An automatic detection system for unmanned aerial vehicles for lightning protection

Abstract

The present invention provides an automatic detection system for lightning protection by an unmanned aerial vehicle, comprising: a data acquisition module, used for scanning a detected building by using a laser radar device installed on the unmanned aerial vehicle, and obtaining total point cloud data of the detected building; a modeling module, used for performing a modeling operation based on the total point cloud data, and obtaining a building model of the detected building; a marking module, used for marking lightning-receiving objects in the building model, and obtaining marked lightning-receiving objects; a range determination module, used for determining the joint protection range of all marked lightning-receiving objects by using a rolling ball method based on the position of the marked lightning-receiving objects and the lightning protection category of the detected building; and an object screening module, used for detecting all objects in the building model that are outside the joint protection range, and screening out unprotected objects therein.

Description

Technical Field

[0001] The present invention relates to the technical field of building lightning protection device detection, and in particular to an automatic detection system of a drone for lightning protection. Background Art

[0002] The inspection of building lightning protection devices is divided into the inspection of external lightning protection devices and the inspection of internal lightning protection devices. The inspection of external lightning protection devices includes the inspection of lightning conductors, down conductors and grounding devices, and the inspection of internal lightning protection devices includes the inspection of lightning protection equipotential connections and the spacing distance from external lightning protection devices.

[0003] Existing inspection activities are completed by visual inspection (for example, installation process) and instrument measurement (for example, ground resistance value) by inspectors. In the LPZ0 area of ​​the building lightning protection zone, the protection range of the lightning rod needs to be calculated by the "rolling ball method" after the lightning protection inspectors measure the number, position and height of the lightning rods. The "rolling ball method" is one of the calculation methods for the protection range of lightning rods recommended by the International Electrotechnical Commission IEC; the building lightning protection code GB50057-2010 currently being implemented in my country also adopts the "rolling ball method". The rolling ball method uses an imaginary sphere of a certain diameter to roll along the ground and around the lightning rod to form an imaginary hourglass-shaped space boundary. All objects outside the space boundary may be directly struck by lightning and conduct all lightning currents, that is: LPZ0 A Area (area not protected by the lightning rod); and the objects within the boundary of this space cannot be directly struck by lightning current greater than the selected rolling ball radius, that is: LPZ0 B Area (the area that the lightning arrester can protect).

[0004] In LPZ0 A There may be objects in the area that need lightning protection, so it is necessary to A To determine the unprotected objects in the area outside the protection range of the lightning arrester at the detection site (area LPZ0 A ) Unprotected objects, the existing method is for inspectors to conduct visual inspections; determine objects that may not be within the protection range of the lightning rod, and measure their height and distance from the lightning rod; determine the number, height and location of existing lightning rods; and calculate through the "rolling ball method" (refer to the contents of Article 5.2.12 and Appendix D of GB50057-2010); determine whether each object that may not be within the protection range of the lightning rod is within the protection range of the lightning rod. (That is, determine which objects are not in the LPZ0 B area, and required them to make rectifications on protection measures against direct lightning strikes).

[0005] However, the existing methods have the following disadvantages:

[0006] (1) All inspections are done manually by inspectors, and calculations need to be performed for each object (so in fact, on-site personnel will ignore this calculation and directly use visual inspection). The inspection and calculation process is complicated and the inspection results are uncontrollable (measurement errors, calculation errors, etc.). Inspections according to national standards (GB50057-2010, GB21431-2015) will greatly increase the workload of lightning protection inspections and greatly reduce the efficiency of lightning protection inspections. In addition, direct visual inspections will lead to misjudgment of the protection range, bringing great potential safety hazards to lightning protection.

[0007] (2) The lightning rods of buildings are all located on the ridges and corners of the roofs of buildings. Most of them are located at high positions, making it difficult for inspection personnel to safely measure the lightning rods or making it inconvenient to measure them.

[0008] (3) Existing technologies basically use lightning rods (lightning rods) as models to construct components, but the actual lightning terminals include lightning rods (lightning rods), lightning strips, lightning wires, lightning nets and metal components, not just lightning rod models. This cannot be applied to actual detection scenarios.

[0009] Therefore, how to accurately and conveniently detect the protection range of the lightning arrester and accurately determine the combined protection range of all lightning arrester devices in the lightning protection zone of the building (i.e., LPZ0 A and LPZ0 B The boundary line of the building lightning protection device is detected, and the objects outside the joint protection range are detected. Then, according to the detection results, the unprotected objects outside the protection range are quickly screened out, and corresponding lightning protection measures are taken in time to avoid related losses caused by untimely lightning protection. This is one of the problems that need to be solved urgently in the field of building lightning protection device detection technology. Summary of the invention

[0010] The present invention aims to solve the technical problems in the above-mentioned technologies to at least a certain extent. To this end, the purpose of the present invention is to provide an automatic detection system for lightning protection by drones, which scans buildings by drones, models buildings according to the scanning results, and determines objects in the building that are not protected by lightning-arresting objects by processing the models, and then protects the objects in a timely and effective manner, thereby accurately and conveniently detecting the protection range of lightning arresters, and accurately determining the joint protection range of all lightning arrester devices in the lightning protection zone of the building according to the range detection results, and detecting objects outside the joint protection range, and then quickly screening out unprotected objects outside the protection range according to the detection results, taking corresponding lightning protection measures in a timely manner, and avoiding the technical effect of related losses caused by untimely lightning protection.

[0011] The present invention provides an automatic detection system for unmanned aerial vehicles for lightning protection, comprising:

[0012] A data acquisition module is used to scan the building under test using a laser radar device installed on the drone to obtain total point cloud data of the building under test;

[0013] A modeling module is used to perform modeling operations based on the total point cloud data to obtain a building model of the measured building;

[0014] A marking module is used to mark the lightning-receiving objects in the building model to obtain marked lightning-receiving objects;

[0015] A range determination module is used to determine the combined protection range of all marked lightning termination objects using a rolling ball method based on the locations of the marked lightning termination objects and the lightning protection category of the building under test;

[0016] The object screening module is used to detect all objects in the building model that are outside the joint protection range and screen out unprotected objects.

[0017] Preferably, the automatic detection system for lightning protection drones, the data acquisition module, comprises:

[0018] The scanning submodule is used to use the laser radar device to emit laser beams to the measured building at different positions, and scan the measured building several times at different angles;

[0019] The data acquisition submodule is used to receive the optical signal reflected by the laser beam after each scan, determine the sub-point cloud data acquired by the current scan according to the optical signal, and compose the total point cloud data of the measured building according to several sub-point cloud data.

[0020] Preferably, the automatic detection system for lightning protection using unmanned aerial vehicles, the modeling module, comprises:

[0021] A data denoising submodule is used to perform data denoising processing on a plurality of sub-point cloud data included in the total point cloud data to obtain a plurality of denoised sub-point cloud data;

[0022] A data registration submodule is used to convert all data points in a number of denoised sub-point cloud data into a target coordinate system. After the conversion, all data points in the target coordinate system constitute the registered point cloud data.

[0023] Data classification submodule, used to:

[0024] Obtain existing building registration point cloud data and corresponding classified point cloud data, and use the existing building registration point cloud data and the corresponding classified point cloud data to train a neural network to obtain a point cloud data classification model; wherein the classified point cloud data is obtained by classifying data points in the registration point cloud data according to the categories of corresponding objects in the building;

[0025] Input the registered point cloud data into the point cloud data classification model to obtain the corresponding classified point cloud data;

[0026] The building modeling submodule is used to perform modeling operations on the tested building based on the classified point cloud data to obtain a building model of the tested building.

[0027] Preferably, the automatic detection system for unmanned aerial vehicles for lightning protection, the data registration submodule, comprises:

[0028] A determination submodule is used to determine any denoised sub-point cloud data as target sub-point cloud data, determine the coordinate system where the target sub-point cloud data is located as the target coordinate system; determine any denoised sub-point cloud data other than the target sub-point cloud data as the sub-point cloud data to be processed;

[0029] A vector determination submodule, used to calculate the feature vector of each data point in the target sub-point cloud data and the sub-point cloud data to be processed by using a fast point feature histogram algorithm;

[0030] A distance calculation submodule is used to select a feature vector of any data point in the sub-point cloud data to be processed as the current feature vector, calculate the Euclidean distance between the current feature vector and the feature vectors of all data points in the target sub-point cloud data, perform the above operation on the feature vectors of all data points in the sub-point cloud data to be processed, and obtain the Euclidean distance between the feature vector of each data point in the sub-point cloud data to be processed and the feature vectors of all data points in the target sub-point cloud data;

[0031] The data point pair determination submodule is used to compare the sizes of all Euclidean distances, sort all Euclidean distances from large to small, obtain a Euclidean distance sorting queue, and select three groups of data point pairs corresponding to the three groups of feature vectors corresponding to the last three Euclidean distances in the Euclidean distance sorting queue as reference data point pairs;

[0032] A coordinate system conversion submodule is used to obtain a coarse registration matrix based on a pair of reference data points using a random sampling consistency algorithm, and to perform a first rigid body transformation on the data points in the sub-point cloud data to be processed using the coarse registration matrix, so as to convert the data points in the sub-point cloud data to be processed from the current coordinate system to the target coordinate system;

[0033] The registration submodule is used to:

[0034] The iterative closest point algorithm is used to iteratively optimize the coarse registration matrix to obtain the fine registration matrix;

[0035] Using the precise registration matrix, a second rigid body transformation is performed on the data points in the transformed sub-point cloud data to be processed in the target coordinate system, and all data points in the sub-point cloud data to be processed are registered to the target coordinate system;

[0036] The above operation is performed on all denoised sub-point cloud data except the target sub-point cloud data, and all data points in the denoised sub-point cloud data are converted to the target coordinate system. After the conversion, all data points in the target coordinate system constitute the registered point cloud data.

[0037] Preferably, the automatic detection system for unmanned aerial vehicles for lightning protection, the data denoising submodule comprises:

[0038] The area division submodule is used to select any sub-point cloud data as the sub-point cloud data to be denoised, determine the three-dimensional coordinate system based on the area occupied by the sub-point cloud data to be denoised, and divide the area occupied by the point cloud into several sub-areas according to the three-dimensional coordinate system;

[0039] Density acquisition submodule, used for:

[0040] Obtain the total number of data points in the sub-point cloud data to be denoised and the number of data points in each sub-area;

[0041] Obtain the total horizontal projection area of ​​the region occupied by the sub-point cloud data to be denoised and the horizontal projection sub-area of ​​each sub-region;

[0042] The ratio of the total number of data points to the total area of ​​horizontal projection is taken as the total density of data points of the sub-point cloud data to be denoised;

[0043] The ratio of the number of data points in each sub-region to the horizontal projection sub-area of ​​each sub-region is taken as the data point sub-density of the sub-region;

[0044] A first deletion submodule is used to calculate the noise point interval of each subregion based on the total density of data points and the sub-density of data points in each subregion by using a preset point cloud data filtering algorithm, delete the data points contained in each noise point interval, and all the remaining data points constitute the first extracted sub-point cloud data;

[0045] A neighboring data point determination submodule is used to select any data point in the first extracted sub-point cloud data as a data point to be determined, determine a spherical area with the data point to be determined as the sphere center based on a preset radius, and use all data points in the spherical area as neighboring data points of the data point to be determined;

[0046] A calculation submodule is used to calculate the Manhattan distance between the data point to be judged and each adjacent data point, and calculate the average Manhattan distance corresponding to the data point to be judged;

[0047] A value determination submodule, used for determining a Gaussian mean and a standard deviation according to an average Manhattan distance and a preset Gaussian distribution probability density function;

[0048] A comparison submodule, used to compare the average Manhattan distance of the data point to be judged with a first preset threshold and a second preset threshold respectively, to obtain a comparison result; wherein the first preset threshold is the Gaussian mean, and the second preset threshold is the sum of the Gaussian mean and two times the standard deviation;

[0049] The second deletion submodule is used to determine that the data point to be judged is a noise data point and delete the data point to be judged when the comparison result is that the average Manhattan distance of the data point to be judged is less than or equal to the first preset threshold or the average Manhattan distance of the data point to be judged is greater than or equal to the second preset threshold; the above operation is performed on all data points in a number of sub-point cloud data to obtain a number of denoised sub-point cloud data.

[0050] Preferably, the automatic detection system for lightning protection using unmanned aerial vehicles, the building modeling submodule, comprises:

[0051] An image acquisition submodule is used to acquire planar images of the top, front, back, left and right sides of the building under test;

[0052] The first enhancer module is used to:

[0053] Select any plane image as the image to be processed, and use a preset bilateral filter kernel to filter the image to be processed to obtain a filtered image;

[0054] Determining an enhancement curve of the filtered image according to the size of the bilateral filter kernel and the pixel intensity value of the filtered image, and enhancing the filtered image based on the enhancement curve to obtain a first enhanced image;

[0055] The second enhancer module is used to:

[0056] Acquire a grayscale image of the first enhanced image;

[0057] Obtain the grayscale value of each pixel in the grayscale image, and determine the grayscale mean according to the grayscale value of each pixel;

[0058] The pixels whose grayscale values ​​in the grayscale image are less than the grayscale mean are taken as the pixels to be enhanced;

[0059] Determine an enhancement multiple of the grayscale value of the pixel to be enhanced according to the grayscale mean, the grayscale value of the pixel to be enhanced and a first preset enhancement coefficient;

[0060] Determine the enhanced grayscale value of the pixel to be enhanced according to the enhancement multiple, the grayscale value of the pixel to be enhanced, and the second preset enhancement coefficient, replace the grayscale value of the pixel to be enhanced with the enhanced grayscale value, implement the enhancement operation of the pixel to be enhanced, and use the image after the enhancement operation as the plane enhanced image;

[0061] Perform the above operations on all plane images to obtain the corresponding plane enhanced images;

[0062] The model generation submodule is used to perform modeling operations on the tested building using 3D modeling software based on all the plane enhanced images and classified point cloud data to obtain an initial building model;

[0063] Model determination submodule, used to:

[0064] Obtain initial building projection images corresponding to the top, front, back, left and right sides of the initial building model;

[0065] Calculate the two-dimensional shape similarity between the initial building projection image and the plane enhanced image of the corresponding surface. When the similarity is greater than a preset similarity threshold, determine that the previous modeling meets the modeling requirements; otherwise, based on the corresponding plane enhanced image, correct the previous modeling;

[0066] Based on the plane enhanced image, the corrected initial building model is color rendered, and the color-rendered initial building model is used as the building model of the measured building.

[0067] Preferably, the automatic detection system for lightning protection drones, the marking module, comprises:

[0068] A position determination submodule is used to determine the position of the lightning-termination object in the building model;

[0069] The object marking submodule is used to mark the lightning-receiving objects at the determined positions by using the preset marking colors to obtain the marked lightning-receiving objects.

[0070] Preferably, the automatic detection system for lightning protection drones, the range determination module, comprises:

[0071] A category determination submodule is used to determine the lightning protection category of the tested building according to the type of the tested building and based on the existing lightning protection category classification regulations for buildings;

[0072] A rolling ball determination submodule, used to determine the radius of the rolling ball based on the lightning protection category, and determine the target rolling ball based on the radius;

[0073] Scoping submodule for:

[0074] Control the target rolling ball to roll all-round on the building model starting from the ground position of the building model;

[0075] Based on the position of the marked lightning-termination objects, the range that the target rolling ball cannot reach due to the obstruction of the marked lightning-termination objects during its rolling process is divided into the joint protection range of all the marked lightning-termination objects.

[0076] Preferably, the automatic detection system for lightning protection drones, the object screening module, comprises:

[0077] The object detection submodule is used to detect all objects in the building model that are outside the joint protection range using a preset detection device to obtain a detection result;

[0078] The object determination submodule is used to compare the detection results with preset parameters according to the lightning protection category of the tested building, and determine the objects outside the joint protection range whose detection results do not meet the preset parameters as unprotected objects.

[0079] Preferably, the automatic detection system for lightning protection by unmanned aerial vehicles also includes a measure generation module for generating lightning protection measures for objects based on object information of unprotected objects.

[0080] The present invention provides an automatic detection system for lightning protection by using an unmanned aerial vehicle, including: a data acquisition module, which is used to scan the building under test by using a laser radar device installed on the unmanned aerial vehicle to obtain the total point cloud data of the building under test; a modeling module, which is used to perform modeling operations based on the total point cloud data to obtain a building model of the building under test; a marking module, which is used to mark the lightning-receiving objects in the building model to obtain the marked lightning-receiving objects; a range determination module, which is used to determine the joint protection range of all marked lightning-receiving objects by using a rolling ball method based on the position of the marked lightning-receiving objects and the lightning protection category of the building under test; and an object screening module, which is used to detect all objects in the building model that are outside the joint protection range and screen out the unprotected objects therein. Thus, the protection range of the lightning-receiving device is accurately and conveniently detected, and the joint protection range of all lightning-receiving devices in the lightning protection zone of the building is accurately determined according to the range detection results, and the objects outside the joint protection range are detected, and then the unprotected objects outside the protection range are quickly screened out according to the detection results, and corresponding lightning protection measures are taken in time to avoid the technical effect of related losses caused by untimely lightning protection.

[0081] Other features and advantages of the present invention will be described in the following description, and partly become obvious from the description, or be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the present application documents.

[0082] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0084] Figure 1 This is a block diagram of an automatic detection system for a UAV for lightning protection according to an embodiment of the present invention;

[0085] Figure 2 is a block diagram of an optional data denoising submodule in an embodiment of the present invention;

[0086] Figure 3 It is a schematic diagram of an optional method of determining a joint protection range by a rolling ball method in an embodiment of the present invention. DETAILED DESCRIPTION

[0087] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0088] Reference Figure 1 The embodiment of the present invention provides an automatic detection system for a UAV for lightning protection, comprising:

[0089] The data acquisition module 10 is used to scan the building under test using a laser radar device installed on the drone to obtain total point cloud data of the building under test;

[0090] A modeling module 20 is used to perform a modeling operation based on the total point cloud data to obtain a building model of the measured building;

[0091] A marking module 30 is used to mark the lightning-proof objects in the building model to obtain marked lightning-proof objects;

[0092] A range determination module 40 is used to determine the joint protection range of all marked lightning termination objects by using a rolling ball method based on the position of the marked lightning termination objects and the lightning protection category of the tested building;

[0093] The object screening module 50 is used to detect all objects in the building model that are outside the joint protection range and screen out unprotected objects therein.

[0094] The technical principle and technical effect of the above technical scheme are: using the laser radar device installed on the drone to scan the building under test and obtain the total point cloud data of the building under test; based on the total point cloud data, modeling operations are performed to obtain the building model of the building under test; marking the lightning-receiving objects in the building model to obtain marked lightning-receiving objects; based on the position of the marked lightning-receiving objects and the lightning protection category of the building under test, the rolling ball method is used to determine the joint protection range of all marked lightning-receiving objects; all objects in the building model that are outside the joint protection range are detected to screen out the unprotected objects. Thus, the protection range of the lightning-receiving device is accurately and conveniently detected, and the joint protection range of all lightning-receiving devices in the lightning protection zone of the building is accurately determined according to the range detection results, and objects outside the joint protection range are detected, and then the unprotected objects outside the protection range are quickly screened out according to the detection results, and corresponding lightning protection measures are taken in time to avoid the technical effect of related losses caused by untimely lightning protection.

[0095] The embodiment of the present invention provides an automatic detection system for a UAV for lightning protection, and a data acquisition module, including:

[0096] The scanning submodule is used to use the laser radar device to emit laser beams to the measured building at different positions, and scan the measured building several times at different angles;

[0097] The data acquisition submodule is used to receive the optical signal reflected by the laser beam after each scan, determine the sub-point cloud data acquired by the current scan according to the optical signal, and compose the total point cloud data of the measured building according to several sub-point cloud data.

[0098] The technical principle and technical effect of the above technical solution are: using a laser radar device to emit a laser beam to the building under test at different positions, and scanning the building under test several times at different angles; receiving the light signal reflected by the laser beam after each scan, determining the sub-point cloud data obtained by the current scan based on the light signal, and composing the total point cloud data of the building under test based on several sub-point cloud data. By scanning the building under test in all directions, the point cloud data of the building under test is obtained in all directions, providing a complete data foundation for modeling, thereby ensuring the technical effect of the complete restoration of the building model obtained through subsequent modeling.

[0099] The embodiment of the present invention provides an automatic detection system for unmanned aerial vehicles for lightning protection, a modeling module, including:

[0100] A data denoising submodule is used to perform data denoising processing on a plurality of sub-point cloud data included in the total point cloud data to obtain a plurality of denoised sub-point cloud data;

[0101] A data registration submodule is used to convert all data points in a number of denoised sub-point cloud data into a target coordinate system. After the conversion, all data points in the target coordinate system constitute the registered point cloud data.

[0102] Data classification submodule, used to:

[0103] Obtain existing building registration point cloud data and corresponding classified point cloud data, and use the existing building registration point cloud data and the corresponding classified point cloud data to train a neural network to obtain a point cloud data classification model; wherein the classified point cloud data is obtained by classifying data points in the registration point cloud data according to the categories of corresponding objects in the building;

[0104] Input the registered point cloud data into the point cloud data classification model to obtain the corresponding classified point cloud data;

[0105] The building modeling submodule is used to perform modeling operations on the tested building based on the classified point cloud data to obtain a building model of the tested building.

[0106] In this embodiment, one of the denoised sub-point cloud data is selected as the target sub-point cloud data, and the coordinate system where the target sub-point cloud data is located is the target coordinate system. The registration is to convert all the remaining denoised sub-point cloud data from the corresponding current coordinate system to the coordinate system where the target sub-point cloud data is located.

[0107] In this embodiment, "classifying point cloud data by classifying data points in the registered point cloud data according to the categories of corresponding objects in the building" is illustrated as follows: when there are objects A, B, and C in the building, the data points corresponding to object A in the registered point cloud data are classified into one category of data points, the data points corresponding to object B are classified into one category of data points, and the data points corresponding to object C are classified into one category of data points, and the classification of the registered point cloud data is completed, and the registered point cloud data is divided into 3 groups in total.

[0108] The technical principle and technical effect of the above technical solution are: perform data denoising on several sub-point cloud data included in the total point cloud data to obtain several denoised sub-point cloud data; convert all data points in several denoised sub-point cloud data to the target coordinate system, and after the conversion, all data points in the target coordinate system constitute the registered point cloud data; input the registered point cloud data into the trained point cloud data classification model to obtain the corresponding classified point cloud data; perform modeling operations on the measured building based on the classified point cloud data to obtain the building model of the measured building. The above technical solution not only removes the noise point cloud data points, but also registers the sub-point cloud data of different coordinate systems and classifies the registered point cloud data after registration, realizing the data optimization of the total point cloud data corresponding to the measured building, thereby improving the efficiency and accuracy of subsequent modeling.

[0109] The embodiment of the present invention provides an automatic detection system for a UAV for lightning protection, and a data registration submodule, including:

[0110] A determination submodule is used to determine any denoised sub-point cloud data as target sub-point cloud data, determine the coordinate system where the target sub-point cloud data is located as the target coordinate system; determine any denoised sub-point cloud data other than the target sub-point cloud data as the sub-point cloud data to be processed;

[0111] A vector determination submodule, used to calculate the feature vector of each data point in the target sub-point cloud data and the sub-point cloud data to be processed by using a fast point feature histogram algorithm;

[0112] A distance calculation submodule is used to select a feature vector of any data point in the sub-point cloud data to be processed as the current feature vector, calculate the Euclidean distance between the current feature vector and the feature vectors of all data points in the target sub-point cloud data, perform the above operation on the feature vectors of all data points in the sub-point cloud data to be processed, and obtain the Euclidean distance between the feature vector of each data point in the sub-point cloud data to be processed and the feature vectors of all data points in the target sub-point cloud data;

[0113] The data point pair determination submodule is used to compare the sizes of all Euclidean distances, sort all Euclidean distances from large to small, obtain a Euclidean distance sorting queue, and select three groups of data point pairs corresponding to the three groups of feature vectors corresponding to the last three Euclidean distances in the Euclidean distance sorting queue as reference data point pairs;

[0114] A coordinate system conversion submodule is used to obtain a coarse registration matrix based on a pair of reference data points using a random sampling consistency algorithm, and to perform a first rigid body transformation on the data points in the sub-point cloud data to be processed using the coarse registration matrix, so as to convert the data points in the sub-point cloud data to be processed from the current coordinate system to the target coordinate system;

[0115] The registration submodule is used to:

[0116] The iterative closest point algorithm is used to iteratively optimize the coarse registration matrix to obtain the fine registration matrix;

[0117] Using the precise registration matrix, a second rigid body transformation is performed on the data points in the transformed sub-point cloud data to be processed in the target coordinate system, and all data points in the sub-point cloud data to be processed are registered to the target coordinate system;

[0118] The above operation is performed on all denoised sub-point cloud data except the target sub-point cloud data, and all data points in the denoised sub-point cloud data are converted to the target coordinate system. After the conversion, all data points in the target coordinate system constitute the registered point cloud data.

[0119] In this embodiment, the specific implementation method of calculating the feature vector of each data point in the target sub-point cloud data and the sub-point cloud data to be processed by using the fast point feature histogram algorithm can be as follows:

[0120] Select any data point in the target sub-point cloud data as the data point P to be calculated, and determine all data points in the K neighborhood of P (set according to actual conditions) as reference data points of P;

[0121] Choose any reference point Q, calculate the relative position vector between Q and P (the vector pointing from point Q to point P, calculated by subtracting the coordinates of point Q from the coordinates of point P) and the normal vector (the vector OQ and the vector OP are obtained by performing a cross product calculation, and O is the origin of the target coordinate system);

[0122] The direction of the normal vector is used as the reference direction, and the angle Q_angle between the relative position vector and the reference direction is calculated using inverse trigonometric functions (arctan, etc.). The coordinates of the relative position vector are determined to be (x, y, z), and the Euclidean distance between the relative position vector and the origin of the coordinate system is

[0123] Take Q_angel and Q_Distance as the feature description values ​​of data point Q;

[0124] Calculate the two feature description values ​​corresponding to all reference points of point P, take the data points with the same two feature description values ​​as the same group of data points, count the feature description values ​​of all reference data points, determine the frequency of occurrence of reference data points with the same feature description values, and obtain the feature-frequency histogram of data point P with the feature description value as the horizontal axis and the frequency of occurrence as the vertical axis;

[0125] The feature-frequency histogram is frequency-weighted to obtain the weighted frequency. The weight depends on the specific situation. For example, when the weight is weight and the frequency is frequency, the corresponding weighted frequency is weight*frequency. It should be noted that the greater the frequency, the greater the weight. All frequency values ​​in the histogram are weighted to obtain the corresponding weighted frequency.

[0126] Each weighted frequency is divided by the number of point cloud data points in the K neighborhood to obtain the corresponding normalized frequency. Based on the feature description value and the corresponding normalized frequency, the feature-normalized histogram of the data point P is obtained;

[0127] Take the information of the first sub-histogram in the feature-normalized histogram as the starting element of the feature vector, and splice the histogram information of each sub-histogram in turn to obtain the feature vector of point P (for example, if the horizontal axis information of the first sub-histogram is A1, A2, and the vertical axis information is A3, and the horizontal axis information of the second sub-histogram is B1, B2, and the vertical axis information is B3, then the spliced ​​feature vector is [A1, A2, A3, B1, B2, B3]), perform the above operation on each data point in the target sub-point cloud data and the sub-point cloud data to be processed to obtain the corresponding feature vector.

[0128] In this embodiment, the specific implementation method for selecting the feature vector of any data point in the sub-point cloud data to be processed as the current feature vector and calculating the Euclidean distance between the current feature vector and the feature vectors of all data points in the target sub-point cloud data can be: setting the current feature vector as vector M, selecting the feature vector N of any data point in the target sub-point cloud data, and the Euclidean distance between M and N is It should be noted that l is Max (the dimension of vector M, the dimension of vector N). When the dimension of the other vector is less than l, 0 is used to fill the position during calculation. For example, when M is (M1, M2, M3) and N is (N1, N2), the Euclidean distance at this time is

[0129] In this embodiment, the specific implementation method for obtaining a coarse registration matrix using a random sampling consistency algorithm based on a reference data point pair is as follows: determine that the required registration model is a rigid body transformation model, and based on the rigid body transformation model and the reference data point pair, use the least squares method to calculate and obtain an initial registration matrix by minimizing the objective function. For data points in the sub-point cloud data to be processed that are not in the reference point pair, determine the corresponding initial transformation data points after transformation to the target coordinate system according to the initial registration matrix, calculate the Euclidean distance between each data point in the sub-point cloud data to be processed and the corresponding initial transformation data point (such as the calculation method of Distance above), when the Euclidean distance is less than a preset Euclidean distance threshold, determine that the initial transformation data point corresponding to the data point is an inlier, and optimize the initial registration matrix based on the data point and the corresponding inlier by the least squares method, repeat this step multiple times, count the number of inliers each time, and select the matrix corresponding to the maximum number of inliers as the coarse registration matrix, wherein the preset Euclidean distance threshold is the minimum Euclidean distance in the above Euclidean distance sorting queue.

[0130] In this embodiment, the rigid body transformation is translation and rotation.

[0131] In this embodiment, a specific implementation method of iteratively optimizing the coarse registration matrix using the iterative nearest point algorithm to obtain a fine registration matrix can be: select any data point after conversion to the target coordinate system as a test data point, determine a sphere with the test data point as the center and a fixed value as the radius, and use all data points in the sphere as neighboring data points of the test data point, calculate the Euclidean distance between the test data point and all neighboring pixel points, determine the average distance between the test data point and all neighboring pixel points based on all Euclidean distances, and use the average distance as an error metric value. Perform the above operation on all data points after conversion to the target coordinate system to obtain multiple error metrics, and based on the multiple error metrics, use the least squares method to obtain the error metric value. The coarse registration matrix is ​​optimized to obtain the optimized matrix. All data points converted to the target coordinate system are transformed based on the optimized matrix. Multiple error metrics are calculated using the above method. The optimized matrix is ​​optimized based on the error metrics. The optimization operation is performed continuously until the multiple error metrics obtained at a certain time are all less than the set threshold. The corresponding matrix is ​​determined to be the fine registration matrix, where the threshold is set to a variable value. During the first calculation, the minimum value of the multiple error metrics is determined as the set threshold for this time based on the calculation result. Thereafter, the minimum value of the multiple error metrics obtained this time is compared with the current set threshold each time, and the smaller one is the latest set threshold.

[0132] The technical principle and technical effect of the above technical scheme are: determine any denoised sub-point cloud data as the target sub-point cloud data, and determine the coordinate system where the target sub-point cloud data is located as the target coordinate system; determine any denoised sub-point cloud data other than the target sub-point cloud data as the sub-point cloud data to be processed; calculate the feature vector of each data point in the target sub-point cloud data and the sub-point cloud data to be processed; select the feature vector of any data point in the sub-point cloud data to be processed, calculate the Euclidean distance between the feature vector and the feature vectors of all data points in the target sub-point cloud data, and determine the Euclidean distance between the feature vector of each data point in the sub-point cloud data to be processed and the feature vectors of all data points in the target sub-point cloud data based on this method; determine the three groups of feature vectors corresponding to the three groups with the smallest Euclidean distance The data point pairs are used as reference data point pairs; based on the reference data point pairs, the random sampling consistency algorithm is used to obtain the rough registration matrix, and the data points in the sub-point cloud data to be processed are subjected to the first rigid body transformation using the rough registration matrix, and converted to the target coordinate system; the iterative closest point algorithm is used to iteratively optimize the rough registration matrix to obtain the fine registration matrix; the fine registration matrix is ​​used to perform the second rigid body transformation on the data points in the converted sub-point cloud data to be processed in the target coordinate system, and all data points in the sub-point cloud data to be processed are registered to the target coordinate system; the above operations are performed on all denoised sub-point cloud data except the target sub-point cloud data, and all data points in the denoised sub-point cloud data are converted to the target coordinate system. After the conversion, all data points in the target coordinate system constitute the registered point cloud data. The multiple sub-point cloud data corresponding to the measured building are registered to the same coordinate system, ensuring the integrity and correctness of the point cloud data, thereby ensuring the integrity of the three-dimensional scene information corresponding to the point cloud data, and then achieving the subsequent technical effect of precise modeling.

[0133] Reference Figure 2 The embodiment of the present invention provides an automatic detection system for unmanned aerial vehicles for lightning protection, a data denoising submodule, including:

[0134] The area division sub-module 2011 is used to select any sub-point cloud data as the sub-point cloud data to be denoised, determine a three-dimensional coordinate system based on the area occupied by the sub-point cloud data to be denoised, and divide the area occupied by the point cloud into a plurality of sub-areas according to the three-dimensional coordinate system;

[0135] The density acquisition submodule 2012 is used for:

[0136] Obtain the total number of data points in the sub-point cloud data to be denoised and the number of data points in each sub-area;

[0137] Obtain the total horizontal projection area of ​​the region occupied by the sub-point cloud data to be denoised and the horizontal projection sub-area of ​​each sub-region;

[0138] The ratio of the total number of data points to the total area of ​​horizontal projection is taken as the total density of data points of the sub-point cloud data to be denoised;

[0139] The ratio of the number of data points in each sub-region to the horizontal projection sub-area of ​​each sub-region is taken as the data point sub-density of the sub-region;

[0140] The first deletion submodule 2013 is used to calculate the noise point interval of each subregion based on the total density of data points and the sub-density of data points in each subregion by using a preset point cloud data filtering algorithm, delete the data points included in each noise point interval, and all the remaining data points constitute the first extracted sub-point cloud data;

[0141] The adjacent data point determination submodule 214 is used to select any data point in the first extracted sub-point cloud data as the data point to be determined, determine a spherical area with the data point to be determined as the sphere center based on a preset radius, and use all data points in the spherical area as adjacent data points of the data point to be determined;

[0142] The calculation submodule 2015 is used to calculate the Manhattan distance between the data point to be determined and each adjacent data point, and calculate the average Manhattan distance corresponding to the data point to be determined;

[0143] The value determination submodule 2016 is used to determine the Gaussian mean and standard deviation according to the average Manhattan distance and a preset Gaussian distribution probability density function;

[0144] The comparison submodule 2017 is used to compare the average Manhattan distance of the data point to be judged with the first preset threshold and the second preset threshold respectively to obtain a comparison result; wherein the first preset threshold is the Gaussian mean, and the second preset threshold is the sum of the Gaussian mean and two times the standard deviation;

[0145] The second deletion submodule 2018 is used to determine that the data point to be judged is a noise data point and delete the data point to be judged when the comparison result is that the average Manhattan distance of the data point to be judged is less than or equal to the first preset threshold or the average Manhattan distance of the data point to be judged is greater than or equal to the second preset threshold; the above operation is performed on all data points in a plurality of sub-point cloud data to obtain a plurality of denoised sub-point cloud data.

[0146] In this embodiment, a specific implementation method of dividing the area occupied by the point cloud into a plurality of sub-areas according to the three-dimensional coordinate system may be: dividing the area occupied by the point cloud into a plurality of cuboid areas evenly according to the three-dimensional coordinate system.

[0147] In this embodiment, based on the total density of data points and the sub-density of data points in each sub-region, a specific implementation method of calculating the noise point interval of each sub-region using a preset point cloud data filtering algorithm may be: determining the ordinate of each data point in the sub-region, taking the smallest ordinate as the first elevation value z1; taking the average ordinate value as the second elevation value z2, and determining the noise point interval F(z) based on the following formula:

[0148]

[0149] Among them, α and β are constants, and their value range is greater than 0 but less than the difference between the maximum elevation value (the largest vertical coordinate) and the second elevation value. The specific values ​​are determined according to the denoising level that needs to be achieved. While ensuring the integrity of the point cloud of the sub-area, the values ​​of α and β are as large as possible; ρ1 is the total density of data points, and ρ2 is the sub-density of data points in the sub-area.

[0150] In this embodiment, when determining to evenly divide the area occupied by the point cloud into a plurality of rectangular areas, the length of the rectangular area is length, the width is width, and the height is height, and the preset radius is Min(length / 2, width / 2, height / 2).

[0151] In this embodiment, the specific implementation method of calculating the Manhattan distance between the data point to be determined and each adjacent data point, and calculating the average Manhattan distance corresponding to the data point to be determined can be: determine the coordinates of the data point to be determined as (x, y, z), determine the coordinates of an adjacent data point as (a, b, c), then the Manhattan distance between the two data points is D=|xa|+|yb|+|zc|, add all the Manhattan distances and divide them by the number of adjacent data points to obtain the average Manhattan distance D 平均 .

[0152] In this embodiment, the specific implementation method of determining the Gaussian mean and standard deviation according to the average Manhattan distance and the preset Gaussian distribution probability density function is as follows: the data point to be determined and the corresponding average Manhattan distance satisfy the Gaussian distribution probability density function relationship Among them, μ is the Gaussian mean, σ is the standard deviation, and the Gaussian mean and standard deviation can be solved according to the relationship. The specific solution process can be completed according to the mean and variance solution method of the Gaussian distribution, which will not be repeated in this embodiment.

[0153] The technical principle and technical effect of the above technical scheme are as follows: the area occupied by the point cloud is divided into several sub-areas according to the three-dimensional coordinate system; the total number of data points in the sub-point cloud data to be denoised and the number of data points in each sub-area are obtained; the total horizontal projection area of ​​the area occupied by the sub-point cloud data to be denoised and the horizontal projection sub-area of ​​each sub-area are obtained; the ratio of the total number of data points to the total horizontal projection area is used as the total density of data points in the sub-point cloud data to be denoised; the ratio of the number of data points in each sub-area to the horizontal projection sub-area of ​​each sub-area is used as the sub-density of data points in the sub-area; the elevation interval of each sub-area is determined based on the elevation filtering method and the data points in the elevation interval are deleted to obtain the corresponding first extracted point cloud data; the adjacent points of each data point in the first extracted point cloud data are determined, the average Manhattan distance between the data point and the adjacent point is compared with the preset threshold to determine the data point to be deleted, and the denoised sub-point cloud data is obtained after the data point is deleted. The noise data is removed, thereby improving the data quality and accuracy of the point cloud data, thereby improving the efficiency of data processing in the subsequent modeling process, saving modeling time, and improving the technical effect of model authenticity.

[0154] The embodiment of the present invention provides an automatic detection system for unmanned aerial vehicles for lightning protection, and a building modeling submodule, including:

[0155] An image acquisition submodule is used to acquire planar images of the top, front, back, left and right sides of the building under test;

[0156] The first enhancer module is used to:

[0157] Select any plane image as the image to be processed, and use a preset bilateral filter kernel to filter the image to be processed to obtain a filtered image;

[0158] Determining an enhancement curve of the filtered image according to the size of the bilateral filter kernel and the pixel intensity value of the filtered image, and enhancing the filtered image based on the enhancement curve to obtain a first enhanced image;

[0159] The second enhancer module is used to:

[0160] Acquire a grayscale image of the first enhanced image;

[0161] Obtain the grayscale value of each pixel in the grayscale image, and determine the grayscale mean according to the grayscale value of each pixel;

[0162] The pixels whose grayscale values ​​in the grayscale image are less than the grayscale mean are taken as the pixels to be enhanced;

[0163] Determine an enhancement multiple of the grayscale value of the pixel to be enhanced according to the grayscale mean, the grayscale value of the pixel to be enhanced and a first preset enhancement coefficient;

[0164] Determine the enhanced grayscale value of the pixel to be enhanced according to the enhancement multiple, the grayscale value of the pixel to be enhanced, and the second preset enhancement coefficient, replace the grayscale value of the pixel to be enhanced with the enhanced grayscale value, implement the enhancement operation of the pixel to be enhanced, and use the image after the enhancement operation as the plane enhanced image;

[0165] Perform the above operations on all plane images to obtain the corresponding plane enhanced images;

[0166] The model generation submodule is used to perform modeling operations on the tested building using 3D modeling software based on all the plane enhanced images and classified point cloud data to obtain an initial building model;

[0167] Model determination submodule, used to:

[0168] Obtain initial building projection images corresponding to the top, front, back, left and right sides of the initial building model;

[0169] Calculate the two-dimensional shape similarity between the initial building projection image and the plane enhanced image of the corresponding surface. When the similarity is greater than a preset similarity threshold, determine that the previous modeling meets the modeling requirements; otherwise, based on the corresponding plane enhanced image, correct the previous modeling;

[0170] Based on the plane enhanced image, the corrected initial building model is color rendered, and the color-rendered initial building model is used as the building model of the measured building.

[0171] In this embodiment, the plane image of the measured building can be collected by the shooting equipment of the drone.

[0172] In this embodiment, the size of the bilateral filter kernel may be 11*11 pixels.

[0173] In this embodiment, the enhancement curve of the filtered image is determined according to the size of the bilateral filter kernel and the pixel intensity value of the filtered image, and the filtered image is enhanced based on the enhancement curve. A specific implementation method of obtaining the first enhanced image may be: determining that the size of the selected bilateral filter kernel is 11*11pixel, and at this time, the most preferred enhancement curve is determined to be a standard S curve, and the corresponding expression is: Enhance=0.26 / (1+exp(-0.26*(S+S max ), where Enhance is the enhancement coefficient of the pixel, S is the intensity value of the pixel, and S max The maximum pixel intensity in the filtered image is obtained by multiplying the enhancement coefficient of the pixel point by the intensity value of the pixel value to obtain the enhanced pixel point. The above operation is performed on all the pixels in the filtered image to obtain the first enhanced image.

[0174] In this embodiment, the specific implementation method of determining the enhancement multiple of the grayscale value of the pixel to be enhanced according to the grayscale mean, the grayscale value of the pixel to be enhanced and the first preset enhancement coefficient can be: enhancement multiple = first preset enhancement coefficient * (1-grayscale mean / grayscale value of the pixel to be enhanced), wherein the first preset enhancement coefficient is between 10-20, and the specific value can be determined according to the actual situation. If the enhancement multiple needs to be larger, the first preset enhancement coefficient can take a larger value, such as 18, 19, etc. If the enhancement multiple needs to be smaller, the first preset enhancement coefficient can take a smaller value, such as 11, 12, etc.

[0175] In this embodiment, the specific implementation method of determining the enhanced grayscale value of the pixel to be enhanced according to the enhancement multiple, the grayscale value of the pixel to be enhanced and the second preset enhancement coefficient can be: the enhanced grayscale value of the pixel to be enhanced = (grayscale value of the pixel to be enhanced + the second preset enhancement coefficient) * enhancement multiple - the second preset enhancement coefficient, wherein the value of the second preset enhancement coefficient is 5.

[0176] In this embodiment, the 3D modeling software may be 3ds max, CAD, SolidWorks, etc.

[0177] In this embodiment, based on all the plane enhanced images and classified point cloud data, the measured building is modeled using 3D modeling software to obtain the initial building model. The specific implementation method can be: using the plane enhanced images as a reference, using 3D modeling software, and performing modeling operations based on the classified point cloud data to obtain the initial building model.

[0178] In this embodiment, a specific implementation method for calculating the two-dimensional shape similarity between the initial building projection image and the plane enhanced image of the corresponding surface may be: obtaining the initial building projection image and the plane enhanced image of the corresponding surface, performing binarization and edge detection operations to obtain a corresponding binary edge image, using a shape description algorithm (for example: an edge point chain code algorithm and an edge point distance transformation algorithm, etc.) to convert the binary edge image into a mathematical representation to obtain a corresponding shape description, using a similarity calculation method (for example: an Euclidean distance calculation method and a Hamming distance calculation method, etc.) to calculate the similarity between the corresponding two shape descriptions, and using the similarity as the two-dimensional shape similarity between the initial building projection image and the plane enhanced image of the corresponding surface.

[0179] In this embodiment, the preset similarity threshold is the minimum value of the multiple two-dimensional shape similarities obtained by the first calculation.

[0180] In this embodiment, a specific implementation method of color rendering the modified initial building model may be: rendering the outer surface and each object on the outer surface of the modified initial building into the color of the outer surface of the building and each object on the outer surface.

[0181] The technical principle and technical effect of the above technical scheme can be: collecting plane images of the top, front, back and left and right sides of the measured building; performing a first enhancement and a second enhancement on the plane image to obtain a corresponding plane enhanced image; based on all the plane enhanced images and classified point cloud data, using 3D modeling software to perform modeling operations on the measured building to obtain an initial building model; obtaining the initial building projection map corresponding to the top, front, back and left and right sides of the initial building model; calculating the two-dimensional shape similarity between the initial building projection map and the plane enhanced image of the corresponding surface, when the similarity is greater than a preset similarity threshold, determining that the previous modeling meets the modeling requirements; otherwise, based on the corresponding plane enhanced image, correcting the previous modeling; based on the plane enhanced image, color rendering is performed on the corrected initial building model, and the initial building model that has completed color rendering is used as the building model of the measured building. The following technical effects are achieved: the plan views of each face of the measured building are enhanced to improve image clarity and authenticity, thereby achieving the technical effect of ensuring the correctness of the building structure when modeling is performed based on the plan views of the measured building; the two-dimensional shape similarity of the projection views of each face of the initial building model and the plan enhancement views of the corresponding faces is calculated, thereby achieving the technical effect of ensuring that the building model has high restoration; the building model is rendered to achieve the technical effect of ensuring the realism of the building model.

[0182] The embodiment of the present invention provides an automatic detection system for a UAV for lightning protection, a marking module, including:

[0183] A position determination submodule is used to determine the position of the lightning-termination object in the building model;

[0184] The object marking submodule is used to mark the lightning-receiving objects at the determined positions by using the preset marking colors to obtain the marked lightning-receiving objects.

[0185] In this embodiment, the lightning-termination objects include lightning-termination strips, lightning-termination rods, and the like.

[0186] The technical principle and technical effect of the above technical solution are: determining the position of the lightning-termination object in the building model; marking the lightning-termination object at the determined position using a preset marking color to obtain a marked lightning-termination object. The marking of the lightning-termination object is achieved, so that the lightning-termination object is highlighted in the model, thereby ensuring the accuracy of the subsequently determined joint protection range.

[0187] Reference Figure 3 , an embodiment of the present invention provides an automatic detection system for unmanned aerial vehicles for lightning protection, a range determination module, including:

[0188] A category determination submodule is used to determine the lightning protection category of the tested building according to the type of the tested building and based on the existing lightning protection category classification regulations for buildings;

[0189] A rolling ball determination submodule, used to determine the radius of the rolling ball based on the lightning protection category, and determine the target rolling ball based on the radius;

[0190] Scoping submodule for:

[0191] Control the target rolling ball to roll all-round on the building model starting from the ground position of the building model;

[0192] Based on the position of the marked lightning-termination objects, the range that the target rolling ball cannot reach due to the obstruction of the marked lightning-termination objects during its rolling process is divided into the joint protection range of all the marked lightning-termination objects.

[0193] In this embodiment, the lightning protection categories of buildings are divided into three categories, for example: the first category of lightning protection buildings includes dangerous buildings where explosives and their products are manufactured, used or stored; the second category of lightning protection buildings includes national key cultural relics protection buildings; the third category of lightning protection buildings includes provincial key cultural relics protection buildings and provincial archives; for specific classification, please refer to Part 3 (Lightning Protection Classification of Buildings) in the Building Lightning Protection Design Code GB50057-2010.

[0194] In this embodiment, when the lightning protection category of the building is the first category, the rolling ball radius is 30 meters, when the lightning protection category of the building is the second category, the rolling ball radius is 45 meters, and when the lightning protection category of the building is the third category, the rolling ball radius is 60 meters.

[0195] In this embodiment, based on the position of the marked lightning-receiving object, the range that the target rolling ball cannot reach due to the obstruction of the marked lightning-receiving object during rolling is divided into the joint protection range of all the marked lightning-receiving objects. Figure 3 As shown in the shaded part, the circle in the figure is the target rolling ball, and the figure shows the plane rolling path of the target rolling ball, such as Figure 3 As shown, the target rolling ball starts to roll from the ground / earth, and when rolling over the building on the right, it touches some protruding facade objects and part of the right parapet wall. Then, it is blocked by the lightning strip (lightning terminal) of the right building, the two lightning rods on the roof of the left building, and the lightning strip on the roof of the left building, and does not touch the roofs of the two buildings, the parapet wall of the left building, and the left parapet wall of the right building. Finally, it rolls back to the ground / earth. The range that cannot be reached due to the obstruction of the lightning strips and lightning rods during the rolling process is the joint protection range.

[0196] The technical principle and technical effect of the above technical solution are as follows: according to the type of the tested building, the lightning protection category of the tested building is determined based on the existing building lightning protection category classification regulations; the radius of the rolling ball is determined based on the lightning protection category, and the target rolling ball is determined based on the radius; the target rolling ball is controlled to roll all-round on the building model starting from the ground position of the building model; based on the position of the marked lightning-receiving object, the range that the target rolling ball cannot reach during the rolling process due to the obstruction of the marked lightning-receiving object is divided into the joint protection range of all marked lightning-receiving objects. After determining the lightning protection category of the building, the joint protection range of the building's lightning-receiving objects is determined based on the rolling ball method, and then the unprotected range is determined based on the joint protection range, thereby accelerating the efficiency and accuracy of the subsequent unprotected object screening process.

[0197] Still refer to Figure 3 The embodiment of the present invention provides an automatic detection system for unmanned aerial vehicles for lightning protection, an object screening module, including:

[0198] The object detection submodule is used to detect all objects in the building model that are outside the joint protection range using a preset detection device to obtain a detection result;

[0199] The object determination submodule is used to compare the detection results with preset parameters according to the lightning protection category of the tested building, and determine the objects outside the joint protection range whose detection results do not meet the preset parameters as unprotected objects.

[0200] In this embodiment, the preset detection device may be an object electromagnetic field tester and a ground potential measuring instrument, etc.

[0201] In this embodiment, the preset parameters can be obtained according to the provisions of the building lightning protection design code GB50057-2010. For example, when the object outside the joint protection range is an isolated metal object on the roof, according to the provisions in 4.5.7, when its size exceeds the following values: no more than 0.3 meters above the roof plane, the total area of ​​the upper surface does not exceed 1.0 square meters or the length of the upper surface does not exceed 2.0 meters, the object is determined to be an unprotected object. At this time, the corresponding preset parameters are 0.3, 1.0 and 2.0; when the object outside the joint protection range is a non-conductive roof object, according to the provisions in 4.5.7, when it protrudes more than 0.5 meters from the plane formed by the lightning rod, the object is determined to be an unprotected object. At this time, the corresponding preset parameter is 0.5. The parameters in other specific cases can refer to the provisions in the building lightning protection design code GB50057-2010.

[0202] In this embodiment, the unprotected object is Figure 3 The protruding facade object shown is an unprotected object because part of it is not within the scope of joint protection.

[0203] The technical principle and technical effect of the above technical solution are: using a preset detection device to detect all objects in the building model that are outside the joint protection range to obtain detection results; according to the lightning protection category of the tested building, the detection results are compared with preset parameters, and objects outside the joint protection range whose detection results do not meet the preset parameters are determined as unprotected objects. The unprotected objects outside the joint protection range are determined according to the object detection results, and subsequent object protection is implemented according to the object determination results, thereby ensuring the technical effect of object safety and building safety.

[0204] An embodiment of the present invention provides an automatic detection system for a drone for lightning protection, which also includes a measure generation module for generating lightning protection measures for objects based on object information of unprotected objects.

[0205] In this embodiment, the lightning protection measures for objects may include setting up lightning termination devices (lightning termination strips, lightning termination rods, lightning termination wires, lightning termination nets, metal components), installing lightning protection grounding devices, etc.

[0206] The technical principle and technical effect of the above technical scheme are: generating lightning protection measures for objects based on the object information of unprotected objects, realizing the determination of adaptive lightning protection measures based on the object information, avoiding objects from being struck by lightning due to being not within the joint protection range of lightning-receiving objects, and thus ensuring the technical effect of building safety.

[0207] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An automatic detection system for unmanned aerial vehicles for lightning protection, characterized in that: include: A data acquisition module is used to scan the building under test using a laser radar device installed on the drone to obtain total point cloud data of the building under test; A modeling module is used to perform modeling operations based on the total point cloud data to obtain a building model of the measured building; A marking module is used to mark the lightning-receiving objects in the building model to obtain marked lightning-receiving objects; A range determination module is used to determine the combined protection range of all marked lightning termination objects using a rolling ball method based on the locations of the marked lightning termination objects and the lightning protection category of the building under test; The object screening module is used to detect all objects in the building model that are outside the joint protection range and screen out unprotected objects; Modeling modules, including: The data denoising submodule is used to perform data denoising on a number of sub-point cloud data included in the total point cloud data to obtain a number of denoised sub-point cloud data; wherein a double-layer point cloud denoising algorithm of density filtering and Manhattan statistical filtering is used; The data registration submodule is used to convert all data points in several denoised sub-point cloud data into the target coordinate system. After the conversion, all data points in the target coordinate system constitute the registered point cloud data. Among them, the coarse registration matrix is ​​obtained by using the random sampling consistency algorithm, and then the coarse registration matrix is ​​iteratively optimized by using the iterative closest point algorithm to obtain the multi-level registration strategy of the fine registration matrix. Data classification submodule, used to: Obtain existing building registration point cloud data and corresponding classified point cloud data, and use the existing building registration point cloud data and the corresponding classified point cloud data to train a neural network to obtain a point cloud data classification model; wherein the classified point cloud data is obtained by classifying data points in the registration point cloud data according to the categories of corresponding objects in the building; Inputting the registered point cloud data into a point cloud data classification model to obtain corresponding classified point cloud data; The building modeling submodule is used to perform modeling operations on the measured building based on the classified point cloud data to obtain the building model of the measured building; wherein, plane images of the top, front, back and left and right sides of the measured building are collected, a first enhancement is performed using a bilateral filter kernel, a grayscale image of the first enhanced image is obtained, a second enhancement is performed on the grayscale image to obtain a corresponding plane enhanced image, and the measured building is modeled based on all the plane enhanced images and the classified point cloud data.

2. The automatic detection system for unmanned aerial vehicle for lightning protection according to claim 1, characterized in that: Data acquisition module, including: The scanning submodule is used to use the laser radar device to emit laser beams to the measured building at different positions, and scan the measured building several times at different angles; The data acquisition submodule is used to receive the optical signal reflected by the laser beam after each scan, determine the sub-point cloud data acquired by the current scan according to the optical signal, and compose the total point cloud data of the measured building according to a plurality of sub-point cloud data.

3. The automatic detection system for unmanned aerial vehicle for lightning protection according to claim 1, characterized in that: Data registration submodule, including: A determination submodule is used to determine any denoised sub-point cloud data as target sub-point cloud data, determine the coordinate system where the target sub-point cloud data is located as the target coordinate system; determine any denoised sub-point cloud data other than the target sub-point cloud data as the sub-point cloud data to be processed; A vector determination submodule, used for calculating the feature vector of each data point in the target sub-point cloud data and the sub-point cloud data to be processed by using a fast point feature histogram algorithm; A distance calculation submodule is used to select a feature vector of any data point in the sub-point cloud data to be processed as a current feature vector, calculate the Euclidean distance between the current feature vector and the feature vectors of all data points in the target sub-point cloud data, perform the above operation on the feature vectors of all data points in the sub-point cloud data to be processed, and obtain the Euclidean distance between the feature vector of each data point in the sub-point cloud data to be processed and the feature vectors of all data points in the target sub-point cloud data; A data point pair determination submodule is used to compare the sizes of all Euclidean distances, sort all Euclidean distances from large to small, obtain a Euclidean distance sorting queue, and select three groups of data point pairs corresponding to three groups of feature vectors corresponding to the last three Euclidean distances in the Euclidean distance sorting queue as reference data point pairs; A coordinate system conversion submodule is used to obtain a coarse registration matrix based on the reference data point pair using a random sampling consistency algorithm, and to perform a first rigid body transformation on the data points in the sub-point cloud data to be processed using the coarse registration matrix, so as to convert the data points in the sub-point cloud data to be processed from the current coordinate system to the target coordinate system; The registration submodule is used to: The iterative closest point algorithm is used to iteratively optimize the coarse registration matrix to obtain the fine registration matrix; Using the precise registration matrix, a second rigid body transformation is performed on the data points in the transformed sub-point cloud data to be processed in the target coordinate system, and all data points in the sub-point cloud data to be processed are registered to the target coordinate system; The above operation is performed on all denoised sub-point cloud data except the target sub-point cloud data, and all data points in the denoised sub-point cloud data are converted to the target coordinate system. After the conversion, all data points in the target coordinate system constitute the registered point cloud data.

4. The automatic detection system for unmanned aerial vehicle for lightning protection according to claim 1, characterized in that: Data denoising submodule, including: A region division submodule is used to select any sub-point cloud data as the sub-point cloud data to be denoised, determine a three-dimensional coordinate system based on the region occupied by the sub-point cloud data to be denoised, and divide the region occupied by the point cloud into a plurality of sub-regions according to the three-dimensional coordinate system; Density acquisition submodule, used for: Obtain the total number of data points in the sub-point cloud data to be denoised and the number of data points in each sub-area; Obtain the total horizontal projection area of ​​the region occupied by the sub-point cloud data to be denoised and the horizontal projection sub-area of ​​each sub-region; The ratio of the total number of data points to the total area of ​​horizontal projection is taken as the total density of data points of the sub-point cloud data to be denoised; The ratio of the number of data points in each sub-region to the horizontal projection sub-area of ​​each sub-region is used as the data point sub-density of the sub-region; A first deletion submodule is used to calculate the noise point interval of each subregion based on the total density of data points and the sub-density of data points in each subregion by using a preset point cloud data filtering algorithm, delete the data points contained in each noise point interval, and all the remaining data points constitute the first extracted sub-point cloud data; A neighboring data point determination submodule is used to select any data point in the first extracted sub-point cloud data as a data point to be determined, determine a spherical area with the data point to be determined as the sphere center based on a preset radius, and use all data points in the spherical area as neighboring data points of the data point to be determined; A calculation submodule is used to calculate the Manhattan distance between the data point to be judged and each adjacent data point, and calculate the average Manhattan distance corresponding to the data point to be judged; A value determination submodule, used for determining a Gaussian mean and a standard deviation according to the average Manhattan distance and a preset Gaussian distribution probability density function; A comparison submodule, used to compare the average Manhattan distance of the data point to be judged with a first preset threshold and a second preset threshold respectively, to obtain a comparison result; wherein the first preset threshold is the Gaussian mean, and the second preset threshold is the sum of the Gaussian mean and two times the standard deviation; The second deletion submodule is used to determine that the data point to be judged is a noise data point and delete the data point to be judged when the comparison result is that the average Manhattan distance of the data point to be judged is less than or equal to the first preset threshold or the average Manhattan distance of the data point to be judged is greater than or equal to the second preset threshold; the above operation is performed on all data points in a plurality of sub-point cloud data to obtain a plurality of denoised sub-point cloud data.

5. The automatic detection system for unmanned aerial vehicle for lightning protection according to claim 1, characterized in that: Building modeling submodule, including: An image acquisition submodule is used to acquire planar images of the top, front, back, left and right sides of the building under test; The first enhancer module is used to: Select any plane image as an image to be processed, and filter the image to be processed using a preset bilateral filter kernel to obtain a filtered image; Determining an enhancement curve of the filtered image according to the size of the bilateral filter kernel and the pixel intensity value of the filtered image, and enhancing the filtered image based on the enhancement curve to obtain a first enhanced image; The second enhancer module is used to: Acquire a grayscale image of the first enhanced image; Obtaining the grayscale value of each pixel in the grayscale image, and determining the grayscale mean according to the grayscale value of each pixel; The pixels whose grayscale values ​​in the grayscale image are less than the grayscale mean are taken as the pixels to be enhanced; Determine an enhancement multiple of the grayscale value of the pixel to be enhanced according to the grayscale mean, the grayscale value of the pixel to be enhanced and a first preset enhancement coefficient; Determine the enhanced grayscale value of the pixel to be enhanced according to the enhancement multiple, the grayscale value of the pixel to be enhanced, and the second preset enhancement coefficient, replace the grayscale value of the pixel to be enhanced with the enhanced grayscale value, implement the enhancement operation of the pixel to be enhanced, and use the image after the enhancement operation as the plane enhanced image; Perform the above operations on all plane images to obtain the corresponding plane enhanced images; The model generation submodule is used to perform modeling operations on the tested building using 3D modeling software based on all the plane enhanced images and classified point cloud data to obtain an initial building model; Model determination submodule, used to: Obtaining initial building projection images corresponding to the top, front, back, and left and right sides of the initial building model; Calculating the two-dimensional shape similarity between the initial building projection image and the plane enhanced image of the corresponding surface, and when the similarity is greater than a preset similarity threshold, determining that the previous modeling meets the modeling requirements; otherwise, correcting the previous modeling based on the corresponding plane enhanced image; Based on the plane enhanced image, the corrected initial building model is color rendered, and the color-rendered initial building model is used as the building model of the measured building.

6. The automatic detection system for unmanned aerial vehicle for lightning protection according to claim 1, characterized in that: Marking module, including: A position determination submodule is used to determine the position of the lightning-termination object in the building model; The object marking submodule is used to mark the lightning-receiving objects at the determined positions by using the preset marking colors to obtain the marked lightning-receiving objects.

7. The automatic detection system for unmanned aerial vehicle for lightning protection according to claim 1, characterized in that: Scoping module, including: A category determination submodule is used to determine the lightning protection category of the tested building according to the type of the tested building and based on the existing lightning protection category classification regulations for buildings; A rolling ball determination submodule, used to determine the radius of the rolling ball based on the lightning protection category, and determine the target rolling ball based on the radius; Scoping submodule for: Control the target rolling ball to roll all-round on the building model starting from the ground position of the building model; Based on the position of the marked lightning-termination objects, the range that the target rolling ball cannot reach due to the obstruction of the marked lightning-termination objects during its rolling process is divided into the joint protection range of all the marked lightning-termination objects.

8. The automatic detection system for unmanned aerial vehicle for lightning protection according to claim 6, characterized in that: Object screening module, including: The object detection submodule is used to detect all objects in the building model that are outside the joint protection range using a preset detection device to obtain a detection result; The object determination submodule is used to compare the detection result with the preset parameters according to the lightning protection category of the tested building, and determine the objects outside the joint protection range whose detection results do not meet the preset parameters as unprotected objects.

9. The automatic detection system for unmanned aerial vehicle for lightning protection according to claim 1, characterized in that: It also includes a measure generation module, which is used to generate lightning protection measures for objects based on object information of unprotected objects.

Images