EECS507_Thermal_Smart_City_Sensing_Survey

Survey of Thermal Sensing for Smart Cities Ping-Huai Tseng [email protected] University of Michigan Ann Arbor, Michigan, USA Yen-Cheng Lin [email protected] University of Michigan Ann Arbor, Michigan, USA Grant Wang [email protected] University of Michigan Ann Arbor, Michigan, USA Anna Huang [email protected] University of Michigan Ann Arbor, Michigan, USA Dimitri Mili [email protected] University of Michigan Ann Arbor, Michigan, USA Abstract Thermal sensing has become an increasingly important component of modern urban sensing systems as cities face growing challenges in management. While traditional approaches to city sensing have relied heavily on RGB imagery, environmental sensors, and fixed or mobile sensing platforms, these methods remain limited by lighting conditions, privacy concerns, and incomplete visibility into heat-related phenomena. Recent advances in thermal hardware, radiometric modeling, image enhancement, and AI-driven interpretation have significantly expanded the potential of thermal sensing for city-scale applications. This survey documents the evolution of thermal sensing from early infrared measurement systems to contemporary smart-city deployments, highlighting deployments in sensor technology, calibration, machine learning, and multimodal data fusion. Many different use cases for modern thermal sensing are surveyed, including Urban Heat Island analysis and mobile UAV-based heat monitoring. This survey also intends to discuss the emerging challenges and future directions for establishing thermal sensing as a foundational component for next-generation city-sensing infrastructure. 1 Introduction The rapid growth of urban populations and the increasing complexities of metropolitan infrastructures in the 21st century have intensified the need for data-driven approaches to city management. As cities aim to become more sustainable and livable, sensing technologies have shown to be cornerstones of these initiatives. Among these technologies, thermal sensing, which captures spatial and temporal heat patterns through infrared, thermal imaging, and temperature-monitoring systems, plays a uniquely important role. Unlike conventional visual or environmental sensors, thermal sensors provide a continuous insight into the heat distribution, energy usage, human activity, and environmental conditions of a city. This enables a broad range of applications from public safety monitoring to traffic and emissions analysis, which impacts policy decisions that affect millions of people daily. In recent years, advances in low-power embedded platforms, wireless sensor networks, and machine learning have expanded the capabilities and affordability of thermal sensing. Despite the progress, thermal sensing technologies and their applications in city sensing remains fragmented. Most of the existing studies focus on specific subtopics in city sensing applications without uniformity in deployment strategies and approaches. This survey provides an overview of previous methodologies used in city sensing, encompassing an overview of general methods and computer vision applications in city sensing. The following sections discuss the current work in thermal sensing and its applications in city sensing before identifying challenges to overcome in this field. 2 Previous Methods Used in City Sensing 2.1 Fixed Infrastructure Sensing Early urban sensing efforts centered on fixed wireless sensor networks designed to deliver continuous, localized measurements of environmental and infrastructural conditions. These systems demonstrated how large collections of static nodes could be networked into reliable, long-lived deployments, often focusing on issues such as calibration consistency, communication reliability, fault tolerance, and energy management [ 1 ]. By placing sensors at key locations throughout a city, researchers were able to capture persistent data streams that offered unprecedented insight into urban dynamics [18]. As these deployments scaled, citywide testbeds revealed broader operational challenges associated with maintaining dense sensing infrastructure in real environments [ 10 ]. Node placement, remote reconfiguration, and multi-application resource sharing became central concerns, shaping the architectural principles used in later smart-city platforms [ 37 ]. Surveys of this early period further emphasized the tension between achieving fine-grained spatial coverage and managing the substantial installation and maintenance demands of fixed infrastructure [ 28 , 47 ]. Despite these limitations, fixed sensor networks established the foundational sensing and networking paradigms on which subsequent urban-sensing methods would build. 2.2 Mobile, Opportunistic, and Participatory Sensing As sensing needs expanded beyond the reach of fixed infrastructure, researchers turned to mobile and participatory approaches that leveraged the widespread use of smartphones and other personal devices. Early work distinguished between opportunistic sensing—where data are collected passively during everyday device use—and participatory sensing, which relies on explicit user involvement, outlining clear tradeoffs in coverage and user burden [ 30 ]. These methods showed that urban sensing could scale naturally with human mobility, enabling richer spatial–temporal Ping-Huai Tseng, Yen-Cheng Lin, Grant Wang, Anna Huang, and Dimitri Mili data collection. Parallel research demonstrated that mobile phone data could support large-scale inference about population movement, environmental conditions, and urban activity, significantly extending the capabilities of static sensor networks [17]. The growing reliance on mobile contributors also raised challenges related to data privacy, security, and reliability. Efforts to address these concerns introduced privacy-preserving aggregation techniques and frameworks for evaluating the trustworthiness of user-provided information [ 65 ]. In addition to personal devices, studies leveraged buses, taxis, and other vehicles as mobile sensing platforms, demonstrating substantial improvements in coverage and reducing the need for extensive fixed infrastructure [ 8 , 58 ]. Mobile sensing further proved effective for environmental monitoring, where vehicle-mounted low-cost sensors produced fine-grained air-quality measurements beyond what static deployments could achieve [ 27 , 44 ]. Together, these approaches established mobile and participatory sensing as scalable, flexible methods for capturing diverse urban phenomena. 2.3 Aerial and Distributed Autonomous Sensing Aerial sensing emerged as a way to overcome many of the spatial and visibility limitations inherent in ground-based methods. Drone-based platforms demonstrated that unmanned aerial vehicles could rapidly survey large areas, capture data from elevated vantage points, and access locations that static or mobile ground sensors could not reach. These systems highlighted advantages in dynamic repositioning, flexible coverage, and responsiveness to real-time sensing needs, making them well-suited for applications such as traffic observation, parking detection, and environmental inspection [ 11 ]. Their ability to operate above urban obstructions provided a richer spatial context and improved continuity in data collection across complex city landscapes. Building on these capabilities, research into distributed autonomous sensing advanced the use of micro-aerial vehicles and coordinated swarms for more sophisticated urban monitoring tasks. These platforms explored how lightweight aerial robots could selforganize, navigate GPS-limited environments, and perform coverage tasks with minimal human intervention. Work in this area investigated challenges such as swarm coordination, localization, robustness, and energy-efficient coverage planning, demonstrating how autonomous aerial systems could adapt to changing conditions and sensing requirements [ 50 ]. Together, these aerial and autonomous approaches expanded the operational envelope of urban sensing, offering highly flexible, resilient, and scalable alternatives to both fixed and mobile ground-based sensing infrastructures. 3 Computer Vision and Deep Learning use in City Sensing The use of computer vision and deep learning models for urban sensing is a largely explored and expanding area. They have become a central tool for extracting actionable urban information from visual data. Whether the data is in the form of remote sensing or street-level sensing, computer vision has been at the forefront. Today, deep learning algorithms can automatically analyze vast amounts of visual data from satellites, drones, street cameras, and mobile devices to extract meaningful insights about urban conditions. The combination of readily available imagery, powerful deep learning frameworks, and cloud computing infrastructure has made sophisticated urban analysis accessible to cities of all sizes. 3.1 Remote Sensing Applications Remote sensing has experienced significant improvements through deep learning adoption. Most deep learning architectures designed for multimedia vision can be successfully applied to remote sensing optical images, resulting in notable improvements in vehicle detection, semantic labeling, and land cover classification [ 6 , 36 ]. Building detection has become one of the most advanced areas of remote sensing deep learning applications. Convolutional neural networks have demonstrated remarkable effectiveness in automatically identifying and extracting building boundaries from high-resolution aerial and satellite images [ 42 ]. Recent research has achieved impressive accuracy rates in this detection [ 25 ], allowing for the support of critical applications including urban planning, disaster management, and 3D city modeling. 3.2 Street Level Image Analysis and Applications Street level image analysis represents a major shift from the traditional aerial remote sensing by providing a ground level perspective that captures the urban environment that is experienced by people. Advances in computer vision research has allowed for researchers to gain meaningful insights from existing data sets. These street-level approaches complement satellite imagery by providing detailed information about vegetation coverage, trash levels, and human activities that are not visible from above [12, 16]. Deep learning models have demonstrated remarkable capabilities in predicting socioeconomic characteristics from street-level visual features alone. Notable research has shown that computer vision algorithms can extract car types from street imagery and correlate vehicle values with neighborhood demographics, achieving statistically significant relationships with census data and voting patterns [ 4 ]. Others have used similar techniques of street level sensing to find tents and other makeshift shelters in order to locate and track homeless encampments [ 3 ]. Another crucial task for computer vision that has emerged is urban scene segmentation. Deep learning models can automatically divide urban landscapes into semantically significant regions such as buildings, roads, vegetation, and infrastructure components [12]. 3.3 Challenges of Using Computer Vision Despite significant advances in both remote sensing and street level analysis, they both face persistent challenges. In the context of remote sensing, variations caused by atmospheric conditions, weather, cloud cover, scale, and spatial resolution can cause inconsistencies that can cause a decrease in model performance. [ 25 ]. Street level sensing presents some different challenges including clutter and variability introduced by weather, seasonality, and camera placement. For example, one trash detection model was shown to misidentify objects when scattered light present in shadows, scattered leaves were present, or there were abnormal stains or marks on trees or the road [24]. Survey of Thermal Sensing for Smart Cities Across both approaches of city sensing, one of the most prevalent challenges is inconsistent lighting. This challenge, whether caused by time of day, weather, or artificial lighting, can significantly impact object detection and segmentation accuracy. These challenges highlight the limitations of relying solely on RGB based computer vision methods in urban environments, motivating the exploration of alternative sensing methods such as thermal sensing. 4 Current Work in Thermal Sensing Thermal sensing refers to the measurement and interpretation of long-wave infrared (LWIR) radiation naturally emitted by objects above absolute zero. Unlike visible-light (RGB) imaging, thermal sensing does not depend on ambient illumination. Instead, it captures spatial temperature patterns that reflect material properties, human activity, and environmental conditions. Figure 1: Comparison of (a) Visible-Light (RGB) and (b) Thermal Images [21] Thermal sensing technologies have advanced significantly in recent years, evolving from basic infrared radiation measurement into a sophisticated fusion of sensor hardware, radiometric modeling, image processing, and machine learning. These developments aim to improve temperature accuracy, spatial resolution, stability under varying environmental conditions, and scalability for largearea deployment. Thermal imaging is inherently non-intrusive and does not rely on background light, offering superior performance in low-light conditions and providing an important advantage in privacy protection compared to visible light (RGB) imaging [ 61 ]. This section reviews contemporary advances in thermal sensing from a technological perspective, providing the foundation upon which thermal-based smart city applications are later built. 4.1 Radiometric Modeling Modern thermal sensing relies on detecting long-wave infrared emissions naturally radiated by objects. The measured radiance is converted into temperature through physical models governed by Planck’s law and the Stefan–Boltzmann equation [ 51 ]. Accurate temperature reconstruction requires addressing two core challenges: emissivity variability and atmospheric attenuation. Emissivity modeling has become a critical research direction, as surface materials such as concrete, asphalt, vegetation, glass, and metals exhibit drastically different emissivities. Recent work develops emissivity databases, multispectral IR sensors, and datadriven emissivity estimation to minimize temperature bias [ 39 , 43 ]. Tian et al. [ 54 ] further addressed the core issue by proposing a data processing method for simultaneous estimation of temperature and emissivity in multispectral thermometry, removing the need for a separate emissivity assumption model. In parallel, atmospheric correction methods—both physics-based and learning-based compensators, correct for humidity, aerosols, scattering, and absorption that distort the measured radiance [ 5 ]. These radiometric advances form the essential basis for reliable downstream thermal analysis. 4.2 Advances in Thermal Sensing Hardware Current thermal sensing hardware can be broadly categorized into two dominant groups: uncooled microbolometers and cooled photon detectors. A key metric for assessing the performance of thermal cameras is the Noise Equivalent Temperature Difference (NETD), which reflects the minimum temperature difference the sensor can reliably resolve. [35] Uncooled microbolometers, typically fabricated using VOx or aSi materials, represent the mainstream solution for commercial and embedded sensing [ 48 ]. They offer LWIR detection, low power consumption, and increasingly compact wafer-level packaging. Contemporary research focuses on reducing NETD, increasing detector fill-factor, improving frame rate, and reducing manufacturing cost to enable large-scale deployment. These sensors commonly appear in portable devices, mobile systems, and emerging IoT-based thermal networks. Cooled photon detectors—such as InSb or HgCdTe (MCT)—provide superior sensitivity and high spatial resolution, suitable for scientific and industrial inspection tasks [ 49 ]. However, their reliance on cryogenic cooling significantly increases cost and limits deployment flexibility. In addition, low-cost emerging sensors, including thermopile arrays [ 31 ], MEMS-based IR detectors [ 62 ], and CMOS-integrated IR pixels, have gained attention for city-scale sensing. Although their resolution is modest, recent calibration and compensation approaches have substantially improved measurement stability, making them promising for distributed thermal monitoring in smart infrastructure. 4.3 Thermal Image Enhancement and Signal Processing Raw thermal images often suffer from low contrast, high noise, and blurred structural boundaries due to intrinsic sensor limitations. Signal processing and image enhancement techniques therefore play a central role in modern thermal sensing systems. Furthermore, the non-contact nature of thermal imaging makes it vital in public health and sanitation applications, such as non-contact physiological measurement during pandemics [35]. Denoising algorithms, ranging from classical BM3D approaches [ 13 ] to deep-learning-based denoisers such as UNet [ 60 ] and transformer models have been developed to reduce fixed-pattern noise and temporal drift. Concurrently, super-resolution reconstruction has emerged as an important direction: deep convolutional networks [ 22 ] and generative adversarial networks (GANs) [ 26 , 63 ] can upscale low-resolution LWIR images, and cross-modal methods leverage RGB images to guide high-quality thermal detail restoration. Ping-Huai Tseng, Yen-Cheng Lin, Grant Wang, Anna Huang, and Dimitri Mili Another key development is thermal–RGB fusion, which combines complementary modalities to enhance object boundaries, improve visibility, and strengthen recognition performance. Stateof-the-art fusion techniques employ attention mechanisms, multiscale representation learning, and transformer-based cross-spectral correspondence models [ 33 ]. These enhancement pipelines significantly improve thermal data interpretability and serve as essential components in downstream detection, segmentation, and anomaly analysis [67]. 4.4 Calibration and Stability in Thermal Sensors Thermal sensors commonly exhibit pixel-wise drift and spatial nonuniformity, making calibration indispensable for reliable temperature measurements. Research in this area focuses on three major techniques: • Shutter-based calibration, traditionally used to perform periodic reference measurements, ensures consistent temperature mapping. Recent work seeks to reduce mechanical shutter usage to avoid operational interruption [9, 34]. • Shutter-less non-uniformity correction (NUC) methods [ 53 ] employ temporal statistics, scene-based correction, or deeplearning models to estimate pixel offsets without hardware intervention, enabling continuous sensing. • Low-cost IoT sensor calibration, including two-point calibration, drift modeling [ 23 ], and environmental compensation [ 38 ], improves the stability of distributed sensing networks with inexpensive sensor 4.5 Thermal Image Analysis with AI With the rise of machine learning (ML), thermal sensing has expanded from temperature estimation to semantic interpretation. Modern research applies deep learning to detection, segmentation, classification, anomaly detection, and simulation tasks on thermal data. Deep convolutional networks, anchor-free detectors, and lightweight networks optimized for edge devices have substantially advanced thermal-based object detection, especially for low-texture LWIR images [ 14 , 45 , 55 ]. However, aerial thermal image detection in urban or wooded areas faces unique challenges, such as the relatively small target size, occlusion caused by the environment, and the confusion between human thermal signatures and background thermal noise, necessitating sophisticated detection models beyond simple thermal thresholding [ 19 ]. Similarly, thermal semantic segmentation using architectures such as UNet and DeepLab enables detailed structural and environmental analysis, including façade defect detection and heat-leak localization [ 7 ]. For industrial and infrastructure applications, anomaly detection frameworks—often based on autoencoders or contrastive learning—identify abnormal heat patterns indicative of faults or degradation [32]. AI-driven interpretation significantly extends the role of thermal sensing from passive monitoring to active decision-making and predictive analytics. 4.6 Transition Toward Large-Scale and Smart City Applications The technological advances described above collectively enable the emergence of thermal sensing as a key component of smart city infrastructures. Improved radiometric modeling supports accurate surface temperature mapping; enhanced hardware and calibration allow reliable long-term deployment; and AI-backed interpretation transforms thermal data into actionable insights. These developments directly enable applications such as urban heat island analysis, energy-efficient building diagnostics, pedestrian and traffic monitoring, and large-scale environmental sensing. In summary, current thermal sensing research is characterized by multimodal integration, AI-driven enhancement, hardware miniaturization, and scalable calibration, forming the foundation upon which city-level thermal intelligence systems can be constructed. The next section will discuss how these technologies are applied within the broader context of urban sensing and smart city management. 5 Thermal Sensing in City Detection Applications For thermal sensing for Urban city applications, majority of them are thermal infrared sensing, which can be viewed as multi-scale and multi-source of observation and derivation. It can range from UAV-based thermal imaging to ground-based sensors in order to support applications such as urban heat island identification, heatrisk management, also building’s energy regulation and infrastructure health monitoring. 5.1 Thermal Sensing Studies for Urban Heat Island Phenomenon Urban Heat Island is a common phenomenon that happened in modern cities, and currently the monitoring and evaluations are largely focused on remote sensing derived from the land surface temperature, which are reviewed by Zhao et al. [ 66 ]. They reviewed how UHI is measured and showed that using collected results for real applications like hotspots mapping or decision supporting, and they also provide future improvement targets like higher resolution and better validation methods. Ahmed et al. [ 2 ] further focuses on the research of AI models’ prediction on UHI, which shows that the research trend has shifted from measuring and describing UHI to using AI for practical prediction, in order to help urban planning and evaluate relevant work’s effectiveness. Since thermal conditions in urban areas in the night are highly associated with human comfort and heat exposure, Kim et al. [ 29 ] has reviewed the relevant works and difficulties in urban applications of nighttime sensing, and they points out that although satellites are suitable for nighttime urban sensing, the results are limited because the satellites cannot capture differences inside cities and weather can also restrict observations. Therefore, higher quality satellites and AI-based reconstruction are needed to improve data quality in the future. Survey of Thermal Sensing for Smart Cities 5.2 City Heat Management on Big Data Heat management and decision making are important factors to achieve higher efficiency in heat sensing applications. Pan et al. [ 41 ] has pointed out that urban heat risk shouldn’t only look at the hottest spot, instead, the number of people can be a critical factor. If heat monitoring can know when and where most people are exposed to extreme high heat, it will be easier for heat management to give out better decisions. At the meantime, from big data aspect, Wang [ 59 ] proposed using neural network for UHI monitoring and smart prediction, and translate the result into more readable results, which highlights the research direction that emphasize data driven approach for UHI monitoring and prediction. When looking at big data applications at larger scale, Overeem et al. [ 40 ] first proposed using groups of people’s smartphone battery temperature to estimate urban air temperature, which can serve as a solution to solve sparse sensor allocation. Later, Droste et al. [ 15 ] further improved the model through real world cast study, showing that under sufficient amount of data, the derived results can show temperatures in block or street dimension, and also can update the result in every hour. In addition, Seong et al. [ 52 ] points out that relying on small scale environmental sensors can map heat exposure at a higher spatial resolution and refresh dynamically, capturing the neighborhood scale temperature differences better than the weather stations. 5.3 High Mobility Thermal Sensors for City Heat Monitoring For different levels of cities’ magnitude, when the scale of research is shrinking from city down to blocks and buildings, thermal sensors with high resolution and mobility can provide higher resolution thermal patterns and hot spot identification. Therefore, Henn and Peduzzi’s [ 20 ] research is a great example by using high-resolution UAV thermal imaging in smaller urban environments, which emphasize the feasibility of high-resolution evaluation and application, also showing UAV’s thermal image can be applied to small scale heat monitoring and providing acceptable results under most conditions. For UAV’s application on city thermal sensing, Rakah et al. [ 46 ] demonstrated using UAV which equipped with a thermal camera to scan the building in order to quickly identify thermal defects and then turn the results into a heat map for decision making. Furthermore, Zheng et al. [ 64 ] has texturized the thermal image information to produce data that can be directly used on Building Information Model, which is more useful for building’s energy saving evaluation. 5.4 Thermal Sensing for City Facilities Inspection Thermal sensing can also support urban utilities, providing another method for facilities inspections. Vollmer et al. [ 56 ] focusing on locating leaking pipelines, which is done by comparing thermal-image in real world cases to improve the reliability of leak localization, and in 2025 [ 57 ] they further using deep-learning strategies against traditional computer vision methods, figuring out the best model structure and producing finer boundaries on images, proofing the feasibility of deep learning integration in thermal defect detection. Combing above, the research for heat sensing in smart cities is shifting from signal source toward integration across different scales, handling lots of different implementations and different tasks. 6 Future Direction of Thermal Sensing 6.1 Thermal Detection Future research in thermal sensing for smart cities is expected to advance toward higher-resolution, multi-modal, and more scalable systems that transform raw thermal data into actionable urban intelligence. Continued improvements in sensor hardware, emissivity modeling, and atmospheric correction will enable more accurate temperature estimation across complex urban surfaces. In parallel, AI-driven enhancement methods—such as thermal super-resolution, denoising, and RGB–thermal fusion—will strengthen the reliability and interoperability of thermal imagery under diverse lighting and weather conditions. At the systems level, integrated sensing networks combining UAVs, IoT thermal nodes, vehicle-mounted sensors, and satellite platforms will allow continuous, multi-scale monitoring of heatrelated urban phenomena. Edge-efficient learning models and privacy-preserving thermal analytics will further support real-time applications in UHI mitigation, infrastructure diagnostics, and public safety. Ultimately, the future of thermal sensing lies in developing autonomous, citywide thermal intelligence frameworks that operate seamlessly within smart-city infrastructure to enhance resilience, sustainability, and data-driven decision-making. 6.2 Applications in City Sensing In the future, research in this field will focus on several key technical problems. First is the usage of Generative AI to handle the difficulty that traditional methods cannot solve, for example, trying to generate realistic thermal images. What’s more, the development of lightweight edge computing models is really important to satisfy the high computation demands of UAVs and real-time urban city monitoring. More importantly is that, future optimization goals will no longer be solely for visualization optimization, and it is possible will shift towards a multi-task direction. 7 Conclusion This document goes through the evolution of urban heat sensing, introducing its development history from early static monitoringbased structures to mobile approaches using smartphones and vehicles, and finally to aerial systems capable of overcoming spatial limitations. Although the integration of deep learning has made computer vision a key tool for extracting useful data from satellite and street-level images, some visual methods are still remain limited by poor lighting, extreme weather conditions, and growing privacy concerns. These limitations have shifted technical focus toward thermal sensing, which offers the advantages of all-day operation. To enhance the feasibility of thermal sensing in smart cities, the technology has evolved from simple infrared radiation measurement into integrated systems combining advanced hardware, physical radiometric models, and artificial intelligence. Improvements in uncooled microbolometers have lowered hardware costs, Ping-Huai Tseng, Yen-Cheng Lin, Grant Wang, Anna Huang, and Dimitri Mili while deep learning algorithms are now employed for image denoising, super-resolution, and the fusion of thermal and visible light data. These software advancements effectively resolve issues with low resolution and blurred textures. Furthermore, integrating AI enables continuous, more stable calibration and precise semantic analysis, enabling thermal sensors to perform object detection and anomaly identification with reliability comparable to traditional cameras. In real world, thermal sensing shows the value through multiscale integration. At the higher level, it is used to monitor the Urban Heat Island (UHI) effect, combining environmental data with population distribution to assess actual heat exposure risks, and even utilizing smartphone battery temperatures to solve the sparsity of collected data. At the smaller level, high-mobility UAVs equipped with thermal imaging are widely used for building energy inspection and infrastructure defect detection. Overall, the field is moving toward multi-source data fusion, integrating thermal signals across different levels of scales with AI prediction models in order to provide real-time, precise data for urban decision-making. References [1] Tarek Abdelzaher, Yaw Anokwa, Peter Boda, Jeff Burke, Deborah Estrin, Leonidas Guibas, Aman Kansal, Samuel Madden, and Jim Reich. 2007. Mobiscopes for Human Spaces. Pervasive Computing, IEEE 6 (05 2007), 20–29. doi:10.1109/MPRV.2007.38 [2] Ali Najah Ahmed, Nouar AlDahoul, Nurhanani A. Aziz, Y. F. Huang, Mohsen Sherif, and Ahmed El-Shafie. 2025. The urban heat Island effect: A review on predictive approaches using artificial intelligence models. City and Environment Interactions 28 (2025), 100234. doi:10.1016/j.cacint.2025.100234 [3] Abdullah Alfarrarjeh, Seon Ho Kim, Utkarsh Baranwal, and Yash Bitla. 2023. Object Detection and Counting Challenges in Real Street Monitoring: Case Study of Homeless Encampments. In 2023 IEEE International Conference on Image Processing (ICIP). 2785–2789. doi:10.1109/ICIP49359. 2023.10222473 [4] Fahad Alhasoun and Marta C. González. 2019. Streetify: Using Street View Imagery And Deep Learning For Urban Streets Development. 2019 IEEE International Conference on Big Data (Big Data) (2019), 2001–2006. https://api.semanticscholar.org/CorpusID:208157968 [5] Gail P Anderson, Alexander Berk, Prabhat K Acharya, Michael W Matthew, Lawrence S Bernstein, James H Chetwynd, H. Dothe, Steven M Adler-Golden, Anthony J Ratkowski, Gerald W Felde, James A Gardner, Michael L Hoke, Steven C Richtsmeier, Brian Pukall, Jason B Mello, and Laila S Jeong. 1999. MODTRAN4: radiative transfer modeling for remote sensing. Proceedings of SPIE (12 1999). doi:10.1117/12.371318 [6] N. Audebert, B. L. Saux, and Sébastien Lefèvre. 2017. Beyond RGB: Very High Resolution Urban Remote Sensing With Multimodal Deep Networks. ArXiv abs/1711.08681 (2017). https: //api.semanticscholar.org/CorpusID:10008238 [7] Konstantinos Bacharidis, Froso Sarri, and Lemonia Ragia. 2020. 3D Building Façade Reconstruction Using Deep Learning. ISPRS International Journal of Geo-Information 9 (05 2020), 322. doi:10.3390/ijgi9050322 [8] Pedro Caminha, Felipe Silva, Roberto Pacheco, Rodrigo Couto, Pedro Braconnot Velloso, Miguel Campista, and Luís Costa. 2018. SensingBus: Using Bus Lines and Fog Computing for Smart Sensing the City. IEEE Cloud Computing 5 (09 2018), 58–69. doi:10.1109/MCC.2018.053711667 [9] Yanpeng Cao and Christel-Loic Tisse. 2013. Shutterless solution for simultaneous focal plane array temperature estimation and nonuniformity correction in uncooled long-wave infrared camera. Applied Optics 52 (08 2013), 6266. doi:10.1364/ao.52.006266 [10] Paolo Casari, Angelo Castellani, Angelo Cenedese, Claudio Lora, Michele Rossi, L. Schenato, and Michele Zorzi. 2009. The “Wireless Sensor Networks for City-Wide Ambient Intelligence (WISE-WAI)” Project. Sensors 9 (05 2009), 4056–4082. doi:10.3390/s90604056 [11] Xuecheng Chen, Haoyang Wang, Zuxin Li, Wenbo Ding, Fan Dang, Chenye Wu, and Xinlei Chen. 2022. DeliverSense: Efficient Delivery Drone Scheduling for Crowdsensing with Deep Reinforcement Learning. In Proceedings of the 2022 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp/ISWC ’22 Adjunct). 6. doi:10.1145/3544793. 3560412 [12] Yuhua Chen, Wen Li, and Luc Van Gool. 2017. ROAD: Reality Oriented Adaptation for Semantic Segmentation of Urban Scenes. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (2017), 7892–7901. https://api.semanticscholar.org/CorpusID:4715607 [13] Kostadin Dabov, Alessandro Foi, Vladimir Katkovnik, and Karen Egiazarian. 2007. Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering. IEEE Transactions on Image Processing 16 (08 2007), 2080–2095. doi:10.1109/tip.2007.901238 [14] Huyen Trang Dinh and Eung-Tae Kim. 2025. A Lightweight Network Based on YOLOv8 for Improving Detection Performance and the Speed of Thermal Image Processing. Electronics 14 (02 2025), 783. doi:10.3390/electronics14040783 [15] A. M. Droste, J. J. Pape, A. Overeem, H. Leijnse, G. J. Steeneveld, A. J. van Delden, and R. Uijlenhoet. 2017. Crowdsourcing Urban Air Temperatures through Smartphone Battery Temperatures in São Paulo, Brazil. Journal of Atmospheric and Oceanic Technology 34, 9 (2017), 1853–1866. doi:10.1175/JTECH-D-16-0150.1 [16] Tomas Fernandez, Yen Cheng Chang, Jesse Codling, Yiwen Dong, Jiale Zhang, Carlee JoeWong, Hae Young Noh, and Pei Zhang. 2024. Poster: Drive-by City Wide Trash Sensing for Neighborhood Sanitation Need. In Proceedings of the 22nd Annual International Conference on Mobile Systems, Applications and Services (Minato-ku, Tokyo, Japan) (MOBISYS ’24). Association for Computing Machinery, New York, NY, USA, 704–705. doi:10.1145/3643832.3661431 [17] Mohammadhossein Ghahramani. 2020. Urban sensing based on mobile phone data: Approaches, applications, and challenges. IEEE/CAA Journal of Automatica Sinica 7 (04 2020), 627 – 637. doi:10.1109/JAS.2020.1003120 [18] William Grosky, Aman Kansal, Suman Nath, Jie Liu, and Feng Zhao. 2007. SenseWeb: An Infrastructure for Shared Sensing. Multimedia, IEEE 14 (11 2007), 8 – 13. doi:10.1109/MMUL. 2007.82 [19] Ali Haider, Furqan Shaukat, and Junaid Mir. 2021. Human detection in aerial thermal imaging using a fully convolutional regression network. Infrared Physics & Technology 116 (08 2021), 103796. doi:10.1016/j.infrared.2021.103796 [20] Katrina Ariel Henn and Alicia Peduzzi. 2024. Surface Heat Monitoring with High-Resolution UAV Thermal Imaging: Assessing Accuracy and Applications in Urban Environments. Remote Sensing 16, 5 (2024), 930. doi:10.3390/rs16050930 [21] Steven C. Hespeler, Hamidreza Nemati, and Ehsan Dehghan-Niri. 2021. Non-destructive thermal imaging for object detection via advanced deep learning for robotic inspection and harvesting of chili peppers. Artificial Intelligence in Agriculture 5 (2021), 102–117. doi:10.1016/j. aiia.2021.05.003 [22] Xinrui Hu, Shaojuan Luo, Chunhua He, Wenhao Wu, and Heng Wu. 2023. Infrared thermal image denoising with symmetric multi-scale sampling network. Infrared Physics & Technology 134 (11 2023), 104909. doi:10.1016/j.infrared.2023.104909 [23] Aaron Hurst, Andrey V. Kalinichev, Klaus Koren, and Daniel E. Lucani. 2025. Not all those who drift are lost: Drift correction and calibration scheduling for the IoT. IEEE Internet of Things Journal (2025), 1–1. doi:10.1109/jiot.2025.3628204 [24] Jackelyn Hwang and Nikhil Naik. 2023. Systematic Social Observation at Scale: Using Crowdsourcing and Computer Vision to Measure Visible Neighborhood Conditions. Sociological Methodology 53, 2 (2023), 183–216. arXiv:https://doi.org/10.1177/00811750231160781 doi:10.1177/00811750231160781 [25] Shunping Ji, Shiqing Wei, and Meng Lu. 2019. Fully Convolutional Networks for Multisource Building Extraction From an Open Aerial and Satellite Imagery Data Set. IEEE Transactions on Geoscience and Remote Sensing 57 (2019), 574–586. https://api.semanticscholar.org/CorpusID: 57190346 [26] Nanhe Jiang, Yucun Zhang, Qun Li, and Fang Yan. 2025. An infrared thermal image denoising method focusing on noise feature learning. Optics & Laser Technology 184 (01 2025), 112475. doi:10.1016/j.optlastec.2025.112475 [27] Wan Jiao, Gayle Hagler, Ronald Williams, Robert Sharpe, Ryan Brown, Daniel Garver, Robert Judge, Motria Caudill, Joshua Rickard, Michael Davis, Lewis Weinstock, Susan ZimmerDauphinee, and Ken Buckley. 2016. Community Air Sensor Network (CAIRSENSE) project: Evaluation of low-cost sensor performance in a suburban environment in the southeastern United States. Atmospheric Measurement Techniques 9 (11 2016), 5281–5292. doi:10.5194/amt9-5281-2016 [28] Burak Kantarci and Sema Oktug. 2018. Special Issue: Wireless Sensor and Actuator Networks for Smart Cities. Journal of Sensor and Actuator Networks 7 (11 2018), 49. doi:10.3390/jsan7040049 [29] Youngseok Kim, Cheolhee Yoo, and Jungho Im. 2025. Nighttime satellite land surface temperature for urban applications: achievements, challenges, and future prospects. GIScience & Remote Sensing 62, 1 (2025), 2527990. doi:10.1080/15481603.2025.2527990 [30] Nicholas Lane, Shane Eisenman, Mirco Musolesi, Emiliano Miluzzo, and Andrew Campbell. 2008. Urban sensing systems: Opportunistic or participatory? 11–16. doi:10.1145/1411759. 1411763 [31] Hongbo Li, Chenchen Zhang, Gaobo Xu, Xuefeng Ding, Yue Ni, Guidong Chen, Dapeng Chen, Na Zhou, and Haiyang Mao. 2022. A Thermopile Infrared Sensor Array Pixel Monolithically Integrated with an NMOS Switch. Micromachines 2022 13 (02 2022), 258–258. doi:10.3390/ mi13020258 [32] Zhiyuan Li, Yu Sun, Laihao Yang, Zhibin Zhao, and Xuefeng Chen. 2022. Unsupervised Machine Anomaly Detection Using Autoencoder and Temporal Convolutional Network. IEEE Transactions on Instrumentation and Measurement 71 (01 2022), 1–13. doi:10.1109/tim.2022. 3212547 [33] Mingjian Liang, Junjie Hu, Chenyu Bao, Hua Feng, Fuqin Deng, and Tin Lun Lam. 2023. Explicit Attention-Enhanced Fusion for RGB-Thermal Perception Tasks. arXiv (Cornell University) (03 2023). doi:10.48550/arxiv.2303.15710 [34] Dongming Lu, Longyin Teng, Jianle Ren, Jiangyun Tan, Mengke Wang, Liping Wang, and Guohua Gu. 2023. Scene-Based Nonuniformity Correction Method Using Principal Component Analysis for Infrared Focal Plane Arrays. Applied Sciences 13 (12 2023), 13331. doi:10.3390/ app132413331 [35] Martin Clinton Tosima Manullang, Yuan-Hsiang Lin, Sheng-Jie Lai, and Nai-Kuan Chou. 2021. Implementation of Thermal Camera for Non-Contact Physiological Measurement: A Systematic Review. Sensors 21 (11 2021), 7777. doi:10.3390/s21237777 [36] Dimitrios Marmanis, Jan Dirk Wegner, Silvano Galliani, Konrad Schindler, Mihai Datcu, and Uwe Stilla. 2016. SEMANTIC SEGMENTATION OF AERIAL IMAGES WITH AN ENSEMBLE OF CNNS. ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences III-3 (2016), 473–480. https://api.semanticscholar.org/CorpusID:26376005 [37] Rohan Murty, Geoffrey Mainland, Ian Rose, Atanu Roy Chowdhury, Abhimanyu Gosain, Josh Bers, and Matt Welsh. 2008. CitySense: An Urban-Scale Wireless Sensor Network and Testbed. 2008 IEEE International Conference on Technologies for Homeland Security, HST’08, 583 – 588. doi:10.1109/THS.2008.4534518 [38] Naga, Ismaila Abimbola, Tasneem Ahmed, Iulia Anton, Khurram Riaz, Qusai Ibrahim, Arghadyuti Banerjee, Ananya Tiwari, and Salem Gharbia. 2024. Challenges and Opportunities in Calibrating Low-Cost Environmental Sensors. Sensors 24 (06 2024), 3650. doi:10.3390/s24113650 [39] Kenta Ogawa, Thomas Schmugge, Frederic Jacob, and Andrew French. 2002. Estimation of broadband land surface emissivity from multi-spectral thermal infrared remote sensing. Agronomie 22 (09 2002), 695–696. doi:10.1051/agro:2002055 [40] A. Overeem, J. C. R. Robinson, H. Leijnse, G. J. Steeneveld, B. K. P. Horn, and R. Uijlenhoet. 2013. Crowdsourcing urban air temperatures from smartphone battery temperatures. Geophysical Research Letters 40, 15 (2013), 4081–4085. doi:10.1002/grl.50786 [41] Xiyu Pan, Dimitris Mavrokapnidis, Hoang T. Ly, Neda Mohammadi, and John E. Taylor. 2024. Assessing and forecasting collective urban heat exposure with smart city digital twins. Scientific Reports 14, 1 (2024), 9653. doi:10.1038/s41598-024-59228-8 [42] Miral Jerambhai Patel and Hasmukh P. Koringa. 2024. Deep Learning based Semantic Segmentation for Buildings Detection from Remote Sensing Images. International Journal of Survey of Thermal Sensing for Smart Cities Next-Generation Computing (2024). https://api.semanticscholar.org/CorpusID:269817978 [43] Corentin Poissenot-Arrigoni, Bertrand Marcon, Frédéric Rossi, and Guillaume Fromentin. 2023. In-Situ Pixel-wise Emissivity Measurement Using a Multispectral Infrared Camera. Journal of Imaging 9 (09 2023), 198. doi:10.3390/jimaging9100198 [44] Aveek Purohit, Zheng Sun, and Pei Zhang. 2013. SugarMap: Location-less Coverage for Micro-Aerial Sensing Swarms. In Proceedings of the 12th ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN 2013). ACM/IEEE, 253–264. doi:10.1145/ 2461381.2461412 [45] Shreyansh Raj, Soumya Ranjan Pradhan, Pushpam Anand, Daksh Deswal, Syed, and Vinod Kumar. 2025. Deep Learning-Based Object Detection in Thermal Imaging. 2025 Global Conference in Emerging Technology (GINOTECH) (05 2025), 1–6. doi:10.1109/ginotech63460.2025.11076813 [46] Tarek Rakha, Amanda Liberty, Alice Gorodetsky, Burak Kakillioglu, and Senem Velipasalar. 2018. Heat Mapping Drones: An Autonomous Computer-Vision-Based Procedure for Building Envelope Inspection Using Unmanned Aerial Systems (UAS). Technology|Architecture + Design 2, 1 (2018), 30–44. doi:10.1080/24751448.2018.1420963 [47] Bushra Rashid and Mubashir Husain Rehmani. 2015. Applications of wireless sensor networks for urban areas: A survey. Journal of Network and Computer Applications 60 (12 2015). doi:10. 1016/j.jnca.2015.09.008 [48] Antoni Rogalski. 2003. Infrared detectors: status and trends. Progress in Quantum Electronics 27 (01 2003), 59–210. doi:10.1016/s0079-6727(02)00024-1 [49] A Rogalski. 2005. HgCdTe infrared detector material: history, status and outlook. Reports on Progress in Physics 68 (08 2005), 2267–2336. doi:10.1088/0034-4885/68/10/r01 [50] Miroslav Saska. 2015. MAV-swarms: unmanned aerial vehicles stabilized along a given path using onboard relative localization. In International Conference on Unmanned Aircraft Systems (ICUAS). [51] John R Schott. 2007. Remote sensing : the image chain approach. Oxford University Press. [52] Kijin Seong, Seung Jun Choi, and Junfeng Jiao. 2025. IoT sensors as a tool for assessing spatiotemporal risk to extreme heat. Journal of Environmental Planning and Management 68, 11 (09 2025), 2621–2643. doi:10.1080/09640568.2024.2320257 [53] A. Tempelhahn, H. Budzier, V. Krause, and G. Gerlach. 2016. Shutter-less calibration of uncooled infrared cameras. Journal of Sensors and Sensor Systems 5 (01 2016), 9–16. doi:10.5194/jsss-5-92016 [54] Zhuangtao Tian, Kaihua Zhang, Yanfen Xu, Kun Yu, and Yufang Liu. 2022. Data processing method for simultaneous estimation of temperature and emissivity in multispectral thermometry. Optics Express 30 (09 2022), 35381. doi:10.1364/oe.470056 [55] Pei-Fen Tsai, Chia-Hung Liao, and Shyan-Ming Yuan. 2022. Using Deep Learning with Thermal Imaging for Human Detection in Heavy Smoke Scenarios. Sensors 22 (07 2022), 5351. doi:10. 3390/s22145351 [56] Elena Vollmer, Julian Ruck, Rebekka Volk, and Frank Schultmann. 2024. Detecting district heating leaks in thermal imagery: Comparison of anomaly detection methods. Automation in Construction 168 (2024), 105709. doi:10.1016/j.autcon.2024.105709 [57] Elena Vollmer, Julian Ruck, Rebekka Volk, and Frank Schultmann. 2025. Leak detection using thermal imagery: Deep learning versus traditional computer vision state-of-the-art. ISPRS Journal of Photogrammetry and Remote Sensing 228 (2025), 505–518. doi:10.1016/j.isprsjprs. 2025.06.006 [58] Jingjing Wang, Chunxiao Jiang, Zhang Kai, Tony Q.S. Quek, Yong Ren, and L. Hanzo. 2018. Vehicular Sensing Networks in a Smart City: Principles, Technologies and Applications. IEEE Wireless Communications 25 (02 2018), 122–132. doi:10.1109/MWC.2017.1600275 [59] Lin Wang. 2025. Urban heat island monitoring and intelligent prediction method based on big data analysis. Sustainable Energy Research 12 (2025), 55. doi:10.1186/s40807-025-00206-7 [60] Minghe Wang, Pan Yuan, Su Qiu, Weiqi Jin, Li Li, and Xia Wang. 2025. Dual-Encoder UNetBased Narrowband Uncooled Infrared Imaging Denoising Network. Sensors 25 (02 2025), 1476. doi:10.3390/s25051476 [61] A. N. Wilson, Khushi Anil Gupta, Balu Harshavardan Koduru, Abhinav Kumar, Ajit Jha, and Linga Reddy Cenkeramaddi. 2023. Recent Advances in Thermal Imaging and its Applications Using Machine Learning: A Review. IEEE Sensors Journal 23 (02 2023), 3395–3407. doi:10.1109/ jsen.2023.3234335 [62] Dehui Xu, Yuelin Wang, Bin Xiong, and Tie Li. 2017. MEMS-based thermoelectric infrared sensors: A review. Frontiers of Mechanical Engineering 12 (06 2017), 557–566. doi:10.1007/s11465017-0441-2 [63] Pengfei Yang, Heng Wu, Lianglun Cheng, and Shaojuan Luo. 2023. Infrared image denoising via adversarial learning with multi-level feature attention network. Infrared Physics & Technology 128 (01 2023), 104527. doi:10.1016/j.infrared.2022.104527 [64] Cheng Zhang, Yang Zou, Jay Dimyadi, and Ruidong Chang. 2023. Thermal-textured BIM generation for building energy audit with UAV image fusion and histogram-based enhancement. Energy and Buildings 301 (2023), 113710. doi:10.1016/j.enbuild.2023.113710 [65] Rui Zhang, Jing Shi, Yanchao Zhang, and Chi Zhang. 2013. Verifiable Privacy-Preserving Aggregation in People-Centric Urban Sensing Systems. Selected Areas in Communications, IEEE Journal on 31 (09 2013), 268–278. doi:10.1109/JSAC.2013.SUP.0513024 [66] Lili Zhao, Xuncheng Fan, and Tao Hong. 2025. Urban Heat Island Effect: Remote Sensing Monitoring and Assessment—Methods, Applications, and Future Directions. Atmosphere 16, 7 (2025), 791. doi:10.3390/atmos16070791 [67] Heng Zhou, Chunna Tian, Zhenxi Zhang, Qizheng Huo, Yongqiang Xie, and Zhongbo Li. 2022. Multispectral Fusion Transformer Network for RGB-Thermal Urban Scene Semantic Segmentation. IEEE Geoscience and Remote Sensing Letters 19 (01 2022), 1–5. doi:10.1109/lgrs. 2022.3179721