CN-121095817-B - Unmanned aerial vehicle point cloud original film data processing system with grading naming function

CN121095817BCN 121095817 BCN121095817 BCN 121095817BCN-121095817-B

Abstract

The invention relates to the technical field of point cloud data processing, and particularly discloses an unmanned aerial vehicle point cloud original sheet data processing system with a grading naming function; the system comprises a boundary dividing module for generating rectangular boundaries adapting to terrains and screening raw sheets to be classified, a classification recognition module for constructing geometric constraints based on projection distances, a naming adaptation module for integrating multi-source data to construct scene feature vectors, a clustering and dividing risk distances, a multi-source filling and normalization ordering to realize differential naming, a feedback verification module for analyzing and positioning a priority analysis interval through deviation, constructing a space neighborhood group and judging high-risk areas, and a space index-hash mapping method for replacing an original algorithm to complete classification renaming and closed loop verification.

Inventors

LI HONGLIANG
SU PING
YU KUI
LU WENBIN
ZHENG HAIBO
LIU WENTAO
Shen Mennan
ZHANG KEXIN
FU YU
QIU YUE
CHEN JUBIN
MO WENBIN

Assignees

国网湖北省电力有限公司襄阳供电公司

Dates

Publication Date: 20260512
Application Date: 20250929

Claims (10)

1. An unmanned aerial vehicle point cloud original film data processing system with a grading naming function is characterized by comprising the following modules: The boundary dividing module is used for acquiring the point cloud original sheets of the unmanned aerial vehicle and the line ledger data, extracting a tower span interval of the line ledger data, and carrying out boundary shaping to obtain a rectangular boundary; The grading recognition module is used for constructing a grading recognition model, inputting the original sheets to be graded into the grading recognition model, and carrying out tower distribution treatment on the original sheets to be graded; the method for constructing the grading recognition model is as follows: S201, obtaining a to-be-segmented original sheet of each tower span interval, and initializing a minimum projection distance; Obtaining the minimum projection distance of the original sheet to be segmented in each tower span interval, initializing the minimum projection distance to be a maximum value, and recording the minimum distance from the original sheet to be segmented to a tower line segment; Initializing a minimum projection distance to a maximum value through a normal infinity constant in a programming language; s202, constructing geometric constraint based on the minimum projection distance, and determining candidate matching line segments; Marking a standardized seat of a raw sheet to be classified as a point P, marking a starting tower of a tower line segment as a point A, marking an ending point as a point B, and marking a tower line segment as an AB; Calculating the vertical distance d from the original sheet to be classified to the tower line segment AB; by projection ratio equation: obtaining a projection proportion coefficient t; the physical meaning of the projection proportion coefficient is that the relative projection position quantized value of the standardized coordinates of the original sheet to be classified in the direction of the pole tower line segment is calculated by dividing the dot product of the vector AP and the vector AB by the dot product of the vector AB, when the coefficient is in a section, the projection of the P point is shown to fall on the pole tower line segment AB, when the coefficient is smaller than 0, the projection falls outside the A point, and when the coefficient is larger than 1, the projection falls outside the B point, so that a position judgment basis is provided for the subsequent calculation of the projection distance or the endpoint distance and the construction of the geometric constraint; Wherein, the A vector pointing to point P for point a, A vector pointing to the point B for the point A; judging whether the projection proportion coefficient is in a constraint range, if so, determining a projection point P1 of the point P on the AB according to the projection proportion coefficient, and calculating the projection distance between the point P and the projection point P1 ; The coordinate of the projection point P1 is determined by multiplying the coordinate of the point A by t and the vector AB, and the projection distance is the Euclidean distance between the standardized coordinates of the point P and the point P1; If not, the description point P is in the extension line area of the pole tower line segment AB, and the complement matching logic is calculated through the end point distance, if t is less than 0, the straight line distance d_A from the point P to the pole tower A is calculated; If t >1, calculating the straight line distance d_B from the point P to the tower B, taking a smaller value in d_A or d_B as an equivalent projection distance, and comparing the equivalent projection distance with the current minimum projection distance, wherein if the equivalent projection distance is less than or equal to the current minimum projection distance, updating the minimum projection distance as the value, and marking the corresponding end point of the current tower line segment as a candidate matching reference point; comparing the projection distance with the minimum projection distance to realize the construction of geometric constraint; the comparison processing is performed by the following steps that if the projection distance is Below or equal to the minimum projection distance, updating the minimum projection distance to be Taking the current pole tower line segment as a candidate matching line segment; If the projection distance is If the projection distance is higher than the minimum projection distance, continuing to traverse the next tower line segment; S203, traversing all tower line segments, and distributing the tower to the point cloud original pieces; selecting a tower line segment corresponding to the minimum projection distance after all the tower line segments are traversed, preferentially selecting a line segment with smaller tower numbers if the minimum projection distances of a plurality of tower line segments are equal, and distributing the current point cloud original piece to a small side-bar tower of the line segment; After the matching of the current point cloud original sheets is completed, selecting the next point cloud original sheet to be processed, and repeating the steps S201-203 until all the original sheets to be classified complete the tower distribution; The method comprises the steps that a pole and tower distribution point cloud original sheet-related pole and tower account list is completed; The effect of carrying out pole tower distribution on the point cloud original sheet is as follows: The method comprises the steps of firstly, establishing association between an original film to be classified, a specific tower span interval and tower account information, constructing scene feature vectors for subsequent integration of multi-source data, and establishing a dynamic naming template to provide basic data with space attributes; Secondly, determining the attribution of the tower span intervals corresponding to the cloud primary sheets of each point, so that statistics and deviation analysis of the number of the deviation primary sheets and the total number of the interval primary sheets in each interval in a monitoring period are carried out, and the positioning of a high-risk neighborhood group is facilitated; The naming adaptation module is used for acquiring the association information corresponding to the point cloud original pieces distributed by the pole tower, constructing scene feature vectors based on the association information, dividing scenes, and establishing a dynamic naming template; the method comprises the steps of obtaining association information corresponding to point cloud original sheets distributed by a pole tower, constructing scene feature vectors based on the association information, dividing scenes, and establishing a dynamic naming template: Simultaneously acquiring the topographic scene information in the line ledger data, the acquisition height of a point cloud original film in the unmanned aerial vehicle log and the historical maintenance rate of faults in the pole tower section from the ledger maintenance log; integrating the line number, the tower span, the terrain scene information, the acquisition height and the historical maintenance rate of the point cloud original film distributed by a single finished tower to carry out numerical processing and normalization processing, and constructing a scene feature vector; acquiring scene feature vectors of point cloud original sheets which are completely distributed by the pole tower, and constructing a scene feature data set; Constructing a scene classification model based on a K-means clustering algorithm, inputting a scene characteristic dataset into the scene classification model, and performing scene risk type division on a tower span interval to obtain a high-risk tower span; When a scene classification model is constructed based on a K-means clustering algorithm, determining the number K of clusters through an elbow rule, and then carrying out iterative clustering on a normalized scene feature vector data set integrating the original line number of a point cloud, the tower span, the numerical terrain scene information, the acquisition height and the historical maintenance rate, wherein K feature vectors selected randomly are taken as initial clustering centers, the distance between each vector and the center is calculated and distributed to the nearest clusters, the clustering centers are updated and then repeated iteration is carried out until the center is stable, a plurality of clusters are finally formed, and the multi-class scene risk division of the tower span interval is correspondingly realized, so that scene classification basis is provided for the establishment of a dynamic naming template; acquiring voltage class and scene risk type division of a line, and establishing a dynamic naming template, namely a preset naming rule base, by combining a grading identification result; acquiring a tower account list of the point cloud original sheet association which completes the step identification, and extracting association matching information; Checking field integrity and compliance of the associated matching information and the dynamic naming template, screening point cloud original sheets meeting the compliance and integrity, naming the point cloud original sheets according to the dynamic naming template, and taking the point cloud original sheets as initial naming original sheets; if the point cloud original sheet does not meet the compliance and the integrity, filling information into the point cloud original sheet; the information filling method includes that the information filling is carried out according to the priority order and conflict judging rules: The data source priority is that line ledger data > unmanned plane log data > history grading data; conflict judging, namely when the information of multiple data sources conflicts, determining a filling value by adopting a weight voting method; in extreme scenes, marking the data source as to be completed manually, suspending the naming process and pushing the alarm; the method for adaptively naming the point cloud original film based on the dynamic naming template and combining the scene comprises the following steps: acquiring initial named original sheets in the same tower span interval, and calculating the space distance from the initial named original sheets to the small-sized side-lever towers in the tower span interval; If the initial named original pieces are in the low-risk tower span, sequencing the initial named original pieces in the same tower span section according to the sequence from small space distance to large space distance to obtain the basic serial numbers of the initial named original pieces; Supplementing the basic serial number to the naming of the original point cloud film to realize the grading naming of the original point cloud film; if the initial named original piece is in the high-risk tower span, calculating the distance proportion weight of the space distance and the tower span; Acquiring the historical maintenance rate of the small-size side-bar tower of the initial named original piece in the high-risk pole tower span, and acquiring the historical maintenance rate of the small-size side-bar tower from the standing account maintenance log; The maintenance frequency of the small-sized side pole tower in the history monitoring period is the maintenance frequency of the small-sized side pole tower in the history maintenance rate; Respectively executing Min-Max normalization processing on the reciprocal of the historical maintenance rate, the historical maintenance rate and the distance proportion weight, and performing product processing on the reciprocal, the historical maintenance rate and the distance proportion weight to obtain a cloud chip sequencing value; Ascending sequence sorting is carried out on the point cloud original sheets in the high-risk tower span according to the numerical value of the cloud sheet sorting value, so that the risk serial numbers of the original sheets which are initially named are obtained; The risk sequence number and the basic sequence number are supplemented to the original point cloud film naming so as to realize the grading naming of the original point cloud film; it should be noted that, the purpose of constructing the risk sequence number and the basic sequence number is: The method comprises the steps of firstly, providing an ordered identification basis for the grading naming of the point cloud original sheets, wherein a basic sequence number is ordered through a space distance to standardize the naming sequence of the original sheets in a low-risk tower span, and the risk sequence number is combined with operation and maintenance data and space information to form a priority identification of the original sheets in a high-risk span, so that the naming confusion of the original sheets in the same span is avoided, and the regularity of naming results is ensured; The risk sequence number highlights the original sheets needing to be focused in the high-risk file distance, the basic sequence number definitely identifies the space relative position of the original sheets, and after the two are supplemented in naming, operation and maintenance personnel can conveniently and quickly locate the original sheets of the key area through naming, identify operation and maintenance major points, and clear data index is provided for hidden trouble investigation and standing account management; The feedback verification module is used for acquiring point cloud raw sheets with grading naming deviation in fault verification, establishing a deviation raw sheet group, acquiring a tower span interval corresponding to the deviation raw sheet group, performing deviation analysis to obtain a priority analysis interval, establishing a high-risk neighborhood group based on the priority analysis interval, and carrying out grading renaming on the point cloud raw sheets of the high-risk neighborhood group.
2. The unmanned aerial vehicle point cloud raw sheet data processing system with the grading naming function according to claim 1, wherein the boundary shaping method is as follows: program analysis is carried out by calling overhead line files in the power system to obtain a pole tower account list; Extracting a line number, a pole number and a pole longitude and latitude in a pole ledger list, and constructing a pole coordinate library; calculating a basic boundary of a tower span interval based on a tower coordinate base; and acquiring a dynamic error range of a historical GPS deviation determination boundary, and carrying out coordinate dynamic expansion processing in combination with the dynamic error range to determine a rectangular boundary of a tower span interval.
3. The unmanned aerial vehicle point cloud raw sheet data processing system with the grading naming function according to claim 1, wherein the method for constructing the grading recognition model is as follows: Obtaining a to-be-classified original sheet of each tower span interval, and initializing a minimum projection distance; constructing a geometric constraint based on the minimum projection distance, and determining candidate matching line segments; traversing all tower line segments, and distributing the tower to the point cloud original pieces.
4. The unmanned aerial vehicle point cloud raw sheet data processing system with the hierarchical naming function according to claim 3, wherein the geometric constraint is constructed in the following way: The method comprises the steps of obtaining standardized coordinates of a raw sheet to be classified, marking the standardized coordinates as a point P, marking a starting point tower of a tower line segment as a point A, marking an ending point as a point B, and marking the tower line segment as an AB; establishing a vector pointing to point P at point A Vector of point A pointing to point B ; Constructing a projection proportion equation and vector 、 Inputting a projection proportion equation to obtain a projection proportion coefficient; Judging whether the projection proportion coefficient is in a preset constraint range, if so, determining a projection point P1 of the point P on the AB according to the projection proportion coefficient, and calculating the projection distance between the point P and the projection point P1; and performing comparison processing based on the projection distance and the minimum projection distance, and constructing geometric constraint.
5. The unmanned aerial vehicle point cloud original sheet data processing system with the hierarchical naming function of claim 1, wherein the dynamic naming template is established by the following steps: acquiring scene feature vectors of point cloud original sheets which are completely distributed by the pole tower, and constructing a scene feature data set; Constructing a scene classification model based on a clustering algorithm, inputting a scene characteristic data set into the scene classification model, and performing scene risk type division on a tower span interval to obtain a high-risk tower span; and obtaining the voltage grade and scene risk type division of the line, and establishing a dynamic naming template by combining the grading recognition result.
6. The unmanned aerial vehicle point cloud original data processing system with the hierarchical naming function according to claim 1, wherein the adaptive naming is performed in the following manner: acquiring initial named original sheets in the same tower span interval, and calculating the space distance from the initial named original sheets to the small-sized side-lever towers in the tower span interval; if the initial named original pieces are in the low-risk tower span, sequencing the initial named original pieces in the same tower span interval according to the space distance to obtain the basic serial numbers of the initial named original pieces; Supplementing the basic serial number to the naming of the original point cloud film to realize the grading naming of the original point cloud film; the cloud piece ordering value is obtained, and the point cloud original pieces in the high-risk tower span are ordered according to the cloud piece ordering value, so that the risk serial numbers of the original named pieces are obtained; and supplementing the risk sequence number and the basic sequence number to the original point cloud film naming so as to realize the grading naming of the original point cloud film.
7. The unmanned aerial vehicle point cloud original data processing system with the grading naming function according to claim 1, wherein the cloud ordering value is obtained by the following steps: if the initial named original piece is in the high-risk tower span, calculating the distance proportion weight of the space distance and the tower span; Acquiring the historical maintenance rate of the small-size side-bar tower of the initial named original piece in the high-risk pole tower span, and acquiring the historical maintenance rate of the small-size side-bar tower from the standing account maintenance log; and respectively carrying out normalization processing on the reciprocal of the historical maintenance rate, the historical maintenance rate and the distance proportion weight, and carrying out product processing on the reciprocal, the historical maintenance rate and the distance proportion weight to obtain a cloud chip ordering value.
8. The unmanned aerial vehicle point cloud original sheet data processing system with the grading naming function according to claim 1, wherein the method for establishing the high-risk neighborhood group is as follows: combining M tower span intervals adjacent to the front and rear of the priority analysis interval to obtain a space neighborhood group; If the high risk tower span exists in the space neighborhood group, constructing a neighborhood first judging equation, acquiring the sum of the relative deviation ratios of all the intervals in the space neighborhood group, the total number of the intervals in the neighborhood group and the high risk tower span ratio, and inputting the sum, the total number and the high risk tower span ratio into the neighborhood first judging equation to acquire a neighborhood first judging value; constructing a neighborhood second judging equation, and inputting the sum of the relative comparison of the deviations of all the regions in the space neighborhood group and the total number of the regions in the neighborhood group into the neighborhood second judging equation to obtain a neighborhood second judging value; Constructing a comparison criterion, acquiring a neighborhood first judgment value or a neighborhood second judgment value of all the spatial neighborhood groups in the scene, and marking the spatial neighborhood as a high-risk neighborhood group if the neighborhood first judgment value or the neighborhood second judgment value of the spatial neighborhood groups respectively meet the corresponding comparison criterion.
9. The unmanned aerial vehicle point cloud original data processing system with the grading naming function according to claim 8, wherein the method for acquiring the priority analysis interval is as follows: Acquiring the number of the point cloud deviation raw pieces and the total number of the interval raw pieces of each tower span interval point cloud in a monitoring period from the deviation raw piece group; Carrying out ratio processing on the number of the original point cloud deviation pieces of each tower span interval in the monitoring period and the total number of the original point cloud deviation pieces of the interval to obtain an interval deviation rate; Acquiring the interval deviation rate of each tower span interval and the average value of the interval deviation rates of scenes to which each tower span interval belongs to, and obtaining a scene reference rate; Calculating the ratio of the interval deviation rate to the scene reference rate to obtain the deviation relative comparison; Carrying out statistical significance test on the relative comparison of the deviations of all tower span sections in the scene, and judging that the deviation of the tower span section of the current scene is systematic abnormality if the statistical significance test is satisfied; If the system is abnormal, the tower span sections are ordered in descending order according to the relative deviation in the same scene, and the tower span section corresponding to the maximum value of the relative deviation is used as the priority analysis section.
10. The unmanned aerial vehicle point cloud original sheet data processing system with the grading naming function according to claim 1, wherein the grading renaming is performed in the following manner: and replacing the original interval edge detection traversal method in the high-risk neighborhood group by using a space index-hash mapping coordinate matching method, and carrying out re-grading naming on the point cloud original sheet.

Description

Unmanned aerial vehicle point cloud original film data processing system with grading naming function Technical Field The invention relates to the technical field of point cloud data processing, in particular to an unmanned aerial vehicle point cloud original sheet data processing system with a grading naming function. Background With the expansion of the power grid scale, the inspection maintenance pressure of the power transmission line channel increases sharply, and the unmanned aerial vehicle becomes a core operation and maintenance tool by virtue of the advantages of high efficiency and flexibility, and plays a key role in the establishment of the mountain area inspection and the 'first-gear-first-trouble' standing book. However, the traditional fast inspection mode has extremely low efficiency, a single line needs a long time from photo acquisition to uploading and archiving, and multiple persons and multiple devices are needed to cooperate. Meanwhile, the problems of resource waste and technical blank exist in the power transmission operation and maintenance, namely, key line point cloud scanning analysis tree barriers are firstly carried out in each quarter, the scanning and the quick inspection fly along the lines, the execution modes are highly similar, and independent operation is realized. A large number of optical raw sheets (point cloud raw sheets) synchronously acquired by point cloud scanning contain channel environment information, but are idle only as modeling auxiliary data. Because of the lack of technical tools, the original point cloud film cannot finish the step recognition and the standard naming, and is difficult to convert into a quick patrol photo, so that the quick patrol still needs to be repeatedly collected, and resource waste is formed. Therefore, the invention provides the unmanned aerial vehicle point cloud original film data processing system with the grading naming function. Disclosure of Invention The invention aims to provide an unmanned aerial vehicle point cloud original sheet data processing system with a grading naming function, so as to solve the background problem. The aim of the invention can be achieved by the following technical scheme: An unmanned aerial vehicle point cloud original film data processing system with a grading naming function comprises the following modules: The boundary dividing module is used for acquiring the point cloud original sheets of the unmanned aerial vehicle and the line ledger data, extracting a tower span interval of the line ledger data, and carrying out boundary shaping to obtain a rectangular boundary; The grading recognition module is used for constructing a grading recognition model, inputting the original sheets to be graded into the grading recognition model, and carrying out tower distribution treatment on the original sheets to be graded; The naming adaptation module is used for acquiring the association information corresponding to the point cloud original pieces distributed by the pole tower, constructing scene feature vectors based on the association information, dividing scenes, and establishing a dynamic naming template; The feedback verification module is used for acquiring point cloud raw sheets with grading naming deviation in fault verification, establishing a deviation raw sheet group, acquiring a tower span interval corresponding to the deviation raw sheet group, performing deviation analysis to obtain a priority analysis interval, establishing a high-risk neighborhood group based on the priority analysis interval, and carrying out grading renaming on the point cloud raw sheets of the high-risk neighborhood group. The method for carrying out the boundary shaping comprises the following steps: program analysis is carried out by calling overhead line files in the power system to obtain a pole tower account list; Extracting a line number, a pole number and a pole longitude and latitude in a pole ledger list, and constructing a pole coordinate library; calculating a basic boundary of a tower span interval based on a tower coordinate base; and acquiring a dynamic error range of a historical GPS deviation determination boundary, and carrying out coordinate dynamic expansion processing in combination with the dynamic error range to determine a rectangular boundary of a tower span interval. The method for constructing the grading recognition model is as follows: Obtaining a to-be-classified original sheet of each tower span interval, and initializing a minimum projection distance; constructing a geometric constraint based on the minimum projection distance, and determining candidate matching line segments; traversing all tower line segments, and distributing the tower to the point cloud original pieces. As a further proposal of the invention, the method for constructing the geometric constraint is as follows: The method comprises the steps of obtaining standardized coordinates of a raw sheet to be classified, marking th