Outdoor space pedestrian thermal comfort detection, evaluation and improvement method based on unmanned aerial vehicle low-altitude remote sensing technology

Abstract

The invention discloses an outdoor space pedestrian thermal comfort detection, evaluation and improvement method based on an unmanned aerial vehicle low-altitude remote sensing technology, and relates to the field of outdoor thermal environment detection and transformation. A thermal infrared image and a multispectral image of a three-dimensional surface of an urban complex scene are synchronously obtained from multiple angles, original images are processed in batches, and characteristic parameter space distribution representing multi-scene surface optics, openness and thermophysical attributes and a thermophysical field representing ideal three-dimensional diffusion surface long and short wave radiation are obtained. In combination with synchronously observed sun and sky radiation and near-earth meteorological parameters, an index view ball method and a ray tracing technology are utilized to generate pedestrian average radiation temperature and outdoor thermal comfort index distribution with high spatial resolution. Aiming at pedestrian thermal comfort evaluation results in different scenes, the invention provides a method for improving the thermal comfort current situation of a specific case by using automatic machine learning so as to improve the thermal comfort of a human body near the ground.

Description

Technical Field

[0001] The present invention relates to the field of outdoor thermal environment detection and modification, and in particular to a method for detecting, evaluating and improving thermal comfort of pedestrians in outdoor spaces based on unmanned aerial vehicle (UAV) low-altitude remote sensing technology. Background Art

[0002] Currently, the observation and evaluation of thermal environments in urban outdoor spaces primarily relies on on-site measurements of wind and heat environments and numerical simulations. From the perspective of "scientific construction," numerical simulations, through simplified modeling and parametric simulations, supplemented by on-site measurements, can effectively complete basic research tasks such as phenomenon analysis, distribution causes, and design strategies. However, from the perspectives of "accurate perception" and "comprehensive evaluation," on-site measurements are limited by the distribution of test sites and measurement points. Furthermore, the discrepancy between theoretical models and real-world scenarios, as well as the simplification of boundary conditions, makes it difficult to achieve accurate and comprehensive evaluations of thermal safety and comfort within spaces where thermal environmental factors are complex and ever-changing. Furthermore, the lack of rapid, accurate, and comprehensive observation methods makes it impossible to carry out "optimization and transformation" for specific cases.

[0003] With the rapid development of low-altitude remote sensing technology using drones, the spatial resolution of ground-based observation data has significantly increased. This has greatly facilitated the precise interpretation of surface characteristic parameters and provided new possibilities for the refined observation of the thermal environment in urban micro-scale outdoor spaces. Current research indicates that the application of low-altitude remote sensing technology using drones in the field of urban outdoor thermal environments is primarily limited to the research and application of mechanisms for the inversion of micro-scale surface emissivity and temperature. Due to the lack of mature mechanisms for quantifying anisotropic radiation received by pedestrians, the application of low-altitude remote sensing technology in the refined measurement of thermal comfort in micro-scale outdoor spaces requires further research. No studies have yet evaluated pedestrian thermal comfort in specific cases based on low-altitude remote sensing observations and provided quantifiable improvement recommendations. Summary of the Invention

[0004] To address these challenges, this paper provides a method for detecting, evaluating, and improving thermal comfort in outdoor spaces for pedestrians based on low-altitude drone remote sensing technology. This method, tailored to the built environment and thermal comfort status of specific cases, identifies specific modification targets and quantitative indicators to enhance thermal comfort near the ground. This research results enhance a comprehensive and accurate system for observing and evaluating the thermal environment during the post-construction phase of a project, providing guidance and reference data for optimizing thermal comfort in micro-scale outdoor spaces.

[0005] To achieve the above objectives, the present invention provides a method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on low-altitude remote sensing technology using drones, which specifically includes the following steps:

[0006] S1: Image acquisition: screening multiple scenes with different vegetation coverage, hard pavement, and openness as observation objects, determining the location and area of the scene, judging the area type, investigating the technical parameters and image capture format of the dual sensors, determining the payload, determining the shooting angle and interval based on the observation range and requirements, and planning the UAV's flight trajectory;

[0007] S2: Arrangement of meteorological stations and near-ground measurement points;

[0008] S3: Collect the coordinates of the ground control points. According to the scene size and the resolution requirements of the collected image, determine the number, placement and size of the target cloth. The target size should be greater than three times the resolution, and the material of the target cloth should be set to tin foil.

[0009] S4: Quantify the spatial distribution of characteristic parameters that characterize the optical and openness of the surface. Import the captured multi-scene multispectral images into the photogrammetry software. Through image alignment, camera distortion correction, control point calibration, high-density point cloud and 3D mesh model construction, digital surface model (DSM) generation and reflectance spectral orthophoto reconstruction, generate orthophotos and DSMs representing the reflectance of different bands of the surface in multiple scenes.

[0010] S5: Quantify the spatial distribution of characteristic parameters that characterize the surface thermophysical properties and obtain the orthophoto of the surface inversion temperature;

[0011] S6: Constructing short-wave radiation fields of ideal diffuse surfaces in three dimensions for multiple scenarios;

[0012] S7: Constructing multi-scenario three-dimensional ideal diffuse surface long-wave radiation fields;

[0013] S8: Establish the radiation component sampling and average radiation temperature field received by pedestrians near the ground;

[0014] S9: Establish a mechanism for evaluating and improving pedestrian thermal comfort based on UAV low-altitude remote sensing technology;

[0015] S10: Based on the surface characteristic parameters extracted in S4 and S5 that are closely related to the thermal environment and affected by design elements, explore their relationship with the average radiant temperature of pedestrians near the ground and the outdoor thermal comfort (OTC) index, and quantify the impact of design elements on human thermal comfort near the ground.

[0016] Preferably, in S2, a meteorological station is arranged in an open area of the scene or on the roof of a higher floor to measure direct radiation, short-wave diffuse horizontal radiation from the sky, long-wave horizontal radiation, air temperature and humidity, and wind speed; a number of near-ground measurement points are evenly arranged in the scene to synchronously measure the air temperature and humidity, wind speed, and black globe temperature at the height of pedestrians.

[0017] Preferably, in S4, the following steps are specifically included:

[0018] S41: Based on the reflectance of the near-infrared and red bands, the Normalized Difference Vegetation Index (NDVI) is obtained through band calculation to characterize the surface vegetation coverage and growth in multiple scenarios;

[0019] S42: To obtain the albedo within the multispectral camera response band (400-900nm), the surface reflectance of the five narrow channels (blue, green, red, red-edge, and near-infrared bands) is weighted averaged according to the spectral response function of the multispectral camera. The surface albedo within the range of 400-900nm is obtained through band calculation to characterize the shortwave radiation reflection characteristics of the surface in multiple scenes.

[0020] S43: Based on the DSM imagery output by photogrammetry software, a sky view factor (SVF) is used to quantify the surface occlusion conditions in multiple scenes. A pixel in the DSM represents the absolute surface height of the corresponding feature. A search is performed in multiple directions within a hemisphere of a self-defined radius, centered on the DSM pixel. The proportion of light reaching the sky is quantified to determine the SVF of each pixel.

[0021]

[0022] Where θ svf is the zenith angle measured vertically upward; φ svf is the azimuth measured along the ground plane.

[0023] Preferably, S5 specifically includes the following steps:

[0024] S51: Based on the wide-band emissivity inversion model, the NDVI image is selected for band calculation to obtain the emissivity (ε 8～14 ), characterizes the surface thermal radiation release capacity in multiple scenarios;

[0025]

[0026] Where, ε 8～14 is the surface broad-band emissivity, DNVI is the surface normalized difference vegetation index;

[0027] S52: For the multi-scene thermal infrared images captured, they are also imported into the photogrammetry software. Through image alignment, camera distortion correction, control point calibration, high-density point cloud construction, DSM generation and orthophoto reconstruction, grayscale orthophotos representing the surface thermal radiation characteristics of the multi-scene are generated.

[0028] S53: Based on radiation calibration, the grayscale orthophoto representing the surface thermal radiation characteristics is converted into the surface brightness temperature (T b) orthophoto, using the equivalent wavelength of the thermal imager response band and the Planck function to perform band calculations, the surface brightness temperature orthophoto is converted into the radiation brightness of the surface received by the thermal imager at the equivalent wavelength orthophotos;

[0029]

[0030] Where DN is the grayscale value that characterizes the surface thermal radiation characteristics of multiple scenes, T b is the surface brightness temperature;

[0031] S54: Building a Lookup Table Loop Based on an Equation ε 8～14 , SVF image, use the least squares method to find the minimum difference between the theoretical value in the lookup table and the image measured value, record the serial number of the corresponding working condition in the table, and extract the corresponding unknown parameter (T s_URTE ), as the inversion result of the corresponding pixel;

[0032]

[0033] Where, Characterizes the radiance of the ground surface received by the thermal imager at its equivalent wavelength, SVF represents the sky viewing angle factor, ε 8～14 represents the surface broadband emissivity, τ 8～14 They represent the low-altitude atmospheric upward radiation, atmospheric downward radiation and transmittance respectively.

[0034] S55: Circulate pixel by pixel and perform quantitative inversion to finally obtain the orthophoto of the surface inverted temperature, which represents the surface thermal conditions in multiple scenarios.

[0035] Preferably, S6 specifically includes the following steps:

[0036] S61: Based on multispectral images from different channels acquired by drones, batch read image metadata, including location, time, and camera parameters. Use these parameters to correct distortion for each lens, ensuring that the coordinates of pixels in the image accurately reflect their actual locations.

[0037] S62: Perform radiation calibration, batch convert the grayscale value of each channel of the image into the corresponding shortwave radiation brightness value (L sw_3D_surface );

[0038] S63: Based on the assumption that the reflection of an ideal diffuse surface is isotropic, the radiance value of each channel is multiplied by π to obtain the shortwave radiation value of the lower hemisphere integral of the channel. The radiation values of the five channels are summed up to obtain the shortwave radiation value (E) in the range of 400-900 nm on the urban three-dimensional space surface. sw_3D_surface), using machine learning to establish a model between shortwave radiation values and the corresponding five channel grayscale values:

[0039] E sw_3D_surface =π*L sw_3D_surface

[0040] Where, E sw_3D_surface Represents the shortwave radiation value from the three-dimensional surface of the city; L sw_3D_surface It represents the shortwave radiation brightness of the urban three-dimensional space surface obtained from a specific observation angle;

[0041] S64: Extract pixel data to form an input table, randomly select 70% of the pixels as a training data set, and the remaining 30% as a validation data set. After the photogrammetry process of the multispectral data in S4, generate and output 3D point cloud data and a 3D mesh model representing the grayscale values of the five shortwave channels of the entire scene;

[0042] S65: Based on the model between the shortwave radiation value established in S63 and the corresponding five channel grayscale values, generate three-dimensional point cloud data with precise coordinate information and ideal diffuse surface shortwave radiation information. Using the principles of spatial geometric operations, assign the point cloud data representing the ideal diffuse surface shortwave radiation to the triangular faces on the grid model, and then establish a multi-scene three-dimensional ideal diffuse surface shortwave radiation field.

[0043] Preferably, in S7, the following steps are specifically included:

[0044] S71: After the photogrammetry process of the thermal infrared data in S5, the three-dimensional surface brightness temperature (T b ), constructing a lookup table to convert the brightness temperature three-dimensional point cloud data into the integrated radiation brightness three-dimensional point cloud data;

[0045] S72: T b The independent variable is regarded as the lookup table, and the range and step size of the independent variable are determined by T b The measured range and B 8～14 (T b ) sensitivity; the dependent variable of the lookup table is the theoretical value of the radiant brightness after the integration of the thermal imager's spectral response band (B' 8～14 (T b ));

[0046] S73: After the lookup table is established, the brightness temperature in the table is compared with the measured three-dimensional surface T b Compare the point cloud data, follow the principle of minimum difference, record the serial number in the table and the corresponding B' 8～14 (T b ), as the radiant brightness after the integration of the thermal imager spectral response band, and then construct a8～14 (T b ) Point cloud data of information;

[0047] S74: Follow the principle of spatial geometric operation and convert B 8～14 (T b ) Assign the point cloud data to the triangular surface on the mesh model to establish an ideal diffuse surface B representing multiple scenes 8～14 (T b ) a three-dimensional model of information;

[0048] S75: Based on a 3D mesh model, a ray tracing method is used to calculate the SVF of each triangle. The broadband emissivity of each surface triangle is assigned based on the inverted surface emissivity and DSM, following the principles of spatial geometric operations. The emissivity of non-surface triangles is assigned based on the point cloud category attributes and the empirical emissivity values of the corresponding materials.

[0049] S76: Combined with the low-altitude atmospheric parameters, the input parameters for inverting the temperature of each triangle surface based on the urban radiation transfer equation have all been obtained (B 8～14 (T b ), SVF, ε 8～14 、 and τ 8～14 ), based on the Boltzmann formula, the inverted temperature of the triangular surface is converted into a long-wave radiation field, and then the long-wave radiation field of a three-dimensional ideal diffuse surface in multiple scenes is established.

[0050] Preferably, in S8, the following steps are specifically included:

[0051] S81: Extract the reference points for pedestrian thermal comfort calculation. Based on the three-dimensional point cloud data generated by S4 representing the grayscale values of the five shortwave channels of the entire scene, extract the ground point cloud by identifying the elevation differences of the point cloud, and construct an independent ground model. According to the scope of the observation scene and the degree of heterogeneity, the ground model is divided into areas of approximately 1.0m 2 The three-dimensional coordinates of the centroid of each triangle are extracted, and its Z coordinate is raised by 1.1m to represent the height of the center of gravity of the human body. The corrected coordinates are the average radiation temperature sampling value of pedestrians in outdoor space (T mrt_sampled ) The position basis for calculation;

[0052] S82: Based on the established multi-scenario 3D ideal diffuse surface long- and short-wave radiation fields, the indexed view sphere (IVS) method is used to quantify the long- and short-wave radiation from different surface elements reaching the human body surface.

[0053] S83: For the long- and short-wave radiation received by the human body from the sky, during the period when the drone collects data, the weather station simultaneously observes and records the average value of the scene's total horizontal irradiance GHI and normal direct irradiance DNI. Under the assumption that the sky diffuse reflection is isotropic, the direct-scatter separation method is used to obtain the sky diffuse reflection horizontal irradiance DHI during this period:

[0054] DHI=GHI-DNI·sinθ

[0055] Where θ is the solar altitude angle;

[0056] The long-wave radiation L from the sky sky Estimation of the average air temperature and humidity based on the human-environment heat exchange model and simultaneous near-ground measurements within the scene:

[0057]

[0058] Where, L sky represents the long-wave radiation from the sky; T a is the average air temperature measured near the ground in the scene; water vapor pressure v p It is calculated based on the average value of the air temperature and humidity measured near the ground in the scene; N represents the degree of cloud cover in the sky, with a value of (0-8), where 0 represents a completely clear sky and 8 represents a cloudy sky completely covered by clouds; σ is the Boltzmann coefficient, with a value of 5.67×10 -8 W·m -2 ·K -4 ;

[0059] The normal direct irradiance DNI received by a person is quantified by ray tracing. The DNI reaching the human body surface depends on the human body surface projection coefficient (f p ), f p According to the human standing model, it is deduced:

[0060] f p =3.67·10 -7 θ 3 -6.74·10 -5 θ 2 +8.49·10 -4 θ+0.297

[0061] If the absorption coefficient of the human body surface for long-wave and short-wave radiation is the same, the average radiation temperature near the ground at different locations in multiple scenes is calculated based on the sampling results of each radiation component:

[0062]

[0063] Where, T mrt_sampled represents the average radiation temperature of pedestrians calculated based on the sampling values of each radiation component; α swand α lw They refer to the absorption coefficients of the human body surface wearing clothes to short-wave radiation and long-wave radiation respectively; δ shadow Used to indicate whether the human body surface receives direct radiation, δ shadow When it is 0, it means the human body is in the shadow, δ shadow If it is 1, it means that the human body has received direct radiation; i represents the number of the triangle face hit by the radiation emitted by the human body; E sw_sky and E lw_sky They represent the shortwave and longwave radiation received by the human body from the sky; E sw_3D_surface_i and E lw_3D_surface__i They respectively refer to the amount of short-wave radiation and long-wave radiation received by the human body from the i-th triangle surface.

[0064] Preferably, in S9, using T mrt_sampled and synchronously measured near-ground meteorological parameters, including air temperature, relative humidity, and wind speed; based on the human thermal comfort model, high-spatial-resolution pedestrian thermal comfort indices, including SET*, PET, and UTCI, are calculated and obtained, and the pedestrian thermal comfort indices are compared with the sampled values of the pedestrian thermal comfort indices and those calculated based on the near-ground measured meteorological parameters.

[0065] Preferably, in S10, a random forest model is used to quantify the characteristic parameters of each triangle in the ground model and T mrt_sampled The nonlinear relationship between the sampling values of typical OTC indicators is analyzed. SHAP and PDPbox analysis methods are used to quantitatively evaluate the sensitivity of near-ground human thermal comfort to various ground characteristic parameters. Before model training, the data is preprocessed, which specifically includes the following steps:

[0066] S101: Geo-register and resample the orthophoto of surface feature parameters and DSM, extract pixel information, and obtain point cloud data that accurately represents the ground feature parameters and their three-dimensional coordinates;

[0067] S102: Based on the established multi-scenario ground model and following the principles of spatial geometric operations, a ground model representing surface characteristic parameters is constructed;

[0068] S103: Extract T of triangular faces in non-building coverage areas mrt_sampled The sampling values of typical OTC indicators and their corresponding ground feature parameters are used as effective inputs for model training. The input data are normalized. 70% and 30% of the input data are randomly selected for model training and validation. The accuracy is verified based on the coefficient of determination and root mean square error of the validation data set.

[0069] S104: Based on the sensitivity analysis results of the training model, clarify the design elements and corresponding quantitative indicators for improving pedestrian thermal comfort.

[0070] Therefore, this paper employs the aforementioned method for detecting, evaluating, and improving thermal comfort in outdoor spaces for pedestrians based on low-altitude drone remote sensing technology. This method, tailored to the built environment and thermal comfort status of a specific case, identifies specific modification targets and quantitative indicators to enhance thermal comfort near the ground. This research results enhance a comprehensive and accurate system for observing and evaluating the thermal environment during the post-construction phase of a project, providing guidance and reference data for optimizing thermal comfort in microscale outdoor spaces. This method has significant practical potential and can provide valuable insights for research and application in related fields.

[0071] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 This is a workflow diagram of the method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology of the present invention;

[0073] Figure 2 1 is a diagram of a visible light orthophoto, a schematic diagram of a nearby meteorological station, and a schematic diagram of a near-ground meteorological measurement point in an embodiment of the present invention;

[0074] Figure 3 is the surface normalized difference vegetation index (NDVI), albedo (Albedo) and sky view factor (SVF) orthophoto in the embodiment of the present invention;

[0075] Figure 4 is the broadband emissivity of the surface in the embodiment of the present invention (ε 8～14 ) and temperature (T s_URTE ) Orthophotos;

[0076] Figure 5 are the three-dimensional ideal surface short-wave radiation field and long-wave radiation field in the embodiment of the present invention;

[0077] Figure 6 Schematic diagram of pedestrian average radiation temperature field and thermal comfort index field and corresponding accuracy verification in an embodiment of the present invention;

[0078] Figure 7 is a pedestrian T in the embodiment of the present invention mrt_sampled And a schematic diagram of the sensitivity analysis results of typical OTC index sampling values to surface characteristic parameters. DETAILED DESCRIPTION

[0079] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0080] Unless otherwise defined, technical or scientific terms used in the present invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.

[0081] The words “include” or “comprising” and similar words used in the present invention mean that the elements before the word include the elements listed after the word, and do not exclude the possibility of also including other elements. The orientation or position relationship indicated by the terms “inside”, “outside”, “upper”, “lower”, etc. is based on the orientation or position relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation of the present invention. When the absolute position of the described object changes, the relative position relationship may also change accordingly. In the present invention, unless otherwise clearly stipulated and limited, the terms such as “attachment” should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral whole; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0082] Example

[0083] A method for detecting, evaluating, and improving thermal comfort for pedestrians in outdoor spaces based on low-altitude drone remote sensing technology is described. The drone is equipped with dual sensors (a thermal imager and a multispectral camera). A synchronization control program on the gimbal enables dual-sensor compatibility and multi-angle simultaneous observation. The drone's flight path and dual-sensor photography missions are connected to a ground-based remote control system via a wireless transmission system. A ground-based positioning control system ensures the accuracy of data collection locations. During low-altitude drone remote sensing observations, meteorological stations are deployed in open areas or on rooftops within the observation scene to measure direct radiation, shortwave diffuse horizontal radiation from the sky, longwave horizontal radiation, air temperature and humidity, and wind speed. Furthermore, considering spatial variations in factors such as vegetation cover, hard pavement, and openness, several representative perigee points within the observation scene are selected and deployed for simultaneous observation of air temperature and humidity, wind speed, and global temperature. This provides the input data required for sampling and calculating outdoor pedestrian thermal comfort indicators and verification. Based on this collected data, embodiments of the present invention batch-process multi-source drone low-altitude remote sensing images to quantify the spatial distribution of surface characteristic parameters and construct a three-dimensional radiation field for multiple scenes. By sampling the radiation received by pedestrians at different spatial locations near the ground, the spatial distribution of pedestrian thermal comfort is further evaluated. Combined with machine learning, the relationship between surface characteristics and thermal comfort is analyzed, and the impact of design factors on pedestrian thermal comfort is quantified. Based on this, strategies to improve thermal comfort are proposed.

[0084] like Figure 2-Figure 7As shown, the experiment used multiple scenes consisting of basic urban spatial units such as industrial parks, streets, and squares as the experimental subjects. A Yangda N70 drone equipped with a thermal imager and a multispectral camera was used for multi-source observation. The aircraft can carry a maximum load of 8.0 kg, and with a 3.5 kg payload, it has a flight time of 70 minutes. The operating environment temperature range is -10-40°C. The gimbal used was a DJI Ronin MX gimbal (2.77 kg), compatible with two third-party payloads. It was equipped with a WIRIS PRO thermal imager (430 g, 9 mm focal length lens, 640 x 512 focal plane array, thermal radiation response band range of 7.5–13.5 μm, equivalent wavelength of 11.092 mm, temperature measurement accuracy of ±2°C, and temperature range of -25–1500°C) and a Micasense Rededge-P multispectral camera (315 g, 5.5 mm focal length lens, 1456 x 1088 focal plane array, response band range of 0.4–0.9 μm, and five channels (blue, green, red, red-edge, and near-infrared) with center wavelengths of 475, 560, 668, 717, and 840 nm, respectively). Data from both sensors was recorded in TIFF format.

[0085] Utilizing an integrated multi-source data observation system for remote sensing thermal environments, the UAV achieved simultaneous acquisition of multispectral and thermal infrared imagery in five directions (orthogonal, east, south, west, and north) by setting the gimbal's mission angles to 0° and 60°. The UAV operated from 1:25 PM to 2:08 PM, maintaining a fixed altitude of 250 meters. Its flight path formed a crisscross pattern, achieving an 80% overlap between heading and lateral directions. The multispectral and thermal cameras captured images continuously at 2.0-second intervals.

[0086] In terms of near-surface synchronous measurements, six near-surface measurement points were evenly distributed in the study area. Each measurement point was equipped with sensors (i.e., HOBO MX2302A, DeltaAP3203.2, and Delta TP3276.2) that measure air temperature, humidity, wind speed, and globe temperature at a height of 1.1 meters above the ground. The data collection and recording interval was uniformly set to once every 30 seconds. In addition, a weather station close to the study area was selected. The weather station was equipped with horizontally pointing shortwave radiometers (Kipp & Zonen SOLYS2 and Kipp & Zonen CMP3) and longwave radiometers (Kipp & Zonen CGR3) to measure DNI, GHI, and L respectively. sky ,The data collection and recording interval is also 30 seconds.

[0087] Specifically include the following steps, such as Figure 1 As shown:

[0088] S1: Image acquisition, e.g. Figure 2As shown, multiple scenes with different vegetation coverage, hard pavement and openness are selected as observation objects, the location and area of the scene are determined, and the type of area is judged whether it is a no-fly zone or a restricted-fly zone; the technical parameters of the dual sensors (such as weight, response spectrum, focal length, focal plane array, operating temperature, horizontal field of view angle and vertical field of view angle, etc.) and image shooting format (both thermal infrared images and multispectral images are in TIFF format) are investigated, the payload is judged, the shooting angle and shooting interval are determined according to the observation range and needs, and the route trajectory of the UAV is planned; including the heading and lateral overlap rate, flight altitude and speed.

[0089] S2: Arrange weather stations and near-ground measurement points. In S2, select open areas or rooftops of higher floors to arrange weather stations to measure direct radiation, short-wave diffuse horizontal radiation from the sky, long-wave horizontal radiation, air temperature and humidity, and wind speed. Evenly arrange several near-ground measurement points in the scene. The distribution of points is mainly based on the spatial differences of factors such as vegetation coverage, hard pavement, and openness in the scene, such as Figure 2 As shown, the air temperature and humidity, wind speed and black globe temperature at the pedestrian height are measured simultaneously.

[0090] S3: Collect the coordinates of the ground control points. Determine the number, placement, and size of the target cloth based on the scene size and image resolution requirements. The target size should be greater than three times the resolution, and the target cloth should be made of tinfoil. During the flight mission, use an RTK GPS surveying device to synchronously collect the coordinates of the target cloth center position as the basis for subsequent image control point correction.

[0091] S4: The original multispectral images collected by Micasense Rededge-P were imported into Pix4D software. After a series of processing including image alignment, camera distortion correction, control point calibration, high-density point cloud and 3D mesh model construction, DSM generation, and reflectance spectral orthophoto reconstruction (see Table 1 for relevant setting parameters), orthophotos and DSMs representing the reflectance of different bands of the surface in multiple scenes were output.

[0092] Table 1

[0093]

[0094]

[0095] In S4, the following steps are specifically included:

[0096] S41: Based on the reflectance of the near-infrared and red bands, the Normalized Difference Vegetation Index (NDVI) is obtained through band calculation to characterize the surface vegetation coverage and growth in multiple scenarios;

[0097] S42: To obtain the albedo within the multispectral camera response band (400-900 nm), the surface reflectance of the five narrow channels (blue, green, red, red-edge, and near-infrared bands) is weighted averaged according to the spectral response function of the MicasenseRededge-P multispectral camera. The surface albedo within the range of 400-900 nm is obtained through band calculation to characterize the shortwave radiation reflection characteristics of the surface in multiple scenes.

[0098] S43: Based on the DSM image output by the photogrammetry software, the sky view factor (SVF) is quantified to represent the surface occlusion of multiple scenes using the sky view area analysis method. The pixel in the DSM represents the absolute surface height of the corresponding object. With the pixel in the DSM as the center, a search is performed in multiple directions within a hemisphere with a self-set radius to quantify the proportion of light reaching the sky and determine the SVF of different pixels. The orthophoto of NDVI, albedo and SVF is as follows: Figure 3 shown.

[0099]

[0100] Where θ svf is the zenith angle measured vertically upward; φ svf is the azimuth measured along the ground plane.

[0101] To take advantage of the high spatial resolution of UAV imagery and focus on the proximity effect caused by local features of the observed scene, a search radius of 10 DSM pixels was determined for the hemispherical search. At the same time, the light was required to diverge evenly in 32 different directions within the hemisphere, and the proportion of light reaching the sky was quantified.

[0102] S5: Quantify the spatial distribution of characteristic parameters that characterize the surface thermophysical properties to obtain an orthophoto of the surface inversion temperature. S5 specifically includes the following steps:

[0103] S51: Based on the wide-band emissivity inversion model, the NDVI image is selected for band calculation to obtain the emissivity (ε 8～14 ), characterizes the surface thermal radiation release capacity in multiple scenarios;

[0104]

[0105] Where, ε 8～14 is the surface broad-band emissivity, and NDVI is the surface normalized difference vegetation index;

[0106] S52: For the multi-scene thermal infrared images captured, they are also imported into the photogrammetry software. Through image alignment, camera distortion correction, control point calibration, high-density point cloud construction, DSM generation and orthophoto reconstruction, grayscale orthophotos representing the surface thermal radiation characteristics of the multi-scene are generated. The relevant technical parameters are shown in Table 2.

[0107] Table 2

[0108]

[0109]

[0110] S53: Based on radiation calibration, the grayscale orthophoto representing the surface thermal radiation characteristics is converted into the surface brightness temperature (T b ) orthophoto, using the equivalent wavelength of the thermal imager response band and the Planck function to perform band calculations, the surface brightness temperature orthophoto is converted into the radiation brightness of the surface received by the thermal imager at the equivalent wavelength orthophotos;

[0111]

[0112] Where DN is the grayscale value that characterizes the surface thermal radiation characteristics of multiple scenes, T b is the surface brightness temperature;

[0113] S54: Building a lookup table loop based on an equation ε 8～14 , SVF image, use the least squares method to find the minimum difference between the theoretical value in the lookup table and the image measured value, record the serial number of the corresponding working condition in the table, and extract the corresponding unknown parameter (T s_URTE ), as the inversion result of the corresponding pixel;

[0114]

[0115] Where, Characterizes the radiance of the ground surface received by the thermal imager at its equivalent wavelength, SVF represents the sky viewing angle factor, ε 8～14 represents the surface broadband emissivity, τ 8～14 They represent the low-altitude atmospheric upward radiation, atmospheric downward radiation and transmittance respectively.

[0116] S55: Circulate pixel by pixel, quantitatively invert, and finally obtain the orthophoto of the surface inverted temperature, which represents the surface thermal conditions in multiple scenes, ε 8～14 and T s_URTE The orthophoto of Figure 4 shown.

[0117] S6: Constructing a multi-scenario three-dimensional ideal diffuse surface short-wave radiation field; S6 specifically includes the following steps:

[0118] S61: Based on multispectral images from different channels acquired by drones, batch read image metadata, including location, time, and camera parameters. Use these parameters to correct distortion for each lens, ensuring that the coordinates of pixels in the image accurately reflect their actual locations.

[0119] S62: Perform radiation calibration, batch convert the grayscale value of each channel of the image into the corresponding shortwave radiation brightness value (L sw_3D_surface );

[0120] S63: Based on the assumption that the reflection of an ideal diffuse surface is isotropic, the radiance value of each channel is multiplied by π to obtain the shortwave radiation value of the lower hemisphere integral of the channel. The radiation values of the five channels are summed up to obtain the shortwave radiation value (E) in the range of 400-900 nm on the urban three-dimensional space surface. sw_3D_surface ), machine learning LightGBM is used to establish a model between shortwave radiation values and the corresponding five channel grayscale values:

[0121] E sw_3 D _surface =π*L sw_3D_surface

[0122] Where, E sw_3D_surface Represents the shortwave radiation value from the three-dimensional surface of the city; L sw_3D_surface It represents the shortwave radiation brightness of the urban three-dimensional space surface obtained from a specific observation angle;

[0123] S64: Extract pixel data to form an input table. Randomly select 70% of the pixels as the training data set and the remaining 30% as the validation data set. The validation results show that the model performs well, with a coefficient of determination of 0.978 and a root mean square error of 3.58 W·m -2 After the photogrammetry process of multispectral data in S4, 3D point cloud data (1,497,774 points) and 3D mesh model (254,154 triangular faces, each with an area of less than 5m2) representing the grayscale values of the five shortwave channels of the entire scene are generated and output. 2 );

[0124] S65: Based on the model between the shortwave radiation value established in S63 and the corresponding five channel grayscale values, generate three-dimensional point cloud data with precise coordinate information and ideal diffuse surface shortwave radiation information. Using the principle of spatial geometric operation, (the set distance threshold is 10cm, which means that when the spatial distance between the point cloud and the triangular surface is not greater than 10cm, the average value of the physical variables of the relevant point cloud is obtained and assigned to the corresponding triangular surface), the point cloud data representing the shortwave radiation of the ideal diffuse surface is assigned to the triangular surface on the grid model, and then a multi-scene three-dimensional ideal diffuse surface shortwave radiation field is established, such as Figure 5 shown.

[0125] S7: Constructing a multi-scenario three-dimensional ideal diffuse surface long-wave radiation field; S7 specifically includes the following steps:

[0126] S71: After the photogrammetry process of the thermal infrared data in S5, the three-dimensional surface brightness temperature (T b ) point cloud data (1,867,397 points). Since it is impossible to directly obtain the radiant brightness after integration based on the spectral response band of the thermal imager, a lookup table is constructed to convert the brightness temperature 3D point cloud data into the integrated radiant brightness 3D point cloud data.

[0127] S72: T b Consider T as the independent variable of the lookup table. b Considered as the independent variable of the lookup table, T b The variation range (step length) is from 290K to 353K (0.3K), including 210 groups of working conditions. The dependent variable of the lookup table is the theoretical value of the radiant brightness after integration of the spectral response band of the thermal imager (B' 8～14 (T b ));

[0128] S73: After the lookup table is established, the brightness temperature in the table is compared with the measured three-dimensional surface T b Compare the point cloud data, follow the principle of minimum difference, record the serial number in the table and the corresponding B' 8～14 (T b ), as the radiant brightness after the integration of the thermal imager spectral response band, and then construct a 8～14 (T b ) Point cloud data of information;

[0129] S74: Follow the principle of spatial geometric operation and convert B 8～14 (T b ) Assign the point cloud data to the triangular surface on the mesh model to establish an ideal diffuse surface B representing multiple scenes 8～14 (T b ) a three-dimensional model of information;

[0130] S75: Based on the three-dimensional mesh model, a ray tracing method is used to calculate the SVF of each triangle. 200 rays are emitted outward from each triangle of the three-dimensional mesh model. The cosine of the angle between the ray and the surface normal is used as the ray weight coefficient. The perspective relationship between each triangle and its surrounding environment is obtained by analyzing the intersection of the ray with other surrounding surfaces. Subsequently, the sum of the perspective factors of each triangle is subtracted from 1 to deduce the SVF of each triangle. The derivation of the three-dimensional surface emissivity is divided into surface and non-surface: the wide-band emissivity of each triangle on the surface is based on the inverted surface emissivity and DSM, and is assigned according to the principles of spatial geometric operations; the emissivity of the non-surface triangle is assigned according to the point cloud category attributes and the empirical value of the emissivity of the corresponding material;

[0131] Specifically, the 3D point cloud data representing the grayscale values of the five shortwave channels of the entire scene generated in S4 was input into the Lidar360 software. The ground point cloud was removed, and the building and tree point clouds were retained. Next, the interactive classification tool in the profile editor was used to select samples of different building facade materials and vegetation. The main types were dark-coated glass, dark fluorocarbon-sprayed aluminum-plastic panel, plastic grille, and trees. After selecting the samples, the machine learning classification interface was entered. The input data to be classified was multispectral 3D point cloud data. Other categories besides ground points were selected as input categories, and relevant samples were loaded for training. After running the machine learning classification, most point clouds were classified into the correct categories, but some misclassified point clouds existed. Therefore, post-processing was required to improve the classification results. The point clouds of various ground objects and their category attributes were output, and the facade point cloud data were assigned values based on the point cloud data category and the empirical emissivity values corresponding to each category of material (0.88, 0.92, 0.94, and 0.98). Finally, the point cloud data representing the broadband emissivity was also assigned to the corresponding triangular faces based on spatial geometric operations.

[0132] S76: Combined with the low-altitude atmospheric parameters, the input parameters for inverting the temperature of each triangle surface based on the urban radiation transfer equation have all been obtained (B 8～14 (T b ), SVF, ε 8～14 、 and τ 8～14 ), based on the Boltzmann formula, the inverted temperature of the triangular surface is converted into a long-wave radiation field, and then the long-wave radiation field of a three-dimensional ideal diffuse surface in multiple scenes is established.

[0133] S8: Establishing the radiation component sampling and average radiation temperature field received by pedestrians near the ground; S8 specifically includes the following steps:

[0134] S81: Extract the reference points for pedestrian thermal comfort calculation. Based on the three-dimensional point cloud data generated by S4 representing the grayscale values of the five shortwave channels of the entire scene, extract the ground point cloud by identifying the elevation differences of the point cloud, and construct an independent ground model. According to the scope of the observation scene and the degree of heterogeneity, the ground model is divided into areas of approximately 1.0m 2 The three-dimensional coordinates of the centroid of each triangle are extracted, and its Z coordinate is raised by 1.1m to represent the height of the center of gravity of the human body. The corrected coordinates are the average radiation temperature sampling value of pedestrians in outdoor space (T mrt_sampled ) The position basis for calculation;

[0135] S82: Based on the established multi-scenario 3D ideal diffuse surface long- and short-wave radiation fields, the indexed view sphere (IVS) method is used to quantify the long- and short-wave radiation from different surface elements reaching the human body surface.

[0136] Assuming that the human body is a microsphere, according to Lambert's law, the radiation it receives from all directions has the same weight. The human microsphere emits rays uniformly in all directions. When the light hits a triangle, the index of the triangle is recorded to facilitate the calculation of T mrt_sampled The corresponding long- and short-wave radiation data for the surface impacted by the ray were retrieved. Considering the spatial complexity of the test scene, the number of rays uniformly emitted in all directions by the human microsphere at each reference point was set to 401. The entire process was based on parallel computing using a computer (CPU: Intel Processor i7-12700F CPU @ 2.10GHz, 12 cores; GPU: Nvidia GeForce RTX 3060Ti, 8GB, 4864 cores).

[0137] S83: For the long- and short-wave radiation received by the human body from the sky, during the period when the drone collects data, the weather station simultaneously observes and records the average value of the scene's total horizontal irradiance GHI and normal direct irradiance DNI. Under the assumption that the sky diffuse reflection is isotropic, the direct-scatter separation method is used to obtain the sky diffuse reflection horizontal irradiance DHI during this period:

[0138] DHI=GHI-DNI·sinθ

[0139] Where θ is the solar altitude angle;

[0140] The long-wave radiation L from the sky sky Estimation of the average air temperature and humidity based on the human-environment heat exchange model and simultaneous near-ground measurements within the scene:

[0141]

[0142] Where, L sky represents the long-wave radiation from the sky; Ta is the average air temperature measured near the ground in the scene; water vapor pressure v p It is calculated based on the average value of the air temperature and humidity measured near the ground in the scene; N represents the degree of cloud cover in the sky, with a value of (0-8), where 0 represents a completely clear sky and 8 represents a cloudy sky completely covered by clouds; σ is the Boltzmann coefficient, with a value of 5.67×10 -8 W·m -2 ·K -4 ;

[0143] The normal direct irradiance DNI received by a person is quantified by ray tracing. The DNI reaching the human body surface depends on the human body surface projection coefficient (f p ), f p According to the human standing model, it is deduced:

[0144] f p =3.67·10 -7 θ 3 -6.74·10 -5 θ 2 +8.49·10 -4 θ+0.297

[0145] If the absorption coefficient of the human body surface for long-wave and short-wave radiation is the same, the average radiation temperature near the ground at different locations in multiple scenes is calculated based on the sampling results of each radiation component, covering an area of about 94,500m 2 In multiple scenes, high spatial resolution (97,200 computational grids, each covering an area of approximately 1.0m 2 ) takes about three hours to construct the mean radiation temperature field.

[0146]

[0147] Where, T mrt_sampled represents the average radiation temperature of pedestrians calculated based on the sampling values of each radiation component; α sw and α lw They refer to the absorption coefficients of the human body surface wearing clothes to short-wave radiation and long-wave radiation respectively; δ shadow Used to indicate whether the human body surface receives direct radiation, δ shadow When it is 0, it means the human body is in the shadow, δ shadow If it is 1, it means that the human body has received direct radiation; i represents the number of the triangle face hit by the radiation emitted by the human body; E sw_sky and E lw_sky They represent the shortwave and longwave radiation received by the human body from the sky; E sw_3D_surface_i and E lw_3D_surface__i They respectively refer to the amount of short-wave radiation and long-wave radiation received by the human body from the i-th triangle surface.

[0148] S9: Establish a mechanism for evaluating and improving pedestrian thermal comfort based on UAV low-altitude remote sensing technology; In S9, using T mrt_sampled and synchronously measured near-surface meteorological parameters (air temperature: 34°C, relative humidity: 57.82% and wind speed: 1.02m / s), including air temperature, relative humidity and wind speed; based on the human thermal comfort model, calculate and obtain high spatial resolution pedestrian thermal comfort indicators, including SET*, PET and UTCI, by comparing the pedestrian thermal comfort index sampling values with the pedestrian thermal comfort index calculated based on the near-surface measured meteorological parameters. The verification results show that when the height of the scene buildings and trees does not exceed 60m, the density does not exceed 70%, and the average value of the sky visibility factor is not less than 0.40, the T sampled by this method is mrt_sampled The T value is consistent with the T value based on the black globe temperature method. mrt_measured The values are highly consistent, with an RMSE of 3.18°C. Meanwhile, the OTC indicator sampling value verification results show that SET* has the highest sampling accuracy, with an RMSE of 1.96°C.

[0149] S10: Based on the surface characteristic parameters extracted in S4 and S5, which are closely related to the thermal environment and affected by design elements, explore the relationship between them and the average radiation temperature of pedestrians near the ground and the outdoor thermal comfort OTC index, and quantify the impact of design elements on the thermal comfort of human beings near the ground. In S10, a random forest model is used to quantify the characteristic parameters of each triangle in the ground model and T mrt_sampled The nonlinear relationship between the sampling values of typical OTC indicators is analyzed. SHAP and PDPbox analysis methods are used to quantitatively evaluate the sensitivity of near-ground human thermal comfort to various ground characteristic parameters. Before model training, the data is preprocessed, which specifically includes the following steps:

[0150] S101: Geo-register and resample the orthophoto of surface feature parameters and DSM, extract pixel information, and obtain point cloud data that accurately represents the ground feature parameters and their three-dimensional coordinates;

[0151] S102: Based on the established multi-scenario ground model, and following the principles of spatial geometric operations, a ground model representing surface characteristic parameters is constructed;

[0152] S103: Extract T of triangular faces in non-building coverage areas mrt_sampled and typical OTC index sampling values and their corresponding ground feature parameters, a total of 45,666 sets of valid data, as the effective input of model training, in order to ensure the scale consistency of different feature parameters in model training and avoid the large values of some features (such as T s_URTE) and produce excessive weights, the input data is normalized, 70% and 30% of the input data are randomly selected for model training and validation, and the accuracy is verified based on the coefficient of determination and root mean square error of the validation data set; Among them, T mrt_sampled The coefficient of determination (R 2 ) is 0.78, and the root mean square error (RMSE) is 2.39℃; while SET * sampled Model T 2 The value is also 0.78, but its RMSE is only 0.49°C.

[0153] S104: Based on the sensitivity analysis results of the training model, clarify the design elements and corresponding quantitative indicators for improving pedestrian thermal comfort. mrt_sampled The sensitivity analysis results of typical OTC index sampling values to surface characteristic parameters are as follows: Figure 7 As shown in Figure 2. Ground albedo is the main factor affecting outdoor thermal comfort. When Albedo increases from 0.0 to 0.3, T mrt_sampled (SET * sampled ) increased by as much as 9℃ (2℃). The influence of ground inversion temperature and normalized difference vegetation index was relatively minor. s_URTE When the temperature rises from 300K to 335K, T mrt_sampled (SET * sampled ) increased by as much as 14℃ (3℃); and when NDVI increased from 0.2 to 0.9, T mrt_sampled (SET * sampled ) decreased by about 2℃(0.5℃).

[0154] Therefore, this paper employs the aforementioned method for detecting, evaluating, and improving thermal comfort in outdoor spaces for pedestrians based on low-altitude drone remote sensing technology. This method, tailored to the built environment and thermal comfort status of a specific case, identifies specific modification targets and quantitative indicators to enhance thermal comfort near the ground. This research results enhance a comprehensive and accurate system for observing and evaluating the thermal environment during the post-construction phase of a project, providing guidance and reference data for optimizing thermal comfort in microscale outdoor spaces. This method has significant practical potential and can provide valuable insights for research and application in related fields.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting, evaluating and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology, characterized by: The specific steps include: S1: Image acquisition: screening multiple scenes with different vegetation coverage, hard pavement, and openness as observation objects, determining the location and area of the scene, judging the area type, investigating the technical parameters and image capture format of the dual sensors, determining the payload, determining the shooting angle and interval based on the observation range and requirements, and planning the UAV's flight trajectory; S2: Arrangement of meteorological stations and near-ground measurement points; S3: Collect the coordinates of the ground control points. According to the scene size and the resolution requirements of the collected image, determine the number, placement and size of the target cloth. The target size should be greater than three times the resolution, and the material of the target cloth should be set to tin foil. S4: Quantify the spatial distribution of characteristic parameters that characterize the optical and openness of the surface. Import the captured multi-scene multispectral images into the photogrammetry software. Through image alignment, camera distortion correction, control point calibration, high-density point cloud and 3D mesh model construction, digital surface model (DSM) generation and reflectance spectral orthophoto reconstruction, generate orthophotos and DSMs representing the reflectance of different bands of the surface in multiple scenes. S5: Quantify the spatial distribution of characteristic parameters that characterize the surface thermophysical properties and obtain the orthophoto of the surface inversion temperature; S6: Constructing short-wave radiation fields of ideal diffuse surfaces in three dimensions for multiple scenarios; S7: Constructing multi-scenario three-dimensional ideal diffuse surface long-wave radiation fields; S8: Establish the radiation component sampling and average radiation temperature field received by pedestrians near the ground; S9: Establish a mechanism for evaluating and improving pedestrian thermal comfort based on UAV low-altitude remote sensing technology; S10: Based on the surface characteristic parameters extracted in S4 and S5 that are closely related to the thermal environment and affected by design elements, explore their relationship with the average radiant temperature of pedestrians near the ground and the outdoor thermal comfort (OTC) index, and quantify the impact of design elements on human thermal comfort near the ground.

2. The method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology according to claim 1, characterized in that: In S2, meteorological stations are set up in open areas of the scene or on the roofs of higher floors to measure direct radiation, short-wave diffuse horizontal radiation from the sky, long-wave horizontal radiation, air temperature and humidity, and wind speed. Several near-ground measurement points are evenly arranged in the scene to simultaneously measure the air temperature and humidity, wind speed, and black globe temperature at pedestrian height.

3. The method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology according to claim 2, characterized in that: In S4, the following steps are specifically included: S41: Based on the reflectance of the near-infrared and red bands, the Normalized Difference Vegetation Index (NDVI) is obtained through band calculation to characterize the surface vegetation coverage and growth in multiple scenarios; S42: To obtain the albedo within the multispectral camera response band (400-900nm), the surface reflectance of the five narrow channels (blue, green, red, red-edge, and near-infrared bands) is weighted averaged according to the spectral response function of the multispectral camera. The surface albedo within the range of 400-900nm is obtained through band calculation to characterize the shortwave radiation reflection characteristics of the surface in multiple scenes. S43: Based on the DSM imagery output by photogrammetry software, a sky view factor (SVF) is used to quantify the surface occlusion conditions in multiple scenes. A pixel in the DSM represents the absolute surface height of the corresponding feature. A search is performed in multiple directions within a hemisphere of a self-defined radius, centered on the DSM pixel. The proportion of light reaching the sky is quantified to determine the SVF of each pixel. Where θ svf is the zenith angle measured vertically upward; φ svf is the azimuth measured along the ground plane.

4. The method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology according to claim 3 is characterized by: S5 specifically includes the following steps: S51: Based on the wide-band emissivity inversion model, the NDVI image is selected for band calculation to obtain the emissivity (ε 8～14 ), characterizes the surface thermal radiation release capacity in multiple scenarios; Where, ε 8～14 is the surface broad-band emissivity, and NDVI is the surface normalized difference vegetation index; S52: For the multi-scene thermal infrared images captured, they are also imported into the photogrammetry software. Through image alignment, camera distortion correction, control point calibration, high-density point cloud construction, DSM generation and orthophoto reconstruction, grayscale orthophotos representing the surface thermal radiation characteristics of the multi-scene are generated. S53: Based on radiation calibration, the grayscale orthophoto representing the surface thermal radiation characteristics is converted into the surface brightness temperature (T b ) orthophoto, using the equivalent wavelength of the thermal imager response band and the Planck function to perform band calculations, the surface brightness temperature orthophoto is converted into the radiation brightness of the surface received by the thermal imager at the equivalent wavelength orthophotos; Where DN is the grayscale value that characterizes the surface thermal radiation characteristics of multiple scenes, T b is the surface brightness temperature; S54: Building a Lookup Table Loop Based on an Equation ε 8～14 , SVF image, use the least squares method to find the minimum difference between the theoretical value in the lookup table and the image measured value, record the serial number of the corresponding working condition in the table, and extract the corresponding unknown parameter (T s_URTE ), as the inversion result of the corresponding pixel; Where, Characterizes the radiance of the ground surface received by the thermal imager at its equivalent wavelength, SVF represents the sky viewing angle factor, ε 8～14 represents the surface broadband emissivity, τ 8～14 They represent the low-altitude atmospheric upward radiation, atmospheric downward radiation and transmittance respectively. S55: Circulate pixel by pixel and perform quantitative inversion to finally obtain the orthophoto of the surface inverted temperature, which represents the surface thermal conditions in multiple scenarios.

5. The method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology according to claim 4 is characterized in that: S6 specifically includes the following steps: S61: Based on multispectral images from different channels acquired by drones, batch read image metadata, including location, time, and camera parameters. Use these parameters to correct distortion for each lens, ensuring that the coordinates of pixels in the image accurately reflect their actual locations. S62: Perform radiation calibration, batch convert the grayscale value of each channel of the image into the corresponding shortwave radiation brightness value (L sw_3D_surface ); S63: Based on the assumption that the reflection of an ideal diffuse surface is isotropic, the radiance value of each channel is multiplied by π to obtain the shortwave radiation value of the lower hemisphere integral of the channel. The radiation values of the five channels are summed up to obtain the shortwave radiation value (E) in the range of 400-900 nm on the urban three-dimensional space surface. sw_3D_surface ), using machine learning to establish a model between shortwave radiation values and the corresponding five channel grayscale values: AND sw_3D_surface =π*L sw_3D_surface Where, E sw_3D_surface Represents the shortwave radiation value from the three-dimensional surface of the city; L sw_3D_surface It represents the shortwave radiation brightness of the urban three-dimensional space surface obtained from a specific observation angle; S64: Extract pixel data to form an input table, randomly select 70% of the pixels as a training data set, and the remaining 30% as a validation data set. After the photogrammetry process of the multispectral data in S4, generate and output 3D point cloud data and a 3D mesh model representing the grayscale values of the five shortwave channels of the entire scene; S65: Based on the model between the shortwave radiation value established in S63 and the corresponding five channel grayscale values, generate three-dimensional point cloud data with precise coordinate information and ideal diffuse surface shortwave radiation information. Using the principles of spatial geometric operations, assign the point cloud data representing the ideal diffuse surface shortwave radiation to the triangular faces on the grid model, and then establish a multi-scene three-dimensional ideal diffuse surface shortwave radiation field.

6. The method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology according to claim 5, characterized in that: In S7, the following steps are specifically included: S71: After the photogrammetry process of the thermal infrared data in S5, the three-dimensional surface brightness temperature (T b ), constructing a lookup table to convert the brightness temperature three-dimensional point cloud data into the integrated radiation brightness three-dimensional point cloud data; S72: T b The independent variable is regarded as the lookup table, and the range and step size of the independent variable are determined by T b The measured range and B 8～14 (T b ) sensitivity; the dependent variable of the lookup table is the theoretical value of the radiant brightness after the integration of the thermal imager's spectral response band (B' 8～14 (T b )); S73: After the lookup table is established, the brightness temperature in the table is compared with the measured three-dimensional surface T b Compare the point cloud data, follow the principle of minimum difference, record the serial number in the table and the corresponding B' 8～14 (T b ), as the radiant brightness after the integration of the thermal imager spectral response band, and then construct a 8～14 (T b ) Point cloud data of information; S74: Follow the principle of spatial geometric operation and convert B 8～14 (T b ) Assign the point cloud data to the triangular surface on the mesh model to establish an ideal diffuse surface B representing multiple scenes 8～14 (T b ) a three-dimensional model of information; S75: Based on a 3D mesh model, a ray tracing method is used to calculate the SVF of each triangle. The broadband emissivity of each surface triangle is assigned based on the inverted surface emissivity and DSM, following the principles of spatial geometric operations. The emissivity of non-surface triangles is assigned based on the point cloud category attributes and the empirical emissivity values of the corresponding materials. S76: Combined with the low-altitude atmospheric parameters, the input parameters for inverting the temperature of each triangle surface based on the urban radiation transfer equation have all been obtained (B 8～14 (T b ), SVF, ε 8～14 、 and τ 8～14 ), based on the Boltzmann formula, the inverted temperature of the triangular surface is converted into a long-wave radiation field, and then the long-wave radiation field of a three-dimensional ideal diffuse surface in multiple scenes is established.

7. The method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology according to claim 6, characterized in that: In S8, the following steps are specifically included: S81: Extract the reference points for pedestrian thermal comfort calculation. Based on the three-dimensional point cloud data generated by S4 representing the grayscale values of the five shortwave channels of the entire scene, extract the ground point cloud by identifying the elevation differences of the point cloud, and construct an independent ground model. According to the scope of the observation scene and the degree of heterogeneity, the ground model is divided into areas of approximately 1.0m 2 The three-dimensional coordinates of the centroid of each triangle are extracted, and its Z coordinate is raised by 1.1m to represent the height of the center of gravity of the human body. The corrected coordinates are the average radiation temperature sampling value of pedestrians in outdoor space (T mrt_sampled ) The position basis for calculation; S82: Based on the established multi-scenario 3D ideal diffuse surface long- and short-wave radiation fields, the indexed view sphere (IVS) method is used to quantify the long- and short-wave radiation from different surface elements reaching the human body surface. S83: For the long- and short-wave radiation received by the human body from the sky, during the period when the drone collects data, the weather station simultaneously observes and records the average value of the scene's total horizontal irradiance GHI and normal direct irradiance DNI. Under the assumption that the sky diffuse reflection is isotropic, the direct-scatter separation method is used to obtain the sky diffuse reflection horizontal irradiance DHI during this period: DHI=GHI-DNI·sinθ Where θ is the solar altitude angle; Longwave radiation from the sky sky Estimation of the average air temperature and humidity based on the human-environment heat exchange model and simultaneous near-ground measurements within the scene: Where, L sky represents the long-wave radiation from the sky; T a is the average air temperature measured near the ground in the scene; water vapor pressure v p It is calculated based on the average value of the air temperature and humidity measured near the ground in the scene; N represents the degree of cloud cover in the sky, with a value of (0-8), where 0 represents a completely clear sky and 8 represents a cloudy sky completely covered by clouds; σ is the Boltzmann coefficient, with a value of 5.67×10 -8 W·m -2 ·K -4 ; The normal direct irradiance DNI received by a person is quantified by ray tracing. The DNI reaching the human body surface depends on the human body surface projection coefficient (f p ), f p According to the human standing model, it is deduced: f p =3.67·10 -7 i 3 -6.74·10 -5 i 2 +8.49·10 -4 θ+0.297 If the absorption coefficient of the human body surface for long-wave and short-wave radiation is the same, the average radiation temperature near the ground at different locations in multiple scenes is calculated based on the sampling results of each radiation component: Where, T mrt_sampled represents the average radiation temperature of pedestrians calculated based on the sampling values of each radiation component; α sw and α lw They refer to the absorption coefficients of the human body surface wearing clothes to short-wave radiation and long-wave radiation respectively; δ shadow Used to indicate whether the human body surface receives direct radiation, δ shadow When it is 0, it means the human body is in the shadow, δ shadow If it is 1, it means that the human body has received direct radiation; i represents the number of the triangle face hit by the radiation emitted by the human body; E sw_sky and E lw_sky They represent the shortwave and longwave radiation received by the human body from the sky; E sw_3D_surface_i and E lw_3D_surface__i They respectively refer to the amount of short-wave radiation and long-wave radiation received by the human body from the i-th triangle surface.

8. The method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology according to claim 7 is characterized in that: In S9, using T mrt_sampled and simultaneously measured near-surface meteorological parameters, including air temperature, relative humidity, and wind speed; Based on the human thermal comfort model, high-spatial-resolution pedestrian thermal comfort indices, including SET*, PET, and UTCI, are calculated and obtained. The pedestrian thermal comfort index sampling values are compared with the pedestrian thermal comfort index calculated based on near-ground measured meteorological parameters.

9. The method for detecting, evaluating, and improving thermal comfort of pedestrians in outdoor spaces based on UAV low-altitude remote sensing technology according to claim 8, characterized in that: In S10, a random forest model is used to quantify the characteristic parameters of each triangle in the ground model and T mrt_sampled The nonlinear relationship between the sampling values of typical OTC indicators is analyzed. SHAP and PDPbox analysis methods are used to quantitatively evaluate the sensitivity of near-ground human thermal comfort to various ground characteristic parameters. Before model training, the data is preprocessed, which specifically includes the following steps: S101: Geo-register and resample the orthophoto of surface feature parameters and DSM, extract pixel information, and obtain point cloud data that accurately represents the ground feature parameters and their three-dimensional coordinates; S102: Based on the established multi-scenario ground model and following the principles of spatial geometric operations, a ground model representing surface characteristic parameters is constructed; S103: Extract T of triangular faces in non-building coverage areas mrt_sampled The sampling values of typical OTC indicators and their corresponding ground feature parameters are used as effective inputs for model training. The input data are normalized. 70% and 30% of the input data are randomly selected for model training and validation. The accuracy is verified based on the coefficient of determination and root mean square error of the validation data set. S104: Based on the sensitivity analysis results of the training model, clarify the design elements and corresponding quantitative indicators for improving pedestrian thermal comfort.

Images