DOCUMENT IMAGE BINARIZATION WITH FEEDBACK FOR IMPROVING CHARACTER SEGMENTATION

Two-step AI-aided Bayesian source identification of urban-scale pollution

The electricity tariffs available to customers in Poland depend on the connection voltage level and contracted capacity, which reflect the customer demand profile. Therefore, before connecting to the power grid, each consumer declares the demand for maximum power. This amount, referred to as the contracted capacity, is used by the electricity provider to assign the proper connection type to the power grid, including the size of the security breaker. Maximum power is also the basis for calculating fixed charges for electricity consumption, which is controlled and metered through peak meters. If the peak demand exceeds the contracted capacity, a penalty charge is applied to the exceeded amount, which is up to ten times the basic rate. In this article, we present several solutions for entrepreneurs based on the implementation of two-stage and deep learning approaches to predict maximal load values and the moments of exceeding the contracted capacity in the short term, i.e., up to one month ahead. The forecast is further used to optimize the capacity volume to be contracted in the following month to minimize network charge for exceeding the contracted level. As confirmed experimentally with two datasets, the application of a multiple output forecast artificial neural network model and a genetic algorithm (two-stage approach) for load optimization delivers significant benefits to customers. As an alternative, the same benefit is delivered with a deep learning architecture (hybrid approach) to predict the maximal capacity demands and, simultaneously, to determine the optimal capacity contract.

Hindi is the most widely spoken language in India, with more than 300 million speakers. As there is no separation between the characters of texts written in Hindi as there is in English, the Optical Character Recognition (OCR) systems developed for the Hindi language carry a very poor recognition rate. In this paper we propose an OCR for printed Hindi text in Devanagari script, using Artificial Neural Network (ANN), which improves its efficiency. One of the major reasons for the poor recognition rate is error in character segmentation. The presence of touching characters in the scanned documents further complicates the segmentation process, creating a major problem when designing an effective character segmentation technique. Preprocessing, character segmentation, feature extraction, and finally, classification and recognition are the major steps which are followed by a general OCR. The preprocessing tasks considered in the paper are conversion of gray scaled images to binary images, image rectification, and segmentation of the documents textual contents into paragraphs, lines, words, and then at the level of basic symbols. The basic symbols, obtained as the fundamental unit from the segmentation process, are recognized by the neural classifier. In this work, three feature extraction techniques-: histogram of projection based on mean distance, histogram of projection based on pixel value, and vertical zero crossing, have been used to improve the rate of recognition. These feature extraction techniques are powerful enough to extract features of even distorted characters/symbols. For development of the neural classifier, a back-propagation neural network with two hidden layers is used. The classifier is trained and tested for printed Hindi texts. A performance of approximately 90% correct recognition rate is achieved.

We examine the use of character image analysis coupled with contextual information in complex data gathering forms to identify and correct optical character recognition (OCR) system rejection and substitution errors. Segmented characters from a complex data gathering form are initially classified using an OCR engine based on a combination of Karhunen-Loeve transforms and a back-propagation neural network. Systems of equations are derived from the data gathering form to determine the values of characters rejected by the OCR engine and to verify the consistency of the data captured. If the OCR results for a single form are determined to be inconsistent with respect to the form's data relationships, a set of decision algorithms which incorporates a second neural network and uses additional character features is used to tag characters according to their likelihood of substitution error. Potential substitution errors are incrementally added to the set of OCR reject errors and are processed through dynamically selected systems of equations and search techniques which correct both error classes. We provide experimental results and determine the extent to which errors can be detected and corrected for various OCR error rates.

Convolutional neural network based methods have dominated object detection in recent years, which can be divided into the one-stage approach and the two-stage approach. In general, the two-stage approach (e.g., Faster R-CNN) achieves high accuracy, while the one-stage approach (e.g., SSD) has the advantage of high efficiency. To inherit the merits of both while overcoming their disadvantages, we propose a novel single-shot based detector, namely RefineDet++, which achieves better accuracy than two-stage methods and maintains comparable efficiency of one-stage methods. The proposed RefineDet++ consists of two inter-connected modules: the anchor refinement module and the alignment detection module. Specifically, the former module aims to (1) filter out negative anchors to reduce search space for the subsequent classifier, and (2) coarsely adjust the locations and sizes of anchors to provide better initialization for the subsequent regressor. The latter module takes (1) the refined anchors as the input from the former module with (2) a newly designed alignment convolution operation to further improve the regression accuracy and predict multi-class label. Meanwhile, we design a transfer connection block to transfer the features in the anchor refinement module to predict locations, sizes and class labels of objects in the object detection module. The multi-task loss function enables us to train the whole network in an end-to-end way. Extensive experiments on PASCAL VOC and MS COCO demonstrate that RefineDet++ achieves state-of-the-art detection accuracy with high efficiency.

Character segmentation is a necessary preprocessing step for character recognition in many OCR systems. It is an important step because incorrectly segmented characters will not be recognized correctly. The most difficult case in character segmentation is cursive script. The scripted nature of Arabic written language poses some high challenges for automatic character segmentation and recognition. The authors present a new Character Segmentation Algorithm (ACSA) of Arabic script. The developed segmentation algorithm yields the splitting up of isolated handwritten words in perfectly separated characters. It is based on topological rules, which are constructed at the feature extraction phase. To increase ACSA’s performances, it was combined it with an Arabic characters recognition system, RECAM.

License plate detection and character recognition have unveiled new possibilities and challenges in the field of intelligent transport system. Numerous algorithms have been proposed regarding license plate localization and tracking, character segmentation and character recognition. License plate character recognition is still an active area of research specially in terms of processing complexity. A handful of feature extraction algorithms including vertical and horizontal histogram, center of gravity, structural characteristics, modified edge maps, image projection, multi zoning, adaptive multi zoning, concavities measurement, gradient directional features, median gradient features, longest-run features and momentum features have been proposed. Neural network followed by ensemble classifier shows recognition rate in the high nineties. However, ensemble classifier based on multi-feature is burdened with computational intricacy and lethargic processing. This paper presents a perspective view based license plate character recognition algorithm that copes with some of the inherent problems regarding character recognition. The proposed algorithm is resilient against slant, skewness and obscurity of image while limiting processing time and memory usage. Our algorithm tracks the principal curvature of the segmented image in terms of magnitude and inter-impulse distance. A simple back propagation neural network offers recognition rate of 93% with faster training and validation time. A thorough analysis of the algorithm and meticulous details of implementation are presented throughout this paper.

Automatic Number Plate Recognition system is an application of computer vision and image processing technology that takes photograph of vehicles as input image and by extracting their number plate from whole vehicle image , it display the number plate information into text. Mainly the ANPR system consists of 4 phases: Acquisition of Vehicle Image and PreProcessing, Extraction of Number Plate Area, Character Segmentation and Character Recognition. The overall accuracy and efficiency of whole ANPR system depends on number plate extraction phase as character segmentation and character recognition phases are also depend on the output of this phase. Further the accuracy of Number Plate Extraction phase depends on the quality of captured vehicle image. Higher be the quality of captured input vehicle image more will be the chances of proper extraction of vehicle number plate area. The existing methods of ANPR works well for dark and bright/light categories image but it does not work well for Low Contrast, Blurred and Noisy images and the detection of exact number plate area by using the existing ANPR approach is not successful even after applying existing filtering and enhancement technique for these types of images. Due to wrong extraction of number plate area, the character segmentation and character recognition are also not successful in this case by using the existing method. To overcome these drawbacks I proposed an efficient approach for ANPR in which the input vehicle image is pre-processed firstly by iterative bilateral filtering , adaptive histogram equalization and number plate is extracted from pre-processed vehicle image using morphological operations, image subtraction, image binarization/thresholding, sobel vertical edge detection and by boundary box analysis. Sometimes the extracted plate area also contains noise, bolts, frames etc. So the extracted plate area is enhanced by using morphological operations to improve the quality of extracted plate so that the segmentation phase gives more successful output. The character segmentation is done by connected component analysis and boundary box analysis and finally in the last character recognition phase, the characters are recognized by matching with the template database using correlation and output results are displayed. This approach works well for low contrast, blurred, noisy as well as for dark and light/bright category images. The comparison is done between the ANPR with Adaptive Histogram Equalization and Iterative Bilateral Filtering that is the proposed approach and the existing ANPR approach using metrics: MSE, PSNR and Success rate.

There are more than 1000 languages and 14 scripts used by 112 million people in India. All of these scripts divide the document in three parts: Text block, Image block, and Table block. In 21st century, there is a need, obvious reasons, to convert these old printed documents in digital form. Converting them manually is a huge and difficult task. Further it is prone to human errors. Another automated technique is to use Optical character recognition (OCR) system to convert the entire printed document image into editable document. In this paper, an effort has been made to develop OCR technique which converts the printed document into editable document. Firstly a scanned document is preprocessed for noise and skew correction. It is then followed by text-non text classification. Then text line detection has to be performed in the text area. There is no method available which can detect the text line if the image contains the multicolumn text area. In this paper the main contribution is to detect the blocks and detect the text lines in these detected blocks. The technique which can extract the text lines in image document is presented here. After extraction of text lines, word segmentation, character segmentation, and template matching can be performed.

The objective of this work is the development of a highly effective Wiener-Physically-Informed-Neural-Network (W-PINN) modeling methodology for systems and processes with highly nonlinear dynamic and highly nonlinear static behavior. This approach has two basic stages. The first stage estimates all the unknown linear dynamic modeling coefficients. With these estimates fixed, the second stage estimates all the unknown static coefficients in an artificial neural network (ANN) framework. When the ANN is a nonlinear function, as in this work, the fitted model is a nonlinear dynamic and nonlinear static structure. In a previous one stage, 60-minute forecast modeling case, the dynamic outputs (vij, where i = the subject number and j = the sample number) were estimated in Excel® for eleven (11) type 1, freely-living, diabetes data sets (cases) of approximately two weeks of data, yielding an average input-only sensor glucose concentration (SGC) validation fit () of 0.68. With these vij’s fixed, this new approach obtains final, two-stage (TS) W-PINN, 60-minute forecast fitted models, for each of the 11 cases, in two ways. One two-stage approach uses the JMP® ANN toolbox for second stage modeling and achieves an average input-only SGC of 0.74 and maximum of 0.84. The second two-stage approach uses a novel ANN methodology coded in Python® and achieves an average input-only SGC of 0.82 and maximum of 0.93. Incorporating bias correction, using current and past SGC residuals, the Python® estimator improved the average from 0.82 to 0.87 with the maximum still 0.93.

Optical Character Recognition (OCR) deals with automated recognition of characters that are in the format of digital image. OCR refers to the process by which scanned images are electronically processed and converted to an editable document. Handwritten and printed texts are the primary research areas of an OCR. Many OCR systems are commercially available for English and Arabic characters but there is still no recognition system available which yields higher recognition rate even though the scanned images are of high quality. The general framework of a Tamil OCR in the literature involves: preprocessing, line segmentation, word segmentation, character segmentation, feature extraction and recognition of characters. OCR for printed Tamil documents poses challenge owing to: one line may have different font styles, presence of pictures, multi columns, touching of adjacent characters, presence of broken characters, low print quality and complex layout. Furthermore, when comparing 26 alphabets in English, Tamil language has 247 alphabets which makes the recognition more difficult. There are few OCRs for Tamil language that are freely available with a moderate recognition rate as the performance comparisons of such OCRs are not available on a benchmark dataset. In this paper we compare OCRs for printed Tamil texts on four different types of documents: books, magazines, newspapers and pamphlets. Furthermore we propose a post-processing error correction technique to the tested OCRs which reduces the overall mean error rate by nearly 10% on those four categories.

Due to the presence of high volume of noise in text documents, it becomes very difficult to achieve high accuracy while performing text line segmentation. Hence, approaches which are dependent on the performance of the line segmentation stage, also suffers. Also due to the variability of noise patterns, the text contents get distorted and sometimes it results in broken strokes of a character. Thus, character level recognition becomes a challenging task for noisy documents. Our proposed method is aimed to address the aforesaid challenges. Since, we are considering the input image to be of noisy kind, a convolutional neural network (CNN) architecture is introduced to initially identify whether the input image contains noise or not. We consider this as a two-class problem and try to identify the respective classes, i.e. noisy or clean. Then, we apply the proposed piecewise projection profile feature and an adaptive region growing based two-stage algorithmic approach which initially identifies the text line upper and lower boundaries in a noisy text document image and then try to regroup the broken strokes of a character to enhance the character recognition accuracy. The proposed method has been tested on a large size dataset containing noisy documents and also different quality measures are computed to establish the effectiveness of the proposed method. From the measures, it can be noted that the text line segmentation accuracy is comparable to other similar state-of-the-art results.

This dissertation introduces a new system for handwritten text recognition based on an improved neural network design. Most of the existing neural networks treat mean square error function as the standard error function. The system as proposed in this dissertation utilizes the mean quartic error function, where the third and fourth derivatives are non-zero. Consequently, many improvements on the training methods were achieved. The training results are carefully assessed before and after the update. To evaluate the performance of a training system, there are three essential factors to be considered, and they are from high to low importance priority: 1) error rate on testing set, 2) processing time needed to recognize a segmented character and 3) the total training time and subsequently the total testing time. It is observed that bounded training methods accelerate the training process, while semi-third order training methods, next-minimal training methods, and preprocessing operations reduce the error rate on the testing set. Empirical observations suggest that two combinations of training methods are needed for different case character recognition. Since character segmentation is required for word and sentence recognition, this dissertation provides also an effective rule-based segmentation method, which is different from the conventional adaptive segmentation methods. Dictionary-based correction is utilized to correct mistakes resulting from the recognition and segmentation phases. The integration of the segmentation methods with the handwritten character recognition algorithm yielded an accuracy of 92% for lower case characters and 97% for upper case characters. In the testing phase, the database consists of 20,000 handwritten characters, with 10,000 for each case. The testing phase on the recognition 10,000 handwritten characters required 8.5 seconds in processing time.

Nowadays, short video feed has attracted billions of mobile users all around the world to interact with content effortlessly, yielding an explosive growth of short video commerce. Typically, users watch full-screen short videos of a few seconds one-by-one in a watch-list generated by recommender systems, skipping those they are not interested in. However, the recommender system at the cloud makes a user-interest-specific decision mostly based on the users' behavior data collected within the application itself (e.g., users' view history), without examining the lower-layer network and communication statistics. When the playback choked due to the limited network bandwidth, the user will probably skip the video, leading to a waste of bandwidth and degradation of the user's quality of experience (QoE). Meanwhile, the excessive number of user requests to video contents raises a heavy computational load and communication cost for the recommender system at the cloud to determine which videos to be recommended and delivered to each user in a real-time manner. The advance of edge computing provides a promising avenue of deploying edge nodes with caches (e.g., household devices) beyond cloud and edge servers, such that the recommender system in the cloud can place popular video contents closer to client users, and meanwhile the contents are delivered to client users with good network condition. In this paper, we propose CFP, a cross-layer recommender system for short video streaming with fine-grained preloading technique at the network edge. CFP jointly optimizes the recommendation effect of the video application and the content preloading efficiency under various network conditions at the network edge. CFP takes a two-stage approach: the cloud server first seeks to perform edge-wise instead of user-interest-specific recommendation with neural collaborative filtering recommender, preloading a list of candidate videos to edge nodes, and each edge node, deploying the GRU with attention, then delivers the proper video contents to the client user device according to the user's preference. Trace-driven emulations demonstrate the efficiency of the proposed CFP scheme.

Due to the digitization of fine art collections, pictures of fine art objects stored at museums and art galleries became widely available to the public. It created a demand for efficient software tools that would allow rapid retrieval and semantic categorization of art. This paper introduces a new, two-stage image classification approach aiming to improve the style classification accuracy. At the first stage, the proposed approach divides the input image into five patches and applies a deep convolutional neural network (CNN) to train and classify each patch individually. At the second stage, the outcomes from the individual five patches are fused in the decision-making module, which applies a shallow neural network trained on the probability vectors given by the first-stage classifier. While the first stage categorizes the input image based on the individual patches, the second stage infers the final decision label categorizing the artistic style of the analyzed input image. The key factor in improving the accuracy compared to the baseline techniques is the fact that the second stage is trained independently on the first stage using probability vectors instead of images. This way, the second stage is effectively trained to compensate for the potential mistakes made during the first stage. The proposed method was tested using six different pre-trained CNNs (AlexNet, VGG-16, VGG-19, GoogLeNet, ResNet-50, and Inceptionv3) as the first-stage classifiers, and a shallow neural network as a second-stage classifier. The experiments conducted using three standard art classification datasets indicated that the proposed method presents a significant improvement over the existing baseline techniques.