A highway slope photovoltaic glare evaluation method and device based on a unmanned aerial vehicle

Abstract

The application relates to a highway slope photovoltaic glare evaluation method and device based on a UAV. The method comprises the following steps: controlling the UAV to fly along a flight path of a pre-planned observation point of a highway covering a photovoltaic module area and simulating a driver's visual angle, synchronously collecting multi-source data including images, reflection spectra, illumination intensity, a sun position and three-dimensional coordinate point clouds to form a comprehensive data set; automatically extracting key parameters of the photovoltaic module, such as an orientation, an inclination, surface cleanliness and reflection spectral characteristics, based on the images, the spectra and the point cloud data in the data set; further combining the sun position information and the illumination data of the observation point of the highway, and using a physical optical model to quantitatively calculate the glare intensity, the duration and the influence range of each observation point; finally, comparing the calculation results with preset threshold values, outputting intuitive glare grade evaluation results and spatial distribution information, and realizing accurate evaluation of the highway slope photovoltaic glare, so as to provide a scientific basis for ensuring driving safety and optimizing the photovoltaic system.

Description

Technical Field

[0001] This invention belongs to the field of highway traffic safety assessment, and in particular relates to a method and device for assessing photovoltaic glare on highway slopes based on unmanned aerial vehicles (UAVs). Background Technology

[0002] With the rapid development of the photovoltaic industry, deploying photovoltaic modules on idle land resources such as highway slopes has become an important way to improve energy efficiency. However, the intense glare generated by the surface of photovoltaic modules under sunlight can severely interfere with drivers' visual perception and judgment when projected onto the driving area of ​​the highway, significantly increasing the risk of traffic accidents and posing a real threat to road traffic safety. Therefore, a scientific, accurate, and comprehensive assessment of the glare generated by photovoltaic systems on highway slopes is an indispensable technical link to ensure the safe operation of highways.

[0003] Currently, the industry primarily relies on manual ground-based measurement methods for glare assessment. This involves technicians carrying equipment such as illuminometers to the site to collect data at a limited number of pre-set points. This method faces several challenges in practical application: First, the complex and rugged terrain along highway slopes makes access difficult, severely limiting the measurement range and preventing systematic coverage of the entire photovoltaic array and the entire highway section. Second, manual point-by-point measurement is inefficient, time-consuming, and labor-intensive, failing to meet the needs of rapid assessment for large-scale, long-distance highway projects. Third, measurement results are highly susceptible to variations in operator skill, equipment placement accuracy, and instantaneous environmental changes, making data consistency and reliability difficult to guarantee. While drone technology, with its advantages of maneuverability and a bird's-eye view, has been applied in many inspection fields, a mature technical solution that truly addresses the pain points of highway photovoltaic glare assessment has yet to be developed.

[0004] Existing attempts often only use drones for simple photography or video recording, lacking the ability to acquire multi-sensor synchronous data, obtain three-dimensional spatial information, and perform quantitative analysis based on physical models. This makes it impossible to accurately analyze the generation mechanism, propagation path, and dynamic impact on the driver's vision of glare from a fundamental perspective. Consequently, the assessment results remain one-sided and crude, failing to provide sufficiently accurate and comprehensive basis for safety decisions and engineering optimization.

[0005] Therefore, how to deeply integrate UAV platforms, multi-source sensing technologies, and advanced optical computing models to develop a set of efficient, accurate, and comprehensive glare analysis methods has become a technical challenge that urgently needs to be overcome in this field. Summary of the Invention

[0006] Therefore, it is necessary to provide a method and device for assessing photovoltaic glare on highway slopes based on unmanned aerial vehicles (UAVs) to address the aforementioned technical problems.

[0007] Firstly, this application provides a method for assessing photovoltaic glare on highway slopes based on unmanned aerial vehicles (UAVs), including:

[0008] S1. Control the drone to fly along the preset route, and use the multiple sensors on the drone to simultaneously collect data from the photovoltaic module area and the highway observation point area, generating a comprehensive dataset including image data, reflectance spectrum data, light intensity data, solar position information, and three-dimensional coordinate point cloud data; the preset route includes a detailed survey route covering the photovoltaic module area and a highway viewpoint observation route simulating the driver's perspective;

[0009] S2. Based on the image data, reflectance spectral data, and three-dimensional coordinate point cloud data in the comprehensive acquisition dataset, extract the set of key parameters for the photovoltaic module; wherein, the set of key parameters includes at least the orientation, tilt angle, surface cleanliness level, and reflectance spectral characteristic parameters of the photovoltaic module;

[0010] S3. Based on the key parameter set, the solar position information in the comprehensive data collection dataset, and the light intensity data corresponding to the highway observation point, calculate the glare assessment index for each highway observation point through a physical optics model; among which, the glare assessment index includes glare intensity, glare duration, and glare impact range;

[0011] S4. Compare the glare assessment index with the preset glare level classification threshold, and output the glare level assessment result of the highway slope photovoltaic glare and the corresponding spatial distribution information.

[0012] Secondly, this application also provides a drone-based photovoltaic glare assessment device for highway slopes, used to implement the method described in the first aspect, the device comprising:

[0013] The aerial survey data acquisition module is used to control the UAV to fly along a preset route. It uses multiple sensors on the UAV to simultaneously collect data from the photovoltaic module area and the highway observation point area, generating a comprehensive dataset including image data, reflectance spectrum data, light intensity data, solar position information, and three-dimensional coordinate point cloud data. The preset route includes a detailed survey route covering the photovoltaic module area and a highway viewpoint observation route simulating the driver's perspective.

[0014] The photovoltaic characteristic analysis module is used to extract the set of key parameters of photovoltaic modules based on image data, reflectance spectrum data and three-dimensional coordinate point cloud data in the comprehensive acquisition dataset; the set of key parameters includes at least the orientation, tilt angle, surface cleanliness level and reflectance spectrum characteristic parameters of the photovoltaic modules;

[0015] The glare physical modeling module is used to calculate the glare assessment index for each highway observation point based on the key parameter set, the solar position information in the comprehensive data collection dataset, and the light intensity data corresponding to the highway observation point, through a physical optics model. The glare assessment index includes glare intensity, glare duration, and glare impact range.

[0016] The glare classification output module is used to compare the glare evaluation index with the preset glare level classification threshold, and output the glare level evaluation result of the photovoltaic glare on the highway slope and the corresponding spatial distribution information.

[0017] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a method for evaluating photovoltaic glare on highway slopes based on unmanned aerial vehicles as described in the first aspect.

[0018] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for evaluating photovoltaic glare on highway slopes based on unmanned aerial vehicles as described in the first aspect.

[0019] The aforementioned method and device for assessing photovoltaic glare on highway slopes based on unmanned aerial vehicles (UAVs) involves controlling a UAV to fly along a pre-planned route covering the photovoltaic module area and simulating a driver's perspective at highway observation points. Simultaneously, it collects multi-source data, including images, reflectance spectra, illuminance, solar position, and 3D coordinate point clouds, to form a comprehensive dataset. Based on the image, spectral, and point cloud data in this dataset, key parameters such as the orientation, tilt angle, surface cleanliness, and reflectance spectral characteristics of the photovoltaic modules are automatically extracted. Then, combining solar position information and illuminance data from the highway observation points, a physical optics model is used to quantitatively calculate the glare intensity, duration, and impact range at each observation point. Finally, the calculation results are compared with preset thresholds to output intuitive glare level assessment results and spatial distribution information. This achieves efficient, comprehensive, and automated accurate assessment of photovoltaic glare on highway slopes, providing a scientific basis for ensuring driving safety and optimizing photovoltaic systems. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart illustrating a method for evaluating photovoltaic glare on highway slopes based on unmanned aerial vehicles (UAVs) provided by this invention.

[0022] Figure 2 This is a schematic diagram of the process for extracting the key parameter set of a photovoltaic module in an optional embodiment of the present invention;

[0023] Figure 3 This invention provides a schematic diagram of a drone-based photovoltaic glare assessment device for highway slopes. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0025] refer to Figure 1 The document presents a flowchart illustrating a method for assessing photovoltaic glare on highway slopes based on unmanned aerial vehicles (UAVs), which includes the following steps:

[0026] S1. Control the drone to fly along the preset route, and use the multiple sensors carried by the drone to simultaneously collect data from the photovoltaic module area and the highway observation point area, and generate a comprehensive collection dataset including image data, reflectance spectrum data, light intensity data, solar position information and three-dimensional coordinate point cloud data; among them, the preset route includes a detailed survey route covering the photovoltaic module area and a highway viewpoint observation route simulating the driver's perspective.

[0027] Specifically, the first step is to plan pre-set flight routes. An industrial-grade drone equipped with ground station software is used as the flight platform. Before planning, two types of basic geographic data need to be imported: one is a vector map of the distribution of photovoltaic modules on highway slopes exported from a geographic information system, including the geographic coordinates of the module boundaries and digital elevation model data of the installation area; the other is a highway route design map, including information such as the highway centerline, lane distribution, and station range. The detailed survey flight route covering the photovoltaic module area adopts a grid-like flight mode. The flight altitude is determined based on the maximum elevation difference within the photovoltaic module installation area to ensure that the high-definition camera can clearly capture module details. The flight speed and overlap rate settings must avoid data omissions in the module area. The highway viewpoint observation flight route simulating the driver's perspective adopts a parallel flight mode along the highway. The flight altitude strictly matches the driver's eye level, the flight route maintains a reasonable horizontal distance from the highway centerline, and the flight direction is consistent with the direction of traffic flow. Observation points are set at fixed intervals to ensure that the collected data accurately reflects the driver's visual experience during driving.

[0028] After the flight path is planned, the UAV is launched and flight commands are sent through the ground station, simultaneously activating the multi-sensor synchronous data acquisition function. The multi-sensor module includes a high-definition camera, spectrometer, illuminance sensor, lidar, and a GPS and inertial measurement unit combined navigation module. Each sensor achieves synchronized data acquisition through a synchronization controller, specifically through a pulse triggering mechanism, ensuring that all sensors record data at the same timestamp. The high-definition camera is equipped with a suitable lens, set with appropriate resolution and frame rate, and stores image data in its original format. The shooting mode is set to automatic exposure, with exposure time dynamically adjusted according to ambient light intensity to ensure neither overexposure nor underexposure. The spectrometer has a suitable spectral detection range and resolution, and the integration time is determined through pre-acquisition testing to avoid spectral signal saturation or excessively low signal-to-noise ratio. During acquisition, reflectance spectral data is recorded at a fixed frequency, and a standard white board is included for subsequent spectral calibration. The illuminance sensor has a suitable range and accuracy, collecting the illuminance intensity of reflected light from the photovoltaic modules and the ambient background illuminance at a fixed sampling frequency to facilitate subsequent background interference removal. The lidar has a suitable ranging range and accuracy, continuously collecting 3D coordinate point cloud data during flight to reconstruct the spatial relationship between the photovoltaic modules and the highway observation point. The GPS and inertial measurement unit combined navigation module adopts a high-precision positioning mode, synchronously recording the UAV's 3D coordinates, attitude angles, and velocity information at a fixed sampling frequency, providing a benchmark for the spatial correlation of various data.

[0029] The sun's position information is obtained through astronomical algorithms. Input parameters include the average latitude and longitude of the evaluation area, the data collection time, altitude, and atmospheric pressure. The algorithm's specific calculation process involves first calculating the sun's ecliptic coordinates using the Julian day, then converting it to equatorial coordinates, and finally correcting for the sun's horizontal coordinates by combining local geographical longitude and time zone. The final output is the sun's altitude angle and azimuth angle. The formula for calculating the sun's position is as follows:

[0030]

[0031]

[0032] in, Indicates the solar altitude angle. Indicates the solar declination angle. Indicates the latitude of the assessment area. Indicates solar hour angle, This indicates the solar azimuth angle. The solar declination angle is determined by the time of data collection and reflects the latitude of the subsolar point. The solar hour angle is calculated based on local solar time, with noon as zero, negative values ​​in the morning, and positive values ​​in the afternoon. The latitude of the evaluation area is the average of the geographical latitudes of the area, used to characterize the north-south location information of the area.

[0033] The generation of the comprehensive data acquisition dataset requires the standardization and association of various sensor data formats. Image data is saved in an adapted format with accompanying exchangeable image file format information, including the capture timestamp and GPS coordinates. Reflectance spectral data is saved in a structured format, including wavelength, spectral intensity, and timestamp. Illumination intensity data is saved in text format, including timestamp, reflected light intensity, and background light intensity. 3D coordinate point cloud data is saved in an adapted format, including the 3D coordinates of the points, timestamp, and reflection intensity. Solar position information is saved in a structured format, including timestamp, elevation angle, and azimuth angle. All data are associated one-to-one with GPS coordinates through timestamps to form a structured comprehensive data acquisition dataset, which is stored on the UAV's local storage medium and simultaneously transmitted back to the ground control center for backup in real time via a wireless transmission module.

[0034] S2. Based on the image data, reflectance spectral data, and three-dimensional coordinate point cloud data in the comprehensive acquisition dataset, extract the set of key parameters for the photovoltaic module; wherein, the set of key parameters includes at least the orientation, tilt angle, surface cleanliness level, and reflectance spectral characteristic parameters of the photovoltaic module.

[0035] Specifically, for the orientation and tilt angle of photovoltaic modules, the three-dimensional coordinate point cloud data is first preprocessed. A statistical filtering algorithm is used to remove outliers. By setting a threshold for the number of neighboring points and the standard deviation, discrete points deviating from the normal range are identified and removed. Then, a radius filtering algorithm is used to remove sparse noise points. By setting a search radius and a threshold for the number of points in the neighborhood, noise interference points are removed. Subsequently, an iterative nearest-point algorithm is used to register multiple frames of point cloud data to the same geographic coordinate system to ensure spatial consistency. Based on the preprocessed point cloud data, a random sampling consensus algorithm is used to segment the point cloud clusters corresponding to the photovoltaic modules. This algorithm fits a plane equation by randomly selecting three non-collinear points, calculates the distance from other points to the plane, and sets a distance threshold to match module installation errors. When the proportion of points meeting the distance condition reaches a set percentage, the plane is determined as the surface plane of the photovoltaic module, and then the plane's normal vector is calculated.

[0036] The equation of the plane is expressed as:

[0037]

[0038] In the formula, , , These are the three components of the plane normal vector. For plane constants, , , These are the three-dimensional coordinate components of a point in space. The plane normal vector is... , , It consists of three components, used to characterize the spatial orientation of the photovoltaic module surface.

[0039] In the geodetic coordinate system, the unit vector in the true north direction is the reference vector. The orientation of the component is calculated by the angle between the projection of the normal vector onto the horizontal plane and the true north direction. First, the normal vector is projected onto the horizontal plane to obtain the horizontal projection vector. Then, the orientation angle is obtained using the vector angle calculation formula. If the calculation result is negative, the angle is corrected. Finally, the azimuth angle within the range of 0°-360° is obtained, which increases clockwise with true north as the reference. The tilt angle of the component is the angle between the normal vector and the horizontal plane, calculated by the vector dot product formula, and finally converted to angle units.

[0040] The core formula for calculating orientation and tilt angle is:

[0041]

[0042]

[0043] In the above two formulas, Indicates the azimuth angle of the photovoltaic module. Indicates the tilt angle of the photovoltaic module. , , These are the components of the normal vector to the surface plane of the photovoltaic module. The `arctan2` function is used to calculate the angle between the horizontal projection vector and the due north direction, and the `arccos` function is used to calculate the angle between the normal vector and the horizontal plane.

[0044] To improve accuracy, verification was performed using image data. A target detection algorithm was employed to identify components within the image data, outputting the pixel coordinate bounding boxes of the components. Combining the intrinsic and extrinsic parameters of the high-definition camera, a perspective n-point algorithm was used to calculate the 3D coordinates of the four corner points of the components, verifying the accuracy of the point cloud-fitted plane. If the average distance from the corner points to the plane exceeded a set accuracy threshold, point cloud segmentation and plane fitting were repeated until the accuracy requirements were met. Camera intrinsic parameters, including focal length and principal point coordinates, were obtained through a camera calibration algorithm. Camera extrinsic parameters, including camera position and attitude, were obtained from GPS and inertial measurement unit data and point cloud registration results.

[0045] The surface cleanliness level of photovoltaic modules was determined through the fusion analysis of reflectance spectral data and image data. First, the reflectance spectral data was preprocessed, employing dark current correction. Dark current data was collected under shading conditions and subtracted from the original spectral intensity to eliminate dark current interference. Baseline correction was then performed, using a polynomial fitting algorithm to eliminate spectral baseline drift. Finally, a smoothing filtering algorithm was used to smooth the spectral curve and remove random noise. After preprocessing, the spectral reflectance in the visible-near-infrared band was extracted. This band represents the main concentrated area of ​​solar radiation and is also the key band for glare reflection by photovoltaic modules. The spectral reflectance was calculated as the ratio of the spectral intensity of the photovoltaic module to that of a standard white board.

[0046] The formula for calculating spectral reflectance is:

[0047]

[0048] In the formula, Indicates wavelength Spectral reflectance at that location Indicates the wavelength of photovoltaic modules Spectral intensity at that location Indicates the standard whiteboard at wavelength Spectral intensity at that location This indicates the spectral wavelength. A standard whiteboard is used to provide a reference reflectance intensity; its reflectance is known and stable. The ratio of the two can eliminate the influence of external factors such as ambient light on the spectral intensity.

[0049] The spectral reflectance curve of a standard clean photovoltaic module was selected as a reference benchmark. This benchmark curve was obtained by collecting spectral data from the same model of clean, uncontaminated modules in a laboratory environment, ensuring that the surface was free of any contaminants. The correlation coefficient between the actual collected spectrum and the benchmark curve was calculated; the closer the correlation coefficient was to 1, the cleaner the module surface. Simultaneously, the grayscale characteristics of image data were used to aid in the judgment. The image data was grayscaled, and the mean grayscale value and texture entropy value of the module area were calculated. The mean grayscale value reflects the overall brightness of the module surface, while the larger the entropy value, the more complex the surface texture and the more severe the contamination. A weighted evaluation system was established, incorporating the correlation coefficient, mean grayscale value, and entropy value. Cleanliness levels were classified based on the weighted calculation results, with each level corresponding to a clear quantitative indicator to ensure the objectivity of the assessment. The core of the weighted evaluation system is to achieve a quantitative assessment of the surface cleanliness of photovoltaic modules by clearly defining the weight allocation of indicators, normalizing the indicators, calculating the comprehensive score, and classifying the grade thresholds. The specific setting method is as follows: First, determine the weights of the three evaluation indicators (correlation coefficient, gray mean, and entropy value). Combining the core needs of cleanliness assessment, construct a judgment matrix using the analytic hierarchy process (AHP). Through expert scoring and consistency checks, determine the weight ratio of each indicator (among which, the spectral correlation coefficient has the highest weight because it directly reflects the fit between the reflective characteristics of the module surface and the cleanliness benchmark; the gray mean has the second highest weight, assisting in characterizing the brightness attenuation caused by surface contamination; and the entropy value has a relatively low weight, used to correct the assessment bias caused by the complexity of surface texture). The sum of the weights is 1 to ensure that the weight allocation is scientific and reasonable. Second, normalize the three indicators to eliminate dimensional differences: normalize the spectral correlation coefficient and gray mean in the forward direction (the larger the value, the higher the cleanliness), and normalize the entropy value in the reverse direction (the smaller the value, the higher the cleanliness), so that the three indicators are all within the same numerical range after normalization. Subsequently, a weighted comprehensive scoring model was constructed. The normalized values ​​of each indicator were multiplied by their corresponding weights, and the summation yielded the comprehensive cleanliness evaluation score for each component. Finally, based on sample data of components with different levels of contamination calibrated in the laboratory, the comprehensive score ranges corresponding to each contamination level were statistically analyzed. The comprehensive scores were then matched one by one with the five cleanliness levels (extremely clean, clean, slightly contaminated, moderately contaminated, and heavily contaminated) to clarify the score threshold range for each level, forming a quantifiable and repeatable standard for level determination, ensuring the objectivity and consistency of cleanliness assessment.

[0050] The extraction of reflectance spectral characteristic parameters needs to focus on the core features that reflect the reflectance characteristics of the component. First, feature points are identified in the preprocessed reflectance spectral curve. A peak-valley detection algorithm is used to extract the characteristic peak wavelengths and valley wavelengths in the spectral curve, and the reflectance corresponding to the peak and valley is calculated. The half-width at half-maximum (WHM) of the peak is calculated using the formula to characterize the sharpness of the spectral peak. Subsequently, the integral reflectance in the visible-near-infrared band is calculated. The area enclosed by the spectral curve and the wavelength axis is calculated using an integral algorithm to reflect the overall reflectance capability of the component in this band. To simplify the feature dimension, principal component analysis is used to reduce the dimensionality of the extracted original feature parameters. The covariance matrix is ​​calculated and the eigenvalues ​​and eigenvectors are solved. The top few principal components that meet the cumulative contribution rate requirements are selected as the core parameters of the reflectance spectral characteristics. These parameters reflect core information such as the overall reflectance intensity, the positional characteristics of the spectral peaks, and the shape characteristics of the spectral peaks, and are directly used for the construction of the subsequent physical optics model.

[0051] S3. Based on the key parameter set, the solar position information in the comprehensive data collection dataset, and the light intensity data corresponding to the highway observation point, calculate the glare assessment index for each highway observation point through a physical optics model; among which, the glare assessment index includes glare intensity, glare duration, and glare impact range.

[0052] Specifically, the form of the physical optics model is first determined. Considering that the surface of a photovoltaic module exhibits both diffuse and specular reflection characteristics, an improved bidirectional reflection distribution function is adopted as the core model, and its expression is as follows:

[0053]

[0054] In the formula, This indicates the reflected radiance of the photovoltaic module in the direction of observation. This represents the observation angle between the observation direction and the normal vector of the component surface. Indicates the observed azimuth angle. Indicates the diffuse reflectance coefficient. Indicates the specular reflection coefficient. Indicates the highlight index. This represents the actual solar incident irradiance on the surface of the photovoltaic module. This represents the angle of incidence between sunlight and the normal vector on the surface of the photovoltaic module. Represents the surface normal vector of the component. The term represents a half-vector. The diffuse reflectance coefficient, specular reflectance coefficient, and specular index are all obtained by fitting the reflectance spectral characteristic parameters through a multiple linear regression model. The regression coefficients are calibrated using laboratory reflectance test data of similar components. The half-vector is the average unit vector between the direction of sunlight and the direction of the viewing line, used to characterize the specular reflectance direction.

[0055] The actual solar incident irradiance on the surface of the photovoltaic module is obtained by combining the ambient solar irradiance from the comprehensive dataset with atmospheric transmittance correction. The atmospheric transmittance is calculated using an atmospheric radiative transfer model, with input parameters including the altitude of the assessment area, atmospheric visibility, and water vapor content. The correction formula is as follows:

[0056]

[0057] In the formula, Indicates ambient solar irradiance. This represents atmospheric transmittance. Atmospheric transmittance reflects the degree of attenuation of solar radiation after it passes through the atmosphere. The higher the value, the less the atmosphere attenuates solar radiation, and the closer the incident irradiance received by the photovoltaic module surface is to the ambient solar irradiance.

[0058] The calculation of glare intensity requires three steps. The first step is to calculate the incident angle and incident azimuth. Based on the sun's position information and the photovoltaic module's orientation and tilt angle, the angle between the sunlight and the normal vector of the module surface is obtained by substituting these values ​​into the incident angle calculation formula. The incident azimuth is calculated by the difference between the sun's azimuth and the module's orientation; if the result is negative, an angle correction is performed. The formula for calculating the incident angle is:

[0059]

[0060] In the formula, Indicates the solar altitude angle. Indicates the tilt angle of the photovoltaic module. Indicates the azimuth angle of the sun. This indicates the azimuth angle of the photovoltaic module. This formula uses trigonometric functions to correlate the sun's position with the spatial orientation of the photovoltaic module, accurately calculating the angle of incidence of sunlight relative to the module's surface.

[0061] The second step is to calculate the observation angle and azimuth. The three-dimensional coordinates of the highway observation point are extracted from the three-dimensional coordinate point cloud data. The three-dimensional coordinates of the center of the photovoltaic module are calculated from the module point cloud clusters after point cloud segmentation. The average value of all point coordinates is taken as the coordinates of the module center. The unit vector of the observation line of sight is calculated by the difference between the coordinates of the observation point and the module center. The unit vector of the module surface normal vector is obtained by normalizing the plane normal vector. The observation angle is calculated by the dot product of the unit vector of the normal vector and the unit vector of the observation line of sight. The observation azimuth is calculated by the difference between the horizontal coordinates of the observation point and the module center.

[0062] The third step is to calculate the half-vector and glare intensity. The unit vector of the sunlight direction is determined based on the solar altitude angle and solar azimuth angle. The half-vector is calculated by averaging the unit vectors of the sunlight direction and the observation line-of-sight direction. Substituting these parameters into the bidirectional reflectance distribution function model yields the reflected radiance. Then, combining the visual weighting coefficient, the effective reflective area of ​​the photovoltaic module, and the straight-line distance from the observation point to the module, the glare intensity is finally obtained using the glare intensity calculation formula. The glare intensity calculation formula is:

[0063]

[0064] In the formula, Indicates glare intensity, Indicates the visual weighting coefficient. Indicates the effective reflective area of ​​the photovoltaic module. This represents the straight-line distance from the observation point to the module. The visual weighting coefficient is calculated based on the human eye's photopic vision characteristic curve, reflecting the human eye's sensitivity to different wavelengths of light. The effective reflective area is the area of ​​the photovoltaic module that actually participates in reflection, excluding non-reflective areas such as the frame. The straight-line distance from the observation point to the module reflects the degree of attenuation during glare propagation; the greater the distance, the smaller the glare intensity.

[0065] The calculation of glare duration is based on the determination of the solar trajectory and glare intensity threshold. First, an astronomical algorithm is used to calculate the solar altitude angle and azimuth angle at fixed intervals within the evaluation period to form solar trajectory data. For each highway observation point, the solar position parameters at different times are substituted into the glare intensity calculation formula to obtain a glare intensity sequence that changes over time. A glare intensity threshold is set, which is determined based on relevant design standards and driving safety test data. The glare intensity sequence is traversed, and the start time when the threshold requirement is first met and the end time when the threshold requirement is last met are recorded. If a start time and an end time exist, the duration of a single occurrence is calculated. If there are multiple discontinuous periods that meet the threshold within a day, the durations of all periods are accumulated to obtain the total glare duration for that observation point.

[0066] Determining the glare impact range requires combining spatial interpolation and highway segment division. First, the highway is discretized using a fixed grid, with grids divided at fixed intervals along and perpendicular to the highway direction, covering all lanes and reasonable areas on both sides. The center point of each grid is used as an evaluation sub-point, and the three-dimensional coordinates of all sub-points are extracted and the corresponding glare intensity is calculated. A spatial interpolation algorithm is then used to spatially interpolate the discrete glare intensity values, generating a glare intensity raster map along the highway. Criteria for determining the glare impact range are set, and regions meeting the criteria are extracted from the raster map based on glare intensity thresholds, forming continuous polygonal vector boundaries. Combined with station information from the highway route design map, the corresponding highway station intervals for the impact range are determined, and the length along the highway and the width perpendicular to the highway are statistically analyzed. Simultaneously, the number of affected lanes is determined based on lane distribution information, ultimately forming glare impact range data that includes spatial boundaries, length, width, and affected lanes.

[0067] S4. Compare the glare assessment index with the preset glare level classification threshold, and output the glare level assessment result of the highway slope photovoltaic glare and the corresponding spatial distribution information.

[0068] Specifically, the first step is to define the preset threshold for glare level classification. This threshold is based on relevant highway traffic safety facility design standards, building glare limitation standards, and real vehicle driving simulation test data. A multi-index weighted judgment method is used to classify the glare level into four levels. The threshold corresponding to each level is determined through statistical analysis. Multiple sets of measured data from different photovoltaic module installation scenarios are selected, and combined with driver visual comfort scores, correlation analysis is used to verify the correlation between glare intensity, glare duration, glare impact range, and comfort scores. Then, a clustering algorithm is used to determine the threshold range for each level to ensure that the level classification can accurately reflect the actual degree of impact.

[0069] The grade assessment results are calculated using a weighted scoring method. First, the three glare assessment indicators for each highway observation point are normalized. The purpose of normalization is to eliminate the influence of different indicator dimensions, ensuring that all indicators are on the same order of magnitude, facilitating weighted calculation. The normalization calculation formula is:

[0070]

[0071] In the formula, This represents the normalized value of a certain indicator. This represents the actual calculated value of the indicator. This represents the maximum possible value of the indicator under extreme conditions. This formula converts glare intensity, glare duration, and glare range into normalized values, ensuring the rationality of the weight allocation for each indicator.

[0072] Then, based on the weight of each indicator's impact on driving safety, a comprehensive score is calculated. The weights are determined using the analytic hierarchy process (AHP), comprehensively considering the degree of visual interference each indicator causes to the driver. Glare intensity has the highest weight, followed by glare duration, while glare range has a relatively lower weight. The formula for calculating the comprehensive score is:

[0073]

[0074] In the formula, This indicates the overall score. The weighting coefficients representing glare intensity The normalized value representing the glare intensity. The weighting coefficients representing the duration of the glare. The normalized value representing the duration of the glare. The weighting coefficients represent the range of glare's influence. This represents the normalized value indicating the range of glare's impact. The sum of the weighting coefficients is 1 to ensure the overall score remains within a reasonable range.

[0075] Finally, the glare level is determined based on the comprehensive score. Different comprehensive score ranges correspond to different glare levels. If a single indicator at a certain observation point reaches the threshold of a higher level, the higher level is used to determine the level, ensuring that no severely affected scenes are overlooked. The output of spatial distribution information adopts a combination of geographic information system visualization and structured data. First, the glare level assessment results are associated with the three-dimensional coordinates of the highway observation points to generate an attribute data table containing station number, three-dimensional coordinates, glare index, and level results. Based on this data table, a glare level distribution map along the highway is drawn in the geographic information system software, using different colors to identify each level and clearly showing the glare impact level of each road segment. At the same time, the spatial boundaries corresponding to each level are extracted to generate a vector-format impact range file, including boundary coordinates, area, affected lanes, and other information. The spatial distribution of glare intensity is output in raster format, with raster values ​​representing glare intensity, supporting subsequent overlay analysis.

[0076] The final output consists of two parts: first, a glare assessment report, which uses a standard patent document format and includes a project overview, data collection details, key parameter extraction results, glare index calculation process, grading basis, and optimization suggestions. The project overview covers information such as the assessment area, photovoltaic module parameters, and data collection time. The data collection details include sensor parameters and flight path information. The optimization suggestions propose appropriate improvement measures for different grading areas. Second, a spatial distribution data package, which includes a glare level vector map, a glare intensity raster map, and an observation point attribute table. This package supports subsequent engineering design software calls, ensuring that the assessment results can directly guide the installation optimization and safety management of photovoltaic modules on highway slopes, achieving a closed loop of assessment-optimization-safety.

[0077] The aforementioned method for assessing photovoltaic glare on highway slopes using unmanned aerial vehicles (UAVs) involves controlling a UAV to fly along a pre-planned route covering the photovoltaic module area and simulating a driver's perspective at highway observation points. Simultaneously, it collects multi-source data, including images, reflectance spectra, illuminance, solar position, and 3D coordinate point clouds, to form a comprehensive dataset. Based on the image, spectral, and point cloud data in this dataset, key parameters such as the orientation, tilt angle, surface cleanliness, and reflectance spectral characteristics of the photovoltaic modules are automatically extracted. Then, combining solar position information and illuminance data from the highway observation points, a physical optics model is used to quantitatively calculate the glare intensity, duration, and impact range at each observation point. Finally, the calculation results are compared with preset thresholds to output intuitive glare level assessment results and spatial distribution information. This achieves efficient, comprehensive, and automated accurate assessment of photovoltaic glare on highway slopes, providing a scientific basis for ensuring driving safety and optimizing photovoltaic systems.

[0078] refer to Figure 2 In one optional embodiment, based on image data, reflectance spectral data, and three-dimensional coordinate point cloud data from a comprehensive acquisition dataset, a set of key parameters for the photovoltaic module is extracted, including the following steps:

[0079] S11. Preprocess the image data to obtain an orthophoto; extract the contour mask of each photovoltaic module by performing semantic segmentation on the orthophoto.

[0080] Specifically, the core of this step is to eliminate interference factors through image preprocessing and then accurately locate the individual photovoltaic module area through semantic segmentation, providing a precise spatial range benchmark for subsequent parameter extraction. The image preprocessing process needs to sequentially complete distortion correction, image registration, and orthorectification to ensure that the processed orthorectified image accurately reflects the spatial position and shape of the photovoltaic module. First, distortion correction is performed. Addressing the inherent radial and tangential distortions of the high-definition camera's optical system, the intrinsic parameter matrix and distortion coefficients obtained from camera calibration are used to correct the original image data through a distortion correction model, eliminating image distortion caused by lens optical characteristics. The camera intrinsic parameter matrix includes core parameters such as focal length and principal point coordinates, and the distortion coefficients include radial and tangential distortion coefficients, both pre-obtained in a laboratory environment using the Zhang Zhengyou calibration method to ensure correction accuracy.

[0081] After distortion correction, image registration is performed. Based on the timestamps and GPS coordinates attached to the image data, feature points are matched among multiple overlapping frames of the same area. The SIFT algorithm is used to extract scale-invariant feature points from each frame, and a FLANN matcher is used to achieve fast feature point matching. Then, a random sampling consensus algorithm is used to remove mismatched points to ensure matching accuracy. Based on the matched feature points, the homography matrix is ​​used to calculate the spatial transformation relationship between the images, unifying the multiple frames of images into the same coordinate system, thus achieving image registration. Finally, orthorectification is performed. Combining the terrain elevation information in the 3D coordinate point cloud data, an inverse distance weighted interpolation algorithm is used to correct the elevation of the registered image, eliminating image distortion caused by terrain undulations, and obtaining an orthorectified image with a viewpoint perpendicular to the horizontal plane. This image can accurately reflect the actual size and spatial distribution of photovoltaic modules, avoiding contour distortion caused by tilted viewing angles.

[0082] After the orthophoto is generated, the contour mask of a single photovoltaic module is extracted through semantic segmentation. The improved U-Net semantic segmentation network is selected as the core algorithm. This network extracts multi-scale features of the image through the encoder and realizes feature upsampling and spatial localization through the decoder, which can accurately segment the photovoltaic module from the background (slope terrain, vegetation, road, etc.). The semantic segmentation process includes three stages: training dataset construction, network training, and inference. The training dataset uses orthophotos of photovoltaic (PV) modules on similar highway slopes. Labeled images are generated by manually annotating PV module regions, with PV module regions labeled as target categories and background regions as non-target categories. The training dataset is divided into training and validation sets proportionally and input into an improved U-Net network for training. The difference between the predicted results and the labeled images is measured using a cross-entropy loss function. The network parameters are adjusted using an adaptive moment estimation optimizer until the segmentation accuracy of the validation set reaches a preset threshold, completing network training. The preprocessed orthophotos are input into the trained semantic segmentation network for inference. In the semantic segmentation results output by the network, the binarized image of the target category region serves as the contour mask of the PV module. Regions with a pixel value of 1 correspond to PV modules, and regions with a pixel value of 0 correspond to the background. Each continuous target category region corresponds to the contour of a single PV module, achieving accurate separation of individual modules.

[0083] S12. Based on the three-dimensional coordinate point cloud data corresponding to the contour mask, fit the surface plane equation of the photovoltaic module using the least squares method, and obtain the surface normal vector of the photovoltaic module based on the surface plane equation; calculate the orientation and tilt angle of the photovoltaic module based on the surface normal vector.

[0084] Specifically, this step achieves accurate fitting of the photovoltaic module's surface plane by spatially associating the point cloud with the contour mask and using the least squares method. Then, the orientation and tilt angle of the module are derived based on the plane's normal vector. All calculations are based on rigorous mathematical models and spatial geometric relationships. First, spatial association is performed between the contour mask and the 3D coordinate point cloud data. Utilizing the timestamp synchronization characteristics of image data and point cloud data, as well as the extrinsic parameters of the high-definition camera (including camera position and attitude), the contour mask in the orthophoto is projected onto the 3D point cloud coordinate system through perspective projection transformation. Point cloud data within the mask's projection range are then selected, forming the 3D coordinate point cloud subset corresponding to a single photovoltaic module. This process accurately removes interference from background point clouds, ensuring the purity of the fitted data.

[0085] Based on the selected point cloud subset, the least squares method is used to fit the surface plane equation of the photovoltaic module. The core idea of ​​the least squares method is to minimize the sum of squared distances from all points in the point cloud subset to the fitting plane, thus achieving optimal plane fitting. Let the surface plane equation of the photovoltaic module be: In the formula , , These are the three components of the plane normal vector. For plane constants, , , Let be the three-dimensional coordinate components of any point in the point cloud subset. The point cloud subset contains... The point, the first The coordinates of the points are The distance from the point to the fitting plane The calculation formula is: The goal of the least squares method is to minimize the sum of squared distances between all points. Through the , , , By taking the partial derivatives and setting them to zero, a system of linear equations is constructed, and solving it yields the optimal parameters of the plane equations. , , , To avoid trivial solutions, constraints must be imposed. This ensures the uniqueness of the solution.

[0086] Once the plane equation is determined, the surface normal vector of the photovoltaic module can be obtained. This vector, perpendicular to the surface of the photovoltaic module, is the core parameter characterizing the spatial attitude of the module. When calculating the orientation and tilt angle of the photovoltaic module based on the surface normal vector, the geodetic coordinate system must be used as the reference (north is the azimuth angle of 0°, increasing clockwise; the horizontal plane is the tilt angle of 0°, with upward being positive).

[0087] tilt angle Let the surface normal vector and the horizontal plane normal vector (the vertically upward vector in the geodetic coordinate system, denoted as ) be the vectors of the surface and the horizontal plane. The angle between the component surface and the horizontal plane reflects the degree of inclination of the component surface relative to the horizontal plane. The calculation formula is: In the formula The dot product of the surface normal vector and the horizontal plane normal vector is calculated as follows: Therefore, the formula for the tilt angle can be simplified to: The angle range is from 0° to 90°, where 0° indicates that the component surface is horizontal and 90° indicates that the component surface is vertical.

[0088] Orientation Azimuth Let the angle between the projection of the surface normal vector onto the horizontal plane and the due north direction be the angle between the surface normal vector and the due north direction. Projecting onto the horizontal plane yields the horizontal projection vector. The unit vector for true north in the geodetic coordinate system is: The azimuth angle is obtained by calculating the angle between the horizontal projection vector and the true north direction vector. The calculation formula is as follows: The output range of the arctan2 function in the formula is from -180° to 180°, which is a unified azimuth angle range of 0° to 360°. When the calculation result is negative, 360° needs to be added for correction. Finally, the orientation azimuth angle with due north as the reference and increasing clockwise is obtained, which accurately represents the orientation of the photovoltaic module.

[0089] S13. Extract the average reflectance spectral curve of the photovoltaic module region from the spatially matched reflectance spectral data corresponding to the contour mask.

[0090] Specifically, this step achieves precise correlation between reflectance spectral data and photovoltaic module areas through spatial matching, and then obtains a spectral curve reflecting the overall reflectance characteristics of the module through averaging processing, providing reliable data for subsequent cleanliness assessment and extraction of reflectance spectral characteristic parameters. First, spatial matching of reflectance spectral data and contour masks is performed. Based on the synchronization characteristics of the integrated acquisition dataset, all sensor data have a unified timestamp. Combining the 3D coordinate point cloud data acquired by the lidar and the extrinsic parameters of the high-definition camera, a sensor spatial position and attitude model at the time of reflectance spectral data acquisition is established. The field of view for the reflectance spectral data acquisition is a known parameter. Using this field of view and the sensor's spatial position and attitude, the 3D spatial detection range (conical region) corresponding to each reflectance spectral data can be calculated.

[0091] The spatial intersection of the conical detection range with the corresponding 3D point cloud subset of the contour mask is determined. Reflectance spectral data whose detection range is completely contained within the 3D point cloud subset of the photovoltaic module are selected as the spatially matched reflection spectral data for that photovoltaic module. This spatial matching process effectively eliminates spectral data whose detection range includes background areas, ensuring that the selected spectral data originates solely from the target photovoltaic module and avoiding interference from background reflection on spectral characteristics.

[0092] Subsequently, the average reflectance spectral curves are extracted. First, each selected reflectance spectral data point is preprocessed, including dark current correction and baseline correction (consistent with the aforementioned reflectance spectral data preprocessing method), resulting in a single effective reflectance spectral curve. Each curve contains reflectance values ​​across a continuous wavelength dimension. For multiple effective reflectance spectral curves of the same photovoltaic module, the average reflectance is calculated using a wavelength-wise averaging method, i.e., for each specific wavelength... Extract the reflectance values ​​of all valid spectral curves at that wavelength, and calculate their arithmetic mean as the average reflectance corresponding to that wavelength. By traversing all wavelength dimensions, the average reflectance corresponding to each wavelength is arranged in wavelength order to form the average reflectance spectrum curve of the photovoltaic module. This curve can eliminate random noise and local reflection differences in individual spectral data, accurately reflect the overall reflectance spectral characteristics of the photovoltaic module, and provide a stable and reliable spectral benchmark for subsequent analysis.

[0093] S14. Compare the average reflectance spectral curve with the preset standard clean component spectral curve, and evaluate the surface cleanliness level of the photovoltaic module by calculating the spectral difference index.

[0094] Specifically, this step achieves an objective assessment of surface cleanliness levels by quantitatively comparing the spectral differences between the target module and the standard clean module. The core of this step lies in constructing a scientific index of spectral difference and a standard for classifying levels. First, the method for obtaining the preset standard clean module's spectral curve is clarified. This curve requires selecting a photovoltaic module of the same model and specifications as the target module. Under a clean laboratory environment (free from dust, oil, stains, or any other contaminants), reflectance spectral data is collected using an instrument of the same model and parameter settings as the spectrometer carried by the drone. After preprocessing and averaging, the curve is obtained to ensure that it accurately reflects the inherent reflectance spectral characteristics of the module under clean conditions, providing a unified benchmark for difference comparison.

[0095] The spectral difference index is calculated by combining spectral angle matching degree and spectral root mean square error, quantifying spectral differences from two dimensions: spectral shape similarity and numerical difference.

[0096] Spectral angular matching This method is used to characterize the shape similarity of two spectral curves. The core idea is to treat the spectral curves as vectors in a high-dimensional space, and measure shape consistency by calculating the angle between the vectors. A smaller angle indicates greater shape similarity and a cleaner component surface. The calculation formula is as follows: In the formula The average reflectance spectrum of the target photovoltaic module at wavelength Reflectivity at that location The standard cleanroom component spectral curve at wavelength The reflectance at a given point is summed within the effective detection wavelength range of the spectrometer. The spectral angle matching degree ranges from 0 to 90°. The closer the value is to 0°, the more consistent the shapes of the two spectra are, and the lower the degree of surface contamination of the component.

[0097] Root mean square error of spectrum This value is used to characterize the average difference in reflectance values ​​between two spectral curves. A smaller value indicates that the values ​​are closer and the component surface is cleaner. The calculation formula is: In the formula To effectively detect the number of wavelength points within a wavelength range, and The definition is consistent with the formula for spectral angle matching. The root mean square error of the spectrum ranges from 0 to 1. The closer the value is to 0, the smaller the difference in reflectance values ​​between the two spectra, and the lower the degree of surface contamination of the component.

[0098] Based on the two spectral difference indices mentioned above, a cleanliness level assessment system is constructed: First, the weights of the two indices are determined using the analytic hierarchy process (AHP), with spectral angle matching having a higher weight (because it better reflects spectral shape distortion caused by contamination), followed by spectral root mean square error (RMSE), with the sum of the weights being 1. Then, spectral angle matching and spectral RMS are inversely normalized (the smaller the value, the larger the normalized value), respectively, to obtain the normalized values ​​of the two indices. Finally, the comprehensive evaluation score is calculated. In the formula The weights for spectral angle matching. This is the normalized value of the spectral angle matching degree. The weights of the root mean square error of the spectrum, This is the normalized value of the root mean square error of the spectrum. Cleanliness levels are determined based on the comprehensive evaluation score; a higher score indicates higher cleanliness. Specifically, it can be divided into five levels: extremely clean, clean, slightly contaminated, moderately contaminated, and heavily contaminated. Each level corresponds to a specific comprehensive evaluation score range, which is calibrated using a large amount of component sample data with different levels of contamination to ensure the objectivity and repeatability of the level assessment.

[0099] S15. Extract characteristic reflectance parameters from the average reflectance spectrum curve as reflectance spectrum characteristic parameters of photovoltaic modules.

[0100] Specifically, this step requires extracting parameters from the average reflectance spectral curve that accurately characterize the core properties of the component's reflectance spectrum, providing crucial input for subsequent physical optics model construction. The extracted feature parameters must comprehensively reflect the intensity, shape, and positional characteristics of the spectrum. First, the extraction dimensions of the feature parameters are determined, covering three core parameters: characteristic wavelength reflectance, spectral peak / valley characteristics, and spectral slope, ensuring a comprehensive characterization of the reflectance spectral properties.

[0101] Extracting reflectance at characteristic wavelengths requires selecting key wavelengths that reflect the reflective characteristics of photovoltaic modules, including the center wavelength of the visible light band and characteristic wavelengths of the near-infrared band. The reflectance at these wavelengths directly reflects the module's reflectivity in the main areas where solar radiation is concentrated. For the average reflectance spectral curve, the reflectance values ​​corresponding to each characteristic wavelength are extracted one by one, denoted as […]. , ... ,in , ... Each characteristic wavelength is selected based on experimental data of the reflection spectral characteristics of photovoltaic module materials, ensuring that the reflection differences of different modules can be effectively distinguished.

[0102] Peak / valley feature extraction employs a peak-valley detection algorithm, traversing the average reflectance spectral curve to identify local maxima (peaks) and local minima (valleys). For each peak, the corresponding feature parameters are extracted: peak wavelength. (Wavelength corresponding to the peak), peak reflectance (Reflectivity value corresponding to peak value), peak half-width at half-maximum (The wavelength range width corresponding to half the peak reflectance), where the peak half-width at half-maximum reflects the sharpness of the spectral peak and indirectly reflects the uniformity of the component surface material; for each valley value, the valley wavelength is extracted. (Wavelength corresponding to the valley value) and valley reflectivity (Reflectance value corresponding to the valley value). If the spectral curve has multiple peaks or valleys, select the two peaks and valleys with the largest amplitudes as core feature parameters to ensure the representativeness of the features.

[0103] Spectral slope extraction is used to characterize the variation trend of the spectral curve within a specific wavelength range, reflecting the variation of component reflectance with wavelength. Two key continuous wavelength ranges are selected (e.g., low to mid-visible light and mid-visible to near-infrared light), and the spectral slope within each range is calculated. For wavelength ranges... spectral slope The calculation formula is: In the formula The average reflectance spectrum at wavelength Reflectivity at that location wavelength The reflectance at a given wavelength. The sign of the spectral slope reflects the direction of reflectance change with wavelength, while the absolute value reflects the rate of change. It is an important supplementary parameter for characterizing the spectral properties of reflectance.

[0104] By integrating the extracted characteristic wavelength reflectance, spectral peak / valley characteristic parameters, and spectral slope, a set of reflectance spectral characteristic parameters of photovoltaic modules is formed. This set can comprehensively and accurately characterize the essential features of the reflectance spectrum of the modules, providing core data support for fitting parameters such as diffuse reflectance coefficient and specular reflectance coefficient in subsequent physical optics models.

[0105] In one optional embodiment, the average reflectance spectral curve is compared with a preset standard clean module spectral curve, and the surface cleanliness level of the photovoltaic module is evaluated by calculating the spectral difference index, including the following steps:

[0106] S21. Normalize the average reflectance spectral curve and the standard clean module spectral curve respectively to obtain the normalized reflectance values ​​of the photovoltaic module in each band and the normalized reflectance values ​​of the standard clean module in each band.

[0107] Specifically, the core of this step is to eliminate the interference caused by the difference in absolute reflectance values ​​between the two spectral curves, allowing the comparison to focus on the shape characteristics and band distribution patterns of the spectra, thus providing a unified scale of basic data for subsequent calculations of spectral difference indices. The normalization process employs a vector normalization algorithm, which converts each spectral curve into a unit vector, ensuring that the two curves have the same modulus reference in high-dimensional band space. This avoids the impact of absolute reflectance deviations caused by minor differences in component materials or slight fluctuations in the measurement environment on the comparison results.

[0108] During normalization, the effective bands are first screened. Based on the average reflectance spectral curve and the standard cleanroom component spectral curve, invalid bands outside the spectrometer's detection range and abnormal bands with a noise signal ratio exceeding a set proportion are jointly removed (determined through signal-to-noise ratio analysis; bands with a signal-to-noise ratio below a set threshold are considered abnormal). The effective detection bands that overlap between the two are retained to ensure band consistency in subsequent calculations. Let the set of effective bands after screening be { },in For the number of effective bands, This represents the spectral band number, and calculations are performed by traversing all bands within this set.

[0109] For any spectral curve (including the average reflectance spectral curve of a photovoltaic module and the spectral curve of a standard clean module), the formula for calculating its normalized reflectance value is:

[0110]

[0111] In the formula, This indicates that a certain spectral curve is in the band The normalized reflectance value at that location, This indicates that the spectral curve is in the band. The original reflectivity value at the location (for photovoltaic modules) For standard cleanroom components ), The L2 norm of the spectral curve, which is the square root of the sum of the squares of the original reflectance values ​​across all effective bands, is calculated using the following formula:

[0112]

[0113] This norm characterizes the length of the spectral curve as a high-dimensional vector. After normalizing the vector by dividing by the norm, the normalized vector magnitude of each spectral curve is 1, ensuring that the comparison between the two curves only reflects the difference in direction (i.e., the difference in shape and distribution), rather than the difference in length (i.e., the difference in absolute reflectance).

[0114] During the implementation process, the L2 norm of the two spectral curves is first calculated separately. and ,in , Then, for each effective band Calculate the normalized reflectance value of the photovoltaic modules respectively. Normalized reflectance value compared to standard cleanroom components ,in This represents the original reflectance value of the average reflectance spectrum curve of the photovoltaic module. This represents the original reflectance value of the spectral curve of the standard cleanroom component. The calculation process uses double-precision floating-point arithmetic to ensure numerical accuracy and avoid deviations in normalization results caused by rounding errors, thus laying a precise data foundation for subsequent calculations of the difference index.

[0115] S22. Calculate the spectral angle based on the normalized reflectance values ​​of photovoltaic modules and standard cleanroom modules in each wavelength band; wherein, the formula for calculating the spectral angle is:

[0116]

[0117] in, The value represents the spectral angle, which measures the overall similarity in shape between the average reflectance spectral curve and the spectral curve of a standard cleanroom component. The smaller the value, the more similar the components are. Indicates the spectral band number, traversing all valid bands; Indicates the photovoltaic module in the wavelength band The normalized reflectance value; Indicates the standard cleanroom components in the wavelength band The normalized reflectance value.

[0118] Specifically, this step quantifies the shape similarity of two spectral curves by calculating the spectral angle. As the angle between two vectors in a high-dimensional vector space, the spectral angle directly reflects the consistency of the vector directions and thus characterizes the similarity of the spectral curve shapes. It is one of the core difference indicators for evaluating the surface cleanliness of components.

[0119] The physical essence of the spectral angle is to reflect the shape similarity of the spectral curve through the directional consistency of high-dimensional spatial vectors. Surface contamination of a photovoltaic module causes changes in reflectivity at specific wavelengths, thereby altering the shape of the spectral curve. This results in an angle between the spectral curve of the photovoltaic module and that of a standard clean module. The more severe the contamination, the greater the shape difference, and the larger the spectral angle. The higher the value, the more accurately the indicator can quantify the impact of pollution on the spectral shape, providing a core basis for cleanliness level assessment.

[0120] S23. Based on the normalized reflectance values ​​of photovoltaic modules and standard cleanroom modules in each wavelength band, calculate the spectral information divergence; wherein, the formula for calculating the spectral information divergence is:

[0121]

[0122] in, It represents the spectral information divergence, used to measure the difference in probability distribution between the average reflectance spectral curve and the spectral curve of a standard cleanroom component; This indicates the value of the normalized reflectance of the photovoltaic module calculated in the band. The probability distribution value; This indicates the normalized reflectance value calculated based on standard cleanroom components in the spectral band. The probability distribution value.

[0123] Specifically, this step quantifies the difference between the two spectral curves from the perspective of probability distribution by calculating the spectral information divergence, which complements the spectral angle (focus shape similarity) and comprehensively characterizes the spectral changes caused by surface contamination of the component. This indicator is based on the concept of relative entropy in information theory and can accurately reflect the degree of deviation between the two spectra in the reflectance distribution patterns of each band.

[0124] Representing spectral information divergence, it is a core output indicator, with a value range of [value range missing]. to A smaller value indicates that the reflectance probability distributions of the two spectral curves are closer, and the higher the surface cleanliness of the photovoltaic module; a larger value indicates a greater difference in distribution and a higher degree of contamination. When the two spectra are completely identical, ... .

[0125] Indicates the photovoltaic module in the wavelength band The reflectance probability distribution value is obtained by treating the normalized reflectance values ​​of each wavelength band of the photovoltaic module as probability densities and converting them into a probability distribution through normalization processing to ensure that all wavelength bands are... The sum is 1, and its calculation formula is... In the middle, molecules For photovoltaic modules in the band The normalized reflectance value, denominator This is the sum of the normalized reflectance values ​​of all effective bands of the photovoltaic module. The essence of this calculation process is to convert the relative intensity of reflectance into a probability proportion, reflecting the contribution weight of each band in the overall reflection.

[0126] Indicates the standard cleanroom components in the wavelength band The probability distribution value of reflectance, its calculation logic is the same as Consistent, calculation formula In the middle, molecules For standard clean components in the band The normalized reflectance value, denominator The sum of the normalized reflectance values ​​of all effective bands of the standard cleanroom components is used as the basis for the probability distribution under clean conditions.

[0127] Spectral information divergence can effectively capture spectral angles The subtle differences in band distribution that cannot be reflected, such as situations where contamination causes a significant increase in reflectivity in some bands while significantly decreasing it in others, thus affecting the spectral angle. This complements each other and comprehensively improves the accuracy of cleanliness assessment.

[0128] S24. Based on the spectral angle and spectral information divergence, and referring to the preset cleanliness grading threshold table, determine the surface cleanliness level of the photovoltaic module.

[0129] Specifically, the core of this step is to achieve an objective determination of cleanliness level based on two complementary spectral difference indicators and scientifically calibrated grading thresholds, ensuring the consistency and repeatability of the assessment results. The grading process needs to take into account both the synergistic effect of the indicators and the needs of actual application scenarios.

[0130] First, the process of developing the preset cleanliness level classification threshold table will be explained. This threshold table was calibrated using a large amount of experimental data to ensure the scientific validity and applicability of the classification standards. The specific development process is as follows: Samples of the same model of photovoltaic modules were collected under different levels of contamination (including five gradients: extremely clean, clean, lightly contaminated, moderately contaminated, and heavily contaminated). A sufficient number of samples were collected for each gradient to ensure statistical significance. Calculations were then performed for each sample. and Combining manually labeled cleanliness levels, the K-means clustering algorithm was used to cluster the samples. and Cluster analysis was performed on the data to determine the corresponding cleanliness levels. Threshold range and Threshold ranges; the effectiveness of the threshold ranges is verified using cross-validation, and the threshold boundaries are adjusted to ensure that the grading accuracy meets the preset requirements, ultimately forming a cleanliness grading threshold table. In this threshold table, each cleanliness level corresponds to a unique threshold. Scope and The range, and the higher the level (the better the cleanliness), the corresponding and The smaller the threshold.

[0131] The cleanliness level is determined using a "dual-indicator synergistic judgment" logic, meaning that only when the cleanliness level of the photovoltaic module is determined... and Only when all criteria are met within the threshold range corresponding to a certain cleanliness level can a sample be classified as belonging to that level, thus avoiding misjudgments that may result from a single indicator. The specific judgment process is as follows: First, extract the data from the photovoltaic module to be evaluated. Calculation results and The calculation results are as follows: The second step is to query the preset cleanliness grading threshold table and compare the results with the corresponding values ​​for each grade. interval and The third step is to evaluate the component within the range; Falling into a certain level Within the threshold range, and At the same level If the threshold range is within the specified range, the surface cleanliness level of the component is determined to be that level. In the fourth step, if one indicator meets the high-level threshold while another indicator meets the low-level threshold, the principle of "choosing the lower level" is adopted to determine the lower cleanliness level, ensuring the conservatism and safety of the evaluation results and avoiding ignoring the pollution problem reflected by another indicator due to the excellence of a single indicator.

[0132] For example, the corresponding level of cleanliness Threshold range is , Threshold range is Cleanliness level corresponding to Threshold range is , Threshold range is And so on. When the component to be evaluated and At that time, it was judged to be extremely clean; if but If it is, then it is determined to be of a cleanliness level.

[0133] Furthermore, the threshold table is scalable, allowing for targeted adjustments based on the reflection characteristics of different photovoltaic module models and the types of pollution in different usage environments (such as dust, oil, bird droppings, etc.). By supplementing sample data for corresponding scenarios, the threshold range can be recalibrated to ensure the applicability and accuracy of the method in various application scenarios. The final output surface cleanliness level will serve as an important component of the photovoltaic module's key parameter set, providing accurate input data for subsequent glare assessment models and ensuring that the glare assessment results truly reflect the glare generation potential of the modules under actual conditions.

[0134] In one optional embodiment, based on a set of key parameters, solar position information from a comprehensive dataset, and illumination intensity data corresponding to the highway observation point, a glare assessment index for each highway observation point is calculated using a physical optics model, including the following steps:

[0135] S31. Based on the set of key parameters, establish a surface reflection model that includes specular reflection components and diffuse reflection components.

[0136] Specifically, the key parameter set includes reflectance spectral characteristic parameters (such as characteristic reflectance parameters, principal component parameters), orientation, etc. Inclination angle The surface cleanliness level is the core input for model construction. The reflectance spectral characteristics directly determine the magnitude of the reflectance coefficient, while the orientation and tilt angle affect the propagation direction of reflected light. The surface cleanliness level modifies the reflectance coefficient to reflect the weakening effect of contamination on the reflectance effect. The surface reflectance model is constructed using a modified bidirectional reflectance distribution function (BRDF), whose expression is: In the formula This indicates the reflected radiance of the photovoltaic module in the direction of observation. The observation direction and the normal vector of the component surface The observation angle To observe the azimuth angle, The diffuse reflectance coefficient is... The specular reflection coefficient, Highlight index, This represents the solar incident irradiance on the surface of the photovoltaic module. The direction of solar incidence and the normal vector of the component surface The angle of incidence, Unit vector of the direction of solar incidence unit vector of the observation direction half vector (i.e. Diffuse reflectance coefficient With specular reflection coefficient The highlight index was obtained by fitting the reflectance spectral characteristic parameters from the key parameter set with a laboratory-calibrated multiple regression model. Adjust according to the surface cleanliness level (the higher the cleanliness level, the better). The larger the surface area, the more pronounced the specular reflection; solar incident irradiance. The data is obtained by combining light intensity data from the comprehensive dataset with atmospheric transmittance. The corrected incident angle Using solar position information (solar altitude angle) Sun azimuth ) and component orientation Inclination angle Calculation yields ( This ensures that the model parameters closely match the actual working conditions.

[0137] S32. Based on the surface reflection model, solar position information, and visibility analysis results between highway observation points, assess the glare radiation contribution of each photovoltaic module to each highway observation point during the assessment period.

[0138] Specifically, this step requires combining a reflection model, dynamic changes in the sun's position, and spatial visibility to accurately quantify the glare radiation contribution of a single photovoltaic module to a single highway observation point. The core is to eliminate shading interference and dynamically match the sun's position with the reflection effect. First, a visibility analysis is performed. A digital surface model (DSM) of the evaluation area is constructed based on three-dimensional coordinate point cloud data. A ray tracing algorithm is used to determine whether there are obstructions (such as slope protrusions, vegetation, other buildings, etc.) between the center of each photovoltaic module and each highway observation point. If the ray does not intersect with a non-target object in the digital surface model during its propagation path, it is considered to have visibility and is marked as such. Otherwise, it is not visible and is marked as such. Under non-line-of-sight conditions, the photovoltaic module contributes no glare radiation to the observation point. During the evaluation period, the solar elevation angle is calculated at fixed time intervals (consistent with the data acquisition time step) using a solar position algorithm. With solar azimuth ( To evaluate a specific moment within a given time period, the reflected radiance of the photovoltaic module in the direction of the observation point is dynamically calculated using a surface reflection model. Then, based on the spatial distance between the observation point and the photovoltaic module (Calculated from the three-dimensional coordinates of both) ,in Here are the coordinates of the observation point. (The coordinates of the photovoltaic module's center) convert the reflected radiation brightness into glare intensity at the observation point. The conversion formula is: In the formula The effective reflective area of ​​the photovoltaic module (calculated from contour mask and point cloud data). For the observation angle. Ultimately, the individual photovoltaic module at a single observation point at time [time value missing]. Glare radiation contribution The product of visibility marker and glare intensity ( When the visibility is poor or the reflected radiation brightness is below the detection threshold, This ensures the authenticity and accuracy of contribution assessments.

[0139] S33. For each highway observation point, the glare radiation contribution of all photovoltaic modules is accumulated according to the time step within the evaluation period to generate the glare illuminance time series of the highway observation point.

[0140] Specifically, the core of this step is to integrate the dynamic contributions of all photovoltaic modules to form a complete glare intensity variation curve for the observation point, providing basic data for subsequent calculations of duration and impact range. First, the time step for the evaluation period is determined, and this time step is consistent with the time step for solar position calculation and data acquisition to ensure synchronization across time dimensions and avoid cumulative errors caused by mismatched time steps. For each highway observation point, each time node within the evaluation period is traversed. ( , To assess the total number of time points within the time period, the contribution of all photovoltaic modules to the glare radiation at that time point is extracted. ( Number the photovoltaic modules. (Number the observation point), and sum the contributions of all components according to the time step to obtain the observation point at time. Total glare illuminance ( (Total number of photovoltaic modules). Arrange the total glare intensity at all time points in chronological order to form a glare intensity time series for this highway observation point. This sequence fully records the dynamic changes in glare illuminance at the observation points during the evaluation period, including key information such as the time of glare occurrence, peak time, and decay time, providing continuous data source support for subsequent index calculations.

[0141] S34. Based on the glare illuminance time series, the glare duration is obtained by calculating the cumulative time exceeding the preset glare intensity threshold.

[0142] Specifically, this step quantifies the cumulative duration of glare exceeding the threshold to clarify the degree of continuous interference of glare on the observation point. The core lies in reasonably setting the threshold and accurately calculating the effective duration. Preset glare intensity threshold. The threshold is set based on relevant highway traffic safety standards and driving visual comfort experimental data, combined with the actual impact characteristics of photovoltaic module glare. This threshold needs to be distinguished from ambient illuminance, and only targets glare generated by photovoltaic module reflection, ensuring the threshold's relevance and rationality. For the glare illuminance time series of each highway observation point... Iterate through all time nodes Determine the total glare intensity of the node. Is it greater than or equal to the preset threshold? If satisfied Then record the time interval corresponding to that time node. (The duration of the time step) is included in the effective cumulative time. During the statistical process, periods where the duration of a single exceedance of the threshold is less than the minimum effective duration (as set according to driving safety experiments to avoid instantaneous interference being included) must be excluded; only the sum of periods where the threshold is continuously exceeded and the duration reaches the minimum effective duration is accumulated. Finally, the glare duration at this observation point is... The cumulative duration of all valid time periods exceeding the threshold, i.e. ,in To meet Furthermore, the set of time points where the continuous duration reaches the minimum effective duration, this indicator directly reflects the degree of visual interference of glare to the driver, providing a key basis for the level assessment.

[0143] S35. Based on the glare illuminance time series and glare duration of all highway observation points, determine the geographical range where the glare intensity exceeds the preset impact threshold through spatial interpolation, and obtain the glare impact range.

[0144] Specifically, this step transforms discrete observation point data into a continuous spatial distribution through spatial interpolation, accurately defining the area affected by glare along the highway. The core of this step lies in selecting a suitable interpolation algorithm and setting a reasonable impact threshold. First, the preset impact threshold is defined, which includes two requirements: first, the glare illuminance must exceed the glare intensity threshold. Secondly, the duration of the glare must exceed a preset duration threshold. (Based on traffic safety requirements, ensuring the affected area is the region that actually interferes with driving safety). Based on the glare illuminance time series of all highway observation points, the maximum glare illuminance at each observation point during the assessment period is extracted. Duration of glare Filter out those that simultaneously meet the requirements and The observation points were selected as effective influence observation points. The Kriging interpolation algorithm was used to spatially interpolate the maximum glare intensity at the effective influence observation points. This algorithm fully utilizes the spatial correlation of the observation points and, combined with the topographic features of the highway slope, generates a spatial distribution raster map of glare intensity along the highway and surrounding areas (raster resolution is set according to the evaluation accuracy requirements). The raster value represents the maximum glare intensity at the corresponding location. The values ​​that meet the requirements are extracted from the spatial distribution raster map. All grid cells, combined with geographical information from the highway route design map (such as station numbers, lane distribution, and slope boundaries), are used to determine the continuous geographical area formed by these grid cells through a vector boundary extraction algorithm. This area is the glare impact range. Simultaneously, key geographical parameters of the impact range are statistically analyzed, including the length along the highway (calculated based on station numbers), the width perpendicular to the highway (calculated based on the lateral distribution of grid cells), and the number of affected lanes (determined by combining lane width and the lateral coverage of the impact range). This forms a complete description of the glare impact range, providing precise spatial basis for highway traffic safety management and photovoltaic module installation optimization.

[0145] In an optional embodiment, the formula for calculating the glare radiation contribution in step S32 is:

[0146]

[0147] in, Indicates at time photovoltaic modules Contribution of glare radiation to the target highway observation point; Indicates at time photovoltaic modules The total reflected radiance of the surface in the direction pointing towards the highway observation point is calculated using the surface reflection model; Indicates photovoltaic modules Effective projected area in the observation direction; Represents the observation direction vector and the photovoltaic module The angle between the surface normal vectors; Indicates photovoltaic modules The straight-line distance from the center to the highway observation point; This represents the atmospheric transmittance of light along its propagation path, calculated using an empirical model based on the straight-line distance and atmospheric extinction coefficient.

[0148] Specifically, this step uses a quantitative formula to calculate the glare radiation contribution of a single photovoltaic module to the target highway observation point at a specific moment. This formula comprehensively considers key influencing factors such as reflected radiation brightness, effective projected area, observation angle, propagation distance, and atmospheric attenuation, ensuring that the calculation results accurately reflect the actual glare effect. The overall physical logic of the formula is as follows: the total reflected radiation brightness of the photovoltaic module surface is corrected for the effective projected area, then the angle between the observation direction and the normal vector of the module surface is considered, and further corrected by the square attenuation of the propagation distance and atmospheric transmittance, finally yielding the glare radiation contribution at the observation point.

[0149] Indicates time Photovoltaic modules The contribution of the glare radiation to the target highway observation point, in lux (lx), is the core output of this step and directly characterizes the glare intensity generated by the component at the observation point at that moment. Indicates time Photovoltaic modules The total reflected radiance of the surface in the direction pointing towards the highway observation point, in candela per square meter ( The value is calculated from the surface reflection model containing specular and diffuse reflection components established in step S31. Its magnitude is directly related to the component's reflection spectral characteristic parameters, solar incident angle, and surface cleanliness level.

[0150] Indicates photovoltaic modules The effective projected area along the observation direction, in square meters ( The projection coefficient is calculated by the actual physical area of ​​the component and the projection coefficient of the observation direction. The projection coefficient is determined by the observation angle. Decision, that is ( For photovoltaic modules The physical area (calculated from the contour mask and 3D point cloud data) reflects the actual area of ​​the component involved in reflection in the observation direction; Represents the observation direction vector and the photovoltaic module The angle between the surface normal vectors, in degrees (°), and the observation direction vector points from the observation point on the target highway to the photovoltaic module. The center of the surface normal vector is obtained from the component surface plane equation fitted in step S2, and is calculated using the vector dot product formula, i.e. ( The observation direction vector, (where is the normal vector of the component surface). The smaller this angle is, the larger the projected area of ​​the component in the observation direction, and the stronger the contribution of glare radiation. Indicates photovoltaic modules The straight-line distance from the center to the highway observation point, in meters (m), is calculated using the three-dimensional coordinates of both. ( The three-dimensional coordinates of the highway observation point. For photovoltaic modules (Center three-dimensional coordinates), according to the law of radiation propagation, the glare intensity decreases with the square of the distance. This parameter directly reflects the weakening effect of the propagation distance on the contribution of glare.

[0151] Atmospheric transmittance represents the light's transmissivity along its propagation path. It is dimensionless and ranges from 0 to 1. A value closer to 1 indicates a smaller attenuation effect of the atmosphere on light. (Based on photovoltaic modules) straight-line distance to highway observation point and atmospheric extinction coefficient Calculated through an empirical model, the empirical model can be used. Atmospheric extinction coefficient By comprehensively collecting environmental monitoring data (such as visibility and humidity) from the dataset and querying atmospheric radiative transfer models, we can ensure that the atmospheric attenuation characteristics of the assessment area are accurately reflected.

[0152] The aforementioned method for assessing photovoltaic glare on highway slopes using unmanned aerial vehicles (UAVs) involves controlling a UAV to fly along a pre-planned route covering the photovoltaic module area and simulating a driver's perspective at highway observation points. Simultaneously, it collects multi-source data, including images, reflectance spectra, illuminance, solar position, and 3D coordinate point clouds, to form a comprehensive dataset. Based on the image, spectral, and point cloud data in this dataset, key parameters such as the orientation, tilt angle, surface cleanliness, and reflectance spectral characteristics of the photovoltaic modules are automatically extracted. Then, combining solar position information and illuminance data from the highway observation points, a physical optics model is used to quantitatively calculate the glare intensity, duration, and impact range at each observation point. Finally, the calculation results are compared with preset thresholds to output intuitive glare level assessment results and spatial distribution information. This achieves efficient, comprehensive, and automated accurate assessment of photovoltaic glare on highway slopes, providing a scientific basis for ensuring driving safety and optimizing photovoltaic systems.

[0153] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0154] Based on the same inventive concept, this application also provides an apparatus for implementing the above-mentioned UAV-based method for assessing photovoltaic glare on highway slopes. The solution provided by this apparatus is similar to the implementation described in the above method. Therefore, the specific limitations of one or more UAV-based methods for assessing photovoltaic glare on highway slopes provided below can be found in the above-described limitations of the UAV-based method for assessing photovoltaic glare on highway slopes, and will not be repeated here.

[0155] In one exemplary embodiment, such as Figure 3 As shown, a drone-based photovoltaic glare assessment device 30 for highway slopes is provided to implement the methods described in the above embodiments. The device includes:

[0156] The aerial survey data acquisition module 31 is used to control the UAV to fly along a preset route and simultaneously collect data from the photovoltaic module area and the highway observation point area through multiple sensors carried by the UAV, generating a comprehensive dataset including image data, reflectance spectrum data, light intensity data, solar position information, and three-dimensional coordinate point cloud data; among which, the preset route includes a detailed survey route covering the photovoltaic module area and a highway viewpoint observation route simulating the driver's perspective.

[0157] The photovoltaic characteristic analysis module 32 is used to extract the set of key parameters of the photovoltaic module based on the image data, reflectance spectrum data and three-dimensional coordinate point cloud data in the comprehensive acquisition dataset; wherein, the set of key parameters includes at least the orientation, tilt angle, surface cleanliness level and reflectance spectrum characteristic parameters of the photovoltaic module.

[0158] The glare physical modeling module 33 is used to calculate the glare evaluation index for each highway observation point based on the key parameter set, the solar position information in the comprehensive data collection dataset, and the light intensity data corresponding to the highway observation point, through a physical optics model. The glare evaluation index includes glare intensity, glare duration, and glare impact range.

[0159] The glare classification output module 34 is used to compare the glare evaluation index with the preset glare level classification threshold, and output the level evaluation result of photovoltaic glare on highway slopes and the corresponding spatial distribution information.

[0160] Embodiments of this application also provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the aforementioned method embodiments.

[0161] Embodiments of this application also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method embodiments.

[0162] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0163] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for evaluating road slope photovoltaic glare based on a UAV, characterized in that, The method includes: S1. Control the drone to fly along a preset route, and simultaneously collect data from the photovoltaic module area and the highway observation point area through the multiple sensors carried by the drone, generating a comprehensive collection dataset including image data, reflectance spectrum data, light intensity data, solar position information, and three-dimensional coordinate point cloud data; wherein, the preset route includes a detailed survey route covering the photovoltaic module area and a highway viewpoint observation route simulating the driver's perspective; S2. Based on the image data, reflectance spectral data, and three-dimensional coordinate point cloud data in the comprehensive acquisition dataset, extract the set of key parameters for the photovoltaic module; wherein, the set of key parameters includes at least the orientation, tilt angle, surface cleanliness level, and reflectance spectral characteristic parameters of the photovoltaic module; S3. Based on the set of key parameters, the solar position information in the comprehensive data acquisition dataset, and the illumination intensity data corresponding to the highway observation point, calculate the glare assessment index for each highway observation point using a physical optics model; wherein, the glare assessment index includes glare intensity, glare duration, and glare impact range; S4. Compare the glare evaluation index with the preset glare level classification threshold, and output the glare level evaluation result of the highway slope photovoltaic glare and the corresponding spatial distribution information.

2. The method of claim 1, wherein, The extraction of a set of key parameters for photovoltaic modules based on the image data, reflectance spectral data, and three-dimensional coordinate point cloud data in the comprehensive acquisition dataset includes: S11. The image data is preprocessed to obtain an orthophoto; the contour mask of each photovoltaic module is extracted by semantic segmentation of the orthophoto. S12. Based on the three-dimensional coordinate point cloud data corresponding to the contour mask, fit the surface plane equation of the photovoltaic module using the least squares method, and obtain the surface normal vector of the photovoltaic module based on the surface plane equation; calculate the orientation and tilt angle of the photovoltaic module based on the surface normal vector. S13. Extract the average reflectance spectral curve of the photovoltaic module region from the reflectance spectral data that spatially matches the contour mask; S14. Compare the average reflectance spectral curve with the preset standard clean component spectral curve, and evaluate the surface cleanliness level of the photovoltaic component by calculating the spectral difference index. S15. Extract characteristic reflectance parameters from the average reflectance spectrum curve as the reflectance spectrum characteristic parameters of the photovoltaic module.

3. The method of claim 2, wherein, The step of comparing the average reflectance spectral curve with a preset standard clean component spectral curve and evaluating the surface cleanliness level of the photovoltaic module by calculating the spectral difference index includes: S21. Normalize the average reflectance spectral curve and the standard clean component spectral curve respectively to obtain the normalized reflectance values ​​of the photovoltaic module in each band and the normalized reflectance values ​​of the standard clean component in each band. S22. Calculate the spectral angle based on the normalized reflectance values ​​of the photovoltaic module and the standard cleanroom module in each wavelength band; wherein, the formula for calculating the spectral angle is: in, The spectral angle represents the overall similarity in shape between the average reflectance spectral curve and the standard cleanroom component spectral curve; the smaller the value, the more similar they are. Indicates the spectral band number, traversing all valid bands; This indicates that the photovoltaic module operates in the specified wavelength band. The normalized reflectance value; This indicates that the standard cleanroom component is in the wavelength band. The normalized reflectance value; S23. Based on the normalized reflectance values ​​of the photovoltaic module and the standard cleanroom module in each band, calculate the spectral information divergence; wherein, the formula for calculating the spectral information divergence is: in, The spectral information divergence is used to measure the difference in probability distribution between the average reflectance spectral curve and the standard cleanroom component spectral curve. This indicates the value of the normalized reflectance of the photovoltaic module calculated in the band. The probability distribution value; This indicates the value of the normalized reflectance of the standard cleanroom component calculated in the band. The probability distribution value; S24. Based on the spectral angle and the spectral information divergence, and referring to a preset cleanliness grading threshold table, determine the surface cleanliness level of the photovoltaic module.

4. The method according to any one of claims 1 to 3, characterized in that, Based on the set of key parameters, the solar position information in the comprehensive data acquisition dataset, and the illumination intensity data corresponding to the highway observation points, the glare assessment index for each highway observation point is calculated using a physical optics model, including: S31. Based on the set of key parameters, establish a surface reflection model that includes specular reflection components and diffuse reflection components; S32. Based on the surface reflection model, the solar position information, and the visibility analysis results between highway observation points, assess the glare radiation contribution of each photovoltaic module to each highway observation point during the assessment period. S33. For each of the highway observation points, the glare radiation contribution of all the photovoltaic modules is accumulated according to the time step within the evaluation period to generate the glare illuminance time series of the highway observation points. S34. Based on the glare illuminance time series, the glare duration is obtained by calculating the cumulative time exceeding the preset glare intensity threshold; S35. Based on the glare illuminance time series and glare duration of all the highway observation points, determine the geographical range where the glare intensity exceeds the preset impact threshold by spatial interpolation, and obtain the glare impact range.

5. The method according to claim 4, characterized in that, The formula for calculating the glare radiation contribution in S32 is as follows: in, Indicates at time photovoltaic modules The contribution of the glare radiation to the target highway observation point; Indicates at time photovoltaic modules The total reflected radiance of the surface in the direction pointing to the highway observation point is calculated by the surface reflection model; Indicates photovoltaic modules Effective projected area in the observation direction; Represents the observation direction vector and the photovoltaic module The angle between the surface normal vectors; Indicates photovoltaic modules The straight-line distance from the center to the highway observation point; The atmospheric transmittance, representing the light's propagation path, is calculated using an empirical model based on the straight-line distance and the atmospheric extinction coefficient.

6. A drone-based photovoltaic glare assessment device for highway slopes, used to implement the method described in any one of claims 1 to 5, characterized in that, The device includes: The aerial survey data acquisition module is used to control the UAV to fly along a preset route. The UAV uses multiple sensors to simultaneously collect data from the photovoltaic module area and the highway observation point area, generating a comprehensive dataset including image data, reflectance spectrum data, light intensity data, solar position information, and three-dimensional coordinate point cloud data. The preset route includes a detailed survey route covering the photovoltaic module area and a highway viewpoint observation route simulating the driver's perspective. The photovoltaic characteristic analysis module is used to extract a set of key parameters of the photovoltaic module based on the image data, the reflectance spectrum data and the three-dimensional coordinate point cloud data in the comprehensive acquisition dataset; wherein, the set of key parameters includes at least the orientation, tilt angle, surface cleanliness level and reflectance spectrum characteristic parameters of the photovoltaic module; The glare physical modeling module is used to calculate the glare evaluation index for each highway observation point based on the key parameter set, the solar position information in the comprehensive acquisition dataset, and the illumination intensity data corresponding to the highway observation point, using a physical optics model; wherein, the glare evaluation index includes glare intensity, glare duration, and glare impact range; The glare classification output module is used to compare the glare evaluation index with the preset glare level classification threshold, and output the glare level evaluation result of the highway slope photovoltaic glare and the corresponding spatial distribution information.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

Images