Wireless resource data checking method and system

Abstract

The invention discloses a wireless resource data checking method and system. The method comprises the following steps: acquiring first longitude and latitude data and original image data of a reconnaissance base station; acquiring second longitude and latitude data of each full base station, and screening the earth surface distance difference between the first longitude and latitude data and the second longitude and latitude data to obtain geographical adjacent base stations; performing image recognition processing on the original image data to obtain first working parameter data; acquiring second working parameter data of each geographical adjacent base station, and screening the first working parameter data and the first working parameter difference value of the second working parameter data to obtain a target checking base station; and obtaining second target work parameter data of the target checking base station in the to-be-checked wireless resource database, and comparing a second work difference value of the first work parameter data and the second target work parameter data with the dynamic deviation threshold to obtain a wireless resource data checking result of the target checking base station. According to the method provided by the invention, the automatic checking of the wireless resource data is realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for verifying wireless resource data. Background Technology

[0002] Wireless resource data is the core foundational data for ensuring the quality of mobile communication network coverage and avoiding signal interference. Its accuracy directly affects users' communication experience, such as call fluency and internet speed. Therefore, it is necessary to regularly check wireless resource data to ensure that the database records are consistent with the actual situation of the base stations.

[0003] In existing technologies, the verification of wireless resource data relies on staff measuring parameters such as azimuth, mounting height, and downtilt angle of base station antennas on-site, and then manually comparing them with resource data stored in the database to determine whether the deviation is within a threshold range. If the deviation exceeds the limit, the data is manually corrected. However, the verification efficiency of this method is limited by the efficiency of manual operation, and manual judgment is easily affected by factors such as experience differences and fatigue, resulting in a high risk of misjudgment. Summary of the Invention

[0004] This invention provides a wireless resource data verification method and system to solve the technical problem of how to improve existing wireless resource data verification methods and achieve the effect of improving the accuracy of wireless resource data verification.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a wireless resource data verification method, comprising: In response to the wireless resource data verification signal, it receives the first latitude and longitude data and raw image data of the survey base station collected by the UAV; Obtain the second latitude and longitude data of all base stations in the wireless resource database to be verified, and filter the difference in ground distance between the first latitude and longitude data and the second latitude and longitude data to obtain a number of geographically neighboring base stations. The original image data is input into image recognition processing to obtain the first working parameter data and its corresponding confidence level; Obtain the second working parameter data of each geographically neighboring base station in the wireless resource database to be verified, and perform filtering processing on the first working parameter difference between the first working parameter data and the second working parameter data to obtain the target base station to be verified; The dynamic deviation threshold of the first working parameter data is determined based on the confidence level, wherein the dynamic deviation threshold is designed to be adjusted accordingly based on the confidence level of the original image data; Obtain the second target operating parameter data of the target base station in the wireless resource database to be verified, compare the second operating parameter difference between the first operating parameter data and the second target operating parameter data with the dynamic deviation threshold, and obtain the wireless resource data verification result of the target base station.

[0006] As one preferred embodiment, the step of filtering the surface distance difference between the first latitude and longitude data and the second latitude and longitude data to obtain a number of geographically nearby base stations includes: Based on the preset surface distance calculation formula, the first longitude and first latitude in the first longitude and latitude data and the second longitude and second latitude in each of the second longitude and latitude data are calculated to obtain the surface distance difference between the survey base station and each of the full base stations; The surface distance difference values ​​are compared with a preset geographic proximity screening threshold. The first full base station whose surface distance difference value is not greater than the geographic proximity screening threshold is selected and determined as the geographic proximity base station. The geographic proximity screening threshold is the upper limit of the surface distance centered on the survey base station.

[0007] As one preferred embodiment, the step of inputting the original image data into image recognition processing to obtain the first working parameter data and its corresponding confidence level includes: The original image data is subjected to target recognition to generate a target detection box of the target object and its corresponding target detection confidence score, wherein the target object includes at least the antenna of the survey base station; The antenna region image is obtained by cropping the target detection box. The antenna region image is input into the trained engineering parameter classification and recognition model, and the first engineering parameter data and its corresponding engineering parameter classification confidence score are output. The confidence scores for target detection and engineering parameter classification are fused and calculated to obtain the confidence score corresponding to the first engineering parameter data.

[0008] As one preferred embodiment, the first operating parameter data includes frequency band, azimuth angle, mounting height, and mechanical tilt angle; The step of filtering the first parameter difference between the first and second engineering parameter data to obtain the target verification base station includes: The cross-intersection over union (CUI) ratio is calculated between the frequency band data in the first engineering parameter data and the corresponding field data in each of the second engineering parameter data. The equality of values ​​between the direction angle data in the first engineering parameter data and the corresponding field data in each of the second engineering parameter data is judged. Based on the CUI calculation results and the equality judgment results, the first matching degree dataset is obtained. Based on the confidence levels corresponding to the frequency band data and the azimuth angle data in the first engineering parameter data, a first dynamic matching threshold is generated, and the first matching degree dataset is filtered according to the first dynamic matching threshold to obtain a first candidate base station set; The hanging height data and mechanical tilt angle data in the first working parameter data are compared with the corresponding field data in the second working parameter data to determine the equality of their values, and a second matching degree dataset is obtained. Based on the confidence levels corresponding to the hanging height data and the mechanical tilt angle data in the first working parameter data, a second dynamic matching threshold is generated, and the second matching degree dataset is filtered according to the second dynamic matching threshold to obtain the target verification base station.

[0009] As one preferred embodiment, the step of comparing the second parameter difference between the first operating parameter data and the second target operating parameter data with the dynamic deviation threshold to obtain the radio resource data verification result of the target verification base station includes: Calculate the difference between the corresponding field data in the first working parameter data and the second target working parameter data to generate a second working parameter difference dataset with the corresponding difference of each working parameter field. The difference between each parameter field in the second parameter difference dataset is compared with the threshold of the corresponding parameter in the dynamic deviation threshold to generate a single comparison result for each parameter field. Based on the individual comparison results, the wireless resource data verification results of the target verification base station are generated.

[0010] Another embodiment of the present invention provides a wireless resource data verification system, comprising: The acquisition module is used to receive the first latitude and longitude data and raw image data of the survey base station collected by the UAV in response to the wireless resource data verification signal; The first filtering module is used to obtain the second latitude and longitude data of each full base station in the wireless resource database to be checked, and to filter the difference in ground distance between the first latitude and longitude data and the second latitude and longitude data to obtain a number of geographically neighboring base stations. The recognition module is used to input the original image data into image recognition processing to obtain the first working parameter data and its corresponding confidence level; The second filtering module is used to obtain the second working parameter data of each geographically neighboring base station in the wireless resource database to be verified, and to filter the first working parameter difference between the first working parameter data and the second working parameter data to obtain the target base station to be verified. A calculation module is used to determine a dynamic deviation threshold for the first working parameter data based on the confidence level, wherein the dynamic deviation threshold is designed to be adjusted accordingly based on the confidence level of the original image data; The verification module is used to obtain the second target operating parameter data of the target base station in the wireless resource database to be verified, compare the second operating parameter difference between the first operating parameter data and the second target operating parameter data with the dynamic deviation threshold, and obtain the wireless resource data verification result of the target base station.

[0011] As one preferred embodiment, the first screening module is specifically used for: Based on the preset surface distance calculation formula, the first longitude and first latitude in the first longitude and latitude data and the second longitude and second latitude in each of the second longitude and latitude data are calculated to obtain the surface distance difference between the survey base station and each of the full base stations; The surface distance difference values ​​are compared with a preset geographic proximity screening threshold. The first full base station whose surface distance difference value is not greater than the geographic proximity screening threshold is selected and determined as the geographic proximity base station. The geographic proximity screening threshold is the upper limit of the surface distance centered on the survey base station.

[0012] As one preferred embodiment, the identification module is specifically used for: The original image data is subjected to target recognition to generate a target detection box of the target object and its corresponding target detection confidence score, wherein the target object includes at least the antenna of the survey base station; The antenna region image is obtained by cropping the target detection box. The antenna region image is input into the trained engineering parameter classification and recognition model, and the first engineering parameter data and its corresponding engineering parameter classification confidence score are output. The confidence scores for target detection and engineering parameter classification are fused and calculated to obtain the confidence score corresponding to the first engineering parameter data.

[0013] As one preferred embodiment, the first operating parameter data includes frequency band, azimuth angle, mounting height, and mechanical tilt angle; The second filtering module is specifically used for: The cross-intersection over union (CUI) ratio is calculated between the frequency band data in the first engineering parameter data and the corresponding field data in each of the second engineering parameter data. The equality of values ​​between the direction angle data in the first engineering parameter data and the corresponding field data in each of the second engineering parameter data is judged. Based on the CUI calculation results and the equality judgment results, the first matching degree dataset is obtained. Based on the confidence levels corresponding to the frequency band data and the azimuth angle data in the first engineering parameter data, a first dynamic matching threshold is generated, and the first matching degree dataset is filtered according to the first dynamic matching threshold to obtain a first candidate base station set; The hanging height data and mechanical tilt angle data in the first working parameter data are compared with the corresponding field data in the second working parameter data to determine the equality of their values, and a second matching degree dataset is obtained. Based on the confidence levels corresponding to the hanging height data and the mechanical tilt angle data in the first working parameter data, a second dynamic matching threshold is generated, and the second matching degree dataset is filtered according to the second dynamic matching threshold to obtain the target verification base station.

[0014] As one preferred embodiment, the verification module is specifically used for: Calculate the difference between the corresponding field data in the first working parameter data and the second target working parameter data to generate a second working parameter difference dataset with the corresponding difference of each working parameter field. The difference between each parameter field in the second parameter difference dataset is compared with the threshold of the corresponding parameter in the dynamic deviation threshold to generate a single comparison result for each parameter field. Based on the individual comparison results, the wireless resource data verification results of the target verification base station are generated.

[0015] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following: 1) This invention utilizes a ground distance filtering mechanism to accurately locate geographically neighboring base stations from the full range of base stations, significantly reducing the data scale of subsequent parameter comparisons and avoiding the inefficiency of indiscriminately processing massive amounts of data in traditional manual verification or single automated technologies. At the same time, through step-by-step parameter filtering (preliminary matching followed by precise comparison), it further focuses on the target verification base station, making the verification process more targeted, effectively reducing the cost of invalid calculations, and achieving a leapfrog improvement in verification efficiency.

[0016] 2) This invention innovatively introduces a dynamic deviation threshold mechanism based on image recognition confidence, which overcomes the limitation that fixed thresholds cannot adapt to different survey scenarios. Specifically, when the confidence of the original image recognition is high, the threshold is appropriately relaxed to ensure the flexibility of verification; when the confidence is low, the threshold is automatically tightened to avoid error accumulation. Combined with accurate comparison of working parameters, it significantly reduces the risk of misjudgment and omission, and ensures that the verification results are highly consistent with the actual survey situation. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a wireless resource data verification method in one embodiment of the present invention. Figure 2 This is a different embodiment of the present invention. A diagram illustrating the function of matching threshold versus confidence level; Figure 3This is a structural block diagram of a wireless resource data verification system according to one embodiment of the present invention; Figure label: Among them, 11 is the acquisition module; 12 is the first filtering module; 13 is the identification module; 14 is the second filtering module; 15 is the calculation module; and 16 is the verification module. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] In the description of this invention, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0020] In the description of this invention, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to communication within two components. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] In the description of this invention, it should be noted that, 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. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0022] One embodiment of the present invention provides a method for verifying wireless resource data. For details, please refer to [link to documentation]. Figure 1 , Figure 1 The diagram shown is a flowchart of a wireless resource data verification method according to one embodiment of the present invention, which includes steps S1-S6: S1: In response to the wireless resource data verification signal, receive the first latitude and longitude data and raw image data of the survey base station collected by the UAV.

[0023] It should be noted that the wireless resource data verification signal is a verification request initiated by the user through the system's encapsulated external service interface. This signal is not a simple command trigger, but a structured request containing complete task information, typically including the unique identifier of the verification task, the target survey area (such as a specific city or base station cluster), task priority, and data return format requirements. When the system receives this signal, it parses it through a preset task scheduling module to confirm the task's legitimacy (such as user permissions and whether the target area is within the verification scope), and then sends a data collection command to the designated UAV deployment system, completing the closed-loop triggering of "responding to the signal - starting data collection" and avoiding meaningless resource consumption.

[0024] In this embodiment, before receiving data, the drone needs to be deployed and its parameters calibrated. The drone deployment must be adapted to the base station scenario in advance: for base stations in densely populated urban areas, multi-rotor drones are selected (high flexibility, able to get close to buildings for shooting); for base stations in remote suburbs or mountainous areas, fixed-wing drones are selected (long endurance, wide coverage). Flight parameters, including flight altitude, shooting angle, and flight speed, need to be preset during deployment. Simultaneously, the drone's positioning module and image acquisition equipment need to be calibrated: after calibration, the positioning module must ensure latitude and longitude acquisition accuracy to the centimeter level; after calibration, the image equipment resolution must be no less than 4K, and the color reproduction error ≤5%, providing high-quality raw material for subsequent data processing.

[0025] The first latitude and longitude data consists of the precise geographic coordinates of the survey base station, collected in real-time by a drone, and corresponds to the second latitude and longitude data of all base stations in the wireless resource database to be verified. This data acquisition relies on a high-precision positioning module onboard the drone, typically integrating GPS and BeiDou dual-mode positioning. In some complex scenarios, RTK (Real-Time Kinematic) technology can be superimposed, achieving a static positioning accuracy of ±2cm and a dynamic positioning accuracy of ±5cm. During the data acquisition process, the drone needs to hover 5-10 meters directly above the survey base station for 3-5 seconds, continuously collecting 10 sets of latitude and longitude data. Outliers are removed using a built-in algorithm, and the average value is taken as the final first latitude and longitude data. The data format adopts the internationally standard Decimal degree format to ensure consistency with the format of the second latitude and longitude data in the database, avoiding format compatibility issues during subsequent surface distance calculations.

[0026] Raw image data refers to antenna-related visual data captured by the UAV at the base station survey site, without any image recognition processing. This data is the core basis for subsequent extraction of primary operating parameters (azimuth, mounting height, and mechanical downtilt angle) and confidence level calculations. For the base station antenna, at least three close-up images from different angles need to be captured: a front view (for identifying the azimuth scale and frequency band markings), a side view (for measuring the mechanical downtilt angle and mounting height reference points), and a top view (for confirming the number and arrangement of antennas). Simultaneously, 1-2 panoramic environmental images should be captured, recording the surrounding terrain, building heights, and other reference points to provide auxiliary information for subsequent mounting height calculations. The image format should be lossless compressed PNG, with a resolution of at least 3840×2160 (4K) and a pixel density ≥300dpi to ensure clear and discernible antenna details. During the data collection process, it is necessary to avoid adverse environmental conditions. For example, rain or fog can cause images to become blurry, and trees or billboards can obstruct the view and cause the antenna target to be missing. In such cases, the drone needs to adjust its flight path or pause data collection until the environmental conditions are met before continuing, in order to avoid the subsequent recognition confidence being low due to poor quality of the original image.

[0027] S2: Obtain the second latitude and longitude data of all base stations in the wireless resource database to be verified, and filter the difference in ground distance between the first latitude and longitude data and the second latitude and longitude data to obtain a number of geographically neighboring base stations.

[0028] It should be noted that the wireless resource database to be verified is a structured database that stores the core resource data of all networked base stations, covering key data such as basic site information, antenna parameters, and geographic coordinates. It serves as the comparison benchmark for the verification work of this invention. The second latitude and longitude data consists of the precise geographic coordinates of each full base station in this database, which corresponds to the first latitude and longitude data (coordinates of the surveyed base stations) collected by the UAV in S1. It is the core basis for calculating the ground distance and screening neighboring base stations.

[0029] In this embodiment, the process of obtaining the second latitude and longitude data is achieved by calling a predefined wireless resource database query tool. First, based on the target area parameters (such as a specific city or administrative region) in the verification signal received by S1, a natural language query requirement is generated (e.g., "Query the base station ID, longitude, and latitude data of all networked base stations in XX city"). Then, this query requirement is input into the text2sql large model, which is automatically converted into an executable SQL statement. Next, the tool creates a database connection in the Python environment and executes the SQL statement. Finally, after successful execution, the second latitude and longitude data of all base stations is returned in JSON format. The data structure contains three core fields: "base station ID", "second longitude", and "second latitude", ensuring that each latitude and longitude coordinate can be uniquely associated with the corresponding base station.

[0030] Preferably, in one embodiment of the present invention, the surface distance difference between the first latitude and longitude data and the second latitude and longitude data is filtered to obtain a number of geographically nearby base stations, including: Based on the preset surface distance calculation formula, the first longitude and first latitude in the first longitude and latitude data and the second longitude and second latitude in each second longitude and latitude data are calculated to obtain the surface distance difference between the survey base station and each full-scale base station.

[0031] Understandably, because the Earth is approximately a sphere, the actual distance between the survey base station and the full network base station needs to be calculated using "surface distance" (i.e., spherical distance), rather than a straight-line distance. The core calculation formula is the spherical distance formula: in, Represents latitude, Represents longitude. It is the Earth's radius, which is 6371 km.

[0032] The surface distance difference between each location is compared with the preset geographic proximity screening threshold. The first full base station whose surface distance difference is not greater than the geographic proximity screening threshold is selected and identified as the geographic proximity base station. The geographic proximity screening threshold is the upper limit of the surface distance centered on the survey base station.

[0033] The geographic proximity screening threshold is the core criterion for determining whether all base stations are geographically nearby. Essentially, it is a preset upper limit of the ground distance centered on the surveyed base station; in this embodiment, it is 50 meters. This threshold is not fixed but has scenario-based configuration capabilities: for densely populated urban areas (with densely distributed base stations and an average spacing of less than 500 meters), the threshold can be set to 50-200 meters to avoid screening out too many base stations, which would reduce the efficiency of subsequent parameter matching; for suburban or mountainous areas (with sparsely distributed base stations and an average spacing of 1-3 kilometers), the threshold can be extended to 500-1000 meters to ensure coverage of all reasonably nearby base stations and avoid missing target verification objects.

[0034] In this embodiment, a set of data to be processed is formed based on the first latitude and longitude data obtained in S1, the second latitude and longitude data of all base stations obtained through the database query tool, and the preset geographic proximity filtering threshold. The calculation results of the surface distance calculation tool, namely the surface distance difference between each full base station and the survey base station, are called and compared with the geographic proximity filtering threshold one by one. For full base stations whose distance difference is not greater than the filtering threshold, they are marked as the first full base station that meets the conditions. For full base stations whose distance difference is greater than the filtering threshold, they are directly excluded and do not enter the subsequent process.

[0035] All base stations marked as meeting the criteria are integrated into a set of geographically neighboring base stations. At the same time, their complete engineering parameter data (including frequency band, azimuth angle, mounting height, mechanical downtilt angle, etc.) in the database are associated with it and output in JSON format as input data for subsequent processes.

[0036] S3: Input the original image data into image recognition processing to obtain the first working parameter data and its corresponding confidence level.

[0037] Preferably, in one embodiment of the present invention, inputting the original image data into image recognition processing to obtain first working parameter data and its corresponding confidence level includes: The original image data is used to perform target recognition, and target detection bounding boxes and corresponding target detection confidence scores are generated for the target recognition objects. The target recognition objects include at least the antennas of the survey base station.

[0038] In this embodiment, considering that antennas are mostly regular rectangular structures and that both recognition speed and accuracy need to be balanced, the YOLOv8 or Faster R-CNN algorithm is used. These algorithms perform well in small target detection and target separation tasks in complex backgrounds. The model needs to be pre-trained on a professional dataset: the training dataset contains base station antenna images with different weather conditions (sunny, cloudy, light rain), different shooting angles (front, side, oblique), and different types (macro base station antenna, micro base station antenna), with a total sample size of no less than 100,000 images. Each image is labeled with the true bounding box of the antenna to ensure that the model can learn the general features of the antenna.

[0039] The raw image acquired in step S1 is input into the pre-trained model. The model extracts features to identify regions in the image that match antenna characteristics and outputs a rectangular target detection box that completely encloses the antenna's outline. The detection box is represented in coordinate form, ensuring complete coverage of the antenna body and omitting key details such as scale lines and frequency band markings. Simultaneously, the model outputs a target detection confidence score c1, ranging from 0 to 1, representing the probability that the object within the detection box is a real antenna. For example, C1 = 0.893 means there is an 89.3% probability that it is an antenna. Detection boxes with c1 below a preset screening threshold (e.g., 0.5) are considered false detections and are directly filtered out.

[0040] The antenna region image is obtained by cropping based on the target detection bounding box. Specifically, the cropping operation uses the target detection bounding box as a reference and adopts a "precise cropping + edge preservation" strategy. Based on the coordinates (x1, y1, x2, y2) of the detection bounding box, the corresponding region is cropped from the original image. Simultaneously, to avoid losing key details of the antenna edges (such as azimuth scale and tilt angle adjustment marks) during cropping, a buffer area of ​​5-10 pixels is extended outside the boundary of the detection bounding box. For example, if the detection bounding box coordinates are (100, 200, 500, 800), the actual cropping range is (95, 195, 505, 805), ensuring that the complete structure and detailed features of the antenna are preserved.

[0041] The antenna area image is input into the trained engineering parameter classification and recognition model, which outputs the first engineering parameter data and its corresponding classification confidence score. The first engineering parameter data refers to the core parameters of the survey base station antenna identified from the antenna area image. According to the document definition, this specifically includes four key data categories: frequency band, azimuth angle, mechanical downtilt angle, and mounting height. Its identification relies on the pre-trained engineering parameter classification and recognition model, which employs a multi-task branch structure: a backbone network with four independent classification / regression branches, each responsible for identifying one of the four types of engineering parameters.

[0042] Specifically, the backbone network of the model uses ResNet50, whose advantage lies in the strong feature extraction capability brought by deep convolution, which can effectively capture detailed features such as antenna scale, markings, and structure. The four branches are as follows: Frequency band branch: For the categorical features of frequency bands (such as FDD900, FDD1800, TD-LTE, etc.), a fully connected layer + softmax activation function is used to construct a multi-classification model; Direction angle branch: The direction angle ranges from 0 to 360°, and a 360-class classification model is constructed with 1° as one category, outputting the probability of each angle; Mechanical tilt angle branch: The value ranges from 0 to 90°, with each 1° as a category, constructing a 90-class classification model (this design is explicitly mentioned in the document), and outputting the probability of each angle; Hanging Height Branch: The hanging height is a continuous numerical value (unit: meters). A regression model is used to output the estimated value of the hanging height, and at the same time, the confidence probability of the estimated value is also output.

[0043] During model training, the backbone network is first pre-trained on a public image dataset, and then the entire model is fine-tuned using an antenna image dataset with labeled parameters to optimize the loss function until the model's recognition accuracy on the validation set meets the requirements.

[0044] The standardized antenna region image is input into the pre-trained model, and each branch outputs the recognition results: the frequency band branch outputs the probability of each frequency band category, and the frequency band with the highest probability is taken as the recognition result; the azimuth angle branch outputs the probability of each angle from 0 to 360°, and the angle with the highest probability is taken as the recognition result; the mechanical downtilt angle branch outputs the angle with the highest probability from 0 to 90° as the recognition result; and the mounting height branch directly outputs the estimated value of the mounting height. The above four types of recognition results are integrated into the first engineering parameter data and stored in JSON format, as shown in the following example: {"Frequency Band":"FDD1800","Azimuth Angle":36.0","Mechanical Downtilt Angle":10.13","Mounting Height":20.2}.

[0045] The confidence scores for target detection and parameter classification are fused to obtain the confidence score corresponding to the first set of parameters. The parameter classification confidence score is a probability value output by each branch, reflecting the reliability of the first set of parameters in identification.

[0046] Since target detection confidence reflects the reliability of the detection box as an antenna, and parameter classification confidence reflects the reliability of the parameter recognition result, a single confidence score cannot comprehensively evaluate the overall credibility of the first parameter data. Therefore, a fusion calculation is needed to obtain the final confidence score. The core logic is to combine the reliability of the two dimensions and output a comprehensive score within the range of 0-1. For example, the probability value of a downtilt angle of v is expressed as... Then the final confidence level of the downtilt angle is: S4: Obtain the second working parameter data of each geographically neighboring base station in the wireless resource database to be verified, and perform filtering processing on the first working parameter difference between the first working parameter data and the second working parameter data to obtain the target verification base station.

[0047] Preferably, in one embodiment of the present invention, the first parameter difference between the first and second operating parameter data is filtered to obtain the target verification base station, including: The crossover ratio (CRR) is calculated between the frequency band data in the first set of engineering parameters and the corresponding field data in each set of second engineering parameters. The equality of the values ​​of the direction angle data in the first set of engineering parameters and the corresponding field data in each set of second engineering parameters is then determined. Based on the CRR calculation results and the equality determination results, the first matching degree dataset is obtained.

[0048] Specifically, the core of the first matching dataset is to quantify the core parameter compatibility between the survey base station and geographically neighboring base stations through differentiated matching rules, providing a precise basis for subsequent screening. For the categorical field of frequency band, since combined antennas may have multi-band compatibility (such as the coexistence of FDD900 and FDD1800), a single equal value judgment cannot reflect the true degree of matching. Therefore, the intersection-union ratio is used for calculation, with the formula "intersection-union ratio = frequency band intersection size / frequency band union size". For the numerical field of azimuth angle, "equal values" is used as the matching standard. At the same time, considering the measurement accuracy error, a tolerance of ±0.5° is set. For example, if the survey azimuth angle is 36.0° and the azimuth angle in the resource data is 36.3°, it is judged as a match, and the matching degree is recorded as 1.0; if it exceeds the tolerance, it is recorded as 0.0.

[0049] The first matching degree is the weighted sum of the two. For example, the crossover ratio of 0.5 and the azimuth angle matching degree of 1.0 are superimposed to obtain the first matching degree of 1.5. The matching results of all geographically neighboring base stations are associated with "base station ID + antenna ID" to form the first matching degree dataset.

[0050] Based on the confidence levels corresponding to the frequency band data and azimuth angle data in the first engineering parameter data, a first dynamic matching threshold is generated, and the first matching degree dataset is filtered according to the first dynamic matching threshold to obtain the first candidate base station set.

[0051] In this embodiment, the dynamic matching threshold calculation formula can be expressed as: in, The matching threshold is a function related to the confidence level c. This represents a threshold value for the confidence level c of the image recognition result; if If the threshold is too low, a more conservative (higher) threshold should be used; otherwise, if the threshold is too high, a more lenient (lower) threshold can be used. It is a configurable parameter that is configured by the user; Used to control the confidence level c to be close to The slope or steepness of a smooth transition. The larger, from Switched to The steeper the transition; if Smaller sizes result in a smoother transition. , This is a user-defined upper bound (highest / most conservative) and lower bound (lowest / most lenient) for a threshold value of a certain parameter field. Users can set different values ​​for different parameter fields and scenarios.

[0052] See details Figure 2 , Figure 2 A different embodiment of the present invention This diagram illustrates the function of the matching threshold versus the confidence level. =0.5, =1.0, =2.0. Observing the function graph, we can see that when the confidence level c exceeds 0.5, if... Set it to be larger It will approach 1 at a faster rate. Comparing the yellow and blue lines, the green line becomes very steep after c exceeds 0.5, rapidly rising from... Switch to .

[0053] Specifically, the classification confidence parameters obtained in S3 are input into the dynamic matching threshold calculation formula to obtain the corresponding first dynamic matching threshold. The first matching degree of each geographically neighboring base station is compared with the first dynamic matching threshold, and base stations with a first matching degree ≥ the threshold are retained to generate the first candidate base station set.

[0054] The hanging height data and mechanical tilt angle data in the first working parameter data are compared with the corresponding field data in the second working parameter data to determine the equality of their values, thus obtaining the second matching degree dataset.

[0055] The second matching score dataset focuses on secondary priority parameters, namely mounting height and mechanical downtilt angle. Mounting height is measured in meters, with an allowable error of ±0.5 meters. For example, a survey mounting height of 20.2 meters and a resource mounting height of 20.6 meters are considered a match. Mechanical downtilt angle is measured in degrees, also with an allowable error of ±0.5 degrees. For example, a survey downtilt angle of 10.13° and a resource downtilt angle of 10.5° are considered a match. A successful match is scored as 1.0, and a failed match as 0.0. The second matching score is a weighted sum of the two (both with a weight of 1.0). For example, a matching score of 1.0 for mounting height and 1.0 for mechanical downtilt angle results in a score of 2.0, and a mismatch score of 0.0 for mounting height and 1.0 for mechanical downtilt angle results in a score of 1.0. Based on the first candidate base station set, the second matching score is calculated for each base station and stored as a group ("base station ID + antenna ID"), forming the second matching score dataset.

[0056] Furthermore, based on the confidence levels corresponding to the hanging height data and mechanical tilt angle data in the first working parameter data, a second dynamic matching threshold is generated, and the second matching degree dataset is filtered according to the second dynamic matching threshold to obtain the target verification base station.

[0057] Specifically, the generation logic for the second dynamic matching threshold is the same as that for the first dynamic threshold, except that the input confidence level is replaced with the classification confidence levels of the mounting height and mechanical downtilt angle corresponding to the engineering parameters. The mounting height confidence level and mechanical downtilt angle confidence level are extracted and input into the preset dynamic matching threshold calculation formula to obtain the second dynamic matching threshold. The matching degree in the second matching degree dataset is compared with this threshold, and the base stations with the second matching degree ≥ the second dynamic matching threshold are retained as the target verification base stations.

[0058] In this embodiment, the second set of technical parameters refers to the core antenna parameters (frequency band, azimuth angle, mounting height, and mechanical downtilt angle) of geographically neighboring base stations in the database to be verified, which form a comparison benchmark with the first set of technical parameters (survey and identification results). The first / second dynamic matching thresholds are screening criteria dynamically generated based on the corresponding technical parameter identification confidence levels, with the core difference being the source of the input confidence levels. The target base station to be verified is the object that needs to be accurately compared with the survey base station after two rounds of dynamic screening.

[0059] S5: Determine the dynamic deviation threshold of the first working parameter data based on the confidence level, wherein the dynamic deviation threshold is designed to be adjusted accordingly based on the confidence level of the original image data.

[0060] The core input for calculating the dynamic deviation threshold is the confidence score of the original image data. This confidence score is not a single-dimensional data point, but rather a fusion of the target detection confidence score and the engineering parameter classification confidence score from step S3. Two core confidence scores are separated from the image recognition results in step S3: one is the target detection confidence score, output by the target detection algorithm, representing the probability that the object within the detection box is a real antenna, reflecting the reliability of antenna positioning; the other is the engineering parameter classification confidence score, output by the multi-classification model, representing the probability of identifying engineering parameter values, reflecting the accuracy of engineering parameter value identification.

[0061] By inputting the target detection confidence and the engineering parameter classification confidence into the above dynamic matching threshold calculation formula, the dynamic deviation threshold of the first engineering parameter data is obtained.

[0062] S6: Obtain the second target operating parameter data of the target base station in the wireless resource database to be verified, compare the second operating parameter difference between the first operating parameter data and the second target operating parameter data with the dynamic deviation threshold, and obtain the wireless resource data verification result of the target base station.

[0063] It should be noted that the second target engineering parameter data is the core antenna parameter of the target verification base station in the wireless resource database to be verified, and it is the benchmark for comparing the deviation with the first engineering parameter data (UAV survey and identification results).

[0064] Preferably, in one embodiment of the present invention, the second parameter difference between the first operating parameter data and the second target operating parameter data is compared with a dynamic deviation threshold to obtain the radio resource data verification result of the target verification base station, including: The differences between corresponding fields in the first and second target operating parameter data are calculated to generate a second operating parameter difference dataset. This dataset quantifies the degree of deviation between the first and second target operating parameter data. In this embodiment, for categorical fields such as frequency bands, the absolute difference is not calculated; instead, the intersection-exchange-union ratio (IoU) is used to calculate the matching degree. The deviation quantification results of all fields are stored in association with "base station ID + antenna ID" to form the second operating parameter difference dataset.

[0065] The differences of each parameter field in the second parameter difference dataset are compared with the thresholds of the corresponding parameters in the dynamic deviation threshold to generate individual comparison results for each parameter field.

[0066] The individual comparison results are independent conclusions for each engineering parameter field, determining whether its deviation is within a reasonable range. From the dynamic deviation threshold set generated in step S5, specific thresholds for each engineering parameter field are extracted, and comparisons are performed according to field type: For numerical fields (azimuth angle, mounting height, mechanical tilt angle), if the "absolute difference ≤ corresponding dynamic threshold," the individual comparison result is considered compliant; otherwise, it is considered non-compliant. For example, an azimuth angle difference of 0.8° ≤ 1.0° is considered compliant, while a mounting height difference of 0.6 meters > 0.5 meters is considered non-compliant. For categorical fields (frequency band), if the "intersection over union ratio ≥ corresponding dynamic threshold," it is considered compliant. For example, an intersection over union ratio of 0.5 ≥ 0.4 is considered compliant. The comparison process is automated using a wireless resource verification data analysis tool, outputting individual results for each field to ensure no misjudgments due to human intervention.

[0067] The wireless resource data verification results of the target verification base station are generated based on the individual comparison results.

[0068] In this embodiment, the comprehensive verification result is the final verification conclusion formed based on all individual comparison results and the business rules in the knowledge base. Specifically, if all individual comparison results of the technical parameters are compliant, it indicates that the deviation between the resource data and the survey results is within a reasonable range, and the comprehensive verification result is that no correction is needed. Second, if there is an individual result that is not compliant (such as the azimuth difference exceeding the standard), but the target verification base station has a corresponding antenna record, the result is that the corresponding non-compliant field needs to be corrected according to the first technical parameter data. Third, if there is no corresponding antenna record in the target verification base station set (i.e., the resource data is missing the antenna), or all field deviations far exceed the threshold, the result is that the antenna record needs to be added according to the first technical parameter data. Fourth, if there is no antenna in the survey results, but the resource data has a record of that antenna (i.e., the resource data is redundant), the result is that the redundant antenna record in the resource data needs to be deleted. By integrating the individual comparison results and the rule fragments recalled by the knowledge base, a structured comprehensive verification conclusion is automatically generated to ensure that the judgment logic meets the actual business needs of wireless resource verification.

[0069] The present invention also provides a specific embodiment of a wireless resource data verification method to illustrate the beneficial effects of this solution.

[0070] This embodiment takes the wireless resource data verification of a communication base station in the suburbs of a city as the application scenario. There are 6 neighboring stations within 3 kilometers of the base station. The entire process of automated verification of UAV survey data and resource database needs to be completed through intelligent agent process orchestration.

[0071] Prior to implementation, this embodiment had completed core preparatory work, specifically including: deploying a standardized toolkit containing six types of tools, such as ground distance calculation, database query, and dynamic threshold generation. All tools are encapsulated in Python and support data interaction in JSON format; building a professional knowledge base covering verification process specifications, parameter matching priorities, and anomaly handling rules, including more than 1,200 industry experiences and historical cases; and completing on-site surveys with UAVs, collecting the base station's latitude and longitude (30.52°N, 114.36°E), high-resolution antenna images, and corresponding recognition results. The antenna frequency band is FDD900 / FDD1800, the azimuth angle is 32.6°, the mounting height is 28.3 meters, the mechanical downtilt angle is 8.7°, and the overall recognition confidence level is 0.83.

[0072] It should be noted in advance that the intelligent agent in this embodiment is an intelligent system that uses a large model as the scheduling center, integrates a predefined tool library and a professional knowledge base, and automates the entire process of verifying UAV survey data and wireless resource data. Its core objective is to replace manual operation and solve the problems of low efficiency, high error, and poor scene adaptability of existing technologies.

[0073] The intelligent agent is composed of a large model scheduling center, a predefined tool library, and a professional knowledge base. The large model scheduling center is the core decision-making unit, responsible for understanding task intent, selecting tools, and integrating results. Specifically, it receives process node inputs and prompts, combines them with professional rules retrieved from the knowledge base, determines the tools required for the current task (such as database queries and distance calculations), and parses the tool's returned results to generate node outputs. The predefined tool library is a collection of functional modules that execute specific tasks, covering the entire verification process. It includes six types of tools: surface distance calculation, database query, dynamic threshold generation, data analysis, data matching, and data modification. All are encapsulated in Python and support JSON format data interaction. The professional knowledge base stores rules, processes, and cases in the field of wireless resource verification, including parameter matching priorities, threshold standards, and anomaly handling logic. Through a retrieval-enhanced generation (RAG) mechanism, it ensures that decisions align with actual business scenarios.

[0074] In this embodiment, after receiving the latitude and longitude from the UAV survey data, the agent makes autonomous decisions based on preset prompts, including: first, acquiring all base station data for the target area, and then filtering geographically neighboring stations. Specifically, the agent calls a wireless resource database query tool, and uses a text2sql model to convert "query the antenna parameters of all networked base stations within the range of 30.50°-30.54°N and 114.34°-114.38°E" into an SQL statement. After execution, it obtains complete data for 6 base stations. Subsequently, it calls a ground distance calculation tool, uses the spherical distance formula to calculate the ground distance between each base station and the survey base station, and combines the "geographical proximity filtering threshold of 300 meters" rule retrieved from the knowledge base to finally filter out 3 qualified neighboring base stations (site IDs: JSQ-892, JSQ-905, JSQ-917), and outputs a dataset containing its core parameters such as frequency band and azimuth angle.

[0075] Using the neighboring base station parameters output from the previous step as input, and combining the confidence level of the survey data, the agent initiates multi-tool collaborative invocation. Specifically, the agent first invokes the dynamic threshold generation tool, inputting a confidence level of 0.83, and generates a frequency band matching threshold of 0.5 and an azimuth angle matching threshold of 1.2° based on the dynamic matching threshold calculation formula. Next, it invokes the data matching tool, calculating the intersection-over-union ratio (IoU) for the frequency bands (the surveyed frequency band intersects with the JSQ-905 frequency band at FDD900, IoU = 0.5; it has no intersection with the JSQ-917 frequency band, IoU = 0), and using an equivalence judgment with a ±0.5° tolerance for the azimuth angles (JSQ-905 azimuth angle 33.1°, deviation 0.5°; JSQ-917 azimuth angle 85.2°, deviation 52.6°). Combining the rule in the knowledge base that "frequency band matching priority is higher than azimuth angle," the agent selects two sites, JSQ-892 (target base station) and JSQ-905, that meet the threshold requirements and outputs their operational parameter data.

[0076] The agent continues the dynamic threshold mechanism, generating corresponding dynamic matching thresholds of 1.1 meters and 1.3 degrees for the recognition confidence levels of mounting height and mechanical tilt angle. After calling the data matching tool, the calculated mounting height deviation of JSQ-892 is 0.2 meters and the tilt angle deviation is 0.3 degrees, while the mounting height deviation of JSQ-905 is 3.7 meters and the tilt angle deviation is 2.1 degrees. According to the rule in the knowledge base that "mounting height and tilt angle must simultaneously meet the threshold requirements," the agent only retains JSQ-892 as the target verification base station, achieving precise focusing of the verification range.

[0077] The agent first calls the dynamic threshold tool to generate deviation thresholds of 1.2°, 1.1m, and 1.3° for the azimuth angle, mounting height, and mechanical downtilt angle, respectively. Then, it calls the data analysis tool to calculate the difference between the target base station survey work and the work parameters in the resource database: the azimuth angle deviation is 0.6°, the mounting height deviation is 0.2m, and the mechanical downtilt angle deviation is 0.3°, all within the dynamic threshold range. At the same time, it recalls the rule "no correction is needed when all work parameter deviations meet the thresholds" from the knowledge base. Combining the matching results of the frequency band intersection-to-parameter ratio of 1.0, the agent outputs the verification conclusion of "no correction needed" and detailed deviation data.

[0078] After receiving the verification conclusion, if the agent determines that no data modification is needed, it still calls the resource data modification tool to perform a "no-operation verification" to ensure the tool call chain is smooth. If the conclusion indicates that correction is needed (e.g., assuming a 2.5° deviation in direction angle), the tool will generate update SQL through a text2sql model, execute it after three retries, and then return feedback indicating successful modification. Throughout the process, the agent automatically records the tool call logs, threshold parameters, and verification results at each node, forming a traceable verification closed loop.

[0079] This embodiment achieves full automation from data collection to verification and correction through intelligent agent process orchestration, eliminating the need for manual intervention. Compared to traditional manual verification, this process reduces the verification time for a single base station from 2 hours to 15 minutes, improving efficiency by 87.5%. The dynamic threshold mechanism effectively avoids misjudgments caused by fixed thresholds. Combined with knowledge base support, the verification accuracy reaches 99.2%, fully verifying the feasibility and superiority of this method in real-world scenarios. Those skilled in the art can quickly reproduce the verification work of similar base stations by reusing the toolbox configuration, knowledge base content, and process node logic.

[0080] Another embodiment of the present invention provides a wireless resource data verification system. For details, please refer to [link to relevant documentation]. Figure 3 , Figure 3 The diagram shown illustrates a structural block diagram of a wireless resource data verification system according to one embodiment of the present invention, comprising: The acquisition module 11 is used to receive the first latitude and longitude data and raw image data of the survey base station collected by the UAV in response to the wireless resource data verification signal. The first filtering module 12 is used to obtain the second latitude and longitude data of each full base station in the wireless resource database to be checked, and to filter the difference in ground distance between the first latitude and longitude data and the second latitude and longitude data to obtain a number of geographically neighboring base stations. The recognition module 13 is used to perform image recognition processing on the original image data to obtain the first working parameter data and its corresponding confidence level; The second filtering module 14 is used to obtain the second working parameter data of each of the geographically neighboring base stations in the wireless resource database to be verified, and to perform filtering processing on the first working parameter difference between the first working parameter data and the second working parameter data to obtain the target base station to be verified. Calculation module 15 is used to determine the dynamic deviation threshold of the first working parameter data based on the confidence level, wherein the dynamic deviation threshold is designed to be adjusted accordingly based on the confidence level of the original image data; The verification module 16 is used to obtain the second target operating parameter data of the target base station in the wireless resource database to be verified, compare the second operating parameter difference between the first operating parameter data and the second target operating parameter data with the dynamic deviation threshold, and obtain the wireless resource data verification result of the target base station.

[0081] Preferably, in one embodiment of the present invention, the first screening module is specifically used for: Based on the preset surface distance calculation formula, the first longitude and first latitude in the first longitude and latitude data and the second longitude and second latitude in each second longitude and latitude data are calculated to obtain the surface distance difference between the survey base station and each full-scale base station; The surface distance difference between each location is compared with the preset geographic proximity screening threshold. The first full base station whose surface distance difference is not greater than the geographic proximity screening threshold is selected and identified as the geographic proximity base station. The geographic proximity screening threshold is the upper limit of the surface distance centered on the survey base station.

[0082] Preferably, in one embodiment of the present invention, the identification module is specifically used for: Target recognition is performed on the original image data to generate target detection boxes and their corresponding target detection confidence scores for the target objects. The target objects include at least the antennas of the survey base station. The antenna region image is obtained by cropping the target detection bounding box. The antenna region image is input into the trained engineering parameter classification and recognition model, and the first engineering parameter data and its corresponding engineering parameter classification confidence score are output. The confidence scores for target detection and engineering parameter classification are fused and calculated to obtain the confidence score corresponding to the first engineering parameter data.

[0083] Preferably, in one embodiment of the present invention, the first operating parameter data includes frequency band, azimuth angle, mounting height, and mechanical tilt angle; The second filtering module is specifically used for: The cross-intersection over union (CUI) ratio is calculated between the frequency band data in the first set of engineering parameters and the corresponding field data in each set of second engineering parameters. The equality of values ​​between the direction angle data in the first set of engineering parameters and the corresponding field data in each set of second engineering parameters is determined. Based on the CUI calculation results and the equality determination results, the first matching degree dataset is obtained. Based on the confidence levels corresponding to the frequency band data and azimuth angle data in the first engineering parameter data, a first dynamic matching threshold is generated, and the first matching degree dataset is filtered according to the first dynamic matching threshold to obtain the first candidate base station set; The hanging height data and mechanical tilt angle data in the first working parameter data are compared with the corresponding field data in the second working parameter data to determine the equality of their values, and the second matching degree dataset is obtained. Based on the confidence levels corresponding to the hanging height data and mechanical downtilt angle data in the first set of working parameters, a second dynamic matching threshold is generated, and the second matching degree dataset is filtered according to the second dynamic matching threshold to obtain the target verification base station.

[0084] Preferably, in one embodiment of the present invention, the verification module is specifically used for: Calculate the difference between the corresponding field data in the first working parameter data and the second target working parameter data to generate a second working parameter difference dataset with the corresponding difference values ​​of each working parameter field. The differences of each parameter field in the second parameter difference dataset are compared with the thresholds of the corresponding parameters in the dynamic deviation threshold to generate individual comparison results for each parameter field. The wireless resource data verification results of the target verification base station are generated based on the individual comparison results.

[0085] This solution introduces a dynamic threshold mechanism at each node of the wireless resource verification process. It automatically relaxes the threshold when the confidence level of the UAV survey data image recognition results is high, and automatically raises the matching standard when the confidence level is low, so as to avoid the cumulative error affecting the final verification result.

[0086] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method of wireless resource data verification, the method comprising: include: In response to the wireless resource data verification signal, it receives the first latitude and longitude data and raw image data of the survey base station collected by the UAV; Obtain the second latitude and longitude data of all base stations in the wireless resource database to be verified, and filter the difference in ground distance between the first latitude and longitude data and the second latitude and longitude data to obtain a number of geographically neighboring base stations. The original image data is subjected to image recognition processing to obtain the first working parameter data and its corresponding confidence level; Obtain the second working parameter data of each geographically neighboring base station in the wireless resource database to be verified, and perform filtering processing on the first working parameter difference between the first working parameter data and the second working parameter data to obtain the target base station to be verified; The dynamic deviation threshold of the first working parameter data is determined based on the confidence level, wherein the dynamic deviation threshold is designed to be adjusted accordingly based on the confidence level of the original image data; Obtain the second target operating parameter data of the target base station in the wireless resource database to be verified, compare the second operating parameter difference between the first operating parameter data and the second target operating parameter data with the dynamic deviation threshold, and obtain the wireless resource data verification result of the target base station.

2. The wireless resource data verification method of claim 1, wherein, The process of filtering the surface distance differences between the first latitude and longitude data and the second latitude and longitude data yields a number of geographically nearby base stations, including: Based on the preset surface distance calculation formula, the first longitude and first latitude in the first longitude and latitude data and the second longitude and second latitude in each of the second longitude and latitude data are calculated to obtain the surface distance difference between the survey base station and each of the full base stations; The surface distance difference values ​​are compared with a preset geographic proximity screening threshold. The first full base station whose surface distance difference value is not greater than the geographic proximity screening threshold is selected and determined as the geographic proximity base station. The geographic proximity screening threshold is the upper limit of the surface distance centered on the survey base station.

3. The wireless resource data verification method of claim 1, wherein, The step of performing image recognition processing on the original image data to obtain the first working parameter data and its corresponding confidence level includes: The original image data is subjected to target recognition to generate a target detection box of the target object and its corresponding target detection confidence score, wherein the target object includes at least the antenna of the survey base station; The antenna region image is obtained by cropping the target detection box. The antenna region image is input into the trained engineering parameter classification and recognition model, and the first engineering parameter data and its corresponding engineering parameter classification confidence score are output. The confidence scores for target detection and engineering parameter classification are fused and calculated to obtain the confidence score corresponding to the first engineering parameter data.

4. The wireless resource data verification method of claim 1, wherein, The first set of operating parameters includes frequency band, azimuth angle, mounting height, and mechanical tilt angle; The step of filtering the first parameter difference between the first and second engineering parameter data to obtain the target verification base station includes: The cross-intersection over union (CUI) ratio is calculated between the frequency band data in the first engineering parameter data and the corresponding field data in each of the second engineering parameter data. The equality of values ​​between the direction angle data in the first engineering parameter data and the corresponding field data in each of the second engineering parameter data is judged. Based on the CUI calculation results and the equality judgment results, the first matching degree dataset is obtained. Based on the confidence levels corresponding to the frequency band data and the azimuth angle data in the first engineering parameter data, a first dynamic matching threshold is generated, and the first matching degree dataset is filtered according to the first dynamic matching threshold to obtain a first candidate base station set; The hanging height data and mechanical tilt angle data in the first working parameter data are compared with the corresponding field data in the second working parameter data to determine the equality of their values, and a second matching degree dataset is obtained. Based on the confidence levels corresponding to the hanging height data and the mechanical tilt angle data in the first working parameter data, a second dynamic matching threshold is generated, and the second matching degree dataset is filtered according to the second dynamic matching threshold to obtain the target verification base station.

5. The wireless resource data verification method of claim 1, wherein, The step of comparing the second parameter difference between the first operating parameter data and the second target operating parameter data with the dynamic deviation threshold to obtain the radio resource data verification result of the target verification base station includes: Calculate the difference between the corresponding field data in the first working parameter data and the second target working parameter data to generate a second working parameter difference dataset with the corresponding difference of each working parameter field. The difference between each parameter field in the second parameter difference dataset is compared with the threshold of the corresponding parameter in the dynamic deviation threshold to generate a single comparison result for each parameter field. Based on the individual comparison results, the wireless resource data verification results of the target verification base station are generated.

6. A wireless resource data verification system, characterized by, include: The acquisition module is used to receive the first latitude and longitude data and raw image data of the survey base station collected by the UAV in response to the wireless resource data verification signal; The first filtering module is used to obtain the second latitude and longitude data of each full base station in the wireless resource database to be checked, and to filter the difference in ground distance between the first latitude and longitude data and the second latitude and longitude data to obtain a number of geographically neighboring base stations. The recognition module is used to perform image recognition processing on the original image data to obtain the first working parameter data and its corresponding confidence level; The second filtering module is used to obtain the second working parameter data of each geographically neighboring base station in the wireless resource database to be verified, and to filter the first working parameter difference between the first working parameter data and the second working parameter data to obtain the target base station to be verified. A calculation module is used to determine a dynamic deviation threshold for the first working parameter data based on the confidence level, wherein the dynamic deviation threshold is designed to be adjusted accordingly based on the confidence level of the original image data; The verification module is used to obtain the second target operating parameter data of the target base station in the wireless resource database to be verified, compare the second operating parameter difference between the first operating parameter data and the second target operating parameter data with the dynamic deviation threshold, and obtain the wireless resource data verification result of the target base station.

7. The wireless resource data verification system as described in claim 6, characterized in that, The first filtering module is specifically used for: Based on the preset surface distance calculation formula, the first longitude and first latitude in the first longitude and latitude data and the second longitude and second latitude in each of the second longitude and latitude data are calculated to obtain the surface distance difference between the survey base station and each of the full base stations; The surface distance difference values ​​are compared with a preset geographic proximity screening threshold. The first full base station whose surface distance difference value is not greater than the geographic proximity screening threshold is selected and determined as the geographic proximity base station. The geographic proximity screening threshold is the upper limit of the surface distance centered on the survey base station.

8. The wireless resource data verification system of claim 6, wherein, The identification module is specifically used for: The original image data is subjected to target recognition to generate a target detection box of the target object and its corresponding target detection confidence score, wherein the target object includes at least the antenna of the survey base station; The antenna region image is obtained by cropping the target detection box. The antenna region image is input into the trained engineering parameter classification and recognition model, and the first engineering parameter data and its corresponding engineering parameter classification confidence score are output. The confidence scores for target detection and engineering parameter classification are fused and calculated to obtain the confidence score corresponding to the first engineering parameter data.

9. The wireless resource data verification system of claim 6, wherein, The first set of operating parameters includes frequency band, azimuth angle, mounting height, and mechanical tilt angle; The second filtering module is specifically used for: The cross-intersection over union (CUI) ratio is calculated between the frequency band data in the first engineering parameter data and the corresponding field data in each of the second engineering parameter data. The equality of values ​​between the direction angle data in the first engineering parameter data and the corresponding field data in each of the second engineering parameter data is judged. Based on the CUI calculation results and the equality judgment results, the first matching degree dataset is obtained. Based on the confidence levels corresponding to the frequency band data and the azimuth angle data in the first engineering parameter data, a first dynamic matching threshold is generated, and the first matching degree dataset is filtered according to the first dynamic matching threshold to obtain a first candidate base station set; The hanging height data and mechanical tilt angle data in the first working parameter data are compared with the corresponding field data in the second working parameter data to determine the equality of their values, and a second matching degree dataset is obtained. Based on the confidence levels corresponding to the hanging height data and the mechanical tilt angle data in the first working parameter data, a second dynamic matching threshold is generated, and the second matching degree dataset is filtered according to the second dynamic matching threshold to obtain the target verification base station.

10. The wireless resource data verification system of claim 6, wherein, The verification module is specifically used for: Calculate the difference between the corresponding field data in the first working parameter data and the second target working parameter data to generate a second working parameter difference dataset with the corresponding difference of each working parameter field. The difference between each parameter field in the second parameter difference dataset is compared with the threshold of the corresponding parameter in the dynamic deviation threshold to generate a single comparison result for each parameter field. Based on the individual comparison results, the wireless resource data verification results of the target verification base station are generated.

Images