Given the ubiquity of handwritten documents in human transactions, Optical Character Recognition (OCR) of documents have invaluable practical worth. Optical character recognition is a science that enables to translate various types of documents or images into analyzable, editable and searchable data. During last decade, researchers have used artificial intelligence/machine learning tools to automatically analyze handwritten and printed documents in order to convert them into electronic format. The objective of this review paper is to summarize research that has been conducted on character recognition of handwritten documents and to provide research directions. In this Systematic Literature Review (SLR) we collected, synthesized and analyzed research articles on the topic of handwritten OCR (and closely related topics) which were published between year 2000 to 2019. We followed widely used electronic databases by following pre-defined review protocol. Articles were searched using keywords, forward reference searching and backward reference searching in order to search all the articles related to the topic. After carefully following study selection process 176 articles were selected for this SLR. This review article serves the purpose of presenting state of the art results and techniques on OCR and also provide research directions by highlighting research gaps.

Handwritten Optical Character Recognition (OCR): A Comprehensive Systematic Literature Review (SLR)

There are many applications of the handwritten character recognition (HCR) approach still exist. Reading postal addresses in various states contains different languages in any union government like India. Bank check amounts and signature verification is one of the important application of HCR in the automatic banking system in all developed countries. The optical character recognition of the documents is comparing with handwriting documents by a human. This OCR is used for translation purposes of characters from various types of files such as image, word document files. The main aim of this research article is to provide the solution for various handwriting recognition approaches such as touch input from the mobile screen and picture file. The recognition approaches performing with various methods that we have chosen in artificial neural networks and statistical methods so on and to address nonlinearly divisible issues. This research article consisting of various approaches to compare and recognize the handwriting characters from the image documents. Besides, the research paper is comparing statistical approach support vector machine (SVM) classifiers network method with statistical, template matching, structural pattern recognition, and graphical methods. It has proved Statistical SVM for OCR system performance that is providing a good result that is configured with machine learning approach. The recognition rate is higher than other methods mentioned in this Journal of Information Technology and Digital World (2021) Vol. 03/ No. 02 Pages: 92-107 https://www.irojournals.com/itdw/ DOI: https://doi.org/10.36548/jitdw.2021.2.003 93 ISSN: 2582-418X Submitted: 18.04.2021 Revised: 12.05.2021 Accepted: 1.06.2021 Published: 08.06.2021 research article. The proposed model has tested on a training section that contained various stylish letters and digits to learn with a higher accuracy level. We obtained test results of 91% of accuracy to recognize the characters from documents. Finally, we have discussed several future tasks of this research further.

https://irojournals.com/itdw/V3/I2/03.pdf

Construction of Statistical SVM based Recognition Model for Handwritten Character Recognition

Face is one of the most widely used biometrics for human identity authentication. Facial recognition has remained an interesting and active research area in the past several decades due to its ever growing applications in biometric authentication, content based data retrieval, video surveillance, access control and social media. Unlike other biometric systems, facial recognition based systems work independently without involving the individual, due to which it does not add unnecessary delay. Its ability of recognizing multiple persons at a time further adds to its speed. There are many face recognition methods based on traditional machine learning that are available in the literature. Improvements are being made with the constant developments in computer vision and machine learning. However, most of the traditional methods lack robustness against varying illumination, facial expression, scale, occlusions and pose. With the advent of big data and graphical computing, deep learning has impressively advanced the traditional computer vision systems over the past decade. In this paper, we present a convolutional neural network based face recognition system which detects faces in an input image using Viola Jones face detector and automatically extracts facial features from detected faces using a pre-trained CNN for recognition. A large database of facial images of subjects is created, which is augmented in order to increase the number of images per subject and to incorporate different illumination and noise conditions for optimal training of the convolutional neural network. Moreover, an optimal pretrained CNN model along with a set of hyperparameters is experimentally selected for deep face recognition. Promising experimental results, with an overall accuracy of 98.76%, are obtained which depict the effectiveness of deep face recognition in automated biometric authentication systems.

Deep Face Recognition for Biometric Authentication

Face recognition is diversely used in modern biometric and security applications. Most of the current face recognition techniques show good results in a constrained environment. However, these techniques face many problems in real-world scenarios such as low-quality images, temporal variations and facial disguises creating variations in facial features. The reason for these deteriorating results is the employment of handcrafted features having weak generalization capabilities and neglecting the complexities associated with domain adaption in case of deep learning models. In this paper, we have studied the efficacy of deep learning methods incorporating simple noise-based data augmentation for disguise invariant face recognition (DIFR). The proposed method detects face in an image using Viola Jones face detector and classifies it using a pre-trained Convolutional Neural Network (CNN) fine-tuned for DIFR. During transfer learning, a pre-trained CNN learns generalized disguise-invariant features from facial images of several subjects to correctly identify them under varying facial disguises. We have compared four different pre-trained 2D CNNs, each with different number of learning parameters, based on their classification accuracy and execution time for selecting a suitable model for DIFR. Comprehensive experiments and comparative analysis have been conducted on six challenging facial disguise datasets. Resnet-18 gives the best trade-off between accuracy and efficiency, by achieving an average accuracy of 98.19% with an average execution time of 0.32 seconds. The promising results achieved in these experiments reflect the efficiency of the proposed method and outperforms the existing methods in all aspects.

An automated and efficient convolutional architecture for disguise-invariant face recognition using noise-based data augmentation and deep transfer learning

Low Earth Orbit (LEO) satellite constellations have seen a surge in deployment over the past few years by virtue of their ability to provide broadband Internet access as well as to collect vast amounts of Earth observational data that can be utilized to develop AI on a global scale. As traditional machine learning (ML) approaches that train a model by downloading satellite data to a ground station (GS) are not practical, Federated Learning (FL) offers a potential solution. However, existing FL approaches cannot be readily applied because of their excessively prolonged training time caused by the challenging satellite-GS communication environment. This paper proposes FedHAP, which introduces high-altitude platforms (HAPs) as distributed parameter servers (PSs) into FL for Satcom (or more concretely LEO constellations), to achieve fast and efficient model training. FedHAP consists of three components: 1) a hierarchical communication architecture, 2) a model dissemination algorithm, and 3) a model aggregation algorithm. Our extensive simulations demonstrate that FedHAP significantly accelerates FL model convergence as compared to state-of-the-art baselines, cutting the training time from several days down to a few hours, yet achieving higher accuracy.

FedHAP: Fast Federated Learning for LEO Constellations using Collaborative HAPs

In this article, we have designed a new information confidentiality mechanism based on the combination of Blowfish encryption algorithm along with Henon and Chen chaotic dynamical systems. We have authenticated our proposed encryption algorithm over satellite and other standard digital images of image processing. Our proposed encryption algorithm is equally deployed as an encrtpter and decrypter while at the sending and receiving ends in satellite communication. In satellite communication, typical wireless mediums are used for transferring heavy payloads and usual telemetry. The priciple aim of this article is to design a new mechanism for the privacy of digital contents mainly satellite images. We have authenticated our anticipated mechanism with security performace analyses.

A hybrid chaotic blowfish encryption for high-resolution satellite imagery

Handwritten character recognition is among the most challenging research areas in pattern recognition and image processing. With everything going digital, applications of handwritten character recognition are emerging in offices, educational institutes, healthcare units and banks etc., where the documents that are handwritten are dealt more frequently. In this paper, a recognition system based on neural network that follows offline handwritten characters has been proposed for Latin digits and alphabets. Each of the characters that are extracted through query image is then resized dynamically to 60×40 pixels’ size and is then passed to the neural networks for the process of recognition. Dynamic resizing enables size invariance in the proposed system and also maintains the aspect ratio of the character so that the image is not distorted during resizing. Neural networks are trained with 19,422 English alphabets’ sample and 7,720 digits’ sample that are written through 150 different writers in various styles of handwriting. Experimental study realized very encouraging results which are compared with the modern methods on this subject corridor.

Size Invariant Handwritten Character Recognition using Single Layer Feedforward Backpropagation Neural Networks

Performance Evaluation of Stream Ciphers for Efficient and Quick Security of Satellite Images

Recently CubeSats have gained alot of momentum in educational institutes, as these mini/pico satellites provide the cheapest way to access the space for research purposes. Institute of Space Technology (IST) pays special attention towards the development of CubeSat and initiated the project named as ICUBE. This paper focuses to propose the earth horizon sensor while considering this project by providing the simple reliable attitude determination of a satellite rather than deploying a complex subsystem to determine the satellite's attitude. Earth horizon sensor plays a significant role in attitude (i.e. pitch axis and roll axis) determination of a satellite. In this paper, the algorithm to determine value of the pitch as well as the value of roll associated to satellite relative to earth horizon is presented. The algorithm takes the satellite image as an input and considers the pre-processing techniques (like gray scale conversion) to compute the value of pitch along with the value of roll associated to a satellite. Column-row based edge detection algorithm is proposed to determine the edges of horizon in the converted gray scaled satellite images, and least square line fitting method is applied to the detected edges to extract the earth horizon. The parameters of a detected horizon line is then considered to find the pitch value and roll value of a satellite. Average Peak Signal to Noise Ratio (PSNR) is considered to evaluate the proposed earth horizon sensor. Our performance evaluation reveals that the proposed attitude determination algorithm based on column-row edge detection method outperforms the attitude determination using the edge detection algorithms named as canny and sobel.

Haroon Ibrahim

Papers

A hybrid chaotic blowfish encryption for high-resolution satellite imagery

Size Invariant Handwritten Character Recognition using Single Layer Feedforward Backpropagation Neural Networks

Performance Evaluation of Stream Ciphers for Efficient and Quick Security of Satellite Images

Earth horizon sensor for attitude determination of LEO satellites