The Ontology Learning System (OLS) aims to automate the construction of hazard-specific ontologies, with a particular focus on floods. Using ontology learning techniques and Machine Learning algorithms, OLS undergoes a series of preprocessing steps, including text cleaning, word tokenization, and the removal of stop words and irrelevant characters. Extracting frequent and relevant terms while also classifying disaster-related keywords is a critical component of automating the construction of a flood ontology. Using Bidirectional Encoder Representations from Transformers (BERT), the model achieved 89% accuracy, precision, and recall in classifying keywords as disaster-related or not. Further, the system employs Machine Learning algorithms, namely Naive Bayes, Logistic Regression, and Linear Support Vector Classifier, for multiclass classification of disaster types. The results of the multiclass classification show that the Linear Support Vector Classifier attained the highest accuracy. The identified ontology keywords can be used to construct a flood-specific ontology, playing a crucial role in organizing flood-related information. Integrating these ontologies into broader systems can significantly enhance disaster preparedness and improve overall response capabilities.
Forest fires are among the worst natural disasters in recent years, causing many deaths, serious injuries, and significant damage to economic and environmental resources. Various methods have been proposed for the early detection of forest fire areas, including the use of meteorological data. However, none of these methods provide real-time image processing that extract significant spatial and temporal features to detect forest fire areas with high accuracy. Additionally, forest fire images exposed to unsafe public networks are vulnerable to security threats and attacks. This paper aims to predict forest fires using advanced Artificial Intelligence (AI) and blockchain based on images captured by drones. Blockchain has immense potential to reduce the time and ensure the security of sharing forest fire data. The proposed system is constructed by combining Convolutional Neural Network (CNN) and transformers to support deep feature extraction with drone-based image capture and predictive modeling. This integrated system enables early detection of forest fires, accurate prediction, and effective response to mitigate the impact of wildfires. The proposed system is applied for the detection of burned areas caused by forest fires in northeastern Algeria. The results are highly promising and are based on spatial, temporal and fire features for effective forest fire prediction, enabling timely alerts and effective firefighting planning
Recently, the Internet of Vehicles (IoV) network, a fusion of Internet of Things (IoT) and VANET, has gained prominence. It facilitates rapid communication of critical information to emergency services in the event of incident or emergencies. However, IoV networks are facing a high rate of network congestion. The primary aim of this work is to alleviate congestion by minimizing the redundancy of messages. This is accomplished through the utilization of unicast connections, enabling point-to-point communication between vehicles and nearby fog servers. Through simulations, this work demonstrates the significance of the proposed Vehicle Emergency Message Dissemination (VEMD) algorithm in enhancing safety applications. The simulation results show that the proposed VEMD algorithm outperforms an efficient work.
MRI brain tumor identification and classification is costly and time-consuming owing to the difficulty and complication of tumors. The process is highly influenced by the experience and knowledge of radiologists and neurologists. For this reason, automating this process becomes indispensable to overcome the drawbacks. In this work, we propose a new model, called DRB-BSIF (i.e., Deep Rule Based Classifier using Binarized Statistical Image Features), to enhance classification performances and reduce the complexity implicated in the diagnostic decision. Specifically, the tumor region is augmented by image dilation and used as the Region of Interest (ROI) instead of the initial tumor region. Then, features are extracted using BSIF image descriptor. Furthermore, we have constructed a Bank-BSIF, which is founded by the best parameters of BSIF filters. In the classification step, a deep-rule based classifier (DRB) has been used. The main of the BRD classifier is a self-organized set of IF and THEN fuzzy rules guided by the prototypes. These fuzzy rules are generated by DRB classifier and represented its ‘engine’. We studied classification of different kinds of brain tumors (e.g., Meningioma, Glioma, and Pituitary tumor). The proposed model is evaluated on publicly available brain CE-MRI images via different measures such as performance accuracy, sensitivity and specificity. Experimental results demonstrated that the DRB-BSIF is effective and can be used in computer aided brain tumor classification.
This paper presents a compact, four-element, Quad-band MIMO patch antenna designed for millimeter-wave (mmWave) applications in the Ka-band (26-40 GHz). The design leverages a monopole-printed antenna with a defected ground plane fabricated on a Rogers RO4003 substrate. The antenna achieves a triple-band -10 dB impedance bandwidth of 27.82-30.81 GHz, 33.8-34.6 GHz, and 35.7-36.4 GHz. Each element exhibits acceptable gain ranging from 5.2 dBi to 6.7 dBi across the operating bands (28 GHz, 30.5 GHz, 34 GHz, and 36 GHz). Following the design and analysis of a single antenna element, a four-port MIMO antenna was realized. Both antenna configurations demonstrate characteristics suitable for mmWave 5G and future wireless applications. The MIMO performance evaluation reveals a remarkably low envelope correlation (ECC) of approximately 10.4, a diversity gain of about 10 dB across the bands, and a negligible channel capacity loss (CCL) of less than 0.001 bits/s/Hz. These findings validate the proposed MIMO antenna's exceptional potential for mmWave future fifth-generation (5G) wireless communication systems deployments.
In this paper we propose a method for submerged houses roof detection on a region with a flood natural disaster by a drone on the fly. We try two different object detection architecture, YOLOv6 and DETR, to find out which one has better performance for a drone application. We propose three different image datasets, namely Floodnet dataset, Giannitsa dataset and RedRoofs dataset merged into one large dataset. Both architecture’s house roofs detectors (YOLOv6 and DETR) trained a) on mixed regions images (flooded and non-flooded regions) and tested on flooded region images and b) on non-flooded region images and tested on flooded region images. Our results depicts that both architecture’s house roof detectors performs very well in environments different from the ones they were trained for. Finally, we do not choose the house roofs detector with the best performance because of computational power requirements and electric power consumption limitations that there are on a drone. The selected house roofs detector requires significantly lower computational but is has slightly lower performance than best.
Social media’s real-time nature has transformed it into a critical tool for disaster response, and for that this study explores the use of tweets for classifying disaster types and identifying humanitarian needs in the aftermath of various disaster events.We compare traditional machine learning models like Random Forest and Support Vector Machines with the deep learning technique, BERT. While BERT demonstrates promising results, a key finding lies in the performance of the voting classifier ensemble, a combination of traditional models. This ensemble achieves accuracy comparable to BERT and even surpasses it. Furthermore, the ensemble boasts exceptional training and inference speeds, making it ideal for real-time applications in disaster response scenarios.Our work investigates the continued value of traditional machine learning methods. By "dusting off" these models we can achieve competitive performance while maintaining computational efficiency. Ultimately, this study empowers humanitarian organizations to leverage the power of text classification for extracting crucial insights from social media data, leading to more effective and targeted responses in times of crisis.
Annually, the devastating impact of natural disasters (such as floods, earthquakes, and forest fires) and man-made incidents (including industrial explosions, oil pipeline leaks, and gas production mishaps) results in widespread damage, financial losses, injuries, and substantial loss of life. Recognizing the imperative to evolve beyond traditional monitoring systems, this paper advocates for adopting intelligent prediction systems that actively involve governmental bodies and the communities affected by these calamities. This paper introduces the SADM-SmartObject framework, a cutting-edge, agent-based conceptual and methodological IoT framework designed for self-adaptation and decision-making, specifically applied to the context of natural flood disasters. The primary objective of this framework is to monitor key climate indicators such as rainfall, humidity, temperature, pressure, and water levels and discern their temporal correlations to enhance flood prediction accuracy. It offers a strong layered architecture complemented by efficient tools for constructing IoT systems, integrating machine learning classifiers for data classification, and agent-based approaches for decision-making. In a compelling case study, we apply the SADM-SmartObject framework to proactively prevent and manage a natural flood disaster scenario.
With the boom in fifth-generation (5G) communication technology, task offloading using multi-access edge computing (MEC) technologies has become an effective approach to satisfy the computation demand of various types of users. Nevertheless, most of the existing task offloading schemes overlook the unavailability of computational resources in disaster areas, failing to guarantee the quality of service (QoS) requirements of task computation in such scenarios. Therefore, this paper treats each edge node as an agent and designs a cooperative multi-agent constrained reinforcement learning (CMACRL) framework for learning the adaptive task offloading policy for each agent, ensuring the performance of emergency services and enhancing the utilization of available computational resources in disaster areas. Specifically, to satisfy the QoS requirements of applications, we formulate the constraints of applications’ performance requirements and then adopt the reward constrained policy optimization algorithm to meet the target constraints. Moreover, to mitigate the resource imbalance between disaster and safe areas, we leverage the heterogeneous-agent proximal policy optimization algorithm to generate cooperative resource use strategies for edge nodes. To evaluate the performance of CMACRL, we develop a simulator based on a real-world urban scenario. Extensive experiments are conducted, and experimental results demonstrate that CMACRL significantly improves the task offloading success rate by up to 22.49% and reduces the average service latency by up to 22.69% compared to all the baselines.
Analyzing sentiment in hotel reviews is important for effective disaster management. Understanding and categorizing guests’ opinions and emotions about their hotel experiences can help identify potential issues promptly and enable hotels to respond to concerns, especially during crises such as natural disasters, health emergencies, or other disruptive events. Implicit sentiment analysis, an important task in Natural Language Processing (NLP), involves analyzing sentiments not explicitly stated in text documents, such as hotel comments. This form of sentiment analysis considers the broader context, tone, and linguistic cues present in the text to uncover underlying attitudes and feelings. By analyzing sentiment in guest reviews, hotels can better prepare for and mitigate the impact of disasters by addressing common concerns and improving overall service quality. To explore the application of sentiment analysis techniques and extract valuable insights from customer feedback, we used two datasets: A and B. We collected reviews from Algerian hotels in dataset A. Dataset B contains 38,932 reviews from hotels and restaurants from websites such as Booking and TripAdvisor. Because of the limited size of dataset A, we applied only machine learning models such as Naive Bayes, Decision Trees, Logistic Regression, and Support Vector Machines (SVM). The Support Vector Machine (SVM) achieved the highest accuracy with 90.9%. Additionally, we compared different frameworks, including machine learning, deep learning, and transformer models, using dataset B. The results show that the Long ShortTerm Memory (LSTM) model outperformed other approaches, yielding an accuracy of 93.76%.
Video summarization technology is crucial for video surveillance applications, particularly in open-pit mine security and disaster relief contexts. This study proposes a novel video summarization model architecture that enhances and integrates multi-dimensional visual features. The model utilizes multi-dimensional visual features to predict importance scores. This method plays a crucial role in responding to and conducting rescue operations in open-pit mining disasters. In contrast to previous methods relying on singular or dual features, our approach employs a multi-dimensional attention network to extract profound semantic features from multi-dimensional visual inputs, thus producing concise and representative video summaries. To comprehensively assess the efficacy of our approach, experiments were conducted on two prominent video summarization benchmark datasets, SumMe and TVSum, yielding advancements over state-of-the-art results. In addition to architectural innovations, we introduce a fusion strategy to mitigate the impact of irrelevant or redundant features on model performance. Through this method, rescue teams can swiftly identify critical events, generate summaries, and respond promptly, thereby playing a pivotal role in open-pit mine disaster management.
Event detection and sentiment analysis have received considerable research attention due primarily to the very large amount of contextual data and information provided by online social networks (OSN) about every area of life. Real-time events reported in posted messages from OSN during disasters can help in disaster management by providing useful information. In addition, sentiment and emotion analysis can provide pertinent information about the degree of satisfaction with the services provided. However, event detection faces several challenges, including event boundary detection, ambiguity and context sensitivity. Hence, we proposed an LSTM network to handle long sequences and effectively capture temporal dependencies. For this task, the results show that the proposed model achieves a success accuracy greater than 71.7% on a sub-dataset containing events generated from 20 keywords. For the sentiment analysis task, we use bag of words (BoW) representation and the RoBERTa pretrained model. The proposed systems were used on a large dataset with five types of disasters.
In large and densely deployed IEEE 802.11 (Wi-Fi) networks, a fast and seamless handover scheme is an important aspect in order to provide reliable connectivity for mobile users. However, most of conventional handover triggering mechanisms of mobile terminal (MT) is designed for fixed scenarios and thus could result in negative effects such as frequent handovers, ping-pong handovers, and handover failures on the handover process of MT at other scenarios. These effects degrade the overall network performance. To address these issues, this paper proposes an intelligent handover triggering mechanism for MT based on deep deterministic policy gradient (DDPG) frameworks with adaptation weight vectors and reward penalty mechanism (APM). The input metrics in one episode are converted to weight vectors for reward of DDPG, which can help MT to be applied in different scenarios. Meanwhile, a penalty mechanism for reward is executed in every step, which can reduce ping-pong handovers. Afterward, the DDPG framework learns the optimal handover triggering policy from the environment. The trained DDPG is deployed to MT to trigger the handover process. The results demonstrate that the proposed method can ensure the stronger mobility robustness of MT that is improved by 20%-96% compared to the conventional approach with respect to the number of handovers and ping-ping handover rate while maintaining a relatively proper level of throughput to guarantee of the MT operation.
The integration of advanced communication technologies is crucial for enhancing public safety solutions, especially in emergency scenarios. Ensuring reliable and resilient communication is essential for both civilians and first responders during such critical situations. Multi-connectivity is one of the key techniques to enhance the robustness and reliability of communication systems in the context of ultra-reliable and low latency communications (URLLC), although it requires network resources, specially in the presence of massive number of devices. This paper investigates the use of multi-connectivity in the context of massive-URLLC use cases in emergency situations. By integrating packet duplication within multi-connectivity, we aim at providing a robust communication framework capable of guaranteeing a required level of massive connectivity during critical situations. We conduct comprehensive simulations to compare the performance of multi-connectivity against single connectivity, focusing on key metrics such as block error rate and bit error rate under varying network conditions and signal-to-noise power ratio thresholds. Our results demonstrate the potential of multi-connectivity by highlighting operating conditions in which a significant improvement of network reliability and performance in emergency scenarios is achieved.
In recent years, Wireless Sensor Networks (WSNs) have gained significant attention for their role in forest fire detection, offering early detection and real-time monitoring capabilities. WSNs are tailored to continuously monitor environmental parameters such as temperature, humidity, smoke, and gas concentrations, utilizing various sensors, communication modules, and a central monitoring system. However, their heterogeneous nature and unattended environments render WSNs vulnerable to diverse cyberthreats. Adversaries can exploit these vulnerabilities to manipulate smart devices remotely, posing serious risks to the integrity and resilience of forest ecosystems. Addressing these security concerns is crucial to ensuring uninterrupted monitoring of environmental parameters and effective forest fire detection. To mitigate these risks, we present a robust Intrusion Detection System (IDS) based on a Convolutional Neural Network (CNN) and Genetic Algorithm (GA), specifically designed to enhance the security of WSNs deployed for forest fire detection. Our proposed model involves hyperparameters optimization of the CNN model using GA. The experiments conducted on three public datasets, namely NSL-KDD, Edge-IoTset, and CICIoT-2023, demonstrate the effectiveness of our methodology in enhancing security, achieving a high accuracy of 99.26% and a low FPR of 0.00735% using NSL-KDD while achieving 100% accuracy and zero FPR using the two new realistic datasets Edge-IIoTset and CICIoT-2023.
Medical risk management is a critical process within healthcare institutions that involves identifying, assessing, and mitigating risks to ensure patient safety and improve care quality. In recent years, the Internet of Medical Things (IoMT) has proved effective in monitoring patient health, particularly during disasters and epidemics such as COVID-19. However, the transmission of information over open channels makes such networks vulnerable to potential attacks. In addition, quantum computing presents a significant threat to current cryptographic algorithms. While existing solutions protect against well-known threats, few address the issue of quantum attacks. To address these challenges, this work presents an effective group authentication scheme for IoMT systems, leveraging post-quantum security using Kyber-PKE and Dilithium, Shamir’s secret sharing (SSS) algorithm, and smart contract. The proposed solution was simulated on the Ethereum platform and evaluated using Hyperledger Caliper, demonstrating its efficiency and scalability. A comparative analysis to recent pertinent works in terms of computation cost, power consumption, and security requirements shows that our system is well suited for various IoMT applications due to its robustness and efficiency.
Forest fires significantly threaten ecosystems, human life, and property, necessitating rapid detection and response mechanisms. Traditional detection methods, such as satellite imagery and ground-based observation, often suffer from delays and limited accuracy. This paper presents a novel approach to enhancing forest fire detection and emergency response using crowdsourcing and smartphone sensors. By leveraging the widespread availability of smartphones equipped with various sensors, coupled with the power of collective human effort, this method aims to improve the timeliness and accuracy of fire detection and optimize emergency response. Moreover, the proposed system enables real-time fire detection and reporting. A dedicated mobile application allows users to submit fire sightings, including multimedia evidence and precise location data. The integration of crowdsourcing and smartphone sensors represents a promising solution for addressing the challenges of forest fire management, enhancing both detection and emergency response capabilities.
Human-robot applications that allow for work to be done remotely are largely dependent on the lassitude of the operators. The exhaustion of these operators is a result of work completion and duration. Existing literature focuses on measuring human fatigue in attempt to quantify the effect on the response. This paper explores the impact of a fatigued human on human-robot applications, particularly, search and rescue aided by a unmanned aerial vehicle (UAV) swarm. A model is presented that describes the relationship between the mental fatigue and responsiveness of the operator over extended periods of time. A simulation of the operation is conducted in which the operator is dependent upon the swarm surveying the area and generating reports. Operation efficiency is estimated according to the human-robot feedback loop, which if not managed, can negatively affect the dynamic. The analysis identifies a trade-off between swarm size and operator fatigue on operation timeliness. The results show that for an operator, there exists an optimal swarm size depending on their level of capability. A more capable operator fatigues at a slower rate and hence, can complete work at a faster rate. This paper details how to evaluate these systems so that operation costs and operator fatigue can be minimized.
As natural disasters rise, traditional infrastructure may be damaged or ineffective, complicating the delivery of medical supplies. Using drones to deliver these supplies to affected areas is increasingly seen as an effective and necessary solution. However, managing drones under these constraints quickly becomes a complex problem requiring intelligent methods. This paper presents a novel approach to address the Medical Supply Drone Delivery (MSDD) problem in post-disaster scenarios by modeling it as a multi-objective optimization problem. We introduce the Slime Mould Algorithm (SMA), designed to optimize drone management during post-disaster situations. Using generative AI techniques, specifically ChatGPT, we augment limited COVID-19 datasets from Algeria to create realistic scenarios. We compare the results of the proposed approach with those obtained using advanced versions of the Particle Swarm Optimization (PSO) and Harris Hawks Optimization (HHO) algorithms. The analysis of the experiment indicates that the proposed multi-objective SMA for the MSDD issue is highly competitive and provides significant performance enhancements.
Post-disaster needs assessment plays a critical role in developing emergency response programs and disaster preparedness. An assessment of needs identifies possible interventions or assistance and the necessary resources for emergency responses. This emergency need assessment assists stakeholders in defining possible response options, evaluating the priorities of needs of affected communities, and determining what support will be required immediately. The prevalent use of social media in Indonesia highlights its significance as a crucial means of acquiring prompt, effortless, individualized, and geo-specific information regarding the demands and requirements of communities impacted by a calamity. This study aims to analyze the tweets posted by local communities affected by the 2018 Central Sulawesi earthquakes, tsunami, and liquefaction using the Social Network Analysis (SNA) method based on machine learning and natural language processing. Through this analysis, we intend to observe the spatial interaction and relationship among the affected areas of communities for the needs and concerns after the disaster to enhance the relief operation. We found that social media users, government, and humanitarian organizations effectively shared information during the disaster. This study also underlined that the behavioral responses of individuals to disasters were not limited to the disaster-stricken areas but extended far beyond them. This action suggests that they were willing to exhibit a more empathetic, benevolent, and philanthropic demeanor in the face of such traumatic events.
