expected number of Kaganga characters is 196 letters. Testing the application by applying artificial neural networks with geometric feature extraction succeeded in character recognition of images in testing with a success rate of 92%.

Intelligent generation of combat simulation scenarios based on UML diagram recognition

The traditional generation of combat simulation scenarios often requires a manual understanding of conceptual scenarios and transformation into simulation scenarios. This method has the problems of long development time and high development threshold. Conceptual scenarios are usually visual Unified Modeling Language (UML) diagrams, so we can use artificial intelligence technology to extract key semantics from them, and automatically map the extracted semantics to simulation scenarios. This method is called the intelligent generation of combat simulation scenarios. To extract the key semantics from conceptual scenarios in UML form, we propose the UML diagram recognition method based on Keypoint Region-based Convolutional Neural Network (R-CNN). This method includes three parts: image character recognizer, primitive object detector, and image semantic extractor. First, we use optical character recognition (OCR) technology to achieve image character recognition. Second, we manually annotate the primitive target-detection dataset and propose a new primitive target-detection model—Keypoint R-CNN. This model considers the direction of connecting lines and realizes the detection of symbols and connecting lines. Third, we propose a targeted combined primitive detection and primitive relationship extraction method to extract the high-level semantics of UML diagrams. Then, we carried out experiments and evaluations on the self-made dataset. Compared with other methods, the F1 score of our method is improved by about 7%, and the Jaccard coefficient is improved by about 10%. Finally, we use a case study to implement the intelligent generation process of combat simulation scenarios using the UML recognition method we proposed. This case shows the operability and efficiency of our method. Our method greatly reduces the labor cost and development threshold of combat simulation scenario generation and improves the development efficiency of the combat simulation system.

A Robust Residual Shrinkage Balanced Network for Image Recognition from Japanese Historical Documents

Kuzushiji, as characters written in a cursive style, had been continuously used for both publishing and handwriting in Japan before the end of the 19th century. However, well into the twentieth century, most modern Japanese lost the ability to read Kuzushiji due to changes in writing systems. Therefore, how to develop advanced machine learning algorithms to identify Japanese historical characters has great significance in preserving cultural knowledge and promoting sustainable development in urban living. Unlike traditional digit image recognition, Kuzushiji image recognition is a more challenging task due to the complex script for cursive handwriting, multiple variations between characters, and imbalanced data. To address such issues, a novel deep learning‐based character recognition method called residual shrinkage balanced network (RSBN) is developed for classifying Japanese historical characters. A novel residual shrinkage structure with a soft threshold function is constructed to reduce redundant features using specially designed skip‐connected subnetworks. Then, multiple residual shrinkage modules are stacked together to learn hierarchical features from raw input. Furthermore, a novel class rebalancing strategy is further designed to improve learning performance from imbalanced data. The experiments are conducted on Japanese classical literature datasets. The results demonstrate that the proposed method obtains more than 89.01% testing accuracy on Kuzushiji‐MNIST dataset, which outperforms the existing deep learning‐based character image recognition methods. Moreover, without additional data augmentation techniques, the proposed approach also achieves 97.16% balanced accuracy on Kuzushiji‐49 dataset, which is the best one compared with existing published results. When considering the computational performance, the number of trainable parameters of the proposed method is competitive with that of the typical ResNet‐18.

Line Segmentation of Tibetan Ancient Books Based on A* Algorithm

Line segmentation is an important step in image character recognition. However, due to the problems of interline adhesion, overlapping, and skewing of document images, the effect of text line segmentation is not ideal. Therefore, further research on image line segmentation of Tibetan ancient books is urgently needed. The standard A* algorithm is not ideal for line segmentation of Tibetan ancient books, so this paper proposes a block-based A* algorithm for line segmentation of Tibetan ancient books. The method firstly preprocesses the image such as binarization and tilt correction and then performs horizontal projection. The obtained horizontal projection histogram is subjected to smoothing and peak detection to determine the position and number of text lines. Use the peak value to find the core area of the image text, extract the upper vowel and perform line attribution. Then, the image after removing the upper vowel is divided into 7 blocks, and each block is processed separately by the A* algorithm. When the algorithm finds the text line segmentation path, five cost functions are selected to calculate the cost from the starting point to the endpoint of the line segmentation, and the minimum cost path is found for text line segmentation. Finally, merge the segmented text lines, attribute the upper vowels, and rotate the text lines to obtain the final text line segmentation result. Experimental results show that our methos can effectively deal with the problems of text line adhesion and overlap, and achieve a good effect on text line segmentation.

A Multimedia Learning for Chinese Character Image Recognition via Human-Computer Interaction Network

As science and technology continue to develop, Chinese character image recognition technology is being used in a wide range of fields. This computer-based technology is a practical way of automatically recognizing images of text. Typically used in Chinese character education, it provides a new form of human–computer interaction for students. In addition, multimedia technology can provide a rich learning environment for students, which can present information about Chinese characters in the form of pictures, sounds, and videos, thus compensating for the disadvantages of learning Chinese characters by rote in the traditional educational process. The combination of Chinese character image recognition technology and multimedia technology can not only enrich the process of learning Chinese characters, but also promote students’ motivation to learn, thus providing a new and more modern approach to Chinese character education. Based on the study of Chinese character image recognition technology, this research combines it with multimedia information, to achieve the image recognition of Chinese character and multimedia information representation. The combined technology can provide significant references for course design and Chinese learners.

Full-Circuit Implementation of Transformer Network Based on Memristor

As an emerging in-memory element, memristor has been widely used in various neural network circuits to represent the weights and accelerate the calculation. However, the Transformer Network (TN), one of the most important models for machine vision and natural language processing in recent years, has not yet been full-circuit implemented using memristors due to the complex calculation process and data storage. In order to carry out the computation of the TN more efficiently, this work proposes a memristor-based full-circuit implementation of the TN capable of: 1) a memristor crossbar module to preserve the weights of the TN and perform the vector-matrix multiplications; 2) an analog signal memory module to store the analog signal directly in near-memory mode; 3) function circuit modules to achieve five transformations, namely Softmax, Layer Normalization, ReLU, Multiply-add and Residual; 4) a timing signal generation module to schedule operations of the circuit. The proposed TN circuit can complete all calculations directly based on the analog signal without using any analog-digital converter (ADC), digital-analog converter (DAC) and digital memory. In addition, character image recognition experiments are carried out in PSPICE to verify the functional correctness of the designed circuit. The corresponding signal retention rates of the analog memory, the performance of the whole circuit, and the non-idealities of the memristors are also analyzed. The results indicate that the circuit has advantages in terms of area overhead, energy efficiency and anti-noise.

Analysis of the Impact of Wire Resistance on Nano-scale Memristor Crossbar Array Implementing Perceptron Neural Network

In this work, the impact of wire resistance in pure memristor crossbar array is mathematically analysed and verified by the circuit simulation. The memristor crossbar without CMOS device is utilized for application of character image recognition, in which wire resistance is presented. Memristor crossbar circuit is analysed separately with respect to wire resistance on vertical line and wire resistance on horizontal line. The result shows that wire resistance on vertical line can be eliminated because they can be self-compensated. The simulation result agrees with the analysis. The variation of output voltage caused by wire resistance less affect the recognition rate of memristor circuit. On the other hand, when wire resistance on horizontal line is assumed to be 2.5Ω. The output voltages are varied remarkably. Such variation of output voltage degrades the recognition rate of memristor crossbar circuit. The interesting phenomenon is also investigated. The column that is close to the first column has less variation of output voltage, whereas the one that is far from the first column has much variation of output voltage. The result can be used in improving the memristor crossbar architecture, which can tolerate the impact of wire resistance. For example, if we want to increase the size of crossbar, we should increase the number of rows, rather than the number of columns.

Multi-type web image text detection based on the improved EAST algorithm

Web image character recognition has important application value in the commercial field and text detection is the basis of extracting this text information. Among many text detection algorithms, EAST (efficient and accurate scene text detector) algorithm has a good performance in natural scenes, but it is ordinary in web pictures. Based on this, aiming at the problem that the number of channels in the QUAD output layer of EAST algorithm is redundant, which leads to the low accuracy of web image text detection, it is modified as the distance from the point marked as a positive sample to the four vertices of the text box where it is located instead of the coordinate offset to the vertices. The improvement reduces from the original 8 channels to 4 channels, which restricts the optimization direction of the model, and then by setting a new loss function in order to improve the loss value of the edge area of the text box and the pixels in the image that are difficult to detect, the model is more suitable for the text detection of the web image. The experimental results show that the improved EAST algorithm significantly improves the accuracy of text detection in web images.

EKSTRAKSI FITUR AKSARA BALI MENGGUNAKAN METODE ZONING

Feature extraction is an important process in character recognition system. The purpose of this process is to obtain special feature from a character image. This paper is focuses on how to obtain special feature from a handwritten Balinese character image using zoning. This algorithm dividing Balinese character image into multiple regions, then a special feature on each region resulting the data extracted feature. The test result in this paper generates a various semantic and direction feature data. This is because this paper using handwritten Balinese character. Furthermore, the features that produced in this paper can be used on Balinese character image recognition process

Comparison of Feed Forward Network and Radial Basis Function for Detecting and Recognition of License Plate

One of the most important topics of intelligent transportation system is the License Plate Recognition (LPR). LPR systems have many potential applications in intelligent traffic systems, such as the payment of parking fee, highway toll fee, traffic data collection, traffic monitoring systems, traffic law enforcement, security control of restricted areas and so on. LPR was developed to identify vehicles by the contents of their license plates. The LPR system consists of three major modules: license plate extraction, segmentation and recognition of individual characters. This paper presents a study of applying the neural network approach for character image recognition. The new approach is tested on 400 samples of extracted license plate images captured in outdoor environment. The result yield 99.2% recognition accuracy, the method takes 1.6 seconds to perform the car plate recognition from vehicle's image. In order to decrease problems such as low quality and low contrast in the vehicle images, image recognition is done by the two different methods first is feed forward method and another is Radial Basis function and the best one is selected. The algorithm based on neural network which can quickly and correctly detect the region of license plate and the license plate detecting rate of success is 99.2%.

Exploration of Improved Methodology for Character Image Recognition of Two Popular Indian Scripts using Gabor Feature with Hidden Markov Model

character recognition plays an important role in the modern world. It can solve more complex problems and make the human's job easier. The present work portrays a novel approach in recognizing handwritten cursive character using Hidden Markov Model (HMM) . The method exploits the HMM formalism to capture the dynamics of input patterns, by applying a Gabor filter to a character image, observation feature vector is obtained, and used to form feature vectors for recognition. The HMM model is proposed to recognize a character image. All the experiments are conducted by using the Matlab tool kit. KeywordsCharacter Recognition, Feature Extraction,

Improving classification using a Confidence Matrix based on weak classifiers applied to OCR

This paper proposes a new feature representation method based on the construction of a Confidence Matrix (CM). This representation consists of posterior probability values provided by several weak classifiers, each one trained and used in different sets of features from the original sample. The CM allows the final classifier to abstract itself from discovering underlying groups of features. In this work the CM is applied to isolated character image recognition, for which several set of features can be extracted from each sample. Experimentation has shown that the use of CM permits a significant improvement in accuracy in most cases, while the others remain the same. The results were obtained after experimenting with four well-known corpora, using evolved meta-classifiers with the k-Nearest Neighbor rule as a weak classifier and by applying statistical significance tests.

ID Card Recognition Based on ArabicOCR System

Optical character recognition of Arabic language is a field of research that is socially very relevant and challenging.The social relevance lies on the fact that OCR is very important for many applications that need character recognition of images. Our system is Egyptian ID cards reader system which extracts important data from the ID card image, recognizes data and translates it into editable text on computer so it can be edited and saved. Then, the system can compare between a tested ID card and the database saved before. This paper extensively reviews the base line-based segmentation and DCT based feature extractor approaches used for building this special Arabic OCR system. It also reports the experimental results obtained so far showing the reliability of our system. Finally, we’ll show thatthe system works fast on the scientific Matlab library, as it needs about 16 seconds in average to process one ID card, and the system is expected to do better performance when transferring it from the academic phase to the product phase.

English

English

Research on Automatic Identification of Substation Circuit Breaker Based on Shape Priors

High-pressure circuit breakers are very important and it undertakes the disconnection and connection control of high voltage transmission lines. It is one of the equipment of substation daily inspection. State of breakers are judged by open and close characters label, so a shape-prior active contour model to realize state automatic recognition of breaker images collected by inspection robot is presented in this paper. Shape-prior active contour model combines the shape information with CV model to build energy functional model, then set up initial position curve by a priori knowledge and drives the curve evolution in minimize energy functional process, the curve position is the character label contour when energy functional shows minimum. We do experiment for the algorithm on different images, demonstrate that the algorithm based on known character contour, have good segmentation results of circuit breaker in the image character recognition accuracy and applicability when the circuit breaker character is actually partial occlusion, local deformation, scale changes.

Video image character recognition based on stroke-related weight

In order to extract the subtitle in the video image,a robust method was proposed.First,the image edge feature was adopted in caption location step,and the binarization method of text images with the edge information was given.Then,the method combined with projection and regional generation was used to locate a character.Finally,taking fully account of the topology of the text strokes,the stroke correlation among the adjacent sub-grids was determined and the stroke fuzzy membership was used to complete the elastic grid feature extraction.This method can effectively get the binary image of characters from a complex background image,ensure the stability and robustness in feature extraction.The experimental results show the method is effective,and its recognition rate has been up to 92.1%.

Off-line Handwritten Character Recognition Using Features Extracted from Binarization Technique

The choice of pattern classifier and the technique used to extract the features are the main factors to judge the recognition accuracy and the capability of an Optical Character Recognition (OCR) system. The main focus of this work is to extract features obtained by binarization technique for recognition of handwritten characters of English language. The recognition of handwritten character images have been done by using multi-layered feed forward artificial neural network as a classifier. Some preprocessing techniques such as thinning, foreground and background noise removal, cropping and size normalization etc. are also employed to preprocess the character images before their classification. Very promising results are achieved when binarization features and the multilayer feed forward neural network classifier is used to recognize the off-line cursive handwritten characters.

Character Image Recognition Based on the Improved BP Neural Network

This paper raises a kind of improved BP algorithm in order to compensate for some shortcomings which exist in traditional BP neural network. It has been applied to the recognition of character images. Computer simulation results demonstrate that it does bring about an ideal result.