Deep Reinforcement Learning for Vision-Based Navigation of UAVs in Avoiding Stationary and Mobile Obstacles

Unmanned Aerial Vehicles (UAVs), also known as drones, have advanced greatly in recent years. There are many ways in which drones can be used, including transportation, photography, climate monitoring, and disaster relief. The reason for this is their high level of efficiency and safety in all operations. While the design of drones strives for perfection, it is not yet flawless. When it comes to detecting and preventing collisions, drones still face many challenges. In this context, this paper describes a methodology for developing a drone system that operates autonomously without the need for human intervention. This study applies reinforcement learning algorithms to train a drone to avoid obstacles autonomously in discrete and continuous action spaces based solely on image data. The novelty of this study lies in its comprehensive assessment of the advantages, limitations, and future research directions of obstacle detection and avoidance for drones, using different reinforcement learning techniques. This study compares three different reinforcement learning strategies—namely, Deep Q-Networks (DQN), Proximal Policy Optimization (PPO), and Soft Actor-Critic (SAC)—that can assist in avoiding obstacles, both stationary and moving; however, these strategies have been more successful in drones. The experiment has been carried out in a virtual environment made available by AirSim. Using Unreal Engine 4, the various training and testing scenarios were created for understanding and analyzing the behavior of RL algorithms for drones. According to the training results, SAC outperformed the other two algorithms. PPO was the least successful among the algorithms, indicating that on-policy algorithms are ineffective in extensive 3D environments with dynamic actors. DQN and SAC, two off-policy algorithms, produced encouraging outcomes. However, due to its constrained discrete action space, DQN may not be as advantageous as SAC in narrow pathways and twists. Concerning further findings, when it comes to autonomous drones, off-policy algorithms, such as DQN and SAC, perform more effectively than on-policy algorithms, such as PPO. The findings could have practical implications for the development of safer and more efficient drones in the future.

Data-Driven Analysis of Privacy Policies Using LexRank and KL Summarizer for Environmental Sustainability

Natural language processing (NLP) is a field in machine learning that analyses and manipulate huge amounts of data and generates human language. There are a variety of applications of NLP such as sentiment analysis, text summarization, spam filtering, language translation, etc. Since privacy documents are important and legal, they play a vital part in any agreement. These documents are very long, but the important points still have to be read thoroughly. Customers might not have the necessary time or the knowledge to understand all the complexities of a privacy policy document. In this context, this paper proposes an optimal model to summarize the privacy policy in the best possible way. The methodology of text summarization is the process where the summaries from the original huge text are extracted without losing any vital information. Using the proposed idea of a common word reduction process combined with natural language processing algorithms, this paper extracts the sentences in the privacy policy document that hold high weightage and displays them to the customer, and it can save the customer’s time from reading through the entire policy while also providing the customers with only the important lines that they need to know before signing the document. The proposed method uses two different extractive text summarization algorithms, namely LexRank and Kullback Leibler (KL) Summarizer, to summarize the obtained text. According to the results, the summarized sentences obtained via the common word reduction process and text summarization algorithms were more significant than the raw privacy policy text. The introduction of this novel methodology helps to find certain important common words used in a particular sector to a greater depth, thus allowing more in-depth study of a privacy policy. Using the common word reduction process, the sentences were reduced by 14.63%, and by applying extractive NLP algorithms, significant sentences were obtained. The results after applying NLP algorithms showed a 191.52% increase in the repetition of common words in each sentence using the KL summarizer algorithm, while the LexRank algorithm showed a 361.01% increase in the repetition of common words. This implies that common words play a large role in determining a sector’s privacy policies, making our proposed method a real-world solution for environmental sustainability.

Hybrid Whale and Gray Wolf Deep Learning Optimization Algorithm for Prediction of Alzheimer’s Disease

In recent years, finding the optimal solution for image segmentation has become more important in many applications. The whale optimization algorithm (WOA) is a metaheuristic optimization technique that has the advantage of achieving the global optimal solution while also being simple to implement and solving many real-time problems. If the complexity of the problem increases, the WOA may stick to local optima rather than global optima. This could be an issue in obtaining a better optimal solution. For this reason, this paper recommends a hybrid algorithm that is based on a mixture of the WOA and gray wolf optimization (GWO) for segmenting the brain sub regions, such as the gray matter (GM), white matter (WM), ventricle, corpus callosum (CC), and hippocampus (HC). This hybrid mixture consists of two steps, i.e., the WOA and GWO. The proposed method helps in diagnosing Alzheimer’s disease (AD) by segmenting the brain sub regions (SRs) by using a hybrid of the WOA and GWO (H-WOA-GWO, which is represented as HWGO). The segmented region was validated with different measures, and it shows better accuracy results of 92%. Following segmentation, the deep learning classifier was utilized to categorize normal and AD images. The combination of WOA and GWO yields an accuracy of 90%. As a result, it was discovered that the suggested method is a highly successful technique for identifying the ideal solution, and it is paired with a deep learning algorithm for classification.

Enhanced Preprocessing Approach Using Ensemble Machine Learning Algorithms for Detecting Liver Disease

There has been a sharp increase in liver disease globally, and many people are dying without even knowing that they have it. As a result of its limited symptoms, it is extremely difficult to detect liver disease until the very last stage. In the event of early detection, patients can begin treatment earlier, thereby saving their lives. It has become increasingly popular to use ensemble learning algorithms since they perform better than traditional machine learning algorithms. In this context, this paper proposes a novel architecture based on ensemble learning and enhanced preprocessing to predict liver disease using the Indian Liver Patient Dataset (ILPD). Six ensemble learning algorithms are applied to the ILPD, and their results are compared to those obtained with existing studies. The proposed model uses several data preprocessing methods, such as data balancing, feature scaling, and feature selection, to improve the accuracy with appropriate imputations. Multivariate imputation is applied to fill in missing values. On skewed columns, log1p transformation was applied, along with standardization, min-max scaling, maximum absolute scaling, and robust scaling techniques. The selection of features is carried out based on several methods including univariate selection, feature importance, and correlation matrix. These enhanced preprocessed data are trained on Gradient boosting, XGBoost, Bagging, Random Forest, Extra Tree, and Stacking ensemble learning algorithms. The results of the six models were compared with each other, as well as with the models used in other research works. The proposed model using extra tree classifier and random forest, outperformed the other methods with the highest testing accuracy of 91.82% and 86.06%, respectively, portraying our method as a real-world solution for detecting liver disease.

Machine-Learning-Based Diabetes Mellitus Risk Prediction Using Multi-Layer Neural Network No-Prop Algorithm

Over the past few decades, the prevalence of chronic illnesses in humans associated with high blood sugar has dramatically increased. Such a disease is referred to medically as diabetes mellitus. Diabetes mellitus can be categorized into three types, namely types 1, 2, and 3. When beta cells do not secrete enough insulin, type 1 diabetes develops. When beta cells create insulin, but the body is unable to use it, type 2 diabetes results. The last category is called gestational diabetes or type 3. This happens during the trimesters of pregnancy in women. Gestational diabetes, however, disappears automatically after childbirth or may continue to develop into type 2 diabetes. To improve their treatment strategies and facilitate healthcare, an automated information system to diagnose diabetes mellitus is required. In this context, this paper presents a novel system of classification of the three types of diabetes mellitus using a multi-layer neural network no-prop algorithm. The algorithm uses two major phases in the information system: the training phase and the testing phase. In each phase, the relevant attributes are identified using the attribute-selection process, and the neural network is trained individually in a multi-layer manner, starting with normal and type 1 diabetes, then normal and type 2 diabetes, and finally healthy and gestational diabetes. Classification is made more effective by the architecture of the multi-layer neural network. To provide experimental analysis and performances of diabetes diagnoses in terms of sensitivity, specificity, and accuracy, a confusion matrix is developed. The maximum specificity and sensitivity values of 0.95 and 0.97 are attained by this suggested multi-layer neural network. With an accuracy score of 97% for the categorization of diabetes mellitus, this proposed model outperforms other models, demonstrating that it is a workable and efficient approach.

A Scientometric Analysis of Deep Learning Approaches for Detecting Fake News

The unregulated proliferation of counterfeit news creation and dissemination that has been seen in recent years poses a constant threat to democracy. Fake news articles have the power to persuade individuals, leaving them perplexed. This scientometric study examined 569 documents from the Scopus database between 2012 and mid-2022 to look for general research trends, publication and citation structures, authorship and collaboration patterns, bibliographic coupling, and productivity patterns in order to identify fake news using deep learning. For this study, Biblioshiny and VOSviewer were used. The findings of this study clearly demonstrate a trend toward an increase in publications since 2016, and this dissemination of fake news is still an issue from a global perspective. Thematic analysis of papers reveals that research topics related to social media for surveillance and monitoring of public attitudes and perceptions, as well as fake news, are crucial but underdeveloped, while studies on deep fake detection, digital contents, digital forensics, and computer vision constitute niche areas. Furthermore, the results show that China and the USA have the strongest international collaboration, despite India writing more articles. This paper also examines the current state of the art in deep learning techniques for fake news detection, with the goal of providing a potential roadmap for researchers interested in undertaking research in this field.

Forest Fire Identification in UAV Imagery Using X-MobileNet

Forest fires are caused naturally by lightning, high atmospheric temperatures, and dryness. Forest fires have ramifications for both climatic conditions and anthropogenic ecosystems. According to various research studies, there has been a noticeable increase in the frequency of forest fires in India. Between 1 January and 31 March 2022, the country had 136,604 fire points. They activated an alerting system that indicates the location of a forest fire detected using MODIS sensor data from NASA Aqua and Terra satellite images. However, the satellite passes the country only twice and sends the information to the state forest departments. The early detection of forest fires is crucial, as once they reach a certain level, it is hard to control them. Compared with the satellite monitoring and detection of fire incidents, video-based fire detection on the ground identifies the fire at a faster rate. Hence, an unmanned aerial vehicle equipped with a GPS and a high-resolution camera can acquire quality images referencing the fire location. Further, deep learning frameworks can be applied to efficiently classify forest fires. In this paper, a cheaper UAV with extended MobileNet deep learning capability is proposed to classify forest fires (97.26%) and share the detection of forest fires and the GPS location with the state forest departments for timely action.

IoMT with Deep CNN: AI-Based Intelligent Support System for Pandemic Diseases

The Internet of Medical Things (IoMT) is an extended version of the Internet of Things (IoT). It mainly concentrates on the integration of medical things for servicing needy people who cannot get medical services easily, especially rural area people and aged peoples living alone. The main objective of this work is to design a real time interactive system for providing medical services to the needy who do not have a sufficient medical infrastructure. With the help of this system, people will get medical services at their end with minimal medical infrastructure and less treatment cost. However, the designed system could be upgraded to address the family of SARs viruses, and for experimentation, we have taken COVID-19 as a test case. The proposed system comprises of many modules, such as the user interface, analytics, cloud, etc. The proposed user interface is designed for interactive data collection. At the initial stage, it collects preliminary medical information, such as the pulse oxygen rate and RT-PCR results. With the help of a pulse oximeter, they could get the pulse oxygen level. With the help of swap test kit, they could find COVID-19 positivity. That information is uploaded as preliminary information to the designed proposed system via the designed UI. If the system identifies the COVID positivity, it requests that the person upload X-ray/CT images for ranking the severity of the disease. The system is designed for multi-model data. Hence, it can deal with X-ray, CT images, and textual data (RT-PCR results). Once X-ray/CT images are collected via the designed UI, those images are forwarded to the designed AI module for analytics. The proposed AI system is designed for multi-disease classification. It classifies the patients affected with COVID-19 or pneumonia or any other viral infection. It also measures the intensity level of lung infection for providing suitable treatment to the patients. Numerous deep convolution neural network (DCNN) architectures are available for medical image classification. We used ResNet-50, ResNet-100, ResNet-101, VGG 16, and VGG 19 for better classification. From the experimentation, it observed that ResNet101 and VGG 19 outperform, with an accuracy of 97% for CT images. ResNet101 outperforms with an accuracy of 98% for X-ray images. For obtaining enhanced accuracy, we used a major voting classifier. It combines all the classifiers result and presents the majority voted one. It results in reduced classifier bias. Finally, the proposed system presents an automatic test summary report textually. It can be accessed via user-friendly graphical user interface (GUI). It results in a reduced report generation time and individual bias.

Segmentation and Analysis Emphasizing Neonatal MRI Brain Images Using Machine Learning Techniques

MRI scanning has shown significant growth in the detection of brain tumors in the recent decade among various methods such as MRA, X-ray, CT, PET, SPECT, etc. Brain tumor identification requires high exactness because a minor error can be life-threatening. Brain tumor disclosure remains a challenging job in medical image processing. This paper targets to explicate a method that is more precise and accurate in brain tumor detection and focuses on tumors in neonatal brains. The infant brain varies from the adult brain in some aspects, and proper preprocessing technique proves to be fruitful to avoid miscues in results. This paper is divided into two parts: In the first half, preprocessing was accomplished using HE, CLAHE, and BPDFHE enhancement techniques. An analysis is the sequel to the above methods to check for the best method based on performance metrics, i.e., MSE, PSNR, RMSE, and AMBE. The second half deals with the segmentation process. We propose a novel ARKFCM to use for segmentation. Finally, the trends in the performance metrics (dice similarity and Jaccard similarity) as well as the segmentation results are discussed in comparison with the conventional FCM method.

Smart IoMT Framework for Supporting UAV Systems with AI

The health monitoring system is one of the most innovative technologies that has gained traction in the Internet of Medical Things (IoMT). It allows the connection of multiple sensors and actuators that can capture and monitor the data through the web page or mobile application. IoMT technology not only provides communications but also will provide monitoring, recording, storage, and display. IoMT in healthcare is used for measuring the vital signs of the human body, which allows medical professionals to assess the well-being of a patient. The doctor may recommend lifestyle modifications, prescribe more tests, or diagnose a disorder according to the results. This paper illustrates the remote-control health monitoring system (HMS) with the integration of a UAV, which allows the doctor to access the data and analyze the patient data remotely. Thus, the proposed HMS-UAV system aims to measure the temperature, humidity, blood pressure, heart rate, and SpO2 and stores the data on the UAV. Several sensors were thus used namely DHT11, MAX30102, Myoware and K24C16, and the Raspberry Pi camera. Reduced hospital stays and avoidance of readmissions are benefits of remote patient monitoring with IoMT-based UAVs. Contrary to its advantages, IoMT has flaws in information processing since a huge volume of data are needed to be handled in a single environment. One major novel inclusion in this work is to measure multiple parameters and provide a comparative analysis for all of them. Furthermore, the functionality of video recorded and stored is included where the doctor can surveil the patient.

DocCompare: An Approach to Prevent the Problem of Character Injection in Document Similarity Algorithm

There is a constant rise in the amount of data being copied or plagiarized because of the abundance of content and information freely available across the internet. Even though the systems try to check documents for the plagiarism, there have been trials to overcome these system checks. In this paper, the concept of character injection is used to trick plagiarism checker is presented. It is also showcased that how does the similarity check algorithms based on k-grams fail to detect the character injection. In order to eradicate the problem or error in similarity rates caused due to the problem of character injection, image processing based approach of multiple histogram projections are used. An application is developed to detect the character injection in the document and produce the accurate similarity rate. The results are shown with some test documents and the proposed method eliminates any kind of character injected in the document that tricks plagiarism. The proposed method has addressed the problem of character injection with image processing based changes in the existing methods of document-similarity check algorithms using k-grams. The proposed method can detect 100% injected character be it any alphabet of any language, The processing time for conversion, histogram projections and applying winnowing algorithm takes 1.2 sec per page on average when experimented on multiple types of document varying in size from 2 KB to 10 MB.

Dynamic resource provisioning and secured file sharing using virtualization in cloud azure

Virtual machines (VMs) are preferred by the majority of organizations due to their high performance. VMs allow for reduced overhead with multiple systems running from the same console at the same time. A physical server is a bare-metal system whose hardware is controlled by the host operating system. A physical server runs on a single instance of OS and application. A virtual server or virtual machine encapsulates the underlying hardware and networking resources. With the existing physical server, it is difficult to migrate the tasks from one platform to another platform or to a datacentre. Centralized security is difficult to setup. But with Hypervisor the virtual machine can be deployed, for instance, with automation. Virtualization cost increases as well as a decrease in hardware and infrastructure space costs. We propose an efficient Azure cloud framework for the utilization of physical server resources at remote VM servers. The proposed framework is implemented in two phases first by integrating physical servers into virtual ones by creating virtual machines, and then by integrating virtual servers into cloud service providers in a cost-effective manner. We create a virtual network in the Azure datacenter using the local host physical server to set up the various virtual machines. Two virtual machine instances, VM1 and VM2, are created using Microsoft Hyper-V with the server Windows 2016 R. The desktop application is deployed and VM performance is monitored using the PowerShell script. Tableau is used to evaluate the physical server functionality of the worksheet for the deployed application.

The proposed Physical to Virtual to Cloud model (P2V2C) model is being tested, and the performance result shows that P2V2C migration is more successful in dynamic provisioning than direct migration to cloud platform infrastructure. The research work was carried out in a secure way through the migration process from P2V2C.

Improved COVID-19 detection with chest x-ray images using deep learning

The novel coronavirus disease, which originated in Wuhan, developed into a severe public health problem worldwide. Immense stress in the society and health department was advanced due to the multiplying numbers of COVID carriers and deaths. This stress can be lowered by performing a high-speed diagnosis for the disease, which can be a crucial stride for opposing the deadly virus. A good large amount of time is consumed in the diagnosis. Some applications that use medical images like X-Rays or CT-Scans can pace up the time used in diagnosis. Hence, this paper aims to create a computer-aided-design system that will use the chest X-Ray as input and further classify it into one of the three classes, namely COVID-19, viral Pneumonia, and healthy. Since the COVID-19 positive chest X-Rays dataset was low, we have exploited four pre-trained deep neural networks (DNNs) to find the best for this system. The dataset consisted of 2905 images with 219 COVID-19 cases, 1341 healthy cases, and 1345 viral pneumonia cases. Out of these images, the models were evaluated on 30 images of each class for the testing, while the rest of them were used for training. It is observed that AlexNet attained an accuracy of 97.6% with an average precision, recall, and F1 score of 0.98, 0.97, and 0.98, respectively.

Covid-19 fake news sentiment analysis

'Fake news' refers to the misinformation presented about issues or events, such as COVID-19. Meanwhile, social media giants claimed to take COVID-19 related misinformation seriously, however, they have been ineffectual. This research uses Information Fusion to obtain real news data from News Broadcasting, Health, and Government websites, while fake news data are collected from social media sites. 39 features were created from multimedia texts and used to detect fake news regarding COVID-19 using state-of-the-art deep learning models. Our model's fake news feature extraction improved accuracy from 59.20% to 86.12%. Overall high precision is 85% using the Recurrent Neural Network (RNN) model; our best recall and F1-Measure for fake news were 83% using the Gated Recurrent Units (GRU) model. Similarly, precision, recall, and F1-Measure for real news are 88%, 90%, and 88% using the GRU, RNN, and Long short-term memory (LSTM) model, respectively. Our model outperformed standard machine learning algorithms.

Cloud-based email phishing attack using machine and deep learning algorithm

Cloud computing refers to the on-demand availability of personal computer system assets, specifically data storage and processing power, without the client's input. Emails are commonly used to send and receive data for individuals or groups. Financial data, credit reports, and other sensitive data are often sent via the Internet. Phishing is a fraudster's technique used to get sensitive data from users by seeming to come from trusted sources. The sender can persuade you to give secret data by misdirecting in a phished email. The main problem is email phishing attacks while sending and receiving the email. The attacker sends spam data using email and receives your data when you open and read the email. In recent years, it has been a big problem for everyone. This paper uses different legitimate and phishing data sizes, detects new emails, and uses different features and algorithms for classification. A modified dataset is created after measuring the existing approaches. We created a feature extracted comma-separated values (CSV) file and label file, applied the support vector machine (SVM), Naive Bayes (NB), and long short-term memory (LSTM) algorithm. This experimentation considers the recognition of a phished email as a classification issue. According to the comparison and implementation, SVM, NB and LSTM performance is better and more accurate to detect email phishing attacks. The classification of email attacks using SVM, NB, and LSTM classifiers achieve the highest accuracy of 99.62%, 97% and 98%, respectively.

J, Mohan S, Ahmadian A, Ciano T. A hybrid learning approach for the stage‐wise classification and prediction of COVID‐19 X‐ray images

Background

The COVID‐19 pandemic has precipitated global apprehensions about increased fatalities and raised concerns about gaps in healthcare infrastructure and accessibility the world over. Consequently, the importance of timely prediction and treatment of the disease to reduce transmission and mortality rates cannot be emphasized enough. Various symptoms of the disease have been identified as it progresses from the time it is contracted. COVID‐19 has been found to internally affect the lungs, and the four progressive stages of the infection can be categorized as mild, moderate, severe, and critical. Therefore, an accurate analysis of the current stage of the disease that can help predict its progression has become critical. X‐ray imaging has been found to be an effective screening procedure for predicting the various stages of this epidemic. Although many different approaches using machine learning, as well as deep learning were utilized to predict and classify diseases in general, till date, such an approach has not been used to predict the various stages of COVID‐19 by using X‐ray imaging to identify and classify those stages.

Materials and method

The proposed hybrid method used three public datasets for its implementation. In this work, extensive images were used for the purposes of testing and training. The dataset‐1 consists of 1200 COVID‐19 as well as 1200 Non‐COVID‐19 images, while dataset‐2 used 700 COVID‐19 as well as 700 Non‐COVID‐19 images, and finally, dataset‐III utilized 1900 COVID‐19 as well as 1900 Non‐COVID‐19 images for purposes of testing and training. The proposed work undertook the task of pre‐processing using textual and morphological features, while the segmentation and prediction of COVID‐19 as well as Non‐COVID‐19 images were undertaken using VGG‐16 with light GBM for better prediction and handing of huge datasets, and finally, the classification of the various stages of COVID‐19 images was performed using Deep Belief Network.

Results

The outcomes of the proposed work were subjected to several iterations which were then compared using different parameters such as accuracy, specificity, and sensitivity. In general, the prediction and grouping of the various stages of COVID‐19 by using affected images were found to be 99.2%, 99.4% and 99.5%, respectively. The bacterial pneumonia prediction rates were observed to be 98.5%, 99.4% and 98.3%, respectively. The average classification of the stages were found to be 98.1%, 98.6% and 98.3%, while the combined multi‐classification prediction rates were observed to be 98.6%, 99.1% and 98.7%, respectively.