Thermal Imagery Research Articles

Ecophysiological variables retrieval and early stress detection: insights from a synthetic spatial scaling exercise

ABSTRACT The ability to access physiologically driven signals, such as surface temperature, photochemical reflectance index (PRI), and sun-induced chlorophyll fluorescence (SIF), through remote sensing (RS) are exciting developments for vegetation studies. Accessing this ecophysiological information requires considering processes operating at scales from the top-of-the-canopy to the photosystems, adding complexity compared to reflectance index-based approaches. To investigate the maturity and knowledge of the growing RS community in this area, COST Action CA17134 SENSECO organized a Spatial Scaling Challenge (SSC). Challenge participants were asked to retrieve four key ecophysiological variables for a field each of maize and wheat from a simulated field campaign: leaf area index (LAI), leaf chlorophyll content (C ab), maximum carboxylation rate (V cmax,25), and non-photochemical quenching (NPQ). The simulated campaign data included hyperspectral optical, thermal and SIF imagery, together with ground sampling of the four variables. Non-parametric methods that combined multiple spectral domains and field measurements were used most often, thereby indirectly performing the top-of-the-canopy to photosystem scaling. LAI and C ab were reliably retrieved in most cases, whereas V cmax,25 and NPQ were less accurately estimated and demanded information ancillary to RS imagery. The factors considered least by participants were the biophysical and physiological canopy vertical profiles, the spatial mismatch between RS sensors, the temporal mismatch between field sampling and RS acquisition, and measurement uncertainty. Furthermore, few participants developed NPQ maps into stress maps or provided a deeper analysis of their parameter retrievals. The SSC shows that, despite advances in statistical and physically based models, the vegetation RS community should improve how field and RS data are integrated and scaled in space and time. We expect this work will guide newcomers and support robust advances in this research field.

Integrating Thermal Infrared Imaging and Weather Data for Short-Term Prediction of Building Envelope Thermal Appearance

This study presents a novel deep-learning framework for predicting the thermal appearance of building envelopes under varying weather conditions based on a new dataset collected using a thermal infrared camera at 10 min intervals over a one-and-a-half-year period. Unlike existing studies that rely on simulated data or physical models that do not always accurately reflect the complex heat transfer processes in real buildings, we have collected a large dataset showing how a building behaves under different climatic conditions. We propose a novel deep-learning approach that integrates weather data and thermal imagery to predict the temperature distribution on the building façade for the next 24 and 48 h. The model uses a state-of-the-art recurrent neural network architecture, PredRNN V2, with an action conditioning mechanism to incorporate weather forecasting data into the prediction process. We evaluate this approach in terms of average accuracy, prediction accuracy in specific regions, and visual-perceptual performance of the images. The proposed framework achieves a prediction accuracy of 1.5 °C (root mean square error—RMSE) for the 24 h prediction and 2.04 °C (RMSE) for the 48 h prediction, outperforming baseline models in terms of temperature prediction accuracy and structural similarity of the predicted images.

Water, salt, and heat in raised field agriculture: Using hydrologic modeling and thermal imagery to investigate soil drainage and temperature dynamics

On the north coast of Peru in the Casma Valley exist relict raised agricultural field systems dating to the Late Intermediate Period Chimu Empire (ca. 1300 – 1470 CE). While similar in many ways to other inland and coastal raised fields in South America, these fields are relatively unique in climate, weather patterns, and layout. The topography and hydrology of the Casma Valley provide clues on how these fields operated and why they differ from the regional norm. This paper reports satellite and drone-based aerial reconnaissance results used to collect multiscalar data for flow modeling and thermal photogrammetry. Flow modeling methods are compared to determine the best way to gain insight into surface hydrology using only elevation data, and thermal photogrammetry is used to analyze temperature dynamics in the raised fields. These data results provide insight into the function of the field system and its unique field morphology.

Utility of a Smart Device Infrared Camera in Localizing Acute Pediatric Long Bone Fractures: A Pilot Study.

Musculoskeletal injuries are one of the top 10 reasons children present to the emergency department (ED). Infrared thermal imaging (IRT) is a noninvasive and nonradiating imaging modality that can detect subtle temperature differences. IRT may be used to detect the presence of musculoskeletal injury. This study aimed to assess the utility of a smart device infrared camera attachment's ability to localize acute pediatric long bone fractures in patients less than 5 years of age. This was a prospective pilot study comparing thermal imagery on an injured extremity to radiographs. All IRT images were obtained via an iPad with an infrared camera attachment. Using two different IRT images (Hi/Lo and Span and Level), the area of maximum temperature (Tmax) was identified and compared to radiographs. A total of 31 patients were enrolled in the study. Twenty-four (77.50%) were identified as having fractures, and 7 (22.50%) did not have any fractures. IRT correctly identified an injury in the fracture group 91.67% of the time when using the Span and Level IRT image compared to standard x-rays (P < 0.0002). When using the Span and Level image to identify Tmax to localize a fracture on x-ray, sensitivity is 0.92, specificity is 0.86, positive predictive value is 0.96, and negative predictive value is 0.75. A receiver operating characteristic (ROC) curve was completed with area under the curve (AUC) being 0.89. This pilot study shows that the use of smart device infrared camera attachments is feasible and has promising results in fracture localization. This could allow for a decrease in radiographs and be particularly useful in resource-limited areas.

U-Net architecture for high-resolution thermal image segmentation: Enhancing CFD models in smart HVAC design

This paper presents an enhanced U-Net architecture for accurately segmenting high-resolution thermal images to facilitate the design of intelligent Heating, Ventilation, and Air Conditioning (HVAC) systems using Computational Fluid Dynamics (CFD) models. Indoor thermal comfort for occupants will improve under the same energy load by imposing comfort demands. Segmented data is fed into CFD simulations using a U-Net variant for thermal imagery. Integrating the segmented thermal data into the CFD simulations increases the model’s accuracy by 15 % for segmentation and 20% for prediction when compared to a traditional U-Net architecture. Experimental findings demonstrate the potential to achieve energy savings of up to 25 % in the simulated scenario by using the trained and validated model on a dataset of 1,500 high-resolution thermal images of various building styles. Compared to existing methods, the proposed method saves 30% in total simulation time for complex building models. Deep learning enables building the next generation of HVAC systems by pushing the boundaries of energy-efficient and environmentally friendly buildings. This study shows that advanced computer vision can be leveraged to create more innovative and sustainable built environments, leading to drastic improvements in energy efficiency.

Comparison of thermal cameras and human observers to estimate population density of fallow deer (Dama dama) from aerial surveys in Tasmania, Australia

Context The population of introduced fallow deer (Dama dama) is thought to have increased exponentially across much of the island of Tasmania, Australia, since 2000. Historically, deer management decisions have relied on population trend data from vehicular spotlight surveys. Renewed focus on the contemporary management of the species requires development of more robust and precise population estimation methodology. Aims This study demonstrates two aerial survey methods – conventional counts by trained human observers, and thermal imaging footage recorded during the same flights – to inform future survey practices. Methods Conventional counts were carried out by three observers, two seated on the left side of the helicopter, and one on the right. A high-resolution thermal camera was fitted to the helicopter and was orientated to meet the assumptions of distance sampling methodologies. Both survey methods were used to generate deer population density estimates. Spatial distribution of deer was also analysed in relation to patches of remnant native vegetation across an agricultural landscape. Mark–recapture distance sampling was used to estimate density from human observer counts and provide a comparison to the distance sampling estimates derived from the thermal camera. Key results Human observer counts gave a density estimate of 2.7 deer per km2, while thermal camera counts provided an estimate of 2.8 deer per km2. Deer population density estimates calculated via both methods were similar, but variability of the thermal camera estimate (coefficient of variation (CV) of 36%) was unacceptably high. Human observer data was within acceptable bounds of variability (CV, 19%). The estimated population size in central and north-eastern Tasmania for 2019 approximated 53,000 deer. Deer were primarily congregated within 200 metres of the interface between canopy cover and open pasture. Conclusions The population density estimate provides a baseline for monitoring and managing the Tasmanian deer population. Human observer data was more precise than thermal camera data in this study, but thermal counts could be improved by reducing sources of variability. Implications Improvements for the collection of thermal imagery are recommended. Future control efforts may be more efficient if they preferentially target habitat edges at this time of year, paired with random or grid-based searches where population density is lower.

A novel lightweight deep learning framework with knowledge distillation for efficient diabetic foot ulcer detection

Diabetic Foot Ulcers (DFUs) are a critical healthcare issue requiring early detection. Although deep learning models show promise in DFU diagnosis, their large parameter sizes and high computational demands limit their practical use. This work addresses these challenges by introducing DFU-LWNet, a lightweight, knowledge-distilled model that maintains or even surpasses the performance of existing models while drastically reducing computing costs. The DFU-LWNet model synergistically integrates the MBConv Block, leveraging its capabilities in feature expansion, depthwise convolution for parameter efficiency, and adaptive feature recalibration through the Squeeze-and-Excitation (SE) mechanism. The methodology commences with the fine-tuning of seven pre-trained models, ultimately selecting InceptionV3 as the optimal teacher model based on experimental results. Subsequently, knowledge distillation (KD) is applied, leveraging the insights from this robust teacher model to enhance the performance of the student model, DFU-LWNet. The primary objective is to reduce the student model's parameter size and computational complexity while maintaining diagnostic accuracy. Through comprehensive evaluations, DFU-LWNet achieves an impressive classification accuracy of 96.23 % on a publicly available dataset comprising 1055 foot images. Notably, this achievement is accomplished with a remarkably low parameter size of only 0.49 million, consuming 2MB of disk space. This reduction in parameter size not only improves inference speed but also significantly reduces computational costs compared to pre-trained models. Moreover, the study employs Grad-CAM visualizations to demonstrate the model's enhanced saliency and interpretability in predictions post-KD, thereby offering valuable insights to clinicians. Additionally, the efficacy of DFU-LWNet extends to thermal imagery domains, as validated by experiments on a diverse dataset of thermal images of diabetic foot cases. The proposed approach achieves remarkable accuracy of 93.13 % on a separate thermal image dataset, demonstrating cross-domain success and reinforcing the model's practical utility in real-world scenarios. Compared to existing state-of-the-art methods, this work demonstrates promising practical viability. DFU-LWNet’s minimal parameter count and low prediction times make it well-suited for deployment in resource-limited healthcare settings.

Deep Learning-Driven Thermal Imaging Hotspot Detection in Solar Photovoltaic Arrays Using YOLOv10

The effective management of solar photovoltaic (PV) arrays is vital to maximising energy generation and ensuring long-term performance reliability. The presence of faults, or partial shading, can give rise to hotspots in PV arrays, making their detection critical for timely maintenance and avoiding further impacts on their performance. This paper describes a new approach to the hotspot detection issue in thermal images of solar PV arrays by leveraging deep learning. This study employs the YOLOv10 (You Only Look Once, version 10) model to deliver very high accuracy and speed in identifying and localising hotspots under defined regions of interest (ROIs). The proposed method begins with taking thermal images from multiple solar PV installations while capturing a range of operational conditions and fault types. These images are annotated with hotspot regions to create a robust training dataset. Given YOLO's capability for real-time object detection with high precision and mean average precision (mAP), the YOLOv10 model is then implemented to train on this dataset. Multiple changes were made to the YOLOv10 hyperparameters to optimise them for detecting hotspots in thermal images. The experimental scenario of this paper demonstrates that this approach indeed performs significantly better than standard image processing methods as well as prior deep learning models for detecting both hotspot accuracy as well as speed of processing the images. The YOLOv10 method demonstrated the highest available classification performance with a mAP at intersection over union (IoU) of 0.5 accuracy of 0.91 with inference time suitable for real-time applications. The sample results demonstrated the model's continuous ability to detect hotspots and provide location data under various conditions, such as crossing through different times of day and weather. The results of this study demonstrate that YOLOv10 can be used for improved detection of hotspots in the explicably thermal imagery of solar PV arrays.

Integrating thermal and optical models with legacy excavation data for spatial inventories

Researching at previously excavated archaeological sites comes with its own challenge set, which includes ground-truthing previously unearthed architectural features to correlate old and new trenches. This study presents a comprehensive examination technique to approach architectural features at mound-type archaeological settlements through the synergistic utilization of thermal and optical imagery obtained via remote sensing technologies. Employing an interdisciplinary approach involving various types of imagery (optical, thermal, and radar) embedded in archaeological data, this research aims to reevaluate legacy spatial data and enhance the understanding of subsurface archaeological features by exploiting the complementary advantages of both thermal and optical sensors.This study involves the acquisition, processing, and integration of Unmanned Aerial Vehicle (UAV)-based thermal and optical datasets to derive a complete representation of archaeological landscapes. Ground-truthing and validation exercises are incorporated to ensure the accuracy and reliability of the results. The results contribute to advancing the field of archaeological remote sensing and highlight the significance of integrating diverse sensing modalities for a comprehensive and nuanced interpretation of archaeological landscapes. The techniques exemplified with two mound-type case sites help to generate high-resolution mappings which provide tools to increase the accuracy of on-site legacy data and help digital protection of the sites by generating digital twins. This presents a less costly method than conventional photogrammetric techniques, making it feasible and advantageous for field archaeologists or researchers dealing with legacy spatial data.

The role of building mass configuration and material selection for mitigating the intensity of the urban heat island

The knowledge of urban heat island (UHI) mitigation for the materials and geometry of urban areas requires enrichment. This study compares to the thermal conditions on two different building mass configurations (Superblock and horizontal residential areas), that involve geometry aspect (density, orientation) and materials aspect (softscape, pavement, roof, wall). A surface urban heat island (SUHI) survey was conducted using Landsat thermal imagery, followed by ground surveys in selected areas. Data collected in both areas included air temperature, relative humidity, air velocity, and globe temperature area. The results show that the superblock configuration had a low SUHI intensity due to the proportional orientation of East-West-North-South (0.56–0.44) and large shading (average of 0.31 m2). In contrast, in the dense residential areas, the intensity of SUHI was high because the proportion of East-West walls was wider than North-South (0.8–0.2), and the shading was low (average of 0.15–0.16 m2). Façade and ground heating occurred owing to the use of heavy-weight materials, which was indicated in the thermal imagery, with a solar heating index in a vertical cluster mass configuration of 21.7 and a horizontal grid of 55.9. Convective cooling occurred in a vertical geometry because the buildings generated turbulence that removed warm air, whereas it did not occur in a horizontal geometry. The convective cooling index for the vertical cluster mass configuration was 0.62, whereas that for the horizontal grid configuration was 0.09. This indicates that regional planning must incorporate geometric and material considerations to control the urban thermal environments.

Beyond the visible: thermal data for facial soft biometric estimation

In recent years, the estimation of biometric parameters from facial visuals, including images and videos, has emerged as a prominent area of research. However, the robustness of deep learning-based models is challenged, particularly in the presence of changing illumination conditions. To overcome these limitations and unlock new opportunities, thermal imagery has arisen as a viable alternative. Nevertheless, the limited availability of datasets containing thermal data and the small amount of annotations on them limits the exploration of this spectrum. Motivated by this gap, this paper introduces the Label-EURECOM Visible and Thermal (LVT) Face Dataset for face biometrics. This pioneering dataset includes paired visible and thermal images and videos from 52 subjects along with metadata of 22 soft biometrics and health parameters. Due to the reduced number of existing datasets in this domain, the LVT Face Dataset aims to facilitate further research and advancements in the utilization of thermal imagery for diverse eHealth applications and soft biometric estimation. Moreover, we present the first comparative study between visible and thermal spectra as input images for soft biometric estimation, namely gender age and weight, from face images on our collected dataset.

Can Multiscale Thermal Infrared Imaging Help Validate and Monitor Water Stress in Alluvial Forests?

ABSTRACTAlluvial forests are sensitive to drought induced by climate change and exacerbated by altered flow regimes. Our ability to detect and map their sensitivity to drought is crucial to evaluate the effects of climate change and adjust management practices. Therefore, we explore the potential of multiscale thermal infrared imagery (TIR) to diagnose their sensitivity to droughts. In summer 2022, we sampled leaves and phloem on Populus nigra trees from two sites with contrasted hydrological connectivity along the Ain River (France) to investigate the seasonality of water stress and act as ground truth for airborne TIR images. To map forest sensitivity to drought, we used TIR data from four airborne campaigns and Landsat archives over a larger spatial and temporal extent. Field data showed that stress conditions were reached for both sites but were higher in the site with lower groundwater connectivity, which was also the case for individual tree crown temperatures. At the forest plot scale, canopy temperature was linked to forest connectivity for two of four TIR campaigns, with higher values in the more degraded reaches. Landsat data were used to locate the areas of the riparian forest impacted by a historical drought event and monitor their recovery and proved useful to identify trends. TIR data showed promising results to help detect and map tree water stress in riparian environments. However, stress is not detected in all TIR campaigns, demonstrating that one‐shot TIR acquisitions alone are not enough to diagnose stress and complementary in‐field eco‐physiological measurements are necessary.

Caught between two states: The compromise in acclimation of photosynthesis, transpiration and mesophyll conductance to different amplitudes of fluctuating irradiance.

While dynamic regulation of photosynthesis in fluctuating light is increasingly recognized as an important driver of carbon uptake, acclimation to realistic irradiance fluctuations is still largely unexplored. We subjected Arabidopsis thaliana (L.) wild-type and jac1 mutants to irradiance fluctuations with distinct amplitudes and average irradiance. We examined how irradiance fluctuations affected leaf structure, pigments and physiology. A wider amplitude of fluctuations produced a stronger acclimation response. Large reductions of leaf mass per area under fluctuating irradiance framed our interpretation of changes in photosynthetic capacity and mesophyll conductance as measured by three separate methods, in that photosynthetic investment increased markedly on a mass basis, but only a little on an area basis. Moreover, thermal imagery showed that leaf transpiration was four times higher under fluctuating irradiance. Leaves growing under fluctuating irradiance, although thinner, maintained their photosynthetic capacity, as measured through light- and CO2-response curves; suggesting their photosynthesis may be more cost-efficient than those under steady light, but overall may incur increased maintenance costs. This is especially relevant for plant performance globally because naturally fluctuating irradiance creates conflicting acclimation cues for photosynthesis and transpiration that may hinder progress towards ensuring food security under climate-related extremes of water stress.

Evaluating the utility of combining high resolution thermal, multispectral and 3D imagery from unmanned aerial vehicles to monitor water stress in vineyards

PurposeHigh resolution imagery from unmanned aerial vehicles (UAVs) has been established as an important source of information to perform precise irrigation practices, notably relevant for high value crops often present in semi-arid regions such as vineyards. Many studies have shown the utility of thermal infrared (TIR) sensors to estimate canopy temperature to inform on vine physiological status, while visible-near infrared (VNIR) imagery and 3D point clouds derived from red–green–blue (RGB) photogrammetry have also shown great promise to better monitor within-field canopy traits to support agronomic practices. Indeed, grapevines react to water stress through a series of physiological and growth responses, which may occur at different spatio-temporal scales. As such, this study aimed to evaluate the application of TIR, VNIR and RGB sensors onboard UAVs to track vine water stress over various phenological periods in an experimental vineyard imposed with three different irrigation regimes.MethodsA total of twelve UAV overpasses were performed in 2022 and 2023 where in situ physiological proxies, such as stomatal conductance (gs), leaf (Ψleaf) and stem (Ψstem) water potential, and canopy traits, such as LAI, were collected during each UAV overpass. Linear and non-linear models were trained and evaluated against in-situ measurements.ResultsResults revealed the importance of TIR variables to estimate physiological proxies (gs, Ψleaf, Ψstem) while VNIR and 3D variables were critical to estimate LAI. Both VNIR and 3D variables were largely uncorrelated to water stress proxies and demonstrated less importance in the trained empirical models. However, models using all three variable types (TIR, VNIR, 3D) were consistently the most effective to track water stress, highlighting the advantage of combining vine characteristics related to physiology, structure and growth to monitor vegetation water status throughout the vine growth period.ConclusionThis study highlights the utility of combining such UAV-based variables to establish empirical models that correlated well with field-level water stress proxies, demonstrating large potential to support agronomic practices or even to be ingested in physically-based models to estimate vine water demand and transpiration.

Advanced Unmixing Methodologies for Satellite Thermal Imagery: Matrix Changing and Classification Insights from ASTER and Landsat 8–9

Advanced Unmixing Methodologies for Satellite Thermal Imagery: Matrix Changing and Classification Insights from ASTER and Landsat 8–9

How advection affects the surface energy balance and its closure at an irrigated alfalfa field

Orbiting around the non-closure problem in eddy covariance, a new generation of high-resolution thermal imagery has revealed that advection may be more common than previously expected. To investigate this, we conducted an extensive study over an irrigated alfalfa field that experienced heat and moisture advection. Over the course of five analysis periods (37 days total), multiple tower arrays and profile measurements were deployed to measure the horizontal advection and vertical heat flux divergence. Latent heat flux (λE) measured at the anchor tower showed an enhancement (i.e., increase) due to both local and non-local processes. Locally, as a result of the upwind λE, advection humidified the atmosphere and increased stomatal opening, enhancing the downwind λE. Simultaneously, with lowered atmospheric demand, λE was suppressed downwind. Our results suggest that stomatal regulation played a dominant role in the enhancement, but not by itself. Spectral analysis revealed that low frequency (i.e., large) eddies contributed high heat and moisture via advection. In combination with thermal remote sensing observations from ECOSTRESS and Landsat 8/9, we found that these large eddies were generated over the upwind surface, and they were independent of the local boundary layer conditions. Consequently, spatiotemporal heterogeneity in land-surface conditions induced large eddies, further enhancing λE through non-local transport of heat and moisture. Lastly, by conditionally including the advective fluxes, the energy balance closure improved from 89 % to 97 % (r2 = 0.97, p < 0.001) over the five analysis periods. Results from this improved energy balance closure suggest an alternative approach for developing validation datasets for remote sensing evapotranspiration (ET) models rather than forcing closure with Bowen-ratio. Furthermore, our findings provide insights for algorithms that may improve remote sensing ET products that treat pixels as isolated columns rather than also considering the lateral effects of heat and moisture transport.

Advancing primate surveillance with image recognition techniques from unmanned aerial vehicles.

Using unmanned aerial vehicles (UAVs) for surveys on thermostatic animals has gained prominence due to their ability to provide practical and precise dynamic censuses, contributing to developing and refining conservation strategies. However, the practical application of UAVs for animal monitoring necessitates the automation of image interpretation to enhance their effectiveness. Based on our past experiences, we present the Sichuan snub-nosed monkey (Rhinopithecus roxellana) as a case study to illustrate the effective use of thermal cameras mounted on UAVs for monitoring monkey populations in Qinling, a region characterized by magnificent biodiversity. We used the local contrast method for a small infrared target detection algorithm to collect the total population size. Through the experimental group, we determined the average optimal grayscale threshold, while the validation group confirmed that this threshold enables automatic detection and counting of target animals in similar datasets. The precision rate obtained from the experiments ranged from 85.14% to 97.60%. Our findings reveal a negative correlation between the minimum average distance between thermal spots and the count of detected individuals, indicating higher interference in images with closer thermal spots. We propose a formula for adjusting primate population estimates based on detection rates obtained from UAV surveys. Our results demonstrate the practical application of UAV-based thermal imagery and automated detection algorithms for primate monitoring, albeit with consideration of environmental factors and the need for data preprocessing. This study contributes to advancing the application of UAV technology in wildlife monitoring, with implications for conservation management and research.

Advancements in Li-Ion Battery Safety: A Multiphysics and Deep Learning Approach for Thermal Runaway Prediction

The increasing adoption of Lithium-ion batteries (LIBs) in various applications necessitates an enhanced focus on safety, particularly concerning the thermal runaway (TR) phenomenon. This study introduces a novel framework combining multiphysics modeling and deep learning (DL) to predict TR in cylindrical Li-ion batteries. The primary objective is to prevent catastrophic battery failures by identifying critical thermal conditions leading to TR.Our methodology integrates state-of-the-art convolutional neural networks (CNNs) and the "You Only Look Once" (YOLO) object detection model to classify and detect the evolution of TR stages in LIBs. The CNNs, trained on simulated thermal images generated through comprehensive multiphysics modeling, classify TR into three distinct stages: safe operation, critical condition, and actual TR occurrence. The YOLO model further pinpoints the location of TR heat sources, enhancing the predictive accuracy.The multiphysics model employed captures the coupled thermal and electrochemical behavior of LIBs, focusing on the solid electrolyte interface (SEI) formation/decomposition on the anode as the primary degradation mechanism. The simulated thermal imagery, reflecting real-time battery conditions, serves as a dataset for training our DL models. This combination of physics-based and data-driven approaches allows for a detailed understanding of TR dynamics.In our study, we also explored the efficiency of other neural network architectures, notably ResNet, EfficientNet, and MobileNet for their potential in TR prediction. ResNet, known for its deep residual learning framework, demonstrated proficiency in handling the complex, layered features of thermal images. EfficientNet, optimized for efficiency and accuracy, showed promise in its scalable architecture and balanced depth, width, and resolution, making it a viable alternative for real-time TR detection. The MobileNet known for its lightweight architecture and efficiency on mobile devices, was evaluated for its capability to process thermal images in a resource-constrained environment. It demonstrated a remarkable balance between computational efficiency and predictive accuracy, making it a viable option for real-time TR monitoring in portable LIB applications.While MobileNet, ResNet, and EfficientNet each showed potential in their respective capacities, our primary CNN-YOLO framework emerged as the most effective for TR prediction in this study. Its proficiency in extracting spatial features and detecting anomalies in real-time was critical for early TR identification in LIBs.Results demonstrate high accuracy in predicting TR stages and heat source locations, emphasizing the ability of the integrated approach. The study highlights the importance of thermal management and real-time monitoring in LIBs, offering insights into the thermal behavior under various load conditions. The predictive framework not only advances the fundamental understanding of TR in LIBs but also provides practical implications for enhancing battery safety in applications like electric vehicles and consumer electronics.In conclusion, this work presents a significant stride in battery research, paving the way for safer and more reliable energy storage solutions. The combined multiphysics modeling and DL framework offer a robust tool for predicting and preventing TR in LIBs, contributing to the broader goal of advancing battery technology in the modern energy landscape.

Object-based strategy for generating high-resolution four-dimensional thermal surface models of buildings based on integration of visible and thermal unmanned aerial vehicle imagery

Object-based strategy for generating high-resolution four-dimensional thermal surface models of buildings based on integration of visible and thermal unmanned aerial vehicle imagery

Effects of repeated drawdown flushing on riverbed fine sediment dynamics downstream from a dam

Sediment accumulation in reservoirs is frequently a problem, impelling dam managers to implement strategies such as drawdown flushing to limit siltation. Drawdown and other sediment removal methods may induce riverbed clogging downstream of dams, especially in river sections where water is diverted, thereby reducing transport capacity. In this study, we investigate the effects of drawdown flushing downstream of the Plan d’Arem run-of-the-river dam (Upper Garonne, Spain – France border) using an adaptive inter-site comparison strategy to consider a range of spatial and temporal conditions, which allow us to separate the effects of dam storage and flushing from other potential factors. Drawdown flushing has been undertaken three times during the study period over a short span of time (2 months). We couple bed material sampling, which provides direct information on bed composition, with airborne infrared thermal imaging to better interpret whether fine sediment interstitial storage within the bed is associated with clogging. We also measure bed surface grain size and bed mobility in order to investigate their potential role in controlling fine sediment dynamics. We identify surface grain size and water diversion as the main factors controlling fine sediment spatial distribution, with coarse-grained bed-surfaces and by-passing promoting fine sediment enrichment. As a result, sites located upstream and within the by-passed reached of the Plan d’Arem dam show higher fine sediment interstitial storage than sites downstream from the by-passed reach. Results from thermal imagery demonstrate such interstitial storage does not induce clogging effect. The reaches with the most important sediment storage host a high number of cool-water patches, indicating water exchanges with the hyporheic zone. Post-flushing bed composition indicates systematic export of fine sediments from the bed matrix at all sites after the first operation, afterwards fine levels remain low after the second and the third flushing. The low impact of dam flushing in term of clogging is interpreted by the fact that the interstitial material is very sandy.