Post-disaster Environment Research Articles

Coordinated restoration of interdependent critical infrastructures: A novel distributed decision-making mechanism integrating optimization and reinforcement learning

The proper functioning of any society heavily depends on its critical infrastructures (CIs), such as power grids, road networks, and water and waste-water systems. These infrastructures consist of facilities spread across a community to provide essential services to its residents. Their spatial expansion and functional interdependencies make them highly vulnerable against natural/manmade disasters. Functional interdependencies mean that the functionality of components in one CI relies on the services provided by others. These features, combined with decentralized decision-making structure of CIs and the stochastic nature of post-disaster environments, highly complicate the optimization process for restoring CIs damaged in disasters. Optimizing CI restorations is critical to maximizing the post-disaster resilience of communities.In this paper, we integrate and leverage Reinforcement Learning (RL) and optimization strengths to design a novel distributed modeling and solution approach for advancing the restoration process for interdependent CIs after disasters. The proposed approach (1) bridges the gap between integrative and distinct decision-making, enabling coordinated restoration planning for CIs within a decentralized decision-making context; (2) handles post-disaster uncertainties (e.g., uncertainty in recovery times of disrupted components); (3) generates adaptive solutions that cope with post-disaster dynamics (e.g., varying numbers of recovery teams); and (4) is flexible enough to handle several restoration decisions (e.g., restoration scheduling and resource allocation) simultaneously.To evaluate its performance, we focus on restoring the road and power CIs in Sioux Falls, South Dakota, disrupted by several tornado scenarios. The numerical results show that coordinated policies in the restoration process of interdependent CIs consistently yield higher service for the community. The overperformance of the coordinated restoration policies can be as high as 27.9 %. The impact of coordination is more significant in severe disasters with higher disruptions and in the absence of efficient recovery resources.

River Communities, Disaster Science, and the Politics of Water Safety: Understanding Water Quality Debates in the Aftermath of the Gold King Mine Spill

ABSTRACT After environmental disasters, scientific claims about environmental risk face a paradox. Impacted communities turn to scientific bodies for safety information, yet scientific findings are often challenged publicly in post-disaster debates. This study explores this paradox after the 2015 Gold King Mine spill in the Animas River, USA. We ask how a river is declared “safe” after such a disaster, by whom, and based on what metrics. We examine how scientific claims about water safety were debated in newspapers in three locations: Silverton and Durango, Colorado, and the Navajo Nation. Drawing on Cash et al.’s [Knowledge systems for sustainable development. Proceedings of the National Academy of Sciences, 100(14), 8086–8091. https://do.org/10.1073/pnas.1231332100] key features of science for public decision-making, we find that local debates hinged on issues of salience, credibility, and legitimacy, but these criteria varied in importance across locations. We also argue that these criteria do not sufficiently explain river safety debates. Attention to regulatory and scientific standards used to interpret river safety is necessary for understanding spill responses. Clean Water Act (CWA) standards sidelined public concerns about waterways that fell outside CWA benchmarks. This study highlights the importance of scrutinizing the standards used to deem environments “safe” when producing and communicating disaster science. Key policy highlights Scientific communication after an environmental disaster should be salient, credible, and legitimate. In addition, the scientific and regulatory standards used to deem post-disaster environments “safe” must be scrutinized and their limitations acknowledged. Acknowledging and addressing community-specific concerns and histories – especially those of indigenous communities – is vital in disaster-recovery communication.

Adaptive task migration strategy with delay risk control and reinforcement learning for emergency monitoring

The timely and reliable handling of post-disaster emergency monitoring tasks is crucial for effective rescue operations. UAV-assisted edge computing plays a pivotal role in the rapid deployment of such systems. However, challenges persist due to communication and computation resource bottlenecks when dealing with delay-sensitive monitoring tasks. In dynamic post-disaster environments, effective task scheduling and resource allocation decisions directly impact the system’s ability to process tasks. Therefore, this paper proposes an adaptive task migration decision-making system for emergency monitoring tasks in UAV-assisted edge computing. Firstly, We decomposed the optimization objectives based on the task processing workflow, then devised a stepwise delay risk control and resource recovery mechanism based on early discarding. Secondly, by integrating multi-agent reinforcement learning (MARL), optimal strategies for task offloading, UAV queue scheduling, and communication resource allocation are learned to enhance the decision system’s environmental awareness and maximize the successful completion of emergency monitoring tasks. Simulation experiments demonstrate that the algorithm significantly improves the success rate of migration tasks and data processing capacity, thereby validating its convergence and effectiveness.

Shelter location–allocation problem for disaster evacuation planning: A simulation optimization approach

Immediately after a major earthquake strikes, a segment of the people in the disaster region must evacuate to relief shelters. If the evacuation proceeds in an efficient and safe manner, many lives can potentially be saved as the effects of secondary factors, e.g., secondary collapses, stampedes, and aftershocks, can largely be mitigated, while at the same time the people taking refuge can quickly gain access to vital relief goods. Formulating the best possible protocols for post-earthquake evacuation is a crucial, yet highly challenging problem considering the profound uncertainty which is present in the post-disaster environment, including the location and degree of infrastructural damage, the number and location of people who need to evacuate, the behavior and psychology of evacuees, etc. In this research, we formulate a two-stage stochastic evacuation simulation model capable of credibly accounting for these uncertainties. We develop a solution methodology to solve the shelter location–allocation problem represented by the two-stage model, finding the optimal location of shelters to be open as the first stage solution and the optimal assignment of evacuees to shelters as the second stage solution to minimize the total expected cost. The cost-based objective function consists of the cost of opening shelters, a deprivation cost component, the cost of trapped evacuees, and the cost of shelters exceeding capacity. In developing this framework, we collaborate with the National Science and Technology Center for Disaster Reduction (NCDR) in utilizing GIS-based probabilistic models of road damage. We apply the proposed framework to run an extensive numerical study and present various valuable observations and insights which can be gained.

Development of system-based digital decision support (“Pocket Ark”) for post-flood enhanced response coordination and worker safety: an Intervention Mapping approach

BackgroundThe health and safety of workers who work in areas severely damaged by natural weather events (reconstruction workers) is becoming an increasingly important health problem as these disasters increase in intensity and frequency. An evidence-based, innovative e-learning tool, Pocket Ark, has been developed to meet this need.MethodsIntervention Mapping, an iterative public health programming methodology, was used to create strategies designed to support the development and implementation of the Pocket Ark application for the health and safety of reconstruction workers before, after, and during natural disasters.DiscussionThe development of an evidence-based application, Pocket Ark, uses the concept of the Intervention Mapping planning framework. It can be an effective bridge of pre-deployment education, deployment health, and safety support for the reconstruction work in the immediate post-disaster environment.

Optimization-driven artificial intelligence-enhanced municipal waste classification system for disaster waste management

This research addresses the critical challenge of disaster waste management, a growing concern exacerbated by the increasing frequency and intensity of natural disasters like flooding. Traditional waste systems often struggle with the volume and heterogeneity of disaster waste, highlighting the need for innovative solutions. In this study, we present a novel disaster waste classification model integrating advanced artificial intelligence (AI) and optimization techniques to streamline waste categorization in post-disaster environments. Our approach leverages a dual ensemble deep learning framework. The first ensemble combines various image-segmentation methods, while the second integrates outputs from diverse convolutional neural network architectures. A modified artificial multiple intelligence system serves as a decision fusion strategy, enhancing accuracy at both ensemble points. We rigorously evaluated our model using three datasets: the “TrashNet” dataset for benchmarking against existing methods, as well as two meticulously curated, real-world datasets collected from flood-affected areas in Thailand. The results demonstrate that our method outperforms existing algorithms like VGG19, YoloV5, and InceptionV3 in general solid waste classification, achieving an average improvement of 11.18%. Regarding disaster waste specifically, our model achieves 96.48% and 96.49% accuracy on the curated datasets, consistently outperforming ResNet-101, DenseNet-121, and InceptionV3 by an average of 3.47%. These findings demonstrate the potential of our AI-enhanced model to revolutionize disaster waste management practices. Thus, we advocate integrating such technologies into municipal waste management policies to enhance resilience and optimize disaster responses. Future research will explore scaling the model to diverse disaster types and incorporating real-time data for adaptable waste management strategies.

'School-based screen-and-treat': An effective blueprint for expediating access to care in children experiencing PTSD following disasters.

While 5%-10% of children exposed to natural disasters develop PTSD, few children access support. This paper reports on the proactive 'screen-and-treat' approach deployed following devastating floods in Queensland, Australia, in 2011 and presents results for children in the Lockyer Valley (the most impacted community). Open treatment study (2011-2012) within a government-funded post-disaster service response. One hundred and fifty children (7-12 years) completed pencil-and-paper screening (PTSD, anxiety and depression) at school. Eighty children endorsing either clinical levels of PTSD, or moderate levels of PTSD and clinical levels of either anxiety or depression, and their parents, completed a structured diagnostic interview. Forty-eight children were offered a free trauma-focused CBT intervention. The parents of 19 children accepted this offer. Most clinicians were clinical psychology trainees from local universities. All measures were repeated at post-treatment, 6- and 12-month follow-up. Note: The term 'parents' is used to refer to the wide variety of people serving as a child's primary caregiver. Pre-treatment, all children met diagnostic criteria for full (N = 17) or sub-clinical PTSD. By post-treatment, 10.5% met criteria for PTSD, with 0% meeting criteria at the 12-month follow-up. The incidence of anxiety and depressive disorders also reduced significantly. There were no differences in outcomes for children seen by trainees compared to experienced clinicians. A school-based screen-and-treat approach offers potential as a means of identifying and treating children following natural disaster exposure. However, engagement of families at the outset, and when offering intervention was challenging. Postgraduate trainees represent an effective potential workforce in a post-disaster environment.

Pedestrian evacuation planning under dam-break flood disaster considering road risk and road pedestrian demand

When facing disasters, a well-planned evacuation of people can significantly reduce casualties and disaster losses. Hurricane Katrina struck New Orleans in 2005, causing multiple dam failures and resulting in Lake Pontchartrain flooding back into the city. Taking this disaster as an example, this paper evaluates the risk of pedestrian evacuation on roads during dam-break flood disasters. The evaluation considers three aspects: road evacuation difficulty, disaster risk and road pedestrian demand. The fuzzy VIKOR model was used to evaluate 8 indicators for 50 candidate shelters. A dual-objective planning model was proposed to consider the time-varying and diverse emergency needs of disaster victims. By assessing the risk of pedestrian evacuation on the road and delineating restricted areas, the optimal path distance is calculated. Pareto optimality is then employed to achieve a trade-off between the convenience, cost, and applicability of the evacuation plan. In a complex post-disaster environment, the model can adapt flexibly to emergency rescue situations at various stages by adjusting the parameters in the objective and constraint equation. The Pareto optimal frontier can assist in determining the most suitable solution for the current scenario. In the case of New Orleans, 22 and 18 shelters were respectively selected from 34 candidate shelters for evacuation based on the simulated scenarios.

Use of cross‐training in human–robot collaborative rescue

AbstractHuman–robot collaboration has been widely used in postdisaster investigation and rescue. Human–robot team training is a good way to improve the team rescue efficiency and safety; two common training methods, namely, procedural training and cross‐training, are explored in this study. Currently, relatively few studies have explored the impact of cross‐training on human–robot collaboration in rescue tasks. Cross‐training will be novel to most rescuers and as such, an evaluation of cross‐training in comparison with more conventional procedural training is warranted. This study investigated the effects of these two training methods on rescue performance, situation awareness and workload. Forty‐two participants completed a path‐planning and a photo‐taking task in an unfamiliar simulated postdisaster environment. The rescue performance results showed that cross‐training method had significant advantages over procedural training for human–robot collaborative rescue tasks. During the training process, compared with procedural training, participants were more likely to achieve excellent photo‐taking performance after cross‐training; after training, the length of the route planned by the cross‐training group was significantly shorter than that of the procedural‐training group. In addition, procedural‐training marginal significantly increased the emotion demand, which proves that cross‐training can well control the emotions of the operators and make them more involved in the rescue task. The study also found that arousal level increased significantly after the first cross‐training session, and decreased to the same level as procedural training after multiple sessions. These results contribute to the application of cross‐training in human–robot collaborative rescue teams.

An Evaluation of Design Proposals for Temporary Shelters Focused on Vulnerable Groups in Post-Disaster Situations

Vulnerable group-oriented temporary space design considers the unique needs and challenges specified groups face during emergencies and aims to create a supportive environment that promotes safety, comfort, and accessibility both physically and psychologically. In this study, the vulnerable group was addressed to women, and a literature study was conducted covering the design of temporary spaces in emergencies and the requirements of the identified group. Alternative temporary space designs were analyzed according to criteria determined from the literature study, and examples were examined with the space syntax method via Depthmap software. As a result, it was suggested that the role of architecture is a significant factor in improving the living conditions and welfare of vulnerable groups, especially women in post-disaster environments, considering different socio-demographic structures and many problems can be corrected with the determined design criteria.

Behavioral activation is associated with post-disaster mental health: Secondary longitudinal analysis from a population-based study.

Interventions in post-disaster environments may be accelerated by identifying protective behavioral factors adding incremental value to models of psychopathology using longitudinal methods. One protective behavior applicable to post-disaster contexts is behavioral activation (BA). BA is defined here as a behavioral pattern involving presence of valued activity engagement. While relevant post-disaster, the incremental value of BA behaviors in predicting longitudinal post-disaster outcomes is not well understood. We hypothesized that higher baseline engagement in behaviors consistent with a BA framework would predict decreased posttraumatic stress disorder(PTSD) symptom severity, depression symptom severity, and sleep disturbance approximately 3, 6, and 12 months after hurricane survivors completed baseline measures. The current study is a secondary analysis from a randomized controlled trial of a disaster mental health digital intervention. Participants completed surveys at baseline and approximately 3, 6, and 12 months post-enrollment. Correlations and hierarchical regression analyses were calculated following data screening to predict PTSD symptom severity, depression symptom severity, and sleep disturbances. Controlling for alcohol use, prior trauma, displacement, and intervention condition, higher baseline BA consistently predicted less PTSD symptom severity, depression symptom severity, and sleep disturbances. Results suggest that post-disaster interventions should consider addressing BA. The study provides evidence that BA is potentially an important protective factor longitudinally predicting sleep disturbances and psychopathology after natural disasters.

The pre-design phase in the post-catastrophe intervention process. The case of Chamanga, Ecuador

This article examines the concept of pre-design as a preliminary phase of a research-by-design methodology. The aim is to reveal the importance of this first stage in the development of an architectural project process and to provide new information on the relationships that occur with the rest of the design phases and, ultimately, with the final design result. The research is carried out experimentally through on-site action in a post-disaster environment. The case study took place in the town of Chamanga, Ecuador, after the seismic event of 2016, which required a reconstruction process. The situation was conducive to applying pre-design methods and reflecting on the results. The research results highlight the importance of establishing the criteria for action and how these must be selected based on successive 'observation-projection' iterations. It also shows how the pre-design is already a determining part of the final design due to the decisions taken during the process. This research contributes to identifying the variables and characteristics of a pre-design phase and provides new insights into understanding the architectural project as a research process.

Analysis using a Model-Based Approach of Path Planning for Multiple Uavs in the Context of Surveying Building Damage Following a Disaster

In the wake of disasters, efficient and timely assessment of building damage is of paramount importance for effective disaster response and recovery efforts. Unmanned Aerial Vehicles (UAVs), commonly known as drones, have emerged as indispensable tools for postdisaster surveying due to their ability to rapidly cover large areas and provide high-resolution imagery. The effective utilization of multiple UAVs in such scenarios necessitates optimal path planning to maximize coverage, minimize mission time, and ensure safe navigation. This paper presents a comprehensive model-based analysis of multi-UAV path planning strategies tailored for surveying building damage in postdisaster environments. The study commences by emphasizing the critical role of UAVs in disaster management, especially in the context of assessing structural damage. Acknowledging the challenges posed by complex urban landscapes and the need for swift response, the research focuses on the development and evaluation of path planning strategies that leverage advanced modeling techniques. The proposed model integrates spatial data, building blueprints, and environmental information to generate a realistic representation of the disaster-affected area, facilitating informed decision- making for path planning. A range of path planning approaches is examined within the model-based framework. These include grid-based methods, which discretize the area into cells and deploy UAVs to cover designated regions, and graph-based approaches such as Rapidly-exploring Random Trees (RRT), which generate paths through probabilistic exploration. The paper also delves into machine learning-enhanced strategies, leveraging reinforcement learning algorithms to adapt UAV paths based on real-time observations and evolving disaster dynamics. Furthermore, the study underscores the scalability of the model-based approach, demonstrating its applicability to scenarios involving a varying number of UAVs. Through comprehensive experimentation, the paper provides insights into the optimal deployment of UAV teams, shedding light on the balance between increased coverage and potential communication overhead.

Investigating the aftermath of the Türkiye 2023 earthquake: exploring post-disaster uncertainty among Syrian migrants using social network analysis with public health approach.

On February 6th, 2023, a doublet earthquake struck Türkiye, impacting more than 15 million people including migrants, and resulting in over 50,000 deaths. The Syrian migrants experience multiple uncertainties in their daily lives which are further compounded by multifaceted challenges of the post-disaster environment. Social media was used intensively and with impunity in this environment and thereby provides a window into the explicit and implicit dynamics of daily life after a disaster. We aimed to explore how a post-disaster environment potentially generates new uncertainties or exacerbating pre-existing ones for migrants through social media analysis with an indirect perspective, in the context of 2023-Earthquake in Türkiye and Syrian migrants. Social network analysis was used to analyze Twitter-data with the hashtags 'Syrian' and 'earthquake' during a 10-day period beginning on March 22nd, 2023. We calculated network metrics, including degree-values and betweenness-centrality and clustered the network to understand groups. We analyzed a combination of 27 tweets with summative content analysis using a text analysis tool, to identify the most frequently used words. We identified the main points of each tweet and assessed these as possible contributors to post-disaster uncertainty among migrants by using inductive reasoning. There were 1918 Twitter users, 274 tweets, 124 replies and 1726 mentions. Discussions about Syrian migrants and earthquakes were established across various groups (ngroups(edges > 15) = 16). Certain users had a greater influence on the overall network. The nine most frequently used words were included under uncertainty-related category (nmost_frequently_used_words = 20); 'aid, vote, house, citizen, Afghan, illegal, children, border, and leave'. Nine main points were identified as possible post-disaster uncertainties among migrants. The post-disaster environment has the potential to exacerbate existing uncertainties, such as being an undocumented migrant, concerns about deportation and housing, being or having a child, inequality of rights between being a citizen and non-citizen, being in minority within minority, political climate of the host nation and access to education or to generate new ones such equitable distribution of aid, which can lead to poor health outcomes. Recognizing the possible post-disaster uncertainties among migrants and addressing probable underlying factors might help to build more resilient and healthy communities.

Using the EEG+VR+LEC evaluation method to explore the combined influence of temperature and spatial openness on the physiological recovery of post-disaster populations

Natural disasters have always been a substantial threat to human survival and development. Due to climate change, the number of weather-related disasters has more than doubled worldwide, further emphasizing the need for research on post-disaster environments. However, conducting comprehensive studies in actual disaster settings poses a number of challenges. This study proposes a novel approach that combines electroencephalography (EEG), virtual reality (VR), and laboratory environmental control (LEC) to investigate the joint effects of temperature and spatial openness on the psychological recovery of post-disaster populations. The integration of these techniques provided a controlled and immersive experimental environment for simulating post-disaster scenarios. This study conducted a controlled group experiment focusing on two variables in post-disaster environments: temperature range and spatial openness. Three temperature ranges were selected for the temperature condition: high (30 °C–36 °C), mid (22 °C–28 °C), and low temperature zones (12 °C–20 °C); the temperature environments were controlled by LEC. In terms of spatial openness, three levels were chosen: temporary emergency shelters (enclosed space), the semi-outdoor area of the emergency building (BBBC) designed by the project team (semi-open space), and open natural environments (open space). This research then established a temperature-spatial openness matrix. The study involved emotional recovery experiments for nine patterns of post-disaster environments, using physiological parameters such as EEG, skin conductance, and heart rate, as well as psychological parameters established through subjective questionnaires. The findings of this study will be beneficial for efficiently and reasonably optimizing human-centered spatial environment designs in post-disaster situations featuring limited resources.

SLAM-based Navigation Technology for Rescue Robots in Post-disaster Scenarios

In order to solve the problems of difficulty and high-risk factor in implementing rescue after disasters, this paper designs an intelligent rescue robot autonomous navigation system based on LIDAR synchronous positioning and map building, with a view to achieving autonomous navigation of robots in complex post-disaster scenarios to complete rescue tasks. Firstly, the autonomous navigation system senses the scene by LiDAR and uses gmapping algorithms to construct a map of the post-disaster environment. Secondly, adaptive Monte Carlo localization algorithm is used to achieve robot localization based on radar and odometer data. Then the robot rescue work path is planned to use the Dijkstra algorithm. And TEB local planning path algorithm is used to control the robot. Finally, to verify the reliability of the autonomous navigation system designed in this paper, the ROS system software framework is used as the basis. The SLAM map construction, global path planning, and local real-time obstacle avoidance are tested practically under the scenario to ensure that the autonomous navigation of the mobile robot meets the requirements.

Mixed-reality for quadruped-robotic guidance in SAR tasks

Abstract In recent years, exploration tasks in disaster environments, victim localization and primary assistance have been the main focuses of Search and Rescue (SAR) Robotics. Developing new technologies in Mixed Reality (M-R) and legged robotics has taken a big step in developing robust field applications in the Robotics field. This article presents MR-RAS (Mixed-Reality for Robotic Assistance), which aims to assist rescuers and protect their integrity when exploring post-disaster areas (against collapse, electrical, and toxic risks) by facilitating the robot’s gesture guidance and allowing them to manage interest visual information of the environment. Thus, ARTU-R (A1 Rescue Tasks UPM Robot) quadruped robot has been equipped with a sensory system (lidar, thermal, and RGB-D cameras) to validate this proof of concept. On the other hand, Human-Robot interaction is executed by using the Hololens glasses. This work’s main contribution is the implementation and evaluation of a Mixed-Reality system based on a ROS-Unity solution, capable of managing at a high level the guidance of a complex legged robot through different interest zones (defined by a Neural Network and a vision system) of a post-disaster environment (PDE). The robot’s main tasks at each point visited involve detecting victims through thermal, RGB imaging, and neural networks and assisting victims with medical equipment. Tests have been carried out in scenarios that recreate the conditions of PDE (debris, simulation of victims, etc.). An average efficiency improvement of 48% has been obtained when using the immersive interface and a time optimization of 21.4% compared to conventional interfaces. The proposed method has proven to improve rescuers’ immersive experience of controlling a complex robotic system.

Examining governance of post-earthquake reconstruction planning from an evolutionary resilience perspective

Purpose The politically laden nature of postdisaster recovery calls for more research on its governance, especially at the micro-scale. Apart from engineering-oriented frameworks, researchers need new theoretical underpinnings. This paper aims to review the development of the evolutionary resilience theory and use it as an analytical framework to evaluate the governance of post-earthquake reconstruction planning in China. Design/methodology/approach This paper examines how reconstruction planning is governed in the epicenter town of the 2008 Great Sichuan Earthquake, highlighting three key qualities of evolutionary resilience. The authors draw on site investigations, semistructured interviews and analysis of official and unpublished documents from various sources. Findings This paper finds that despite the absence of specific resilience statements in reconstruction plans of the time, qualities of evolutionary resilience, including social connectedness, flexibility and innovation, were evident in a hybrid and contradictory reconstruction planning system. In this respect, resilience thinking appears in Chinese planning earlier than generally assumed. This paper suggests that this manifestation of resilience was the result of an instrumental utility in addressing socioeconomic uncertainties in the postdisaster environment and, thus, may not be systematic. Originality/value This work enriches the understanding of recovery governance from an evolutionary resilience perspective where existing research is insufficient. It also offers ample practical guidance for similar cases in China and elsewhere.

Post-Colonial Disasters and Narratives of Erasure: Reimagining Testimonies of Toxic Encounter

Mainstream media narratives and the official historiography of the 1984 Bhopal Gas Tragedy have overlooked the testimonies of disaster survivors, culturally discounting the authority of witnessing in both scholarly discourse and public arenas. This has left a space for novelists, as writer-activists, to trace the socio-political, economic and ecological injustices of post-colonial disasters like the Bhopal gas leak. Indra Sinha’s Animal’s People bears witness to such accounts and recreates the night of December 2–3, 1984, through the testimonies of the people surviving the gas disaster. This article identifies Sinha’s narrative technique as ‘eco-testimony’, which strategically revives hitherto undocumented survivor testimonies and their experiences of eco-social exploitation in the post-disaster environment and forges a voice of dissent against the uneven, attritional and necropolitical violence of multinational companies and their chief ally, the neocolonial nation-state.

Autonomous victim detection system based on deep learning and multispectral imagery

Post-disaster environments resulting from catastrophic events, leave sequels such as victims trapped in debris, which are difficult to detect by rescuers in a first inspection. Technological advances in electronics and perception have allowed the development of versatile and powerful optical sensors capable of capturing light in spectrums that humans cannot. new deep learning techniques, such as convolutional neural networks (CNNs), has allowed the generation of network models capable of autonomously detecting specific image patterns according to previous training. This work introduces an autonomous victim detection system to be deployed by using search and rescue robots. The proposed system defines new indexes based on combining the multispectral bands (Blue, Green, Red, Nir, Red Edge) to obtain new multispectral images where relevant characteristics of victims and the environment are highlighted. CNNs have been used as a second phase for automatically detecting victims in these new multispectral images. A qualitative and quantitative analysis of new indexes proposed by the authors has been carried out to evaluate their efficiency in contrast to the state-of-the-art ones. A data set has been generated to train different CNN models based on the best obtained index to analyze their effectiveness in detecting victims. The results show an efficiency of 92% in automatically detecting victims when applying the best multispectral index to new data. This method has also been contrasted with others based on thermal and RGB imagery to detect victims, where it has been proven that it generates better results in situations of outdoor environments and different weather conditions.