An intelligent recovery path planning method for green view recovery

Abstract

The present application relates to the technical field of green view rate supplementing, and particularly relates to an intelligent supplementing path planning method for green view rate supplementing, comprising: dividing a sampling complexity category of a to-be-sampled area; determining a preset green view rate mutation threshold for identifying missing data in each initial image; determining whether to correct a data update rate of GNSS RTK based on an aggregation degree of the missing data; determining a supplementing point based on the missing data, generating a supplementing path based on coordinates corresponding to each supplementing point identified; supplementing each supplementing point by using a UAV, determining an anchor green view rate difference based on supplementing images acquired in real time, determining whether supplementing of the supplementing point is qualified based on the anchor green view rate difference, supplementing a next supplementing point when supplementing of the supplementing point is qualified; and fusing each supplementing image and each initial image to generate a green view rate panorama. When data is missing, sampling parameters are determined, and the supplementing efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of green visibility rate supplementation technology, and in particular to an intelligent supplementation path planning method for green visibility rate supplementation. Background Technology

[0002] Green view rate, a key indicator for measuring green space in three dimensions, has undergone a transformation in its survey technology from manual to intelligent methods. Traditional manual survey methods mainly rely on on-site photography and visual interpretation, which suffers from low efficiency, strong subjectivity, and limited spatiotemporal coverage. In recent years, with the development of remote sensing and mobile measurement technologies, green view rate surveys have gradually entered the automation stage. The current technical approach utilizes vehicle-mounted mobile measurement systems, which use cameras, GPS / IMUs, and other sensors mounted on vehicles to achieve continuous data collection along the road network. This system improves data collection efficiency, but is limited by fixed driving routes and viewing angles, making it difficult to obtain green information in obscured areas.

[0003] Chinese Patent Publication No. CN120913058A discloses a method and system for determining green visibility rate based on a vehicle-mounted system. Utilizing multi-source data, it enhances the model's sensitivity to vegetation through channel-spatial dual attention and pyramid pooling during the prediction of two-dimensional green visibility rate. It accelerates convergence and improves segmentation edge accuracy by combining intermediate loss and model-specific loss. It also uses three-dimensional data to suppress two-dimensional segmentation misjudgments and improve robustness. The decoder is designed with a lightweight approach to balance accuracy and efficiency, adapting to vehicle-mounted edge computing. Furthermore, it uses an energy function to further refine the segmentation results and reduce noise. However, the above technical solution has the following drawback: it does not consider determining sampling parameters when data is missing, affecting the efficiency of data re-sampling. Summary of the Invention

[0004] To address this issue, the present invention provides an intelligent supplementary sampling path planning method for green visibility rate supplementary sampling, which overcomes the problem in the prior art that it does not consider determining sampling parameters when data is missing, thus affecting the supplementary sampling efficiency.

[0005] To achieve the above objectives, the present invention provides an intelligent supplementary sampling path planning method for green visibility rate supplementary sampling, comprising:

[0006] The sampling complexity category of the area to be sampled is determined based on the sampling complexity characterization value of the area to be sampled.

[0007] Based on the sampling complexity category, a preset green visibility mutation threshold is determined to identify missing data in the initial image of each frame;

[0008] Determine whether to adjust the GNSS RTK data update rate based on the clustering degree of missing data;

[0009] Based on the missing data, the supplementary sampling points are determined, and the supplementary sampling path is generated based on the coordinates corresponding to each identified supplementary sampling point.

[0010] By using drones to conduct supplementary image collection at each point, the difference in anchored green visibility is determined based on the real-time acquired supplementary images. Based on this difference in anchored green visibility, the suitability of the supplementary image collection at each point is determined, including:

[0011] Identify anomalies in the supplementary data collection at the supplementary data collection points, and correct the sampling parameters based on the differences in the horizontal comparison, including correcting the number of supplementary data collection points for a single missing data point or correcting the data update rate of GNSS RTK used to determine the location of the UAV.

[0012] Alternatively, if the supplementary sampling at the designated sampling point is deemed satisfactory, supplementary sampling can be carried out at the next sampling point.

[0013] The suitability of each supplementary sampling point is determined based on the difference in the anchored green visibility rate corresponding to each supplementary sampling point, until the supplementary sampling of each supplementary sampling point is completed.

[0014] Furthermore, the process of determining the complex characterization values ​​of the area to be sampled includes:

[0015] The obstacle change characterization value is determined based on the number of buildings at several time points in the area to be mined and the time interval between two time points.

[0016] The clustering influence characterization value is determined based on the total area of ​​each multi-person clustering area captured by the sampling vehicle and the total area of ​​the area to be sampled during the process of identifying and acquiring each frame of the initial image.

[0017] The protrusion influence coefficient is determined based on the obtained 3D model of the area to be mined;

[0018] The dynamic influence coefficient is obtained by summing the corresponding coefficients assigned to the obstacle change characterization value and the clustering influence characterization value;

[0019] The product of the dynamic influence coefficient and the convexity influence coefficient is calculated to obtain the sampled complex characterization value.

[0020] Furthermore, the process of classifying the sampling complexity category of the region to be sampled based on the sampling complexity characterization value of the region to be sampled includes:

[0021] If the sampled complexity value is less than or equal to the preset sampled complexity value, the area to be sampled is determined as a weakly sampled complexity category.

[0022] If the sampled complexity value is greater than the preset sampled complexity value, the area to be sampled will be classified as a strongly sampled complexity category.

[0023] Furthermore, based on the sampling complexity category, a preset green visual rate mutation threshold is determined to identify missing data in the initial images of each frame, including:

[0024] When the area to be sampled is determined to be a weakly sampled complex category, the initial preset green visibility rate mutation threshold is continuously used to identify missing data.

[0025] When the area to be sampled is determined to be a highly complex category, the preset green visibility threshold is adjusted to the corresponding value based on the complex sampling characteristic value.

[0026] Furthermore, based on the sampled complex characterization values, the preset green visibility rate mutation threshold is adjusted to the corresponding value, wherein,

[0027] The reduction in the preset green visibility threshold is positively correlated with the sampled complex characterization value.

[0028] Furthermore, the process of determining whether to adjust the GNSS RTK data update rate based on the clustering degree of missing data includes:

[0029] The clustering degree of missing data is determined based on the distance between the coordinates of each adjacent missing data point.

[0030] When the clustering degree of missing data is greater than the preset clustering degree of missing data, the data update rate of GNSS RTK is adjusted to the corresponding value based on the clustering degree of missing data.

[0031] Furthermore, the GNSS RTK data update rate is adjusted to a corresponding value based on the clustering degree of missing data, wherein,

[0032] The increase in GNSS RTK data update rate is positively correlated with the degree of clustering of missing data.

[0033] Furthermore, the process of determining the anchored green visibility difference based on the real-time acquired supplementary images, and determining whether the supplementary images at the supplementary points are qualified based on the anchored green visibility difference, includes:

[0034] The anchored green view rate difference is determined based on the green view rate of the acquired current supplementary image;

[0035] When the difference in the anchored green view rate is greater than the preset green view rate mutation threshold, the abnormality of the supplementary sampling point is identified, and the sampling parameters are corrected based on the difference value of the horizontal comparison.

[0036] Furthermore, the process of correcting the sampling parameters based on the differences between the two sides includes:

[0037] The horizontal comparison difference value is determined based on the green view rate of the currently acquired image and the green view rate of the corresponding missing data;

[0038] When the horizontal comparison difference value is less than or equal to the preset horizontal comparison difference value, the number of supplementary collection points for a single missing data will be adjusted to the corresponding value based on the horizontal comparison difference value.

[0039] When the difference value of the horizontal comparison is greater than the preset difference value of the horizontal comparison, the data update rate of GNSS RTK will be adjusted to the corresponding value.

[0040] Furthermore, based on the horizontal comparison difference values, the number of supplementary data points for a single missing data point is adjusted to the corresponding value, wherein,

[0041] The increase in the number of missing data points is negatively correlated with the difference value of cross-sectional comparison.

[0042] Compared with existing technologies, the beneficial effects of this invention are that the complexity of the area to be sampled affects the collection and processing of green visibility data. By comprehensively considering factors such as obstacle changes, population gathering, and terrain protrusion, a sampling complexity characterization value is calculated to classify sampling complexity categories, providing a basis for subsequent parameter adjustment and processing. The obstacle change rate reflects the speed at which the number of buildings in the area to be sampled changes over time, reflecting the dynamic changes of the area. The obstacle change characterization value compares the current obstacle change rate of the area to be sampled with the historical average to measure the relative degree of obstacle change in the area. The clustering influence rate quantifies the degree of influence of population gathering on data collection. The clustering influence characterization value compares the current clustering influence rate of the area to be sampled with the historical average to measure the relative degree of population gathering in the area. The convex projected area reflects the protrusion of the terrain. The protrusion influence coefficient quantifies the degree of influence of terrain protrusion on data collection. The dynamic influence coefficient comprehensively considers the impact of obstacle changes and population gathering on data collection. The sampling complexity characterization value comprehensively reflects the overall complexity of the area to be sampled. Areas with different levels of complexity will encounter different problems during the data collection process. By dividing the sampling into complex categories, different processing strategies can be adopted for different categories of areas, thereby improving the accuracy and efficiency of data collection and increasing the efficiency of supplementary sampling.

[0043] Furthermore, a preset green visual rate mutation threshold is determined based on the sampling complexity category to identify missing data in the initial images of each frame. The stability and variation of data differ across regions with different sampling complexity categories. In regions with weak sampling complexity, the data is relatively stable, and the initial preset green visual rate mutation threshold is sufficient for identification. In regions with higher complexity, the green visual rate fluctuates significantly; adjusting the preset green visual rate mutation threshold allows for more sensitive detection of data changes, improving the accuracy of missing data identification. Dynamically adjusting the preset green visual rate mutation threshold according to the sampling complexity category more accurately identifies missing data in regions of varying complexity, improving data processing accuracy and increasing re-sampling efficiency.

[0044] Furthermore, in areas with higher complexity, green visibility changes more frequently and drastically. Lowering the preset green visibility mutation threshold allows for more sensitive capture of these changes, leading to more accurate identification of missing data. Higher sampling complexity values ​​indicate a more complex region and greater green visibility changes; in this case, a more significant reduction in the preset green visibility mutation threshold further enhances identification accuracy. By dynamically adjusting the identification criteria based on a comprehensive consideration of various factors, the accuracy and efficiency of green visibility re-collection are improved.

[0045] Furthermore, the decision to adjust the GNSS RTK data update rate is based on the clustering degree of missing data, which reflects the degree of clustering of missing data. The GNSS RTK data update rate is used to determine the frequency of location data updates for the UAV's position. When the clustering degree of missing data is less than or equal to a preset value, it indicates that the missing data is relatively concentrated, with concentrated obstruction in some areas. Supplementing data in these concentrated areas is sufficient, and the current GNSS RTK data update rate meets the requirements. When the clustering degree of missing data is greater than the preset value, it indicates that the missing data is scattered, with special conditions such as special terrain, special weather, or complex vegetation communities interfering with data collection around the scattered points. In this case, increasing the data update rate obtains more accurate location information to ensure that the UAV can accurately reach the supplementary data collection points. Dynamically adjusting the GNSS RTK data update rate based on the clustering degree of missing data improves the accuracy of supplementary data collection, avoids errors caused by inaccurate positioning, and further improves supplementary data collection efficiency.

[0046] Furthermore, the clustering degree of missing data reflects the concentration of data collection anomalies. A high clustering degree of missing data means that the anomaly data is more scattered and the regional terrain is more complex. Improving the data update rate of GNSS RTK to obtain more frequent and accurate positioning information will allow UAVs to reach the re-collection points more accurately and improve the accuracy of re-collection.

[0047] Furthermore, the anchored green visibility difference is determined based on the real-time acquired supplementary images. This difference is then used to determine whether the supplementary sampling at the corresponding point is satisfactory. The anchored green visibility difference is the absolute value of the difference between the green visibility of the current supplementary image and the green visibility of the initial image in the previous frame corresponding to the missing data, reflecting the degree of difference between the supplementary data and the original continuous data. The purpose of supplementary sampling is to fill in missing data, making the data more complete and accurate. By comparing the green visibility difference, the consistency between the supplementary data and the original data can be intuitively judged. If the difference is too large, it is determined that other factors have interfered with the supplementary sampling process, and the sampling parameters are further adjusted.

[0048] Furthermore, sampling parameters are corrected based on the horizontal comparison difference value. The horizontal comparison difference value is the ratio of the absolute value of the difference between the green view rate of the currently supplemented image and the green view rate of the corresponding missing data to the green view rate of the corresponding missing data, reflecting the relative difference between the supplemented data and the missing data. When the horizontal comparison difference value is small, the same problem still occurs due to the abnormality of the determined supplemented sampling point, resulting in the unqualified image. In this case, the number of sampling points is increased to perform multiple samplings. When the difference value is large, the re-acquired image differs from the original image, but the re-acquired image is still unqualified. In this case, it is determined that the actual location of the supplemented sampling point is deviated due to inaccurate UAV positioning. In this case, the data update rate of GNSS RTK is increased to improve positioning accuracy, further improving the efficiency of green view rate data supplementation. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating the steps of the intelligent supplementary sampling path planning method for green visibility supplementary sampling according to an embodiment of the present invention;

[0050] Figure 2 This is a logic decision diagram for classifying the sampling complexity category of the sampling area based on the sampling complexity characterization value of the sampling area in an embodiment of the present invention.

[0051] Figure 3 This is a logic determination diagram for an embodiment of the present invention, which determines a preset green visual rate mutation threshold for identifying missing data in the initial image of each frame based on a complex category of sampling.

[0052] Figure 4 This is a logic diagram illustrating how an embodiment of the present invention determines whether to correct the data update rate of a GNSS RTK based on the degree of clustering of missing data. Detailed Implementation

[0053] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0054] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0055] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0056] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0057] Please see Figure 1 The diagram shows a flowchart of the intelligent supplementary sampling path planning method for green view rate supplementary sampling according to an embodiment of the present invention. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to the present invention includes:

[0058] S1, classify the sampling complexity category of the area to be sampled based on the sampling complexity characterization value of the area to be sampled;

[0059] S2, based on the sampling complexity category, determine the preset green visibility mutation threshold for identifying missing data in the initial image of each frame;

[0060] S3, determine whether to adjust the GNSS RTK data update rate based on the clustering degree of missing data;

[0061] S4, determine the supplementary sampling points based on the missing data, and generate the supplementary sampling path based on the coordinates corresponding to each identified supplementary sampling point;

[0062] S5 uses drones to perform supplementary image collection at each point, determines the anchored green visibility difference based on the real-time acquired supplementary images, and determines whether the supplementary image collection at each point is qualified based on the anchored green visibility difference, including:

[0063] Identify anomalies in the supplementary data collection at the supplementary data collection points, and correct the sampling parameters based on the differences in the horizontal comparison, including correcting the number of supplementary data collection points for a single missing data point or correcting the data update rate of GNSS RTK used to determine the location of the UAV.

[0064] Alternatively, if the supplementary sampling at the designated sampling point is deemed satisfactory, supplementary sampling can be carried out at the next sampling point.

[0065] S6, repeat S5 until the supplementary sampling of each sampling point is completed;

[0066] S7, fuse each supplementary image with each initial image to generate a panoramic view of green visibility.

[0067] Specifically, the process of identifying missing data in each frame of the initial image based on a preset green visibility rate mutation threshold includes:

[0068] The system receives initial images and GNSS / IMU positioning and attitude data uploaded by the vehicle-mounted acquisition unit in real time via a wireless transmission network; the coordinates of the initial image of a single frame are determined by GNSS, and the angle and height of the initial image of a single frame are determined by IMU.

[0069] The semantic segmentation model identifies green vegetation pixels in each frame of the initial image and generates a preliminary green visibility rate; the rule engine then determines whether there is missing data in a single set of initial images based on the green visibility rate mutation values ​​of consecutive frames.

[0070] When the green view rate mutation value in consecutive frames exceeds the preset green view rate mutation threshold, the initial image of the next frame is identified as missing data, and a geospatial list of missing data is generated based on each missing data.

[0071] A geospatial list of missing data, including the coordinates, altitude, and angle of each missing data point.

[0072] The absolute value of the difference between the green visual rate of the current frame and the green visual rate of the previous frame in a continuous frame is calculated to obtain the green visual rate mutation value of the continuous frames.

[0073] Specifically, the process of generating a supplementary sampling path based on the coordinates corresponding to each identified supplementary sampling point includes:

[0074] Obtain the set of geographic coordinates for the missing data; each missing data point has a corresponding data collection point.

[0075] Obtain the distribution location of each drone base station, the status of each drone (idle / busy / battery), and the coverage radius of each drone base station. The drone status includes idle, busy, and battery status.

[0076] Obtain airspace constraint data for urban geographic information, including no-fly zones and height-restricted zones.

[0077] Based on the spatial distribution of the supplementary sampling points and the relevant data of the UAV base stations, the nearest neighbor clustering algorithm is used to allocate the supplementary sampling points to the corresponding idle base stations so as to carry out supplementary sampling at the supplementary sampling points;

[0078] For each supplementary data collection point corresponding to a single base station, a hybrid path planning algorithm is run, including global optimization using ant colony algorithm to determine the approximate shortest path to all supplementary data collection points; Dijkstra algorithm to ensure that the UAV can avoid real-time obstacles when flying between adjacent supplementary data collection points; and Dijkstra algorithm to access urban dynamic airspace information to determine the flight path of each UAV through the hybrid path planning algorithm.

[0079] Specifically, the process of conducting supplementary sampling at each sampling point includes:

[0080] Based on the coordinates, height, and angle of each missing data point, the preset shooting parameters for each supplementary data point are determined, including latitude and longitude, altitude, and preset camera pitch and yaw angles for each supplementary data point. The preset shooting parameters are then encapsulated into task instructions and sent to the corresponding UAV base station.

[0081] When the drone travels to the re-collection point via GNSS RTK, the drone's position and attitude are perceived in real time through cameras and IMU;

[0082] The camera's orientation is dynamically fine-tuned to ensure that the camera's optical axis is precisely and vertically aligned with the facade of the point to be sampled, so as to obtain an image that can be naturally connected with the vehicle's data perspective.

[0083] Specifically, the process of fusing each supplementary image with each initial image to generate a panoramic view of green visibility includes:

[0084] The sequential images acquired by the vehicle and the oblique images acquired by the UAV are input into the SfM (Structure from Motion) algorithm. The position and attitude information annotated in each image is determined as the strong constraint initial value of the SfM algorithm. The unified three-dimensional point cloud model is matched and reconstructed through the SfM algorithm.

[0085] Render the 3D point cloud model to obtain a panoramic view of green visibility.

[0086] In this embodiment, optionally,

[0087] Vehicle-mounted image sequences and UAV oblique image sequences are input into the SfM (Structure from Motion) algorithm. GNSS / IMU data (coordinates, altitude, angle) from each image are used as strong initial constraints for the SfM algorithm, directly initializing camera extrinsic parameters to reduce feature matching ambiguity. Feature point matching and bundle adjustment are performed using the SfM algorithm to reconstruct a unified 3D point cloud model. A Poisson surface reconstruction algorithm is used to generate a 3D mesh, and image textures are attached to the mesh surface via texture mapping to generate a green view panoramic image; this is existing technology and will not be elaborated further.

[0088] Specifically, in this embodiment, the initial preset threshold for green view rate mutation is optionally set to 0.15, that is, a 15% change in green view rate, which can be determined based on the statistical quantile of green view rate fluctuations in historical data.

[0089] Specifically, the process of determining the complex characterization values ​​of the area to be sampled includes:

[0090] Obtain the number of buildings at several time points within the area to be surveyed;

[0091] The rate of change of obstacles is obtained by calculating the ratio of the absolute value of the difference in the number of buildings between two time points to the time interval between the two time points.

[0092] The obstacle change rate is calculated as the ratio of the current obstacle change rate in the area to be mined to the average obstacle change rate in the historical data of each area to be mined, and the obstacle change characterization value is obtained.

[0093] The total area of ​​each multi-person gathering area captured by the sampling vehicle during the acquisition of each frame of the initial image is identified to obtain the gathering influence area.

[0094] The ratio of the area affected by aggregation to the total area of ​​the area to be mined is used to obtain the aggregation impact rate.

[0095] Calculate the ratio of the current clustering impact rate of the area to be mined to the average clustering impact rate of each area to be mined in the historical data to obtain the clustering impact characterization value;

[0096] Obtain a 3D model of the area to be mined, determine the area of ​​the region in the area with an elevation higher than the average elevation of the area to be mined, projected onto the sea level, and obtain the convex projection area.

[0097] The ratio of the projected area of ​​the protrusion to the total area of ​​the area to be mined projected onto the sea level is used to obtain the protrusion influence coefficient.

[0098] The dynamic influence coefficient is obtained by summing the corresponding coefficients assigned to the obstacle change characterization value and the clustering influence characterization value;

[0099] The product of the dynamic influence coefficient and the convexity influence coefficient is calculated to obtain the sampled complex characterization value.

[0100] Specifically, the coefficients corresponding to the obstacle change characterization value and the clustering impact characterization value are both 0.5 to comprehensively determine the impact of building changes and personnel mobility on supplementary sampling, and the assignment method is multiplication.

[0101] Specifically, the method for identifying the total area of ​​each multi-person gathering area captured by the sampling vehicle during the acquisition of the initial images of each frame is not limited. Images can be captured in real time by an onboard camera, and a pre-trained crowd detection model can be used to identify pedestrian bounding boxes in the images. The area of ​​each bounding box is calculated by the number of pixels, and the pixel coordinates are converted into real-world coordinates in square meters using camera calibration parameters and GNSS / IMU data. The areas of the multi-person gathering areas detected in all frames are summed to obtain the area of ​​the gathering influence. This is existing technology and will not be elaborated further.

[0102] The complexity of the area to be sampled affects the collection and processing of green view rate data. By comprehensively considering factors such as obstacle changes, population gathering, and terrain protrusion, a sampling complexity characterization value is calculated to classify sampling complexity categories, providing a basis for subsequent parameter adjustments and processing. The obstacle change rate reflects the speed at which the number of buildings in the area changes over time, reflecting the dynamic changes of the area. The obstacle change characterization value compares the current obstacle change rate of the area to be sampled with the historical average, measuring the relative degree of obstacle change in that area. The clustering impact rate quantifies the degree of influence of population gathering on data collection. The clustering impact characterization value compares the current clustering impact rate of the area to be sampled with the historical average, measuring the relative degree of population gathering in that area. The projected area reflects the protrusion of the terrain. The protrusion impact coefficient quantifies the degree of influence of terrain protrusion on data collection. The dynamic impact coefficient comprehensively considers the impact of obstacle changes and population gathering on data collection. The sampling complexity characterization value comprehensively reflects the overall complexity of the area to be sampled. Areas with different levels of complexity will encounter different problems during the data collection process. By dividing the sampling into complex categories, different processing strategies can be adopted for different categories of areas, thereby improving the accuracy and efficiency of data collection and increasing the efficiency of supplementary sampling.

[0103] Please see Figure 2 As shown, it is a logical decision diagram for classifying the sampling complexity category of the sampling area based on the sampling complexity characterization value of the sampling area according to an embodiment of the present invention. The process of classifying the sampling complexity category of the sampling area based on the sampling complexity characterization value of the sampling area according to the present invention includes:

[0104] If the sampled complexity value is less than or equal to the preset sampled complexity value, the area to be sampled is determined as a weakly sampled complexity category.

[0105] If the sampled complexity value is greater than the preset sampled complexity value, the area to be sampled will be classified as a strongly sampled complexity category.

[0106] Specifically, the preset sampling complexity characterization value is selected within the range [0.74, 0.82]. Those skilled in the art can select and determine the preset sampling complexity characterization value themselves. It can be determined by statistical distribution of historical data. It is understood that the situation can be divided based on whether the regional complexity is higher than the historical average level. In this embodiment, the preset sampling complexity characterization value is preferably 0.82.

[0107] Please see Figure 3 As shown, this is a logic decision diagram for determining a preset green view rate mutation threshold for identifying missing data in the initial images of each frame based on sampling complex categories, according to an embodiment of the present invention. The process of determining the preset green view rate mutation threshold for identifying missing data in the initial images of each frame based on sampling complex categories includes:

[0108] When the area to be sampled is determined to be a weakly sampled complex category, the initial preset green visibility rate mutation threshold is continuously used to identify missing data.

[0109] When the area to be sampled is determined to be a highly complex category, the preset green visibility threshold is adjusted to the corresponding value based on the complex sampling characteristic value.

[0110] Specifically, a preset green visual rate mutation threshold is determined based on the sampling complexity category to identify missing data in the initial images of each frame. The stability and variation of data differ across regions with different sampling complexity categories. In regions with weak sampling complexity, the data is relatively stable, and the initial preset green visual rate mutation threshold is sufficient for identification. In regions with higher complexity, the green visual rate fluctuates significantly; adjusting the preset green visual rate mutation threshold allows for more sensitive detection of data changes, improving the accuracy of missing data identification. Dynamically adjusting the preset green visual rate mutation threshold according to the sampling complexity category more accurately identifies missing data in regions of varying complexity, improving data processing accuracy and increasing re-sampling efficiency.

[0111] Specifically, based on the sampled complex characterization values, the preset green visibility rate mutation threshold is adjusted to a corresponding value, wherein,

[0112] The reduction in the preset green visibility threshold is positively correlated with the sampled complex characterization value.

[0113] In this embodiment, optionally,

[0114] The sampled complex characterization value is compared with the first preset complex comparison value and the second preset complex comparison value;

[0115] If the sampled complex characterization value is less than or equal to the first preset complex comparison value, the preset green visibility rate mutation threshold will be corrected to 0.95 times the initial preset green visibility rate mutation threshold.

[0116] If the sampled complex characterization value is less than or equal to the second preset complex comparison value and greater than the first preset complex comparison value, then the preset green visibility rate mutation threshold is corrected to 0.86 times the initial preset green visibility rate mutation threshold.

[0117] If the sampled complex characterization value is greater than the second preset complex comparison value, the preset green visibility rate mutation threshold will be corrected to 0.77 times the initial preset green visibility rate mutation threshold.

[0118] The first preset complex comparison value is 1.3, and the second preset complex comparison value is 1.5.

[0119] In areas with higher complexity, green visibility changes more frequently and drastically. Lowering the preset green visibility mutation threshold allows for more sensitive capture of these changes, leading to more accurate identification of missing data. Higher sampling complexity values ​​indicate more complex areas and greater green visibility changes; in such cases, a more significant reduction in the preset green visibility mutation threshold further enhances identification accuracy. By dynamically adjusting the identification criteria based on a comprehensive consideration of various factors, the accuracy and efficiency of green visibility re-collection are improved.

[0120] Please see Figure 4 As shown, this is a logic diagram illustrating the determination of whether to correct the GNSS RTK data update rate based on the clustering degree of missing data in an embodiment of the present invention. The process of determining whether to correct the GNSS RTK data update rate based on the clustering degree of missing data in the present invention includes:

[0121] Identify the distance between the coordinates of adjacent missing data, calculate the average of each distance, and obtain the clustering degree of the missing data.

[0122] If the clustering degree of missing data is less than or equal to the preset clustering degree of missing data, the current GNSS RTK data update rate will be used to continuously determine the drone's position.

[0123] If the clustering degree of missing data is greater than the preset clustering degree of missing data, the data update rate of GNSS RTK will be adjusted to the corresponding value based on the clustering degree of missing data.

[0124] Specifically, in determining the clustering degree of missing data, the data is sorted according to the collection time, the Euclidean distance between adjacent coordinates is calculated, and the average of all adjacent distances is obtained to obtain the clustering degree of missing data.

[0125] The preset missing data clustering degree is selected within the range [100, 150], with the unit being m. Those skilled in the art can select and determine the preset missing data clustering degree themselves, which can be determined by combining the actual area to be sampled and the distance interval of each initial image acquisition. In this embodiment, preferably, the preset missing data clustering degree is 150.

[0126] Specifically, the decision to adjust the GNSS RTK data update rate is based on the degree of clustering of missing data. The degree of clustering reflects the extent to which missing data is concentrated. The GNSS RTK data update rate is used to determine the frequency of location data updates for the UAV's position. When the degree of clustering is less than or equal to a preset value, it indicates that the missing data is relatively concentrated, with some areas experiencing concentrated obstruction. Supplementing data in these concentrated areas is sufficient, and the current GNSS RTK data update rate meets the requirements. When the degree of clustering is greater than the preset value, it indicates that the missing data is scattered, with special conditions such as terrain variations, weather conditions, or complex vegetation hindering data collection around these scattered points. In this case, increasing the data update rate yields more accurate location information, ensuring the UAV can accurately reach the supplementary data collection points. Dynamically adjusting the GNSS RTK data update rate based on the degree of clustering improves the accuracy of supplementary data collection, avoids errors caused by inaccurate positioning, and further improves collection efficiency.

[0127] Specifically, the GNSS RTK data update rate is adjusted to a corresponding value based on the degree of clustering of missing data.

[0128] The increase in GNSS RTK data update rate is positively correlated with the degree of clustering of missing data.

[0129] In this embodiment, optionally,

[0130] The clustering degree of the missing data is compared with the first preset clustering degree comparison value and the second preset clustering degree comparison value;

[0131] If the clustering degree of the missing data is less than or equal to the first preset clustering degree comparison value, the data update rate of the GNSS RTK will be adjusted to 1.13 times the initial data update rate;

[0132] If the clustering degree of missing data is less than or equal to the second preset clustering degree comparison value and greater than the first preset clustering degree comparison value, then the data update rate of GNSS RTK will be adjusted to 1.24 times the initial data update rate.

[0133] If the clustering degree of missing data is greater than the second preset clustering degree comparison value, the data update rate of GNSS RTK will be adjusted to 1.37 times the initial data update rate;

[0134] The first preset aggregation degree comparison value is 300m, and the second preset aggregation degree comparison value is 400m.

[0135] The clustering degree of missing data reflects the concentration of data collection anomalies. A high clustering degree of missing data means that the anomalies are more scattered and the terrain of the area is more complex. Improving the data update rate of GNSS RTK to obtain more frequent and accurate positioning information will allow UAVs to reach the re-collection points more accurately and improve the accuracy of re-collection.

[0136] Specifically, the process of determining the anchored green visibility difference based on the real-time acquired supplementary images, and determining whether the supplementary images at the supplementary points are qualified based on the anchored green visibility difference, includes:

[0137] Obtain the green view rate of the current supplemented image, obtain the green view rate of the previous frame of the initial image corresponding to the missing data of the current supplemented point, calculate the absolute value of the difference between the two green view rates, and obtain the anchored green view rate difference.

[0138] When the difference in the anchored green view rate is less than or equal to the preset green view rate mutation threshold, the supplementary sampling point is deemed qualified and the next supplementary sampling point is sampled.

[0139] When the difference in the anchored green view rate is greater than the preset green view rate mutation threshold, the abnormality of the supplementary sampling point is identified, and the sampling parameters are corrected based on the difference value of the horizontal comparison.

[0140] The anchored green field rate difference is determined based on the real-time acquired supplementary images. This difference is then used to determine whether the supplementary sampling at each point is satisfactory. The anchored green field rate difference is the absolute value of the difference between the green field rate of the current supplementary image and the green field rate of the corresponding missing data in the previous frame's initial image, reflecting the degree of difference between the supplementary data and the original continuous data. The purpose of supplementary sampling is to fill in missing data, making the data more complete and accurate. By comparing the green field rate difference, the consistency between the supplementary data and the original data can be intuitively judged. If the difference is too large, it is determined that other factors have interfered with the supplementary sampling process, and the sampling parameters are further adjusted.

[0141] Specifically, the process of correcting sampling parameters based on the differences between horizontal comparisons includes:

[0142] The absolute value of the difference between the green visibility rate of the currently acquired image and the green visibility rate of the corresponding missing data is calculated as the ratio of the green visibility rate of the corresponding missing data to obtain the horizontal comparison difference value.

[0143] When the horizontal comparison difference value is less than or equal to the preset horizontal comparison difference value, the number of supplementary collection points for a single missing data will be adjusted to the corresponding value based on the horizontal comparison difference value.

[0144] When the difference value of the horizontal comparison is greater than the preset difference value of the horizontal comparison, the data update rate of GNSS RTK will be adjusted to the corresponding value.

[0145] Specifically, the preset horizontal comparison difference value is selected within the range of [0.08, 0.12]. Those skilled in the art can select and determine the preset horizontal comparison difference value themselves. It can be set based on the statistics of green visibility measurement error and experimental experience. In this embodiment, the preset horizontal comparison difference value is preferably 0.12.

[0146] In this embodiment, optionally, the data update rate of the GNSS RTK is adjusted to 1.13 times the initial data update rate.

[0147] The sampling parameters are corrected based on the horizontal comparison difference value. The horizontal comparison difference value is the ratio of the absolute value of the difference between the green view rate of the current supplemented image and the green view rate of the corresponding missing data to the green view rate of the corresponding missing data, reflecting the relative difference between the supplemented data and the missing data. When the horizontal comparison difference value is small, the same problem may still occur due to the abnormality of the identified supplemented sampling points, resulting in unqualified images. In this case, the number of sampling points is increased to perform multiple samplings. When the difference value is large, the re-acquired image differs from the original image, but the re-acquired image is still unqualified. In this case, it is determined that the actual location of the supplemented sampling point is deviated due to inaccurate UAV positioning. In this case, the data update rate of GNSS RTK is increased to improve positioning accuracy, further improving the efficiency of green view rate data supplementation.

[0148] Specifically, based on the horizontal comparison difference values, the number of supplementary data points for a single missing data point is adjusted to a corresponding value.

[0149] The increase in the number of missing data points is negatively correlated with the difference value of cross-sectional comparison.

[0150] In this embodiment, optionally,

[0151] The lateral comparison difference value is compared with the first lateral comparison value and the second lateral comparison value;

[0152] If the difference value of the horizontal comparison is less than or equal to the first horizontal comparison value, the number of supplementary collection points corresponding to a single missing data point will be adjusted to 4 times the initial number of supplementary collection points.

[0153] If the difference value of the horizontal comparison is less than or equal to the second horizontal comparison value and greater than the first horizontal comparison value, then the number of supplementary collection points corresponding to a single missing data point will be adjusted to three times the initial number of supplementary collection points.

[0154] If the difference value of the horizontal comparison is greater than the second horizontal comparison value, the number of supplementary collection points corresponding to a single missing data point will be adjusted to twice the initial number of supplementary collection points.

[0155] The first horizontal comparison value is 0.03, and the second horizontal comparison value is 0.05.

[0156] Specifically, the method for adding supplementary sampling points involves generating several offset points around the original sampling point coordinates. Using the original sampling point as the center and a preset sampling length as the radius, several additional sampling points are randomly added. The drone then sequentially captures images of each of these newly added sampling points. After acquiring images for the newly added sampling points, the average and standard deviation of the green view rate (BP) for all these images are calculated. Images whose BP differs from the average by more than two standard deviations are discarded to filter out outliers. Among the remaining images, the image with the BP closest to the average of the adjacent frames containing the missing data is selected as valid data.

[0157] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0158] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent supplementary data collection path planning for green visibility rate supplementary data collection, characterized in that, include: The sampling complexity category of the area to be sampled is determined based on the sampling complexity characterization value of the area to be sampled. Based on the sampling complexity category, a preset green visibility mutation threshold is determined to identify missing data in the initial image of each frame; The process of identifying missing data in the initial images of each frame includes: The system receives initial images of each frame uploaded by the vehicle-mounted acquisition unit in real time; the semantic segmentation model identifies green vegetation pixels in each initial image frame and generates a preliminary green visibility rate; the absolute value of the difference between the green visibility rate of the current frame and the green visibility rate of the previous frame in consecutive frames is calculated to obtain the green visibility rate mutation value of consecutive frames; the rule engine determines whether there is missing data in a single set of initial images based on the green visibility rate mutation value of consecutive frames; when the green visibility rate mutation value of consecutive frames is greater than the preset green visibility rate mutation threshold, the initial image of the next frame is identified as missing data, and a geospatial list of missing data is generated based on each missing data, including the coordinates and acquisition height and angle of each missing data. Determine whether to adjust the GNSS RTK data update rate based on the clustering degree of missing data; Based on the missing data, the supplementary sampling points are determined, and the supplementary sampling path is generated based on the coordinates corresponding to each identified supplementary sampling point. By using drones to conduct supplementary image collection at each point, the difference in anchored green visibility is determined based on the real-time acquired supplementary images. Based on this difference in anchored green visibility, the suitability of the supplementary image collection at each point is determined, including: Identify anomalies in the supplementary data collection at the supplementary data collection points, and correct the sampling parameters based on the differences in the horizontal comparison, including correcting the number of supplementary data collection points for a single missing data point or correcting the data update rate of GNSS RTK used to determine the location of the UAV. Alternatively, if the supplementary sampling at the designated sampling point is deemed satisfactory, supplementary sampling can be carried out at the next sampling point. Continue until all replenishment points are replenished; Each supplementary image is fused with each initial image to generate a panoramic view of green visibility.

2. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 1, characterized in that, The process of determining the complex characterization values ​​of the area to be sampled includes: The obstacle change characterization value is determined based on the number of buildings at several time points in the area to be mined and the time interval between two time points. The clustering influence characterization value is determined based on the total area of ​​each multi-person clustering area captured by the sampling vehicle and the total area of ​​the area to be sampled during the process of identifying and acquiring each frame of the initial image. The protrusion influence coefficient is determined based on the obtained 3D model of the area to be mined; The dynamic influence coefficient is obtained by summing the corresponding coefficients assigned to the obstacle change characterization value and the clustering influence characterization value; The product of the dynamic influence coefficient and the convexity influence coefficient is calculated to obtain the sampled complex characterization value.

3. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 2, characterized in that, The process of classifying the sampling complexity category of the region to be sampled based on the sampling complexity characterization value of the region to be sampled includes: If the sampled complexity value is less than or equal to the preset sampled complexity value, the area to be sampled is determined as a weakly sampled complexity category. If the sampled complexity value is greater than the preset sampled complexity value, the area to be sampled will be classified as a strongly sampled complexity category.

4. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 3, characterized in that, Based on the sampling complexity category, a preset green visual rate mutation threshold is determined to identify missing data in the initial image of each frame, including: When the area to be sampled is determined to be a weakly sampled complex category, the initial preset green visibility rate mutation threshold is continuously used to identify missing data. When the area to be sampled is determined to be a highly complex category, the preset green visibility threshold is adjusted to the corresponding value based on the complex sampling characteristic value.

5. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 4, characterized in that, Based on the sampled complex characterization values, the preset green visibility rate mutation threshold is adjusted to the corresponding value, whereby, The reduction in the preset green visibility threshold is positively correlated with the sampled complex characterization value.

6. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 5, characterized in that, The process of determining whether to adjust the GNSS RTK data update rate based on the clustering degree of missing data includes: The clustering degree of missing data is determined based on the distance between the coordinates of each adjacent missing data point. When the clustering degree of missing data is greater than the preset clustering degree of missing data, the data update rate of GNSS RTK is adjusted to the corresponding value based on the clustering degree of missing data.

7. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 6, characterized in that, The GNSS RTK data update rate is adjusted to a corresponding value based on the clustering degree of missing data. The increase in GNSS RTK data update rate is positively correlated with the degree of clustering of missing data.

8. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 7, characterized in that, The process of determining the anchored green visibility difference based on real-time acquired supplementary images, and determining whether the supplementary images at the supplementary points are qualified based on the anchored green visibility difference, includes: The anchored green view rate difference is determined based on the green view rate of the acquired current supplementary image; When the difference in the anchored green view rate is greater than the preset green view rate mutation threshold, the abnormality of the supplementary sampling point is identified, and the sampling parameters are corrected based on the difference value of the horizontal comparison.

9. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 8, characterized in that, The process of correcting sampling parameters based on the difference values ​​of horizontal comparison includes: The horizontal comparison difference value is determined based on the green view rate of the currently acquired image and the green view rate of the corresponding missing data; When the horizontal comparison difference value is less than or equal to the preset horizontal comparison difference value, the number of supplementary collection points for a single missing data will be adjusted to the corresponding value based on the horizontal comparison difference value. When the difference value of the horizontal comparison is greater than the preset difference value of the horizontal comparison, the data update rate of GNSS RTK will be adjusted to the corresponding value.

10. The intelligent supplementary sampling path planning method for green view rate supplementary sampling according to claim 9, characterized in that, Based on the horizontal comparison difference values, the number of supplementary data points for a single missing data point is adjusted to the corresponding value, whereby... The increase in the number of missing data points is negatively correlated with the difference value of cross-sectional comparison.

Images