Method for Constructing an Electronic Sand Table for Photovoltaic Arrays Using Drone Aerial Imagery

Abstract

This invention relates to the field of image processing technology, specifically to a method for constructing an electronic sand table for a photovoltaic array using drone aerial imagery. The method involves acquiring drone aerial imagery data of the photovoltaic array, along with wind intensity and POS attitude information obtained during the drone's aerial photography process. Based on the wind intensity and POS attitude information, the drone aerial imagery data is initially screened to determine initial feature point matching pairs and their matching indices for any two frames of the target drone aerial imagery data to be matched. Feature analysis is then performed on the local regions of the aerial images containing the two feature points in each initial feature point matching pair to determine the selection confidence level of each initial feature point matching pair. The matching indices of each initial feature point matching pair are then corrected, thereby completing the image matching of the target drone aerial imagery data. Based on the image matching results, an electronic sand table for the photovoltaic array is constructed. This invention ensures the accuracy of the electronic sand table construction.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method for constructing a photovoltaic array electronic sand table using aerial images taken by drones. Background Technology

[0002] The photovoltaic array electronic sand table is a core tool for the digital operation and maintenance and display of photovoltaic power plants. Based on 3D visualization, it integrates technologies such as digital twins, Internet of Things, and AI simulation to achieve real-time monitoring and simulation of photovoltaic array layout, operating status, and power generation efficiency. It can help users understand complex spatial information more intuitively, simulate scene evolution, and support decision-making.

[0003] In existing technologies, drones are typically used to acquire aerial images of the target photovoltaic scene through oblique photography. The acquired multiple frames of aerial images are then aligned. During the alignment of two frames, feature point matching is performed on the two frames. The feature matching results are then uploaded to a calibration model to achieve coordinate alignment. Finally, based on the aligned image after coordinate correction and combined with information such as the position coordinates of the photovoltaic module, module function, and real-time operating status, a three-dimensional model is created using a high-precision terrain model of the photovoltaic array, thereby obtaining an electronic sand table of the photovoltaic array.

[0004] However, during the alignment process of aerial imagery, the flight attitude of drones is often affected by wind interference in practical applications, leading to fluctuations. This instability results in a large number of noise points in the aerial images, increasing the probability of misselecting noise points during SIFT (Scale-invariant feature transform) feature point matching. Since the quality of the matched feature points directly affects the accuracy of subsequent image alignment and correction, an excessively high proportion of noise points significantly reduces the effectiveness of feature point matching, ultimately leading to increased errors in the electronic sand table construction. Summary of the Invention

[0005] To address the technical problem mentioned above, where wind interference during UAV oblique photography in practical applications leads to numerous noise points in aerial images, resulting in misselection of noise points during feature point matching and ultimately increasing the error in electronic sand table construction, the present invention aims to provide a method for constructing a photovoltaic array electronic sand table using UAV aerial images. The specific technical solution adopted is as follows:

[0006] In a first aspect, the present invention provides a method for constructing a photovoltaic array electronic sand table using aerial imagery from a drone, comprising the following steps:

[0007] Acquire drone aerial imagery data of the photovoltaic array, as well as wind intensity and POS attitude information during the drone aerial photography process;

[0008] Based on the wind intensity information and POS attitude information, interference image filtering is performed on the UAV aerial image data to obtain the target UAV aerial image data.

[0009] Determine the initial feature point matching pairs and their matching indices in any two frames of aerial images to be matched in the target UAV aerial image data, and perform feature analysis on the local areas of the aerial images where the two feature points in each initial feature point matching pair are located, and determine the selection confidence of each initial feature point matching pair. The selection confidence is used to reflect the possibility that the initial feature point matching pair belongs to the real feature points of the image rather than noise points.

[0010] Based on the selected confidence level, the matching index of each initial feature point matching pair is corrected, and based on the corrected matching index of each initial feature point matching pair, each initial feature point matching pair is filtered to obtain the final feature point matching pair.

[0011] Based on the final feature point matching pairs, image matching is performed on the target UAV aerial image data, and a photovoltaic array electronic sand table is constructed based on the image matching results.

[0012] In conjunction with the first aspect above, in some possible implementations, interfering image filtering is performed on the UAV aerial image data to obtain target UAV aerial image data, including:

[0013] Based on the wind intensity information and POS attitude information, the aerial photography stability coefficient of each frame of aerial photography data in the UAV aerial photography data is determined.

[0014] Based on the aerial photography stability coefficient, the drone aerial photography data is filtered to obtain the filtered target drone aerial photography data.

[0015] In conjunction with the first aspect above, in some possible implementations, the POS attitude information includes vibration intensity information and pitch angle information; determining the aerial stability coefficient of each frame of aerial imagery in the UAV aerial imagery data includes:

[0016] Based on the wind intensity corresponding to the shooting time of each frame of the UAV aerial image data in the wind intensity information, and the vibration intensity in the POS attitude information, the aerial interference intensity at the shooting time of each frame of the UAV aerial image data is determined.

[0017] Based on the fluctuation of the pitch angle corresponding to the pitch angle information in the pitch angle information during the local time period in which each frame of the UAV aerial image data is captured, the degree of change in the fuselage attitude at the time of capture of each frame of the UAV aerial image data is determined.

[0018] Based on the intensity of aerial interference and the degree of change in the aircraft's attitude, the aerial stability coefficient of each frame of aerial imagery in the UAV aerial imagery data is determined.

[0019] In conjunction with the first aspect above, in some possible implementations, the UAV aerial image data is filtered based on the aerial photography stability coefficient to obtain filtered target UAV aerial image data, including:

[0020] Based on the shooting time of each aerial image in the UAV aerial image data, several aerial shooting time periods are determined in the aerial shooting process;

[0021] Based on the distance from the center time at each moment in the aerial photography period, the centering performance at each moment in the aerial photography period is determined;

[0022] Based on the centering performance and aerial photography stability coefficient, the matching image confidence of each frame of aerial photography in the UAV aerial photography data is determined.

[0023] Based on the confidence level of the matched images, the drone aerial image data is filtered to obtain the filtered target drone aerial image data.

[0024] In conjunction with the first aspect mentioned above, among some possible implementation methods, the selection confidence of each initial feature point matching pair is determined, including:

[0025] Based on the gradient difference between the two feature points in each initial feature point matching pair and their neighboring pixels in the aerial image, the feature selection fit degree of the two feature points in each initial feature point matching pair is determined.

[0026] Based on the similarity between the grayscale differences between the two feature points in each initial feature point matching pair and the neighboring pixels in the aerial image, the local texture similarity between the two feature points in each initial feature point matching pair is determined.

[0027] Based on the feature selection fit and the local texture similarity, the selection confidence of each initial feature point matching pair is determined.

[0028] In conjunction with the first aspect mentioned above, among some possible implementation methods, determining the feature selection fit between two feature points in each initial feature point matching pair includes:

[0029] Determine the average gradient of the two feature points in each initial feature point matching pair with the eight neighboring pixels in the aerial image.

[0030] Determine the mean value of the gradient mean for all pixels in the local region of the aerial image where the two feature points in each initial feature point matching pair are located, and obtain the baseline value of the gradient mean.

[0031] Based on the difference between the mean gradient value of the two feature points in each initial feature point matching pair and the baseline value of the mean gradient value, the gradient highlighting factor corresponding to the two feature points in each initial feature point matching pair is determined.

[0032] Determine the distance from the center of the aerial image to each initial feature point matching pair;

[0033] Based on the gradient highlighting factor and the distance from the two feature points in each initial feature point matching pair to the center of the image in the aerial image, the feature selection fit degree of the two feature points in each initial feature point matching pair is determined.

[0034] In conjunction with the first aspect mentioned above, among some possible implementations, determining the local texture similarity between two feature points in each initial feature point matching pair includes:

[0035] The grayscale difference between the two feature points in each initial feature point matching pair and the neighboring pixels in the local neighborhood of the aerial image is determined to obtain the image presentation deviation factor.

[0036] The graph rendering deviation factors are arranged in a set order to obtain the graph rendering deviation factor sequence corresponding to the two feature points in each initial feature point matching pair.

[0037] The similarity between the graph presentation deviation factor sequences corresponding to the two feature points in each initial feature point matching pair is determined, and the local texture similarity between the two feature points in each initial feature point matching pair is obtained.

[0038] In conjunction with the first aspect mentioned above, among some possible implementation methods, the selection confidence of each initial feature point matching pair is determined, including:

[0039] Based on the feature selection fitness and the local texture similarity, the matching combination optimization degree of each initial feature point matching pair is determined;

[0040] Determine the mean of the matching combination optimization of all initial feature point matching pairs to obtain the average matching combination optimization.

[0041] Based on the difference between the matching combination preference of each initial feature point matching pair and the average matching combination preference, the selection confidence of each initial feature point matching pair is determined.

[0042] In conjunction with the first aspect mentioned above, in some possible implementations, the SIFT algorithm is used to determine each initial feature point matching pair and its matching index in any two frames of aerial images to be matched in the target UAV aerial image data. The matching index refers to the Euclidean distance between the feature descriptors of the two feature points in each initial feature point matching pair.

[0043] In conjunction with the first aspect above, in some possible implementations, based on the selected confidence level, the matching index of each initial feature point matching pair is corrected, including:

[0044] Based on the matching index of each initial feature point matching pair and the selection confidence, the weighted matching index of each initial feature point matching pair is determined.

[0045] The weighted matching index of each initial feature point matching pair is mapped to the value range of the matching index of each initial feature point matching pair, thereby obtaining the modified matching index of each initial feature point matching pair.

[0046] Secondly, the present invention also provides a photovoltaic array electronic sand table construction device utilizing drone aerial imagery, the device comprising:

[0047] Acquire drone aerial imagery data of the photovoltaic array, as well as wind intensity and POS attitude information during the drone aerial photography process;

[0048] The data acquisition module is used to filter out interfering images from the UAV aerial image data based on the wind intensity information and POS attitude information to obtain the target UAV aerial image data.

[0049] A confidence analysis module is selected to determine the initial feature point matching pairs and their matching indices in any two frames of aerial images to be matched in the target UAV aerial image data, and to perform feature analysis on the local areas of the aerial images where the two feature points in each initial feature point matching pair are located, and to determine the selection confidence of each initial feature point matching pair. The selection confidence is used to reflect the possibility that the initial feature point matching pair belongs to the real feature points of the image rather than noise points.

[0050] The feature point matching pair filtering module is used to correct the matching index of each initial feature point matching pair based on the selected confidence level, and to filter each initial feature point matching pair based on the corrected matching index of each initial feature point matching pair to obtain the final feature point matching pair.

[0051] The electronic sand table construction module performs image matching on the target UAV aerial image data based on the final feature point matching pairs, and constructs a photovoltaic array electronic sand table based on the image matching results.

[0052] Thirdly, the present invention also provides a photovoltaic array electronic sand table construction system utilizing drone aerial imagery, comprising a memory and a processor. The memory stores executable computer program code, and the processor retrieves and runs the executable computer program code from the memory, causing the system to execute the photovoltaic array electronic sand table construction method utilizing drone aerial imagery as described in the first aspect or any possible implementation thereof.

[0053] Fourthly, the present invention also provides a computer program product comprising: computer program code, which, when run on a computer, causes the computer to execute the photovoltaic array electronic sand table construction method using UAV aerial imagery in the first aspect or any possible implementation thereof.

[0054] Fifthly, the present invention also provides a computer-readable storage medium storing computer program code that, when executed on a computer, causes the computer to perform the photovoltaic array electronic sand table construction method using UAV aerial imagery as described in the first aspect or any possible implementation thereof.

[0055] This invention offers the following advantages: By utilizing wind intensity and POS attitude information from drone aerial imagery data acquired during the acquisition of photovoltaic arrays, the invention performs preliminary screening of drone aerial imagery data, thereby eliminating aerial images that are significantly affected by interference due to poor shooting environment stability. For the remaining target drone aerial imagery data after screening, feature analysis is performed on the local regions of the aerial images containing the two feature points in each initial feature point matching pair from any two frames of images to be matched. The selection confidence level of each initial feature point matching pair is determined. This selection confidence level reflects the probability that the initial feature point matching pair belongs to a true feature point in the image rather than a noise point. Based on this selection confidence level, the matching index of each initial feature point matching pair is corrected, thereby achieving the screening of true feature point matching pairs in the image. This effectively avoids the influence of noise points during feature point matching, ensuring the accuracy of image matching and ultimately ensuring the construction accuracy of the electronic sand table based on the image matching results. Attached Figure Description

[0056] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a flowchart illustrating the steps of a photovoltaic array electronic sand table construction method using drone aerial imagery, as described in an embodiment of the present invention.

[0058] Figure 2 This is a flowchart illustrating the steps of removing interfering images from drone aerial image data according to an embodiment of the present invention.

[0059] Figure 3 This is a flowchart illustrating the steps for determining the selection confidence of each initial feature point matching pair in an embodiment of the present invention.

[0060] Figure 4 This is a flowchart illustrating the steps of correcting the matching index of each initial feature point matching pair according to an embodiment of the present invention.

[0061] Figure 5 This is a schematic diagram of the photovoltaic array electronic sand table construction device using drone aerial imagery, according to an embodiment of the present invention. Detailed Implementation

[0062] To clearly illustrate the technical features of this solution, the invention will be described in detail below through specific implementation methods and in conjunction with the accompanying drawings.

[0063] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the invention. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the invention.

[0064] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0065] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0066] It should be noted that the concepts of "first" and "second" mentioned in this invention are only used to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0067] Although operations or steps are described in a specific order in the accompanying drawings in the embodiments of the present invention, this should not be construed as requiring these operations or steps to be performed in the specific order or serial order shown, or requiring all of the shown operations or steps to be performed to obtain the desired result. In the embodiments of the present invention, these operations or steps may be performed serially; they may be performed in parallel; or a portion of these operations or steps may be performed.

[0068] Furthermore, it is understood that the data involved in the technical solutions of this invention (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions. Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains, and all parameters or indicators in the formulas involved in this invention are normalized values ​​that have eliminated the influence of dimensions.

[0069] The following will describe in detail, with reference to the accompanying drawings, a method for constructing a photovoltaic array electronic sand table using drone aerial imagery provided by an embodiment of the present invention.

[0070] Figure 1 This diagram illustrates the basic flowchart of a photovoltaic array electronic sand table construction method using drone aerial imagery provided by an embodiment of the present invention. Figure 1 As shown, the method specifically includes the following steps:

[0071] Step S100: Acquire drone aerial image data of the photovoltaic array, as well as wind intensity information and POS attitude information during the drone aerial photography process.

[0072] Electronic sand tables can deeply integrate geographic data, business information and three-dimensional scenes to achieve an immersive experience of "virtual simulation + real-time interaction". They are often used in photovoltaic array scenarios to realize real-time monitoring and simulation of photovoltaic array layout, operation status and power generation efficiency, and help photovoltaic power plants achieve "precise management, intelligent operation and maintenance and efficient power generation".

[0073] When constructing the electronic sand table of the photovoltaic array, drones were used to collect multi-angle, oblique aerial images of the photovoltaic array area, thus obtaining drone aerial image data of the photovoltaic array. Simultaneously, during the drone aerial photography process, wind intensity information and the drone's POS attitude information were collected at the drone's location. This allows for subsequent analysis of the wind intensity and POS attitude information at the drone's location, enabling preliminary screening of the drone aerial image data and avoiding interference from images acquired under conditions of high interference intensity.

[0074] Specifically, to acquire drone aerial imagery data of the photovoltaic array, as well as wind intensity and POS attitude information during the drone aerial photography process, the following steps are taken: First, a drone flight path is planned to ensure that the path covers the entire area of ​​the target photovoltaic array. Second, a set of multiple oblique aerial images of each photovoltaic array location is obtained using drone oblique photography technology, thus obtaining the drone aerial imagery data of the photovoltaic array. Simultaneously, a wind sensor module is installed on the drone. During the drone aerial photography process, the wind intensity information at the drone's location is read through the wind sensor module, and the POS attitude information of the drone, including pitch angle and fuselage vibration intensity, is read through the situational awareness module.

[0075] Step S200: Based on wind intensity information and POS attitude information, interference images are filtered out from the UAV aerial image data to obtain the target UAV aerial image data.

[0076] Considering the potential interference from wind and changes in drone attitude during tilted camera shooting, which can increase noise levels in aerial footage and consequently increase the analysis error of subsequent aerial images, this embodiment first filters the drone aerial image data based on the ambient wind strength and drone attitude changes during aerial shooting to remove interfering images, thereby obtaining target drone aerial image data under stable shooting conditions. For example, when the wind intensity is high and the POS attitude changes frequently, it indicates poorer shooting stability of the aerial images. In this case, such aerial images should be removed from the drone aerial image data to obtain the target drone aerial image data composed of the remaining images after filtering.

[0077] Step S300: Determine the initial feature point matching pairs and their matching indices in any two frames of aerial images to be matched in the target UAV aerial image data, and perform feature analysis on the local areas of the aerial images where the two feature points in each initial feature point matching pair are located, and determine the selection confidence of each initial feature point matching pair.

[0078] Among them, the confidence level is selected to reflect the probability that the initial feature point matching pair belongs to the real feature points of the image rather than noise points.

[0079] Feature points are extracted from the target UAV aerial imagery data, and image matching is performed based on the extracted feature points to determine initial feature point matching pairs in any two frames of aerial images to be matched, as well as the corresponding matching index for each initial feature point matching pair. This matching index is used to reflect the degree of matching between the two feature points in the initial feature point matching pair.

[0080] Specifically, the SIFT algorithm is used to extract feature points from the target UAV aerial imagery data and perform image matching to determine the initial feature point matching pairs and their matching indices for any two frames of the target UAV aerial imagery data to be matched. The matching indices refer to the Euclidean distance between the feature representations of the two feature points in each initial feature point matching pair. A larger Euclidean distance indicates a lower degree of matching between the two feature points in the corresponding initial feature point matching pair.

[0081] Feature points typically contain more unique and stable information in an image (such as corners and edges), and they usually possess both saliency and neighborhood correlation. Saliency refers to the significant changes in local grayscale or structure that feature points typically exhibit, which are represented by high amplitude and consistent direction in the gradient. Neighborhood correlation means that the gradient changes of feature points are not isolated; their neighboring pixels often contribute to the formation of certain structural patterns (such as edges, corners, and spots). For example, the gradient of a single feature point and the mean of its neighborhood gradients will significantly deviate from the mean of the entire region.

[0082] In aerial imagery, noise points are sometimes misidentified as valid feature points. Due to the randomness of noise, the position, scale, and orientation of such points are often unstable, leading to increased local texture differences between their center point and surrounding pixels. This, in turn, causes matching errors during feature matching, resulting in significant texture inconsistencies between the matched points corresponding to noise points and the images to be matched. Therefore, this embodiment performs feature analysis on the local regions of the aerial image containing the two feature points in each initial feature point matching pair to determine the selection confidence of each initial feature point matching pair. This is used to quantify the probability that the initial feature point matching pair belongs to a real feature point in the image rather than a noise point.

[0083] Step S400: Based on the selected confidence level, the matching index of each initial feature point matching pair is corrected, and based on the corrected matching index of each initial feature point matching pair, each initial feature point matching pair is filtered to obtain the final feature point matching pair.

[0084] A higher confidence level indicates that the feature points in the corresponding initial feature point matching pair are more likely to be true image feature points, and should be retained for the image matching process of the two aerial images to be matched. Therefore, based on the selected confidence level, the matching index of each initial feature point matching pair is corrected to obtain the corrected matching index of each initial feature point matching pair. Then, based on the corrected matching index of each initial feature point matching pair, the SIFT algorithm is used to filter each initial feature point matching pair to obtain the final feature point matching pair. This involves eliminating fuzzy matches using the nearest neighbor distance ratio (NNDR) (retaining matching pairs where the ratio of the nearest neighbor to the second nearest neighbor distance is less than a threshold), and then combining algorithms such as RANSAC to remove outliers, retaining interior points that conform to the geometric transformation model as the final feature point matching pair, ensuring the stability and accuracy of the matching. Since this implementation process is existing technology, it will not be elaborated further here.

[0085] Step S500: Based on the final feature point matching pairs, perform image matching on the target UAV aerial image data, and construct a photovoltaic array electronic sand table based on the image matching results.

[0086] Based on the final feature point matching pairs, the SIFT algorithm is used to perform image matching on the target UAV aerial imagery data, thereby obtaining the image matching results. Using the image matching results, a complete photovoltaic array electronic sand table is constructed. Specifically, firstly, the matched aerial imagery data is imported into an aerial triangulation module for alignment and correction. Then, the corrected aerial imagery data is combined with the terrain elevation data of each point to generate a photovoltaic array terrain model using GIS technology. Finally, information about each photovoltaic module, such as its location coordinates, function, and real-time operating status, is imported into the generated photovoltaic array terrain model, ultimately completing the construction of the photovoltaic array electronic sand table.

[0087] Based on the above technical solution, wind intensity information and POS attitude information obtained during drone aerial photography are used to initially screen drone aerial image data, thereby eliminating aerial images that are significantly affected by interference due to poor shooting environment stability. Then, for the remaining target drone aerial image data after screening, feature analysis is performed on the local regions of the aerial images containing the two feature points in each initial feature point matching pair from any two frames of images to be matched. This determines the selection confidence level of each initial feature point matching pair, reflecting the probability that the initial feature point matching pair belongs to a real feature point in the image rather than a noise point. Based on this selection confidence level, the matching index of each initial feature point matching pair is corrected, thereby achieving the screening of real feature point matching pairs in the image to ensure the accuracy of image matching and thus ensuring the accuracy of electronic sand table construction.

[0088] In one possible implementation, such as Figure 2As shown, in step S200 above, interference image removal is performed on the UAV aerial image data to obtain the target UAV aerial image data, including:

[0089] Step S201: Based on wind intensity information and POS attitude information, determine the aerial stabilization coefficient of each frame of aerial imagery in the UAV aerial imagery data.

[0090] During aerial photography of the target photovoltaic array area by a drone, wind interference may occur in the aerial photography environment, causing the drone to vibrate and reducing its attitude stability, thus increasing the probability of noise in the aerial images. Therefore, based on wind intensity information and POS attitude information, an aerial photography stability coefficient is determined for each frame of the drone aerial image data. This aerial photography stability coefficient is used to characterize the environmental stability during the capture of the corresponding aerial image.

[0091] In one possible implementation, the POS attitude information includes vibration intensity information and pitch angle information. Based on this, the aerial stabilization coefficient for each frame of the UAV aerial imagery data is determined, including:

[0092] First, based on the wind intensity corresponding to the shooting time of each frame of the UAV aerial image data in the wind intensity information and the vibration intensity in the POS attitude information, the aerial interference intensity at the shooting time of each frame of the UAV aerial image data is determined.

[0093] Specifically, for a single frame of drone aerial imagery, the wind intensity at the time the frame was captured is obtained. and vibration intensity Regarding the wind intensity and vibration intensity Perform maximum and minimum value normalization separately, and calculate the wind intensity after normalization. and vibration intensity product The product is then used as the aerial interference intensity at the time the aerial image was captured, and denoted as... The stronger the wind around the drone at the time the aerial image was captured, and the greater the vibration of the drone, the greater the interference with the aerial image, and the corresponding intensity of the aerial interference. The larger the value, the better.

[0094] Secondly, based on the fluctuation of the pitch angle corresponding to the shooting time of each frame of the UAV aerial image data in the pitch angle information of the local time period, the degree of change of the fuselage attitude at the shooting time of each frame of the UAV aerial image data is determined.

[0095] Specifically, due to the influence of drone aerial photography route planning or environmental wind, the drone's body attitude may change in the short term during tilted aerial photography. The more drastic the change in body attitude, the more interference it will cause to the accuracy of aerial image analysis. Therefore, for a single frame of drone aerial imagery, we obtain the pitch angle data of the drone within a 3-second time period adjacent to the time of the image's capture, and calculate the standard deviation of each pitch angle within this time period. Regarding the standard deviation Perform maximum and minimum value normalization, and then calculate the standard deviation after normalization. The degree of aircraft attitude change at the moment the aerial image was captured is recorded as follows: .

[0096] Finally, based on the intensity of aerial interference and the degree of change in the aircraft's attitude, the aerial stability coefficient of each frame of the UAV aerial image data is determined.

[0097] Specifically, for a single frame of drone aerial imagery, the intensity of aerial interference at the time the frame was captured can be calculated. and degree of change in fuselage attitude product and based on that product Determine the aerial stabilization coefficient of this aerial image frame and denote it as . At this time there is The smaller the standard deviation of the drone's body attitude change at the moment the aerial image is captured, the greater the degree of body attitude change. The smaller the value, the lower the aerial interference intensity at that moment, i.e., the lower the aerial interference intensity. The smaller the value, the better the stability of the aerial image at the time of capture; the corresponding aerial stability coefficient is... The larger the value, the better.

[0098] Step S202: Based on the aerial photography stability coefficient, filter the drone aerial photography data to obtain the filtered target drone aerial photography data.

[0099] Based on the aerial stability coefficient of each frame of the UAV aerial imagery data, the aerial images with lower stability coefficients are filtered out, and the remaining aerial images are used as the filtered UAV aerial imagery data, thereby achieving the filtering of UAV aerial imagery data.

[0100] For example, the aerial stability coefficients of all frames in the drone aerial image data can be arranged in descending order, and the aerial images with the lowest stability coefficients in the bottom 20% can be removed from the drone aerial image data. The remaining aerial images can then be used as the filtered drone aerial image data.

[0101] Based on the above technical solution, by analyzing the wind intensity information and POS attitude information during the aerial photography process of UAV aerial image data, the aerial photography stability coefficient of each frame of aerial image data is determined. This coefficient is used to quantify the aerial photography stability at the time of each frame of aerial image and to filter out aerial images with poor aerial photography stability, thereby avoiding interference from aerial images with poor aerial photography stability to the feature matching of subsequent aerial images.

[0102] In one possible implementation, step S202 involves filtering the UAV aerial imagery data based on the aerial photography stability coefficient to obtain the filtered target UAV aerial imagery data, including:

[0103] First, based on the shooting time of each aerial image in the drone aerial image data, several aerial shooting time periods are determined in the aerial shooting process.

[0104] The drone aerial photography process is divided into several aerial photography periods, each corresponding to a specific time of image capture. When performing feature matching on aerial images within the same aerial photography period, it is necessary to ensure that there are overlapping areas among the images being matched. Images captured closer to the center of the same aerial photography period have a higher probability of overlapping with the overall image set compared to images on either side. Therefore, it is necessary to analyze the centering characteristics of the capture times of aerial images within the same aerial photography period.

[0105] Specifically, based on the shooting time of each aerial image in the UAV aerial imagery data, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is used to cluster these shooting times to obtain multiple aerial shooting time periods. It should be understood that in other implementations, if the UAV captures images at equal intervals during the aerial photography process, i.e., the time interval between adjacent aerial images is the same, then 5 seconds can be used as one aerial shooting time period, thereby determining multiple aerial shooting time periods.

[0106] Secondly, the centering performance of each moment in the aerial photography period is determined based on the distance of each moment from the center moment.

[0107] Within the same aerial photography period, determine the center time of that period and the time interval between each moment in the aerial photography period and the center time. Calculate the time interval reciprocal Regarding the reciprocal Perform sum normalization (i.e., determine all time intervals) Calculate the reciprocal of the accumulated value. The ratio of the sum to the sum is used as the normalized value, and the normalized value is used as the centering performance at each time step, denoted as . It should be understood that, for the center moment of an aerial photography session, due to its corresponding time interval... If the value is 0, then the centering performance of that central moment is directly set to 1. The higher the centering performance of a certain moment in the aerial photography period, the better the temporal centering performance of the aerial image at that moment relative to other aerial images in that aerial photography period.

[0108] Next, based on the centering performance and aerial stability coefficient, the matching image confidence of each frame of aerial imagery in the UAV aerial imagery data is determined.

[0109] For a single frame of UAV aerial imagery, a higher aerial stability coefficient indicates lower noise interference and a higher centering performance, suggesting greater overlap with other images captured during the same time period during subsequent image matching. This results in higher confidence for the matched image. Specifically, the product of the centering performance and the aerial stability coefficient for each frame of the UAV aerial imagery is calculated, and this product is used as the confidence score for the matched image of each frame. .

[0110] Finally, based on the confidence level of the matched images, the drone aerial image data is filtered to obtain the filtered target drone aerial image data.

[0111] Specifically, the matching image confidence scores of all frames of drone aerial imagery data are arranged in descending order, and the aerial images with matching image confidence scores in the bottom 20% are removed from the drone aerial imagery data. The remaining aerial images are then used as the filtered drone aerial imagery data.

[0112] Based on the above technical solution, several aerial photography time periods are determined, and the centering performance at each moment within those time periods is assessed. Based on the aerial stability coefficient of each frame of the UAV aerial imagery data, and combined with the centering performance at the corresponding moment, the matching image confidence score is determined. Finally, this matching image confidence score is used to filter the UAV aerial imagery data, resulting in filtered UAV aerial imagery data. This process ensures that aerial images captured during periods of poor aerial stability are removed, while also ensuring that the subsequently filtered aerial images can be effectively matched for features.

[0113] In one possible implementation, such as Figure 3 As shown, step S300, determining the selection confidence level for each initial feature point matching pair, includes:

[0114] Step S301: Based on the gradient difference between the two feature points in each initial feature point matching pair and their neighboring pixels in the aerial image, determine the feature selection fit of the two feature points in each initial feature point matching pair.

[0115] For any feature point in each initial feature point matching pair, the more significant the gradient between the feature point and its neighboring pixels, the clearer the local edge of the feature point and the more stable its local structure. However, since the neighborhood gradient of noise points often exhibits low amplitude and random directions, lacking consistency, it usually cannot show significant gradient characteristics. In this case, it reflects that the feature point is more likely to be a real image feature rather than a noise point, and its corresponding feature selection fit should be higher.

[0116] In one possible implementation, the determination of the feature selection fit between the two feature points in each initial feature point matching pair in step S301 above includes:

[0117] First, determine the average gradient values ​​of the two feature points in each initial feature point matching pair and their eight neighboring pixels in the aerial image. Specifically, for any feature point in each initial feature point matching pair, determine the average gradient value of that feature point and its eight neighboring pixels in the aerial image. This average gradient value is the average of the gradient magnitudes and is denoted as . .

[0118] Secondly, the mean of the gradient mean is determined for all pixels in the local region of the aerial image containing the two feature points in each initial feature point matching pair, thus obtaining the gradient mean baseline value. Specifically, based on the pixel values ​​(i.e., grayscale values) of the pixels in the aerial image containing the two feature points in each initial feature point matching pair, the DIBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is used to cluster the pixels in the aerial image, resulting in multiple pixel clusters, each constituting a local region. For any feature point in each initial feature point matching pair, the mean of the gradient mean for all pixels in the local region of the aerial image containing that feature point is determined, and this mean is used as the gradient mean baseline value, denoted as . .

[0119] Next, based on the difference between the mean gradient of the two feature points in each initial feature point matching pair and the baseline value of the mean gradient, the gradient highlighting factor corresponding to the two feature points in each initial feature point matching pair is determined. Specifically, if the gradient level of the two feature points in each initial feature point matching pair and their neighborhood is larger than the level of other pixels in their local region, it indicates that the feature point is more preferred as a feature point, and the corresponding gradient highlighting factor is larger. Therefore, for any feature point in each initial feature point matching pair, its corresponding mean gradient is calculated. Compared with the gradient mean baseline value The difference And the difference was normalized using the Min-Max method. Normalization is performed to obtain the gradient highlighting factor corresponding to the feature point, and denoted as . .

[0120] Next, the distances from the two feature points in each initial feature point pair to the center of the image in the aerial image are determined. Ideally, during shooting, the image, the focal point, and the actual photovoltaic array scene should be in a straight line. However, in actual shooting, due to the influence of convex lenses within the camera, the further away from the center point in the image, the greater the distortion caused by the convex lens. In other words, pixels closer to the center of the image accurately reflect the location of the actual photovoltaic array scene. Therefore, it is necessary to use the distances from the feature points to the center of the image in the aerial image to suppress the gradient highlighting factor corresponding to feature points at the image edges. Specifically, for any feature point in each initial feature point pair, the distance from that feature point to the center of the image in the aerial image is determined and denoted as . .

[0121] Finally, based on the gradient highlighting factor and the distance from the two feature points in each initial feature point matching pair to the center of the image in the aerial image, the feature selection fit of the two feature points in each initial feature point matching pair is determined. Specifically, for any feature point in each initial feature point matching pair, based on its corresponding gradient highlighting factor... and its distance from the center of the aerial image. The feature selection fit of the feature point is determined by the following formula. :

[0122]

[0123] In the formula: This represents the correction parameter, used to control the gradient salience factor of feature points. The degree of suppression can be calibrated empirically, generally when using a conventional wide-angle camera for drone aerial photography. The value range is 0.8–1.2; The reference distance is half the diagonal length of the aerial image containing the feature point. The closer a feature point is to the center of its corresponding aerial image, and the larger its gradient salience factor, the higher its suitability as a true feature point.

[0124] Step S302: Based on the similarity between the grayscale differences between the two feature points in each initial feature point matching pair and the neighboring pixels in the aerial image, determine the local texture similarity between the two feature points in each initial feature point matching pair.

[0125] Noise points exhibit numerical randomness, which causes unstable changes in the texture representation of noise points and local pixels in aerial images. Specifically, the deviation between the center point of the noise point and the local pixels increases, leading to a greater texture difference between the noise point and its matching pixels. Therefore, analyzing the similarity between the grayscale differences of the two feature points in each initial feature point matching pair and their neighboring pixels in the aerial image determines the local texture similarity of the two feature points in each initial feature point matching pair. A higher local texture similarity indicates that the two feature points in the corresponding initial feature point matching pair are less likely to be noise points.

[0126] In one possible implementation, step S302, determining the local texture similarity between two feature points in each initial feature point matching pair, includes:

[0127] First, the grayscale difference between the two feature points in each initial feature point matching pair and the pixels in their local neighborhood in the aerial image is determined to obtain the image rendering deviation factor. Specifically, for any feature point in each initial feature point matching pair, its 5*5 neighborhood in the aerial image is determined as its local neighborhood, and the grayscale difference between the feature point and the pixels in its local neighborhood is calculated. This difference is recorded as the image rendering deviation factor.

[0128] Secondly, the image rendering deviation factors are arranged in a predetermined order to obtain the image rendering deviation factor sequence corresponding to the two feature points in each initial feature point matching pair. Specifically, for any feature point in each initial feature point matching pair, the image rendering deviation factors corresponding to the neighboring pixels in its local neighborhood are arranged in the order from top to bottom and from left to right, forming an image rendering deviation factor sequence.

[0129] Finally, the similarity between the image rendering bias factor sequences corresponding to the two feature points in each initial feature point matching pair is determined, thus obtaining the local texture similarity between the two feature points in each initial feature point matching pair. Specifically, for each initial feature point matching pair, the DTW distance (i.e., dynamic regularization distance) of the image rendering bias factor sequences corresponding to the two feature points is calculated and denoted as... Using the exponential function to measure the DTW distance Normalized values ​​are obtained by performing negative correlation normalization. The normalized value is used as the local texture similarity between the two feature points. When the DTW distance... The larger the value of , the lower the similarity between the graphs corresponding to the two feature points and the smaller the value of the local texture similarity.

[0130] Step S303: Based on feature selection fit and local texture similarity, determine the selection confidence of each initial feature point matching pair.

[0131] For each initial feature point matching pair, the higher the local texture similarity between the image presentation deviation factor sequences corresponding to the two feature points in the initial feature point matching pair, the smaller the local texture deviation of the two feature points in their respective aerial images. At the same time, the higher the feature selection fit of the two feature points, the more likely the two feature points are to be real feature points in the image rather than noise points, and the higher the selection confidence of the corresponding initial feature point matching pair.

[0132] In one possible implementation, step S303, determining the selection confidence of each initial feature point matching pair, includes:

[0133] First, based on feature selection fitness and local texture similarity, the optimal matching combination for each initial feature point pair is determined. Specifically, for each initial feature point pair, the average feature selection fitness of the two feature points in the pair is taken as the mean feature selection fitness. The product of this mean feature selection fitness and the local texture similarity of the two feature points in the pair is calculated, and this product is taken as the optimal matching combination for the initial feature point pair, denoted as . .

[0134] Secondly, the mean of the matching combination optimization degree of all initial feature point matching pairs is determined to obtain the average matching combination optimization degree. Specifically, the average value of the matching combination optimization degree of all initial feature point matching pairs is determined as the average matching combination optimization degree, and denoted as . .

[0135] Finally, based on the difference between the matching combination optimization degree of each initial feature point matching pair and the average matching combination optimization degree, the selection confidence level of each initial feature point matching pair is determined. Specifically, for any initial feature point matching pair, if the matching combination optimization degree of this initial feature point matching pair is higher than that of other initial feature point matching pairs, it indicates that the feature points in this initial feature point matching pair are more likely to be real feature points in the image, and they are more likely to be used preferentially in the matching process of the corresponding two frames of aerial images to be matched. Therefore, the matching combination optimization degree of each initial feature point matching pair is calculated. Best matching combination with average The difference Regarding this difference Perform maximum and minimum value normalization, and use the normalized value as the selection confidence score for the initial feature point matching pairs, denoted as . .

[0136] Based on the above technical solution, the feature selection fit of the two feature points in each initial feature point matching pair is determined by analyzing the gradient difference between the two feature points and their neighboring pixels in the aerial image. Simultaneously, the similarity between the grayscale differences of the two feature points in each initial feature point matching pair and their neighboring pixels in the aerial image is quantified to determine the local texture similarity of the two feature points in each initial feature point matching pair. Finally, the feature selection fit and local texture similarity of the two feature points in each initial feature point matching pair are fused to determine the selection confidence of each initial feature point matching pair, which reflects the probability that the initial feature point matching pair belongs to a real feature point in the image rather than a noise point.

[0137] In one possible implementation, such as Figure 4 As shown, in step S400, the matching index of each initial feature point matching pair is corrected based on the selected confidence level, including:

[0138] Step S401: Based on the matching index and selection confidence of each initial feature point matching pair, determine the weighted matching index of each initial feature point matching pair.

[0139] In the process of extracting feature points and performing image matching on target UAV aerial image data using the SIFT algorithm, the matching index for each initial feature point matching pair is the Euclidean distance between the feature representations of the two feature points in each initial feature point matching pair. A larger Euclidean distance indicates a lower degree of matching between the two feature points in the corresponding initial feature point matching pair. Based on this, the confidence level for selecting each initial feature point matching pair is determined. Perform negative correlation mapping processing, that is, calculate the selection confidence of the matching pairs between the value 1 and each initial feature point. The difference To select confidence levels Perform negative correlation mapping and assign the difference The matching indices of each initial feature point matching pair are multiplied together as weights, and the multiplied value is used as the weighted matching index of each initial feature point matching pair.

[0140] Step S402: Map the weighted matching index of each initial feature point matching pair to the value range of the matching index of each initial feature point matching pair, thereby obtaining the corrected matching index of each initial feature point matching pair.

[0141] To ensure that the range of values ​​for the matching index of each initial feature point pair remains unchanged before and after correction, the range of values ​​for the matching index of each initial feature point pair before correction is determined. Then, linear normalization (Min-MaxScaling) is used to map the weighted matching index of each initial feature point pair to this range. The larger the weighted matching index, the larger the value after mapping, thus obtaining the corrected matching index of each initial feature point pair.

[0142] Based on the above technical solution, by using the selection confidence of each initial feature point matching pair to weight its matching index, and mapping the obtained weighted matching index to the value range of the matching index of each initial feature point matching pair, the robustness of the matching process and the algorithm compatibility can be effectively ensured.

[0143] Based on the same inventive concept, embodiments of the present invention also provide a photovoltaic array electronic sand table construction device utilizing drone aerial imagery, such as... Figure 5 As shown, the device includes:

[0144] The data acquisition module is used to filter out interfering images from the drone aerial image data based on wind intensity information and POS attitude information, so as to obtain the target drone aerial image data.

[0145] The confidence analysis module is selected to determine the initial feature point matching pairs and their matching indices in any two frames of aerial images to be matched in the target UAV aerial image data. Feature analysis is performed on the local area of ​​the aerial image where the two feature points in each initial feature point matching pair are located to determine the selection confidence of each initial feature point matching pair. The selection confidence is used to reflect the possibility that the initial feature point matching pair belongs to the real feature points of the image rather than noise points.

[0146] The feature point matching pair filtering module is used to correct the matching index of each initial feature point matching pair based on the selected confidence level, and to filter each initial feature point matching pair based on the corrected matching index to obtain the final feature point matching pair.

[0147] The electronic sand table construction module performs image matching on the target UAV aerial image data based on the final feature point matching pairs, and constructs a photovoltaic array electronic sand table based on the image matching results.

[0148] It should be noted that the device provided in the above embodiments is only an example of the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above.

[0149] Based on the same inventive concept, embodiments of the present invention also provide a photovoltaic array electronic sand table construction system using drone aerial imagery. The system includes: a memory, a processor, and computer program code stored in the memory and running on the processor. When the processor executes the computer program code, the system can execute any of the aforementioned methods for constructing a photovoltaic array electronic sand table using drone aerial imagery.

[0150] In this embodiment of the invention, the system can be divided into functional modules according to the above method example. For example, each module can correspond to a separate functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0151] Based on the same inventive concept, embodiments of the present invention also provide a computer program product, which includes: computer program code, which, when run on a computer, causes the computer to execute any of the aforementioned methods for constructing a photovoltaic array electronic sand table using drone aerial imagery.

[0152] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing computer program code, which, when run on a computer, causes the computer to execute any of the aforementioned methods for constructing a photovoltaic array electronic sand table using drone aerial imagery.

[0153] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for constructing a photovoltaic array electronic sand table using drone aerial imagery, characterized in that, Includes the following steps: Acquire drone aerial imagery data of the photovoltaic array, as well as wind intensity and POS attitude information during the drone aerial photography process; Based on the wind intensity information and POS attitude information, interference image filtering is performed on the UAV aerial image data to obtain the target UAV aerial image data. Determine the initial feature point matching pairs and their matching indices in any two frames of aerial images to be matched in the target UAV aerial image data, and perform feature analysis on the local areas of the aerial images where the two feature points in each initial feature point matching pair are located, and determine the selection confidence of each initial feature point matching pair. The selection confidence is used to reflect the possibility that the initial feature point matching pair belongs to the real feature points of the image rather than noise points. Based on the selected confidence level, the matching index of each initial feature point matching pair is corrected, and based on the corrected matching index of each initial feature point matching pair, each initial feature point matching pair is filtered to obtain the final feature point matching pair. Based on the final feature point matching pairs, image matching is performed on the target UAV aerial image data, and a photovoltaic array electronic sand table is constructed based on the image matching results.

2. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 1, characterized in that, Interference image removal is performed on the drone aerial image data to obtain target drone aerial image data, including: Based on the wind intensity information and POS attitude information, the aerial photography stability coefficient of each frame of aerial photography data in the UAV aerial photography data is determined. Based on the aerial photography stability coefficient, the drone aerial photography data is filtered to obtain the filtered target drone aerial photography data.

3. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 2, characterized in that, The POS attitude information includes vibration intensity information and pitch angle information; the aerial stability coefficient of each frame of aerial imagery in the UAV aerial imagery data is determined, including: Based on the wind intensity corresponding to the shooting time of each frame of the UAV aerial image data in the wind intensity information, and the vibration intensity in the POS attitude information, the aerial interference intensity at the shooting time of each frame of the UAV aerial image data is determined. Based on the fluctuation of the pitch angle corresponding to the pitch angle information in the pitch angle information during the local time period in which each frame of the UAV aerial image data is captured, the degree of change in the fuselage attitude at the time of capture of each frame of the UAV aerial image data is determined. Based on the intensity of aerial interference and the degree of change in the aircraft's attitude, the aerial stability coefficient of each frame of aerial imagery in the UAV aerial imagery data is determined.

4. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 2, characterized in that, Based on the aerial photography stability coefficient, the UAV aerial photography data is filtered to obtain the filtered target UAV aerial photography data, including: Based on the shooting time of each aerial image in the UAV aerial image data, several aerial shooting time periods are determined in the aerial shooting process; Based on the distance from the center time at each moment in the aerial photography period, the centering performance at each moment in the aerial photography period is determined; Based on the centering performance and aerial photography stability coefficient, the matching image confidence of each frame of aerial photography in the UAV aerial photography data is determined. Based on the confidence level of the matched images, the drone aerial image data is filtered to obtain the filtered target drone aerial image data.

5. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 1, characterized in that, Determine the selection confidence level for each initial feature point matching pair, including: Based on the gradient difference between the two feature points in each initial feature point matching pair and their neighboring pixels in the aerial image, the feature selection fit degree of the two feature points in each initial feature point matching pair is determined. Based on the similarity between the grayscale differences between the two feature points in each initial feature point matching pair and the neighboring pixels in the aerial image, the local texture similarity between the two feature points in each initial feature point matching pair is determined. Based on the feature selection fit and the local texture similarity, the selection confidence of each initial feature point matching pair is determined.

6. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 5, characterized in that, Determine the feature selection fit between the two feature points in each initial feature point matching pair, including: Determine the average gradient of the two feature points in each initial feature point matching pair with the eight neighboring pixels in the aerial image. Determine the mean value of the gradient mean for all pixels in the local region of the aerial image where the two feature points in each initial feature point matching pair are located, and obtain the baseline value of the gradient mean. Based on the difference between the mean gradient value of the two feature points in each initial feature point matching pair and the baseline value of the mean gradient value, the gradient highlighting factor corresponding to the two feature points in each initial feature point matching pair is determined. Determine the distance from the center of the aerial image to each initial feature point matching pair; Based on the gradient highlighting factor and the distance from the two feature points in each initial feature point matching pair to the center of the image in the aerial image, the feature selection fit degree of the two feature points in each initial feature point matching pair is determined.

7. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 5, characterized in that, Determine the local texture similarity between two feature points in each initial feature point matching pair, including: The grayscale difference between the two feature points in each initial feature point matching pair and the neighboring pixels in the local neighborhood of the aerial image is determined to obtain the image presentation deviation factor. The graph rendering deviation factors are arranged in a set order to obtain the graph rendering deviation factor sequence corresponding to the two feature points in each initial feature point matching pair. The similarity between the graph presentation deviation factor sequences corresponding to the two feature points in each initial feature point matching pair is determined, and the local texture similarity between the two feature points in each initial feature point matching pair is obtained.

8. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 5, characterized in that, Determine the selection confidence level for each initial feature point matching pair, including: Based on the feature selection fitness and the local texture similarity, the matching combination optimization degree of each initial feature point matching pair is determined; Determine the mean of the matching combination optimization of all initial feature point matching pairs to obtain the average matching combination optimization. Based on the difference between the matching combination preference of each initial feature point matching pair and the average matching combination preference, the selection confidence of each initial feature point matching pair is determined.

9. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 1, characterized in that, The SIFT algorithm is used to determine the initial feature point matching pairs and their matching indices in any two frames of aerial images to be matched in the target UAV aerial image data. The matching index refers to the Euclidean distance between the feature descriptors of the two feature points in each initial feature point matching pair.

10. The method for constructing a photovoltaic array electronic sand table using UAV aerial imagery according to claim 9, characterized in that, Based on the selected confidence level, the matching index of each initial feature point matching pair is corrected, including: Based on the matching index of each initial feature point matching pair and the selected confidence level, the weighted matching index of each initial feature point matching pair is determined. The weighted matching index of each initial feature point matching pair is mapped to the value range of the matching index of each initial feature point matching pair, thereby obtaining the modified matching index of each initial feature point matching pair.

Images