An unmanned aerial vehicle point cloud original piece data processing system with grading naming function

Abstract

The application relates to the technical field of point cloud data processing, and specifically discloses a UAV point cloud original piece data processing system with a grading naming function; the system comprises the following modules: a boundary division module generates a rectangular boundary suitable for a terrain and screens a to-be-graded original piece; a grading identification module constructs geometric constraints based on a projection distance, completes tower distribution through extreme value judgment and parallel scene processing rules; a naming adaptation module integrates multi-source data to construct a scene feature vector, divides a risk range through clustering, realizes differential naming by adopting multi-source filling and normalization sorting; a feedback checking module locates a preferential analysis interval through deviation analysis, constructs a space neighborhood group and judges a high-risk area, and replaces an original algorithm to complete grading renaming and closed-loop checking by adopting a space index-hash mapping method; the application realizes dynamic optimization of point cloud grading naming, improves grading precision and naming compliance, and provides data support for power transmission line operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of point cloud data processing technology, and specifically to a UAV point cloud original image data processing system with a file naming function. Background Technology

[0002] With the expansion of the power grid, the pressure of inspecting and maintaining transmission line corridors has increased dramatically. Drones, with their advantages of high efficiency and flexibility, have become a core operation and maintenance tool, playing a crucial role in mountainous inspections and the establishment of "one file, one problem" ledgers. However, the traditional rapid inspection mode is extremely inefficient, requiring a long time from photo collection to uploading and archiving for a single line, and necessitating the collaboration of multiple people and multiple devices.

[0003] Meanwhile, power transmission operation and maintenance also suffers from resource waste and technological gaps: Every quarter, key lines undergo point cloud scanning analysis to identify tree obstacles. Both scanning and rapid inspections fly along the lines, with highly similar execution methods, yet they are independent operations. The large number of optical images (point cloud images) collected simultaneously during point cloud scanning, although containing channel environmental information, are only used as auxiliary data for modeling and remain idle. Due to a lack of technical tools, the point cloud images cannot be properly categorized and named, making it difficult to convert them into rapid inspection photos. This results in repeated collection during rapid inspections, leading to resource waste.

[0004] Therefore, the present invention provides a UAV point cloud original image data processing system with graded naming function. Summary of the Invention

[0005] The purpose of this invention is to provide a UAV point cloud original image data processing system with a tiered naming function to solve the aforementioned background problems.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A UAV point cloud original image data processing system with file-level naming functionality includes the following modules:

[0008] Boundary segmentation module: used to acquire UAV point cloud original images and line ledger data, extract the tower span intervals from the line ledger data and perform boundary shaping to obtain rectangular boundaries; match the point cloud original images with the rectangular boundaries to obtain the original images to be segmented;

[0009] Segmentation and recognition module: Used to build a segmentation and recognition model, input the original film to be segmented into the segmentation and recognition model to perform pole and tower allocation processing on the original film to be segmented;

[0010] Naming adaptation module: used to obtain the association information corresponding to the point cloud original images assigned to the tower, construct scene feature vectors based on the association information and perform scene segmentation, establish dynamic naming templates; and perform adaptive naming of the point cloud original images based on the dynamic naming templates and in combination with the scene.

[0011] Feedback verification module: used to obtain the original point cloud images of the classification and naming deviation in the fault verification, and establish the deviation original image group; obtain the tower span interval corresponding to the deviation original image group for deviation analysis, and obtain the priority analysis interval; establish a high-risk neighborhood group based on the priority analysis interval, and classify and rename the point cloud original images of the high-risk neighborhood group.

[0012] As a further aspect of the present invention, the boundary shaping is performed as follows:

[0013] The list of poles and towers is obtained by calling the overhead line files in the power system and parsing them through the program.

[0014] Extract the line number, tower number, and tower latitude and longitude from the tower ledger list to establish a tower coordinate database;

[0015] Calculate the basic boundary of the tower span interval based on the tower coordinate library;

[0016] The dynamic error range of the boundary is determined by obtaining historical GPS deviations. The coordinates are then dynamically expanded based on the dynamic error range to determine the rectangular boundary of the tower span interval.

[0017] As a further aspect of the present invention, the method for constructing the graded recognition model is as follows:

[0018] Obtain the original images to be divided for each tower span interval and initialize the minimum projection distance;

[0019] Geometric constraints are constructed based on the minimum projection distance to determine candidate matching line segments;

[0020] Traverse all tower segments and assign towers to the original point cloud image.

[0021] As a further aspect of the present invention, the geometric constraints are constructed as follows:

[0022] Obtain the standardized coordinates of the original image to be sorted and mark it as point P. Mark the starting tower of the tower line segment as point A and the ending tower as point B. The tower line segment is AB.

[0023] Establish a vector pointing from point A to point P. The vector pointing from point A to point B ;

[0024] Construct a projection scaling equation, and transform the vector , Input the projection scale equation to obtain the projection scale coefficient;

[0025] Determine whether the projection scale factor is within the preset constraint range. If it is, determine the projection point P1 of point P on AB based on the projection scale factor, and calculate the projection distance between point P and projection point P1.

[0026] Geometric constraints are constructed by comparing the projected distance and the minimum projected distance.

[0027] As a further aspect of the present invention: the dynamic naming template is established in the following way:

[0028] Obtain the scene feature vectors of all point cloud original images after tower allocation is completed, and construct a scene feature dataset;

[0029] A scene classification model is constructed based on a clustering algorithm. The scene feature dataset is input into the scene classification model to classify the scene risk type of the tower span interval and obtain the high-risk tower span.

[0030] Obtain the voltage level and scenario risk type classification of the line, and establish a dynamic naming template based on the classification and identification results.

[0031] As a further aspect of the present invention, the method for performing the adaptation naming is as follows:

[0032] Obtain the initial naming original image within the same tower span interval, and calculate the spatial distance from the initial naming original image to the tower on the smaller side of the tower span interval;

[0033] If the initial naming original is within the low-risk tower span, then the initial naming originals in the same tower span range are sorted according to spatial distance to obtain the basic sequence number of the initial naming original.

[0034] The basic sequence number is added to the naming of the original point cloud image to achieve the file-based naming of the original point cloud image;

[0035] Obtain the cloud image sorting value, sort the point cloud original images within the high-risk tower span according to the cloud image sorting value, and obtain the risk sequence number of the initial named original image;

[0036] The risk number and the basic number are added together to the naming of the point cloud original image to achieve graded naming of the point cloud original image.

[0037] As a further aspect of the present invention: the method for obtaining the cloud sorting value is as follows:

[0038] If the original image is located within the span of a high-risk tower, then the distance ratio weight between the spatial distance and the tower span is calculated.

[0039] Obtain the historical maintenance rate of the small side towers within the initial naming original image of the high-risk tower span, and obtain the historical maintenance rate of the small side towers from the maintenance logbook;

[0040] Normalize the reciprocal of the historical maintenance rate, the historical maintenance rate, and the distance ratio weight, and then multiply the three to obtain the cloud sorting value.

[0041] As a further aspect of the present invention: the high-risk neighborhood group is established in the following way:

[0042] The spatial neighborhood group is obtained by combining the M tower span intervals that are adjacent to the priority analysis interval.

[0043] If there are high-risk tower spans within the spatial neighborhood group, then construct the first neighborhood judgment equation to obtain the sum of the relative deviations of all intervals within the spatial neighborhood group, the total number of intervals in the neighborhood group, and the proportion of high-risk tower spans. Input these three factors into the first neighborhood judgment equation to obtain the first neighborhood judgment value.

[0044] Construct a second neighborhood decision equation, input the sum of the relative deviations of all intervals within the spatial neighborhood group and the total number of intervals in the neighborhood group into the second neighborhood decision equation to obtain the second neighborhood decision value;

[0045] Construct comparison criteria and obtain the first or second neighborhood judgment value of all spatial neighborhood groups in the scene. If the first or second neighborhood judgment value of a spatial neighborhood group satisfies the corresponding comparison criteria, then the spatial neighborhood is marked as a high-risk neighborhood group.

[0046] As a further aspect of the present invention, the method for obtaining the priority analysis interval is as follows:

[0047] Obtain the number of point cloud deviation original images for each tower span interval within the monitoring period from the deviation original image group, as well as the total number of original images for each interval;

[0048] The interval deviation rate is obtained by comparing the number of original point cloud images with the total number of original images in each tower span interval during the monitoring period.

[0049] Obtain the interval deviation rate for each tower span interval, and the average interval deviation rate for the scene to which each tower span interval belongs, to obtain the scene baseline rate;

[0050] Calculate the ratio of the interval deviation rate to the scene reference rate to obtain the relative deviation ratio;

[0051] A statistical significance test is performed on the relative deviations of all tower span intervals in the scenario. If the statistical significance test is satisfied, the deviation of the tower span interval in the current scenario is determined to be a systematic anomaly.

[0052] If it is a systemic anomaly, the tower span intervals are sorted in descending order according to the relative deviation comparison within the same scenario, and the tower span interval corresponding to the maximum value of the relative deviation comparison is taken as the priority analysis interval.

[0053] As a further aspect of the present invention, the method for renaming the file categories is as follows:

[0054] The original point cloud images were re-filed and renamed by replacing the original interval border inspection traversal method within the high-risk neighborhood group with the spatial index-hash mapping coordinate matching method.

[0055] The beneficial effects of this invention are:

[0056] (1) In the boundary shaping stage, the dynamic error range is calculated by combining the historical average GPS deviation, average positioning error, and terrain scene coefficient. The basic boundary is then expanded into a rectangular boundary that adapts to the actual data collection environment, effectively connecting the relationship between GPS positioning data and the spatial location of the tower. By traversing the coordinates of the original point cloud images one by one to filter the original images to be graded, effective data related to the tower span interval can be retained, while irrelevant original images can be filtered out, thus realizing the connection between front-end data collection and back-end grading processing.

[0057] (2) By initializing the maximum value as the minimum projection distance and combining vector operations to calculate the projection ratio coefficient and projection distance, a grading recognition logic based on geometric constraints is constructed. This logic can specifically handle matching scenarios of point cloud original images in the extended line area of ​​tower segments, thus improving the coverage of grading recognition. By traversing all tower segments to complete the original image allocation and associating it with the tower ledger list, the point cloud original images can correspond to clear tower span information, which helps to ensure the orderliness of the grading process and achieve deep binding between the grading results and the ledger information.

[0058] (3) Integrate multi-source data such as grading identification results, line ledgers, UAV logs, and maintenance logs, and construct scene feature vectors through numerical and normalization processing to provide comprehensive data dimensions for scene risk classification. Classify tower spans by risk so that the naming rules can adapt to the different needs of different risk scenarios. Design differentiated sorting strategies for high and low risk spans, and integrate maintenance-related data such as repair rate and maintenance rate into the sorting logic of high-risk tower spans, so that the naming results not only include basic identification information, but also reflect the key points of maintenance, and enhance the relevance of naming to actual business needs.

[0059] (4) Extracting grading and naming deviation data from transmission line fault verification, and preprocessing to remove deviation originals that are not due to algorithmic reasons, ensures the pertinence of deviation analysis. In the neighborhood group construction stage, some processing rules involving scenarios and sorting logic for parallel priority analysis intervals are added to ensure the comprehensiveness of high-risk area location. The neighborhood judgment value is calculated by combining the proportion of high-risk tower spans, so that the identification of high-risk neighborhood groups has both data objectivity and scenario adaptability. The original algorithm is replaced by the spatial index-hash mapping method, and a closed-loop verification mechanism for the optimized deviation rate is established, forming an iterative link of deviation identification - area location - algorithm optimization - effect verification, which is conducive to the continuous optimization of grading and naming quality. Attached Figure Description

[0060] The invention will now be further described with reference to the accompanying drawings.

[0061] Figure 1 This is a block diagram of a UAV point cloud original image data processing system with graded naming function according to the present invention;

[0062] Figure 2 This is a schematic diagram of point cloud scanning in this invention;

[0063] Figure 3 This is a schematic diagram of the parsing architecture line template of the present invention;

[0064] Figure 4 This is a schematic diagram of the interval boundary inspection traversal coordinate matching method of the present invention;

[0065] Figure 5 This is a schematic diagram of the spatial index-hash mapping coordinate matching method of the present invention;

[0066] Figure 6 This is a flowchart of a method for processing UAV point cloud original image data with a file naming function according to the present invention. Detailed Implementation

[0067] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Example 1:

[0069] Please see Figure 1 As shown, this invention is a UAV point cloud original image data processing system with graded naming function, including the following modules:

[0070] Boundary segmentation module: used to acquire UAV point cloud original images and line ledger data, extract the tower span intervals from the line ledger data and perform boundary shaping to obtain rectangular boundaries; match the point cloud original images with the rectangular boundaries to obtain the original images to be segmented;

[0071] The method for acquiring original drone point cloud images and route ledger data, and then classifying and identifying the original point cloud images and route ledger data, is as follows:

[0072] In some embodiments, such as Figure 2 As shown, by equipping a drone with a laser camera, a three-dimensional scan of the power transmission line channel is performed to generate an optical image, which serves as the point cloud image.

[0073] Obtain the GPS latitude and longitude contained in the EXIF ​​(Exchangeable Image File) of the original point cloud image, and convert the GPS latitude and longitude into standardized decimal coordinates;

[0074] Obtain the log trajectory coordinates of the UAV during point cloud image acquisition, and calculate the trajectory deviation between the standardized coordinates and the UAV log trajectory coordinates;

[0075] It should be noted that the trajectory deviation of the drone log trajectory coordinates is calculated by: calculating the Euclidean distance between the log trajectory coordinates and the standardized coordinates as the trajectory deviation.

[0076] If the trajectory deviation exceeds the preset trajectory deviation threshold, a coordinate alarm will be triggered. The EXIF ​​coordinates and log trajectory coordinates of the original point cloud image will be manually reviewed to determine whether the EXIF ​​coordinates are incorrect or the log trajectory coordinates are drifting.

[0077] If the EXIF ​​coordinates are incorrect, the standardized coordinates will be discarded.

[0078] If the trajectory deviation is lower than the preset trajectory deviation threshold, no action will be taken.

[0079] like Figure 3 As shown, by calling the overhead line file in the power system, the program parses the data to obtain the line ledger data containing the pole and tower ledger list, and extracts the line number, pole and tower number, pole and tower latitude and longitude, pole and tower span range and voltage level from the pole and tower ledger list;

[0080] Establish a tower coordinate database based on the line number, tower number, and tower latitude and longitude coordinates in the tower ledger list;

[0081] like Figure 4 As shown, the coordinate matching method of interval boundary inspection is adopted to determine the rectangular boundary of each tower span interval based on the tower coordinate library and historical error range, thereby realizing boundary shaping;

[0082] The method for determining the rectangular boundary of each tower span interval based on the tower coordinate library and historical error range is as follows:

[0083] S101. Calculate the basic boundary of the tower span interval based on the tower coordinate library;

[0084] Preferably, the towers at both ends of the span are... The basic boundary is: ;

[0085] S102. Obtain the dynamic error range of the historical GPS deviation to determine the boundary, and perform dynamic coordinate expansion processing in combination with the dynamic error range to determine the rectangular boundary of the tower span interval.

[0086] Preferably, by formula: Obtain the dynamic error range ;

[0087] in, This represents the historical average GPS deviation. denoted as the mean GPS positioning error, and 'a' as the scene coefficient.

[0088] It should be noted that the scene coefficient is 1.0 for plains and 1.5 for mountainous areas; the mean positioning error and scene coefficient are set by professionals in this field based on experience.

[0089] Extend the latitude and longitude of the basic boundary outward. The corresponding coordinate increments are used as the rectangle boundaries;

[0090] It should be noted that the coordinate increments corresponding to the outward extension of the latitude and longitude of the basic boundary are used as the rectangular boundary; the outward extension is achieved by converting the length unit Δ (meter) to the latitude and longitude increment (degree) using an approximate formula: 1 meter ≈ 1 × 10⁻⁶. -5 degrees (latitude direction), 1 meter ≈ (1×10) -5 ) / cos(latitude value) degrees (longitude direction) to ensure that the extended range matches the error range;

[0091] The standardized coordinates of the original point cloud image are traversed one by one to determine whether the standardized coordinates fall within the rectangular boundary.

[0092] If it falls within the rectangular boundary, it is classified into the tower span range corresponding to the rectangular boundary.

[0093] If the standardized coordinates do not fall within the rectangular boundary, continue the matching process; obtain all the original point cloud images within the tower span interval as the original images to be segmented.

[0094] Segmentation and recognition module: Used to build a segmentation and recognition model, input the original film to be segmented into the segmentation and recognition model to perform pole and tower allocation processing on the original film to be segmented;

[0095] The method for constructing the graded recognition model is as follows:

[0096] S201. Obtain the original image to be divided for each tower span interval and initialize the minimum projection distance;

[0097] Preferably, the minimum projected distance of the original piece to be divided in each tower span interval is obtained, and the minimum projected distance is initialized to a maximum value to record the minimum distance from the original piece to be divided to the tower line segment (the line segment formed by two adjacent towers in the tower span interval).

[0098] It should be noted that the minimum projection distance can be initialized to a maximum value using a positive infinity constant in programming languages, such as float('inf') in Python.

[0099] S202. Construct geometric constraints based on the minimum projection distance to determine candidate matching line segments;

[0100] Preferably, the standardized coordinates of the original image to be graded are marked as point P, the starting tower of the tower line segment is marked as point A, the ending tower is marked as point B, and the tower line segment is AB;

[0101] Calculate the vertical distance d from the original sheet to be graded to the pole segment AB;

[0102] By using the projection ratio equation: Obtain the projection scale factor t;

[0103] It is understandable that the physical meaning of the projection scale coefficient is: the quantified value of the relative projection position of the standardized coordinates (point P) of the original image to be divided in the direction of the tower line segment (AB, where A and B are the towers at both ends of the span). It is calculated by dividing the dot product of vector AP (A points to P) and vector AB (A points to B) by the dot product of vector AB itself. When the coefficient is in the range of [0,1], it indicates that the projection of point P falls on the tower line segment AB. When the coefficient is less than 0, the projection falls outside point A. When the coefficient is greater than 1, the projection falls outside point B. This provides a basis for position judgment for subsequent calculation of projection distance or endpoint distance and construction of geometric constraints.

[0104] in, Let A be the vector pointing from point A to point P. Let A be the vector pointing from point A to point B;

[0105] Determine if the projection scale factor is within the constraint range [0,1]. If it is, determine the projection point P1 of point P on AB based on the projection scale factor, and calculate the projection distance between point P and projection point P1. ;

[0106] It should be noted that the coordinates of the projection point P1 are determined by multiplying the coordinates of point A plus t and the vector AB, and the projection distance is calculated by calculating the Euclidean distance between the standardized coordinates of point P and point P1.

[0107] If not, it means that point P is in the extension area of ​​the tower line segment AB. The matching logic needs to be completed by calculating the endpoint distance: if t<0, calculate the straight-line distance d_A from point P to tower A (the starting point of the line segment);

[0108] If t>1, calculate the straight-line distance d_B from point P to tower B (the endpoint of the line segment); take the smaller value of d_A or d_B as the equivalent projected distance and compare it with the current minimum projected distance: if the equivalent projected distance ≤ the current minimum projected distance, update the "minimum projected distance" to this value and mark the corresponding endpoint (A or B) of the current tower line segment as a candidate matching reference point; if the equivalent projected distance > the current minimum projected distance, continue traversing the next tower line segment;

[0109] The geometric constraints are constructed by comparing the projected distance with the minimum projected distance.

[0110] Preferably, the comparison process is performed as follows: if the projection distance If the minimum projected distance is less than or equal to the minimum projected distance, then update the minimum projected distance to... The current tower segment is selected as a candidate matching segment.

[0111] If the projection distance If the distance is higher than the minimum projection distance, continue traversing the next tower segment;

[0112] S203. Traverse all tower segments and assign towers to the original point cloud image.

[0113] Preferably, after all tower segments have been traversed, the tower segment with the smallest projection distance is selected; if multiple tower segments have the same minimum projection distance, the segment with the smaller tower number is selected first, and the current point cloud image is assigned to the smaller side tower of that segment (i.e., the tower with the smaller number in the segment); if all segments have not been traversed, the matching of the next segment continues.

[0114] After completing the matching of the current point cloud original image, select the next point cloud original image to be processed and repeat steps S201-203 until all original images to be sorted are assigned to towers.

[0115] The original image of the pole and tower allocation point cloud will be associated with the pole and tower ledger list.

[0116] It is understandable that the purpose of assigning poles to the original point cloud image is as follows:

[0117] Function 1: To establish a connection between the original images to be graded and the specific tower span intervals and tower ledger information, providing basic data with spatial attributes for subsequent integration of multi-source data to construct scene feature vectors and establish dynamic naming templates;

[0118] Secondly, it clarifies the tower span intervals corresponding to each point cloud original image, enabling the statistics and deviation analysis of the number of deviation original images in each interval and the total number of original images in each interval within the monitoring period, which is beneficial for locating high-risk neighborhood groups.

[0119] Example 2:

[0120] Please see Figure 1 As shown, the present invention is a UAV point cloud original image data processing system with graded naming function, and also includes the following modules:

[0121] Naming adaptation module: used to obtain the association information corresponding to the point cloud original images assigned to the tower, construct scene feature vectors based on the association information and perform scene segmentation, establish dynamic naming templates; and perform adaptive naming of the point cloud original images based on the dynamic naming templates and in combination with the scene.

[0122] This includes obtaining the association information corresponding to the original point cloud images allocated to the towers, constructing scene feature vectors based on the association information, performing scene segmentation, and establishing a dynamic naming template:

[0123] Preferably, the line number and tower span are read from the grading and identification results; at the same time, terrain scene information (such as plains, mountains, etc.) is obtained from the line ledger data, the acquisition height of the point cloud original image in the UAV log, and the historical maintenance rate of tower section failures is obtained from the ledger maintenance log;

[0124] The line number, tower span, terrain scene information, acquisition height, and historical maintenance rate of the point cloud original images of a single completed tower allocation are integrated and quantified (i.e., the terrain in the terrain scene information is expressed in numerical form, 1 represents low altitude in plains, 2 represents high altitude in mountains, and 3 represents low altitude in mountains) and normalized to construct a scene feature vector.

[0125] Obtain the scene feature vectors of all point cloud original images after tower allocation is completed, and construct a scene feature dataset;

[0126] A scene classification model is constructed based on the K-means clustering algorithm. The scene feature dataset is input into the scene classification model to classify the scene risk type of the tower span interval and obtain the high-risk tower span.

[0127] Those skilled in the art will understand that when constructing a scene classification model based on the K-means clustering algorithm, the number of clusters K is determined by the elbow rule (or the conventional method for determining the number of clusters in this field). Then, iterative clustering is performed on the normalized scene feature vector dataset that integrates the line number of the original point cloud image, the span of the tower, the numerical terrain scene information, the collection height, and the historical maintenance rate: K feature vectors are randomly selected as the initial cluster centers, the distance between each vector and the center is calculated and assigned to the nearest cluster, the cluster centers are updated and the iteration is repeated until the centers are stable, and finally multiple clusters are formed, which correspond to the classification of multiple scene risks in the tower span interval (such as obtaining different clustering results such as high risk and low risk), providing a basis for scene classification for the establishment of dynamic naming templates;

[0128] Obtain the voltage level and scenario risk type classification of the line, and establish a dynamic naming template, i.e., a preset naming rule library, based on the classification and identification results;

[0129] For example, the template for 220kV - Mountainous Area - Low Risk is: Voltage Level - Terrain - Line Name - Tower Number; the template for 220kV - Mountainous Area - High Risk is: Voltage Level - Terrain - Line Name - High Risk - Tower Number.

[0130] Obtain the list of pole and tower ledgers associated with the point cloud original images that have completed the grade recognition, and extract the associated matching information;

[0131] The association matching information is checked against the dynamic naming template for field integrity and compliance. Point cloud originals that meet the compliance and integrity requirements are selected and named according to the dynamic naming template as the initial naming originals.

[0132] If the original point cloud image does not meet the requirements for compliance and completeness, information should be filled into the original point cloud image.

[0133] It should be noted that the information is populated according to priority and conflict resolution rules.

[0134] Data source priority: Line log data (terrain, voltage level) > UAV log data (collection altitude) > Historical data (line number);

[0135] Conflict resolution: When information from multiple data sources conflicts, a weighted voting method (60% for ledgers, 30% for logs, and 10% for historical data) is used to determine the fill value;

[0136] Extreme scenario: When there is no data source, mark it as needing manual completion, pause the naming process and push an alert;

[0137] The method of adaptively naming the original point cloud image based on a dynamic naming template and combined with the scene is as follows:

[0138] Obtain the initial naming original image within the same tower span interval, and calculate the spatial distance from the initial naming original image to the tower on the smaller side of the tower span interval;

[0139] If the initial naming original is within the low-risk tower span, then the initial naming originals in the same tower span range are sorted in order of increasing spatial distance to obtain the basic serial number of the initial naming original.

[0140] The basic sequence number is added to the naming of the original point cloud image to achieve the file-based naming of the original point cloud image;

[0141] If the original image is located within the span of a high-risk tower, then the distance ratio weight between the spatial distance and the tower span is calculated.

[0142] Obtain the historical maintenance rate of the small side towers within the initial naming original image of the high-risk tower span, and obtain the historical maintenance rate of the small side towers from the maintenance logbook;

[0143] It should be noted that the historical maintenance rate is the percentage of maintenance and upkeep performed on the small side towers within the historical monitoring period.

[0144] The reciprocal of the historical maintenance rate, the historical maintenance rate, and the distance ratio weight are all processed by Min-Max normalization (mapped to the [0,1] interval), and the three are multiplied to obtain the cloud sorting value;

[0145] The point cloud original images within the high-risk tower span are sorted in ascending order according to the numerical value of the cloud image sorting value to obtain the risk number of the initial named original image.

[0146] The risk number and the basic number are added together to the naming of the point cloud original image to achieve the graded naming of the point cloud original image;

[0147] It should be noted that the purpose of constructing risk serial numbers and basic serial numbers is as follows:

[0148] Objective 1: To provide an orderly identification basis for the classification and naming of point cloud original images: The basic sequence number is sorted by spatial distance to standardize the naming order of original images within the span of low-risk towers, and the risk sequence number is combined with operation and maintenance data (repair rate, maintenance rate) and spatial information to form a priority identifier for original images within the span of high-risk towers, so as to avoid confusion in the naming of original images within the same span and ensure the regularity of the naming results.

[0149] Objective 2: Connecting naming results with operation and maintenance needs: Risk serial numbers highlight the original images that require key attention within high-risk ranges, while basic serial numbers clarify the spatial relative location of the original images. When both are added to the naming, it is easier for operation and maintenance personnel to quickly locate key areas of original images and identify key operation and maintenance points through naming, providing a clear data index for hidden danger investigation and ledger management.

[0150] Feedback verification module: used to obtain the original point cloud images of the classification and naming deviation in the fault verification, and establish the deviation original image group; obtain the tower span interval corresponding to the deviation original image group for deviation analysis, and obtain the priority analysis interval; establish a high-risk neighborhood group based on the priority analysis interval, and classify and rename the point cloud original images of the high-risk neighborhood group.

[0151] The method for obtaining the original point cloud images with graded naming deviations during fault verification and establishing a deviation image group is as follows:

[0152] Preferably, point cloud original images with graded naming deviations are obtained from the fault verification of the transmission line channel as point cloud deviation original images; at the same time, the associated data of the point cloud deviation original images are obtained to construct a deviation original image group containing the point cloud deviation original images and associated data.

[0153] Among them, the method for obtaining the tower span interval corresponding to the original deviation image group and performing deviation analysis to determine the priority analysis interval is as follows:

[0154] Further preprocessing and filtering of the original image group with deviations will remove point cloud images with deviations that are not due to algorithmic reasons.

[0155] The number of point cloud deviation original images for each tower span interval within the monitoring period is obtained from the deviation original image group, and the number of point cloud original images for all tower span intervals that have been classified and named within the monitoring period is used as the total number of original images for the interval.

[0156] The interval deviation rate is obtained by comparing the number of original point cloud images with the total number of original images in each tower span interval during the monitoring period.

[0157] Obtain the interval deviation rate for each tower span interval, and the average interval deviation rate for the scene to which each tower span interval belongs, to obtain the scene baseline rate;

[0158] Calculate the ratio of the interval deviation rate to the scene reference rate to obtain the relative deviation ratio;

[0159] A statistical significance test is performed on the relative deviations of all tower span intervals in the scene. If the statistical significance test is satisfied, the deviation of the tower span interval in the current scene is determined to be a systematic anomaly and the deviation analysis signal of the current scene is triggered.

[0160] Those skilled in the art will understand that conducting a statistical significance test refers to using the Poisson test to compare the relative deviations of the pole span intervals within a scene, calculating the theoretical random deviation number of the interval (total number of original images in the interval × scene baseline rate), and obtaining the probability (p-value) that the actual deviation number is greater than or equal to the observed value through the cumulative probability formula of the Poisson distribution, which is used to distinguish the nature of the deviation.

[0161] Satisfying the statistical significance test means that the probability (p-value) calculated by the Poisson test is <0.05, indicating that the deviation in the span range of the tower is not random, but a result of a systematic problem.

[0162] If the statistical significance test is not met, the deviation of the tower span interval is determined to be random fluctuation, and no action is taken.

[0163] If the current scene triggers the deviation analysis signal, the tower span intervals within the same scene are sorted in descending order according to the relative deviation ratio, and the tower span interval corresponding to the maximum value of the relative deviation ratio is taken as the priority analysis interval.

[0164] The method for classifying and renaming the original point cloud images of high-risk neighborhood groups based on priority analysis intervals is as follows:

[0165] The spatial neighborhood group is obtained by combining the M tower span intervals that are adjacent to the priority analysis interval.

[0166] Preferably, M=2;

[0167] It should be noted that if a tower span interval is already included in a previous spatial neighborhood group, it will not participate in subsequent spatial neighborhood combinations. If only some of the M adjacent intervals before and after the priority analysis interval are not included, then only the unincluded intervals will be combined with the priority analysis interval to form a spatial neighborhood group (e.g., priority analysis interval #005-#006, the preceding neighbor #004-#005 is included, and the following neighbors #006-#007 and #007-#008 are not included, then the spatial neighborhood groups are #005-#006, #006-#007, and #007-#008). If the deviations of multiple tower span intervals are relatively the same, then the neighborhood combinations of each priority analysis interval will be processed in ascending order of tower number.

[0168] If there are high-risk tower spans within the spatial neighborhood group, the first neighborhood judgment value is obtained by using the neighborhood first judgment equation: the sum of the relative deviations of all intervals within the spatial neighborhood group ÷ the total number of intervals in the neighborhood group × (1 + the proportion of high-risk tower spans).

[0169] The proportion of high-risk tower spans is obtained through the high-risk tower span proportion equation: High-risk tower span proportion = Number of high-risk tower spans in the neighboring group ÷ Total number of intervals in the neighboring group.

[0170] The second neighborhood determination value is obtained by using the neighborhood second determination equation: Neighborhood second determination value = Sum of relative deviations of all intervals within the spatial neighborhood group ÷ Total number of intervals in the neighborhood group;

[0171] Construct comparison criteria and obtain the first or second neighborhood judgment value of all spatial neighborhood groups in the scene. If the first or second neighborhood judgment value of a spatial neighborhood group satisfies the corresponding comparison criteria, then the spatial neighborhood is marked as a high-risk neighborhood group.

[0172] Preferably, the method for constructing the comparison criteria is as follows: obtain the first neighborhood judgment value or the second neighborhood judgment value of all spatial neighborhood groups in the scene, compare the first neighborhood judgment value with the preset first neighborhood judgment threshold, and compare the second neighborhood judgment value with the preset second neighborhood judgment threshold.

[0173] If the first judgment value of the neighborhood is higher than or equal to the preset first judgment threshold of the neighborhood, or the second judgment value of the neighborhood is higher than or equal to the preset second judgment threshold of the neighborhood, then the corresponding spatial neighborhood group is marked as a high-risk neighborhood group.

[0174] If the first determination value of the neighborhood is lower than the preset first determination threshold of the neighborhood, or the second determination value of the neighborhood is lower than the preset second determination threshold of the neighborhood, then the corresponding spatial neighborhood group is marked as a low-risk neighborhood group.

[0175] It should be noted that the second and first neighborhood decision thresholds are derived from historical scene data using the 3σ principle.

[0176] Second threshold for neighborhood determination = mean of relative deviation of historical neighborhood + 3 × standard deviation of relative deviation of historical neighborhood;

[0177] The first threshold for determining the neighborhood is equal to the second threshold for determining the neighborhood by 1.2 (plus a correction factor for high-risk scenarios).

[0178] like Figure 5 As shown, the original interval border inspection traversal method within the high-risk neighborhood group is replaced by the spatial index-hash mapping coordinate matching method. The latitude and longitude intervals corresponding to the transmission lines within the high-risk neighborhood group are divided into multiple small grids by multiple binary divisions. Geohash encoding is used to compress the latitude and longitude into short strings, so that each has its own unique code.

[0179] A hash table is built based on the unique code of each cell, and Geohash is used for coarse screening and then the cells corresponding to the original point cloud image are verified.

[0180] Based on the corresponding grid, the original point cloud image is re-filed and renamed, and the deviation is checked after re-filed and renamed until the file-filed and named original point cloud image meets the accuracy requirements.

[0181] It should be noted that the deviation verification method is as follows: After naming, calculate the "optimized deviation rate" of the high-risk neighborhood group (number of original images with optimized deviation ÷ total number of original images in the interval × 100%).

[0182] If the optimized deviation rate is less than the scene baseline rate × 1.2, then the configuration parameters of the spatial index-hash mapping method are fixed.

[0183] If the optimized deviation rate is greater than or equal to the scene reference rate × 1.2, then reduce the mesh division precision (e.g., adjust from 10m × 10m to 5m × 5m), and repeat the subdivision and naming process until the precision requirements are met.

[0184] Example 3:

[0185] like Figure 6 As shown, the present invention is a method for processing UAV point cloud original image data with graded naming function, including the following steps:

[0186] Step 1: Obtain the original point cloud image of the UAV and the line ledger data. Extract the tower span interval from the line ledger data and shape the boundary to obtain a rectangular boundary. Match the original point cloud image with the rectangular boundary to obtain the original image to be divided into spans.

[0187] Step 2: Construct a grading recognition model. Input the original film to be graded into the grading recognition model to perform pole and tower allocation processing on the original film to be graded.

[0188] Step 3: Obtain the association information corresponding to the point cloud original images assigned to the towers, construct scene feature vectors based on the association information and perform scene segmentation, and establish dynamic naming templates; perform adaptive naming of the point cloud original images based on the dynamic naming templates and in combination with the scene.

[0189] Step 4: Obtain the original point cloud images of the classification and naming deviations during fault verification, and establish a deviation original image group; obtain the tower span intervals corresponding to the deviation original image group for deviation analysis to obtain the priority analysis interval; establish a high-risk neighborhood group based on the priority analysis interval, and classify and rename the point cloud original images of the high-risk neighborhood group.

[0190] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A UAV point cloud original image data processing system with tiered naming function, characterized in that: Includes the following modules: Boundary segmentation module: used to acquire UAV point cloud original images and line ledger data, extract the tower span intervals from the line ledger data and perform boundary shaping to obtain rectangular boundaries; match the point cloud original images with the rectangular boundaries to obtain the original images to be segmented; Segmentation and recognition module: Used to build a segmentation and recognition model, input the original film to be segmented into the segmentation and recognition model to perform pole and tower allocation processing on the original film to be segmented; The method for constructing the graded recognition model is as follows: S201. Obtain the original image to be divided for each tower span interval and initialize the minimum projection distance; Obtain the minimum projected distance of the original piece to be divided in each tower span interval, and initialize the minimum projected distance to a maximum value to record the minimum distance from the original piece to be divided to the tower line segment. The minimum projection distance is initialized to a maximum value using a positive infinity constant in the programming language; S202. Construct geometric constraints based on the minimum projection distance to determine candidate matching line segments; The standardized coordinates of the original image to be graded are marked as point P. The starting tower of the tower line segment is marked as point A, and the ending tower is marked as point B. The tower line segment is AB. Calculate the vertical distance d from the original sheet to be graded to the pole segment AB; By using the projection ratio equation: Obtain the projection scale factor t; The physical meaning of the projection scale factor is: the quantified value of the relative projection position of the standardized coordinates of the original image to be graded in the direction of the tower line segment. It is calculated by dividing the dot product of vector AP and vector AB by the dot product of vector AB itself. When the coefficient is in the range, it indicates that the projection of point P falls on the tower line segment AB. When the coefficient is less than 0, the projection falls outside point A. When the coefficient is greater than 1, the projection falls outside point B. This provides a basis for position judgment for subsequent calculation of projection distance or endpoint distance and construction of geometric constraints. in, Let A be the vector pointing from point A to point P. Let A be the vector pointing from point A to point B; Determine if the projection scale factor is within the constraint range. If it is, determine the projection point P1 of point P on AB based on the projection scale factor, and calculate the projection distance between point P and projection point P1. ; The coordinates of the projection point P1 are determined by multiplying the coordinates of point A plus t and the vector AB. The projection distance is calculated by multiplying the standardized Euclidean distance between the coordinates of point P and point P1. If not, it means that point P is in the extension area of ​​the tower line segment AB. The matching logic needs to be completed by calculating the endpoint distance: if t<0, calculate the straight-line distance d_A from point P to tower A; If t>1, calculate the straight-line distance d_B from point P to tower B; take the smaller value of d_A or d_B as the equivalent projected distance and compare it with the current minimum projected distance: if the equivalent projected distance ≤ the current minimum projected distance, update the "minimum projected distance" to this value and mark the corresponding endpoint of the current tower segment as a candidate matching reference point; if the equivalent projected distance > the current minimum projected distance, continue traversing the next tower segment. The geometric constraints are constructed by comparing the projected distance with the minimum projected distance. The comparison process is as follows: if the projection distance If the minimum projected distance is less than or equal to the minimum projected distance, then update the minimum projected distance to... The current tower segment is selected as a candidate matching segment. If the projection distance If the distance is higher than the minimum projection distance, continue traversing the next tower segment; S203. Traverse all tower segments and assign towers to the original point cloud image. After all tower segments have been traversed, select the tower segment with the smallest projection distance. If multiple tower segments have the same minimum projection distance, select the segment with the smaller tower number and assign the current point cloud image to the smaller side tower of that segment. If all segments have not been traversed, continue matching the next segment. After completing the matching of the current point cloud original image, select the next point cloud original image to be processed and repeat steps S201-203 until all original images to be sorted are assigned to towers. The original image of the pole and tower allocation point cloud will be associated with the pole and tower ledger list. The purpose of assigning towers to the original point cloud image is: Function 1: To establish a connection between the original images to be graded and the specific tower span range and tower ledger information, providing basic data with spatial attributes for subsequent integration of multi-source data to construct scene feature vectors and establish dynamic naming templates; Secondly, it clarifies the tower span intervals corresponding to each point cloud original image, enabling the statistics and deviation analysis of the number of deviation original images in each interval and the total number of original images in each interval within the monitoring period, which is beneficial for locating high-risk neighborhood groups. Naming adaptation module: used to obtain the association information corresponding to the point cloud original images assigned to the tower, construct scene feature vectors based on the association information and perform scene segmentation, establish dynamic naming templates; and perform adaptive naming of the point cloud original images based on the dynamic naming templates and in combination with the scene. This includes obtaining the association information corresponding to the original point cloud images allocated to the towers, constructing scene feature vectors based on the association information, performing scene segmentation, and establishing a dynamic naming template: The line number and tower span are read from the grading and identification results; at the same time, the terrain scene information in the line ledger data, the acquisition height of the point cloud original image in the UAV log, and the historical maintenance rate of tower section failures are obtained from the ledger maintenance log. The line number, tower span, terrain scene information, acquisition height, and historical maintenance rate of the point cloud original images of a single completed tower allocation are integrated, and then numerically and normally processed to construct a scene feature vector; Obtain the scene feature vectors of all point cloud original images after tower allocation is completed, and construct a scene feature dataset; A scene classification model is constructed based on the K-means clustering algorithm. The scene feature dataset is input into the scene classification model to classify the scene risk type of the tower span interval and obtain the high-risk tower span. When constructing a scene classification model based on the K-means clustering algorithm, the number of clusters K is determined by the elbow rule. Then, the normalized scene feature vector dataset, which integrates the original point cloud image line number, tower span, numerical terrain scene information, collection height and historical maintenance rate, is iteratively clustered: K randomly selected feature vectors are used as the initial cluster centers. The distance between each vector and the center is calculated and assigned to the nearest cluster. After updating the cluster centers, the iteration is repeated until the centers are stable, and finally multiple clusters are formed. This corresponds to the classification of multiple scene risks in the tower span range, providing a basis for scene classification for the establishment of dynamic naming templates. Obtain the voltage level and scenario risk type classification of the line, and establish a dynamic naming template, i.e., a preset naming rule library, based on the classification and identification results; Obtain the list of pole and tower ledgers associated with the point cloud original images that have completed the grade recognition, and extract the associated matching information; The association matching information is checked against the dynamic naming template for field integrity and compliance. Point cloud originals that meet the compliance and integrity requirements are selected and named according to the dynamic naming template as the initial naming originals. If the original point cloud image does not meet the requirements for compliance and completeness, information should be filled into the original point cloud image. It should be noted that the information is populated according to priority and conflict resolution rules. Data source priority: Route ledger data > Drone log data > Historical data categorized by grade; Conflict resolution: When information from multiple data sources conflicts, a weighted voting method is used to determine the fill value; Extreme scenario: When there is no data source, mark it as needing manual completion, pause the naming process and push an alert; The method of adaptively naming the original point cloud image based on a dynamic naming template and combined with the scene is as follows: Obtain the initial naming original image within the same tower span interval, and calculate the spatial distance from the initial naming original image to the tower on the smaller side of the tower span interval; If the initial naming original is within the low-risk tower span, then the initial naming originals in the same tower span range are sorted in order of increasing spatial distance to obtain the basic serial number of the initial naming original. The basic sequence number is added to the naming of the original point cloud image to achieve the file-based naming of the original point cloud image; If the original image is located within the span of a high-risk tower, then the distance ratio weight between the spatial distance and the tower span is calculated. Obtain the historical maintenance rate of the small side towers within the initial naming original image of the high-risk tower span, and obtain the historical maintenance rate of the small side towers from the maintenance logbook; It should be noted that the historical maintenance rate is the percentage of maintenance and upkeep performed on the small side towers within the historical monitoring period. The reciprocal of the historical maintenance rate, the historical maintenance rate, and the distance ratio weight are each processed by Min-Max normalization. The three are then multiplied to obtain the cloud sorting value. The point cloud original images within the high-risk tower span are sorted in ascending order according to the numerical value of the cloud image sorting value to obtain the risk number of the initial named original image. The risk number and the basic number are added together to the naming of the point cloud original image to achieve the graded naming of the point cloud original image; It should be noted that the purpose of constructing risk serial numbers and basic serial numbers is as follows: Objective 1: To provide an orderly identification basis for the classification and naming of point cloud original images: The basic sequence number is sorted by spatial distance to standardize the naming order of original images within the span of low-risk towers, and the risk sequence number is combined with operation and maintenance data and spatial information to form a priority identifier for original images within the span of high-risk towers, so as to avoid confusion in the naming of original images within the same span and ensure the regularity of the naming results. Objective 2: Connecting naming results with operation and maintenance needs: Risk serial numbers highlight the original images that require special attention within high-risk ranges, while basic serial numbers clarify the spatial relative location of the original images. When both are added to the naming, it is easier for operation and maintenance personnel to quickly locate key areas of original images and identify key operation and maintenance points through naming, providing a clear data index for hidden danger investigation and ledger management. Feedback verification module: used to obtain the original point cloud images of the classification and naming deviation in the fault verification, and establish the deviation original image group; obtain the tower span interval corresponding to the deviation original image group for deviation analysis, and obtain the priority analysis interval; establish a high-risk neighborhood group based on the priority analysis interval, and classify and rename the point cloud original images of the high-risk neighborhood group.

2. The UAV point cloud original image data processing system with tiered naming function according to claim 1, characterized in that: The boundary shaping is performed as follows: The list of poles and towers is obtained by calling the overhead line files in the power system and parsing them through the program. Extract the line number, tower number, and tower latitude and longitude from the tower ledger list to establish a tower coordinate database; Calculate the basic boundary of the tower span interval based on the tower coordinate library; The dynamic error range of the boundary is determined by obtaining historical GPS deviations. The coordinates are then dynamically expanded based on the dynamic error range to determine the rectangular boundary of the tower span interval.

3. The UAV point cloud original image data processing system with tiered naming function according to claim 1, characterized in that: The method for constructing the graded recognition model is as follows: Obtain the original images to be divided for each tower span interval and initialize the minimum projection distance; Geometric constraints are constructed based on the minimum projection distance to determine candidate matching line segments; Traverse all tower segments and assign towers to the original point cloud image.

4. The UAV point cloud original image data processing system with tiered naming function according to claim 3, characterized in that: The geometric constraints are constructed as follows: Obtain the standardized coordinates of the original image to be sorted and mark it as point P. Mark the starting tower of the tower line segment as point A and the ending tower as point B. The tower line segment is AB. Establish a vector pointing from point A to point P. The vector pointing from point A to point B ; Construct a projection scaling equation, and transform the vector , Input the projection scale equation to obtain the projection scale coefficient; Determine whether the projection scale factor is within the preset constraint range. If it is, determine the projection point P1 of point P on AB based on the projection scale factor, and calculate the projection distance between point P and projection point P1. Geometric constraints are constructed by comparing the projected distance and the minimum projected distance.

5. A UAV point cloud original image data processing system with tiered naming function according to claim 1, characterized in that: The method for establishing the dynamic naming template is as follows: Obtain the scene feature vectors of all point cloud original images after tower allocation is completed, and construct a scene feature dataset; A scene classification model is constructed based on a clustering algorithm. The scene feature dataset is input into the scene classification model to classify the scene risk type of the tower span interval and obtain the high-risk tower span. Obtain the voltage level and scenario risk type classification of the line, and establish a dynamic naming template based on the classification and identification results.

6. A UAV point cloud original image data processing system with tiered naming function according to claim 1, characterized in that: The method for performing the aforementioned adaptation naming is as follows: Obtain the initial naming original image within the same tower span interval, and calculate the spatial distance from the initial naming original image to the tower on the smaller side of the tower span interval; If the initial naming original is within the low-risk tower span, then the initial naming originals in the same tower span range are sorted according to spatial distance to obtain the basic sequence number of the initial naming original. The basic sequence number is added to the naming of the original point cloud image to achieve the file-based naming of the original point cloud image; Obtain the cloud image sorting value, sort the point cloud original images within the high-risk tower span according to the cloud image sorting value, and obtain the risk sequence number of the initial named original image; The risk number and the basic number are added together to the naming of the point cloud original image to achieve graded naming of the point cloud original image.

7. A UAV point cloud original image data processing system with tiered naming function according to claim 1, characterized in that: The method for obtaining the cloud sorting value is as follows: If the original image is located within the span of a high-risk tower, then the distance ratio weight between the spatial distance and the tower span is calculated. Obtain the historical maintenance rate of the small side towers within the initial naming original image of the high-risk tower span, and obtain the historical maintenance rate of the small side towers from the maintenance logbook; Normalize the reciprocal of the historical maintenance rate, the historical maintenance rate, and the distance ratio weight, and then multiply the three to obtain the cloud sorting value.

8. A UAV point cloud original image data processing system with tiered naming function according to claim 1, characterized in that: The high-risk neighborhood group is established as follows: The spatial neighborhood group is obtained by combining the M tower span intervals that are adjacent to the priority analysis interval. If there are high-risk tower spans within the spatial neighborhood group, then construct the first neighborhood judgment equation to obtain the sum of the relative deviations of all intervals within the spatial neighborhood group, the total number of intervals in the neighborhood group, and the proportion of high-risk tower spans. Input these three factors into the first neighborhood judgment equation to obtain the first neighborhood judgment value. Construct a second neighborhood decision equation, input the sum of the relative deviations of all intervals within the spatial neighborhood group and the total number of intervals in the neighborhood group into the second neighborhood decision equation to obtain the second neighborhood decision value; Construct comparison criteria and obtain the first or second neighborhood judgment value of all spatial neighborhood groups in the scene. If the first or second neighborhood judgment value of a spatial neighborhood group satisfies the corresponding comparison criteria, then the spatial neighborhood is marked as a high-risk neighborhood group.

9. A UAV point cloud original image data processing system with tiered naming function according to claim 8, characterized in that: The method for obtaining the priority analysis interval is as follows: Obtain the number of point cloud deviation original images for each tower span interval within the monitoring period from the deviation original image group, as well as the total number of original images for each interval; The interval deviation rate is obtained by comparing the number of original point cloud images with the total number of original images in each tower span interval during the monitoring period. Obtain the interval deviation rate for each tower span interval, and the average interval deviation rate for the scene to which each tower span interval belongs, to obtain the scene baseline rate; Calculate the ratio of the interval deviation rate to the scene reference rate to obtain the relative deviation ratio; A statistical significance test is performed on the relative deviations of all tower span intervals in the scenario. If the statistical significance test is satisfied, the deviation of the tower span interval in the current scenario is determined to be a systematic anomaly. If it is a systemic anomaly, the tower span intervals are sorted in descending order according to the relative deviation ratio within the same scenario, and the tower span interval corresponding to the maximum value of the relative deviation ratio is taken as the priority analysis interval.

10. A UAV point cloud original image data processing system with tiered naming function according to claim 1, characterized in that: The method for renaming the file categories is as follows: The original point cloud images were re-filed and renamed by replacing the original interval border inspection traversal method within the high-risk neighborhood group with the spatial index-hash mapping coordinate matching method.

Images