Scene Parameters Research Articles

Influence of cloudy and clear-sky partitions, aerosols, and geometry on the recent variability in surface solar irradiance components in northern France

Abstract. Surface solar irradiance (SSI) is a fundamental parameter whose components (direct and diffuse) and variabilities are highly influenced by changes in atmospheric content and scene parameters. The respective importance of cloudy-sky conditions and atmospheric aerosols on SSI evolutions is region dependent and only partially quantified. Here we provide a comprehensive analysis of SSI variabilities recorded in northern France, a region with extensive variability in sky conditions and aerosol loads. Through the application of automatic filtering methods to 1 min resolution SSI ground-based measurements over Lille, sky conditions are classified as clear-sky, 11 %; clear-sun-with-cloud, 22 %; and cloudy-sun situations, 67 % from 2010 to 2022, for which we analyze the statistics and variabilities in the global horizontal irradiance (GHI), beam horizontal irradiance (BHI), and diffuse horizontal irradiance (DHI). Coincident photometric measurements of aerosol properties and radiative-transfer simulations provide the means to conduct a multivariate analysis of the SSI observed trends and year-to-year evolutions and to estimate aerosol and cloud forcings under clear-sun conditions. The analysis of the record value of all-sky GHI in spring 2020 attributes 89 % of the changes to the exceptional sunlight conditions (57 % of clear-sun situations). It highlights also for that season the importance of solar zenith-angle changes, whose positive effects on clear-sun conditions surpass those due to aerosols. Our results show all-sky GHI and BHI positive trends of around +4.0 and +4.4 Wm-2yr-1, respectively, in both spring and summer, which are explained by more than 60 % by an increase in clear-sun occurrences of +1 % yr−1. Additional significant BHI increases under clear-sun conditions are mainly explained in spring by the negative trend in aerosol optical depth (−0.011 yr−1) and partly by angular effects in summer. Moreover, we find that clear-sun-with-cloud situations are frequently marked by irradiance enhancement due to clouds, with 13 % more GHI on a monthly average and 10 % additional diffuse proportion than in clear-sky situations. Under such conditions, clouds add on average 25 W m−2 of diffuse irradiance that sets the GHI at the remarkable level of pristine (aerosol-and-cloud-free) conditions or even higher, by more than +10 W m−2 in summer and for low aerosol loads. Overall, our results highlight the dominant and complex influence of cloudy conditions on SSI, which precedes or combines with that of aerosols and geometrical effects, and leads to a remarkable global level of SSI in clear-sun-with-cloud situations.

PesRec: A parametric estimation method for indoor semantic scene reconstruction from a single image

Reconstructing semantic indoor scenes is a challenging task in augmented and virtual reality. The quality of scene reconstruction is limited by the complexity and occlusion of indoor scenes. This is due to the difficulty in estimating the spatial structure of the scene and insufficient learning for object location inference. To address these challenges, we have developed PesRec, an end-to-end multi-task scene reconstruction network for parameterizing indoor semantic information. PesRec incorporates a newly designed spatial layout estimator and a 3D object detector to effectively learn scene parameter features from RGB images. We modify an object mesh generator to enhance the robustness of reconstructing indoor occluded objects through point cloud optimization in PesRec. Using the analyzed scene parameters and spatial structure, the proposed PesRec reconstructs an indoor scene by placing object meshes scaled to 3D detection boxes in an estimated layout cuboid. Extensive experiments on two benchmark datasets demonstrate that PesRec performs exceptionally well for object reconstruction with an average chamfer distance of 5.24 × 10-3 on the Pix3D dataset including 53.61 % mAP for 3D object detection and 79.7 % 3D IoU for the estimation of layout on the commonly-used SUN RGB-D datasets. The proposed computing network breaks through the limitations caused by complex indoor scenes and occlusions, showing optimization results that improve the quality of reconstruction in the fields of augmented reality and virtual reality.

The shading isophotes: Model and methods for Lambertian planes and a point light

Structure-from-Motion (SfM) and Shape-from-Shading (SfS) are complementary classical approaches to 3D vision. Broadly speaking, SfM exploits geometric primitives from textured surfaces and SfS exploits pixel intensity from the shading image. We propose an approach that exploits virtual geometric primitives extracted from the shading image, namely the level-sets, which we name shading isophotes. Our approach thus combines the strength of geometric reasoning with the rich shading information. We focus on the case of untextured Lambertian planes of unknown albedo lit by an unknown Point Light Source (PLS) of unknown intensity. We derive a comprehensive geometric model showing that the unknown scene parameters are in general all recoverable from a single image of at least two planes. We propose computational methods to detect the isophotes, to reconstruct the scene parameters in closed-form and to refine the results densely using pixel intensity. Our methods thus estimate light source, plane pose and camera pose parameters for untextured planes, which cannot be achieved by the existing approaches. We evaluate our model and methods on synthetic and real images.

Target-Aware Image Denoising for Inverse Monte Carlo Rendering

Physically based differentiable rendering allows an accurate light transport simulation to be differentiated with respect to the rendering input, i.e., scene parameters, and it enables inferring scene parameters from target images, e.g., photos or synthetic images, via an iterative optimization. However, this inverse Monte Carlo rendering inherits the fundamental problem of the Monte Carlo integration, i.e., noise, resulting in a slow optimization convergence. An appealing approach to addressing such noise is exploiting an image denoiser to improve optimization convergence. Unfortunately, the direct adoption of existing image denoisers designed for ordinary rendering scenarios can drive the optimization into undesirable local minima due to denoising bias. It motivates us to reformulate a new image denoiser specialized for inverse rendering. Unlike existing image denoisers, we conduct our denoising by considering the target images, i.e., specific information in inverse rendering. For our target-aware denoising, we determine our denoising weights via a linear regression technique using the target. We demonstrate that our denoiser enables inverse rendering optimization to infer scene parameters robustly through a diverse set of tests.

Factorized visual representations in the primate visual system and deep neural networks.

Object classification has been proposed as a principal objective of the primate ventral visual stream and has been used as an optimization target for deep neural network models (DNNs) of the visual system. However, visual brain areas represent many different types of information, and optimizing for classification of object identity alone does not constrain how other information may be encoded in visual representations. Information about different scene parameters may be discarded altogether ('invariance'), represented in non-interfering subspaces of population activity ('factorization') or encoded in an entangled fashion. In this work, we provide evidence that factorization is a normative principle of biological visual representations. In the monkey ventral visual hierarchy, we found that factorization of object pose and background information from object identity increased in higher-level regions and strongly contributed to improving object identity decoding performance. We then conducted a large-scale analysis of factorization of individual scene parameters - lighting, background, camera viewpoint, and object pose - in a diverse library of DNN models of the visual system. Models which best matched neural, fMRI, and behavioral data from both monkeys and humans across 12 datasets tended to be those which factorized scene parameters most strongly. Notably, invariance to these parameters was not as consistently associated with matches to neural and behavioral data, suggesting that maintaining non-class information in factorized activity subspaces is often preferred to dropping it altogether. Thus, we propose that factorization of visual scene information is a widely used strategy in brains and DNN models thereof.

Transforming a Non-Differentiable Rasterizer into a Differentiable One with Stochastic Gradient Estimation

We show how to transform a non-differentiable rasterizer into a differentiable one with minimal engineering efforts and no external dependencies (no Pytorch/Tensorflow). We rely on Stochastic Gradient Estimation, a technique that consists of rasterizing after randomly perturbing the scene's parameters such that their gradient can be stochastically estimated and descended. This method is simple and robust but does not scale in dimensionality (number of scene parameters). Our insight is that the number of parameters contributing to a given rasterized pixel is bounded. Estimating and averaging gradients on a per-pixel basis hence bounds the dimensionality of the underlying optimization problem and makes the method scalable. Furthermore, it is simple to track per-pixel contributing parameters by rasterizing ID- and UV-buffers, which are trivial additions to a rasterization engine if not already available. With these minor modifications, we obtain an in-engine optimizer for 3D assets with millions of geometry and texture parameters.

DermSynth3D: Synthesis of in-the-wild annotated dermatology images

In recent years, deep learning (DL) has shown great potential in the field of dermatological image analysis. However, existing datasets in this domain have significant limitations, including a small number of image samples, limited disease conditions, insufficient annotations, and non-standardized image acquisitions. To address these shortcomings, we propose a novel framework called DermSynth3D. DermSynth3D blends skin disease patterns onto 3D textured meshes of human subjects using a differentiable renderer and generates 2D images from various camera viewpoints under chosen lighting conditions in diverse background scenes. Our method adheres to top-down rules that constrain the blending and rendering process to create 2D images with skin conditions that mimic in-the-wild acquisitions, ensuring more meaningful results. The framework generates photo-realistic 2D dermatological images and the corresponding dense annotations for semantic segmentation of the skin, skin conditions, body parts, bounding boxes around lesions, depth maps, and other 3D scene parameters, such as camera position and lighting conditions. DermSynth3D allows for the creation of custom datasets for various dermatology tasks. We demonstrate the effectiveness of data generated using DermSynth3D by training DL models on synthetic data and evaluating them on various dermatology tasks using real 2D dermatological images. We make our code publicly available at https://github.com/sfu-mial/DermSynth3D.

Automatic Generation System for Autonomous Driving Simulation Scenarios Based on PreScan

The advancement of autonomous driving technology has urgently necessitated enhanced safety testing measures. Traditional road testing methods face significant challenges due to their high costs and prolonged durations. In response to the inefficiencies of manual scenario construction and the difficulties in selecting effective scenarios using common scenario generation methods in autonomous driving safety testing, this study proposes an innovative automatic SG system based on PreScan2021.1.0. The SG process is significantly simplified by this system’s capability to swiftly and accurately generate a vast array of specific scenarios through the input of scene parameters. The results indicate that this system achieves SG at a rate 2.5-fold faster than manual methods, alongside substantial improvements in accuracy. This system introduces a novel approach to virtual simulation, which is vital for the progress of autonomous driving safety. It offers a new paradigm for quickly and precisely generating test scenarios for autonomous driving.

Indoor Synthetic Data Generation: A Systematic Review

Objective:Deep learning-based object recognition, 6D pose estimation, and semantic scene understanding require a large amount of training data to achieve generalization. Time-consuming annotation processes, privacy, and security aspects lead to a scarcity of real-world datasets. To overcome this lack of data, synthetic data generation has been proposed, including multiple facets in the area of domain randomization to extend the data distribution. The objective of this review is to identify methods applied for synthetic data generation aiming to improve 6D pose estimation, object recognition, and semantic scene understanding in indoor scenarios. We further review methods used to extend the data distribution and discuss best practices to bridge the gap between synthetic and real-world data. Methods:We adhered to the guidelines of the systematic PRISMA technique. Three databases, IEEE Xplore, Springer Link, and ACM, and an additional manual search were conducted. In total, we identified 241 studies and included 34 in our systematic review. Conclusion:In summary, synthetic data generation has been performed using crop-out methods, graphic APIs, 3D modeling or authoring tools, or game engine-based methods. To extend the data distribution, varying scene parameters, i.e., lighting conditions or textures and the use of distracting objects in the scene are promising.

Neural Projection Mapping Using Reflectance Fields.

We introduce a high resolution spatially adaptive light source, or a projector, into a neural reflectance field that allows to both calibrate the projector and photo realistic light editing. The projected texture is fully differentiable with respect to all scene parameters, and can be optimized to yield a desired appearance suitable for applications in augmented reality and projection mapping. Our neural field consists of three neural networks, estimating geometry, material, and transmittance. Using an analytical BRDF model and carefully selected projection patterns, our acquisition process is simple and intuitive, featuring a fixed uncalibrated projected and a handheld camera with a co-located light source. As we demonstrate, the virtual projector incorporated into the pipeline improves scene understanding and enables various projection mapping applications, alleviating the need for time consuming calibration steps performed in a traditional setting per view or projector location. In addition to enabling novel viewpoint synthesis, we demonstrate state-of-the-art performance projector compensation for novel viewpoints, improvement over the baselines in material and scene reconstruction, and three simply implemented scenarios where projection image optimization is performed, including the use of a 2D generative model to consistently dictate scene appearance from multiple viewpoints. We believe that neural projection mapping opens up the door to novel and exciting downstream tasks, through the joint optimization of the scene and projection images.

Illumination-guided inverse rendering benchmark: Learning real objects with few cameras

The realm of 3D computer vision and graphics has experienced exponential growth recently, enabling the creation of realistic virtual environments and digital representations of real-world objects. Central to this progression are 3D reconstruction methods that facilitate the virtualization of shape, color, and surface details of real objects. Current methods predominantly employ neural scene representations, which despite their efficacy, grapple with limitations such as necessitating a high number of captured images and the complexity of transforming these representations into explicit geometric forms.An alternative strategy that has gained traction is the deployment of methods such as physically-based differentiable rendering (PBDR) and inverse rendering. These approaches require fewer viewpoints, yield explicit format results, and ensure a smoother transition to other representation methods. To meaningfully assess the performance of different 3D reconstruction methods, it is imperative to utilize benchmark scenes for comparison.Despite the existence of standard objects and scenes within the literature, there is a noticeable deficiency in real-world benchmark data that concurrently captures camera, illumination, and scene parameters — all critical to high-fidelity 3D reconstructions using PBDR and inverse rendering-based methods. In this research, we introduce a methodology for capturing real-world scenes as virtual scenes, integrating illumination parameters alongside camera and scene parameters to enhance the veracity of virtual representations. In addition, we introduce a set of ten real-world scenes, along with their virtual counterparts, designed as benchmarks. These benchmarks encompass a fundamental variety of geometric constructs, including convex, concave, plain, and mixed surfaces. Additionally, we demonstrate the 3D reconstruction results of state-of-the-art 3D reconstruction methods employing PBDR in real-world scenes, using both established methodologies and our proposed one.

Imaging operator in indirect imaging correlography.

Indirect imaging correlography (IIC) is a coherent imaging technique that provides access to the autocorrelation of the albedo of objects obscured from line-of-sight. This technique is used to recover sub-mm resolution images of obscured objects at large standoffs in non-line-of-sight (NLOS) imaging. However, predicting the exact resolving power of IIC in any given NLOS scene is complicated by the interplay between several factors, including object position and pose. This work puts forth a mathematical model for the imaging operator in IIC to accurately predict the images of objects in NLOS imaging scenes. Using the imaging operator, expressions for the spatial resolution as a function of scene parameters such as object position and pose are derived and validated experimentally. In addition, a self-supervised deep neural network framework to reconstruct images of objects from their autocorrelation is proposed. Using this framework, objects with ≈ 250 μ m features, located at 1 mt standoffs in an NLOS scene, are successfully reconstructed.

The InflateSAR Campaign: Developing Refugee Vessel Detection Capabilities with Polarimetric SAR

In the efforts to mitigate the ongoing humanitarian crisis at the European sea borders, this work builds detection capabilities to help find refugee boats in distress. For this paper, we collected dual-pol and quad-pol synthetic aperture radar (SAR) data over a 12 m rubber inflatable in a test-bed lake near Berlin, Germany. To consider a real scenario, we prepared the vessel so that its backscattering emulated that of a vessel fully occupied with people. Further, we collected SAR imagery over the ocean with different sea states, categorized by incidence angle and by polarization. These were used to emulate the conditions for a vessel located in ocean waters. This setup enabled us to test nine well-known vessel-detection systems (VDS), to explore the capabilities of new detection algorithms and to benchmark different combinations of detectors (detector fusion) with respect to different sensor and scene parameters (e.g., the polarization, wind speed, wind direction and boat orientation). This analysis culminated in designing a system that is specifically tailored to accommodate different situations and sea states.

UAV Mission Path Planning Based on Reinforcement Learning in Dynamic Environment

With the rapid development of information technology, various products used in information technology are also constantly optimized. Among them, the task and path planning of UAV in the high-end robot industry has always been the focus of relevant researchers. In the high-end robot industry, in addition to the research and development of UAVs, they also continue to learn and strengthen the task and path planning of UAVs. Nowadays, using unmanned aerial vehicles for real-time shooting has become the trend of this era. Drones have brought great convenience to people’s lives, and more and more people are willing to use drones. Based on the above situation, this paper studies the task and path planning of UAV based on reinforcement learning in dynamic environment. In the case of unpredictable scene parameters, reinforcement learning method can be established by value function. Thus, a more reasonable path can be given to realize the reconnaissance and detection of points of interest. MATLAB simulation experiments show that the algorithm can effectively detect targets in complex terrain composed of terrain restricted areas, and return to the designated end point to complete communication. Firstly, the development of unmanned aerial vehicles in various countries and the social status of unmanned aerial vehicles are discussed. By making UAV build threat model and task allocation in dynamic environment. The path planning and optimization of UAV in dynamic environment is studied, and the path planning algorithm and Hungarian algorithm are added. The optimized UAV has the fastest data transmission and calculation speed, while the other two types of UAVs have slower data transmission and calculation speed. In particular, ordinary UAVs also have data transmission failures, resulting in incomplete experimental results. The results show that the optimized UAV system is better in data calculation and transmission, which also shows that the UAV can quickly plan and process flight paths, which is suitable for practical applications.

Towards 3D Scene Understanding Using Differentiable Rendering

Deep learning methods have achieved significant results in many 2D computer vision tasks. To realize similar results in 3D tasks, equipping deep learning pipelines with components that incorporate knowledge about 2D image generation from the 3D scene description is a promising research direction. Rasterization, the standard formulation of the image generation process is not differentiable, and thus not compatible with the deep learning models trained using gradient-based optimization schemes. In recent years, many new approximate differentiable renderers have been proposed to enable compatibility between deep learning methods and image rendering techniques. Differentiable renderers fit naturally into the render-and-compare framework where the 3D scene parameters are estimated iteratively by minimizing the error between the observed image and the image rendered according to the current scene parameter estimate. In this article, we present StilllebenDR, a light-weight, scalable differentiable renderer built as an extension to the openly available Stillleben library. We demonstrate the usability of the proposed differentiable renderer for the task of iterative 3D deformable registration using a latent shape-space model and occluded object pose refinement using order-independent transparency based on analytical gradients and learned scene aggregation.

The case against probabilistic inference: a new deterministic theory of 3D visual processing.

How the brain derives 3D information from inherently ambiguous visual input remains the fundamental question of human vision. The past two decades of research have addressed this question as a problem of probabilistic inference, the dominant model being maximum-likelihood estimation (MLE). This model assumes that independent depth-cue modules derive noisy but statistically accurate estimates of 3D scene parameters that are combined through a weighted average. Cue weights are adjusted based on the system representation of each module's output variability. Here I demonstrate that the MLE model fails to account for important psychophysical findings and, importantly, misinterprets the just noticeable difference, a hallmark measure of stimulus discriminability, to be an estimate of perceptual uncertainty. I propose a new theory, termed Intrinsic Constraint, which postulates that the visual system does not derive the most probable interpretation of the visual input, but rather, the most stable interpretation amid variations in viewing conditions. This goal is achieved with the Vector Sum model, which represents individual cue estimates as components of a multi-dimensional vector whose norm determines the combined output. This model accounts for the psychophysical findings cited in support of MLE, while predicting existing and new findings that contradict the MLE model. This article is part of a discussion meeting issue 'New approaches to 3D vision'.

HyperGal: Hyperspectral scene modeling for supernova typing with the SED Machine integral field spectrograph

Context. Recent developments in time domain astronomy, such as Zwicky Transient Facility (ZTF), have made it possible to conduct daily scans of the entire visible sky, leading to the discovery of hundreds of new transients every night. Among these detections, 10 to 15 of these objects are supernovae (SNe), which have to be classified prior to cosmological use. The spectral energy distribution machine (SEDM) is a low-resolution (ℛ ~ 100) integral field spectrograph designed, built, and operated with the aim of spectroscopically observing and classifying targets detected by the ZTF main camera. Aims. As the current pysedm pipeline can only handle isolated point sources, it is limited by contamination when the transient is too close to its host galaxy core. This can lead to an incorrect typing and ultimately bias the cosmological analyses, affecting the homogeneity of the SN sample in terms of local environment properties. We present a new scene modeler to extract the transient spectrum from its structured background, with the aim of improving the typing efficiency of the SEDM. Methods. HyperGal is a fully chromatic scene modeler that uses archival pre-transient photometric images of the SN environment to generate a hyperspectral model of the host galaxy. It is based on the cigale SED fitter used as a physically-motivated spectral interpolator. The galaxy model, complemented by a point source for the transient and a diffuse background component, is projected onto the SEDM spectro-spatial observation space and adjusted to observations, and the SN spectrum is ultimately extracted from this multi-component model. The full procedure, from scene modeling to transient spectrum extraction and typing, is validated on 5000 simulated cubes built from actual SEDM observations of isolated host galaxies, covering a broad range of observing conditions and scene parameters. Results. We introduce the contrast, c, as the transient-to-total flux ratio at the SN location, integrated over the ZTF r-band. From estimated contrast distribution of real SEDm observations, we show that HyperGal correctly classifies ~95% of SNe Ia, and up to 99% for contrast c ≳ 0.2, representing more than 90% of the observations. Compared to the standard point-source extraction method (without the hyperspectral galaxy modeling step), HyperGal correctly classifies 20% more SNe Ia between 0.1 < c < 0.6 (50% of the observation conditions), with less than 5% of SN Ia misidentifications. The false-positive rate is less than 2% for c > 0.1 (> 99% of the observations), which represents half as much as the standard extraction method. Assuming a similar contrast distribution for core-collapse SNe, HyperGal classifies 14% additional SNe II and 11% additional SNe Ibc. Conclusions. HyperGal has proven to be extremely effective in extracting and classifying SNe in the presence of strong contamination by the host galaxy, providing a significant improvement with respect to the single point-source extraction.

Differentiable Rendering Using RGBXY Derivatives and Optimal Transport

Traditional differentiable rendering approaches are usually hard to converge in inverse rendering optimizations, especially when initial and target object locations are not so close. Inspired by Lagrangian fluid simulation, we present a novel differentiable rendering method to address this problem. We associate each screen-space pixel with the visible 3D geometric point covered by the center of the pixel and compute derivatives on geometric points rather than on pixels. We refer to the associated geometric points as point proxies of pixels. For each point proxy, we compute its 5D RGBXY derivatives which measures how its 3D RGB color and 2D projected screen-space position change with respect to scene parameters. Furthermore, in order to capture global and long-range object motions, we utilize optimal transport based pixel matching to design a more sophisticated loss function. We have conducted experiments to evaluate the effectiveness of our proposed method on various inverse rendering applications and have demonstrated superior convergence behavior compared to state-of-the-art baselines.

Research on calculation of surface irradiance for infrared extended sources based on CUDA parallel speedup

To correct the error caused by uneven distribution of target surface irradiance in the measurement of emissivity by the irradiation reflection method, a progressive method for calculating the surface irradiance for infrared extended sources was proposed, and its CPU-GPU heterogeneous operation speedup was realized based on the compute unified device architecture (CUDA). The proposed method calculated the radiation transfer at the scale of small surface sources while considering the multiple reflections between the sources, which better reflected the actual physical condition. The CUDA-based parallel speedup enabled calculation within a few seconds even for several data levels, which facilitated pixel-by-pixel analysis of surface emissivity measurements. In addition, the effectiveness of the proposed calculation method was analyzed by adjusting multiple scene parameters. The advantage of the proposed method was analyzed by comparing it with several irradiance calculation methods in terms of the result and correction effect on emissivity measurements. The speedup effect of CPU-GPU heterogeneous operation was analyzed by adjusting the calculation parameters. The results revealed that the proposed method could effectively calculate the irradiance distribution on the target surface regardless of the distance between the radiation sources, and the calculation consumed a very short time. It could correct the emissivity measurement errors caused by uneven irradiance and was of great significance to further improving the accuracy of emissivity measurements by the irradiation reflection method.

Real-Time Construction of Thermal Model Based on Multimodal Scene Data

In commercial buildings, the total consumption of central air conditioning accounts for about 40%–50%. However, at present, the initial design value of building Heating Ventilation and Air Conditioning (HVAC) is usually far greater than the actual refrigeration value of refrigeration demand, which will lead to great energy consumption waste. Moreover, the operation of HVAC affects the thermal comfort of users, so it is necessary to establish a thermal model for the scene to control. The thermal model describes the temperature of the scene in different environments. So it is very important to design a thermal model to calculate the scene in real time. Because the flow of people, the opening of windows, the ventilation of the scene and other parameters influence the change of thermal state in the scene environment, these parameters are complicated to model. Human disturbance will lead to the instability of the state of the scene environment. The inconsistency of its thermal model will lead to energy allocation tracking strategies in different regions. To solve this problem, We propose a thermal model for building thermal comfort using a multimodal analysis framework. This paper analyzes multiple temperature and humidity sensors and area image by multimodal combination and processes the image and sensor data by combining CNN and LSTM. Our results show that when the thermal model analyzed by this method is deployed in a building in the south of China, the MSE accuracy of the local effect of temperature field prediction reaches 99%, and its AMAX reaches 94%, so the running stability of the model in the scene is high. In addition, the research shows that the thermal model analysis framework can make the Internet of Things (IoT) in buildings more intelligent, and it can be combined with this thermal model to improve human comfort, make it easier to deploy in each hot zone, and have a better overall energy-saving effect.