Lightning location error correction method, system and medium in coordination of optical and vlf data

Abstract

This invention relates to the field of lightning location technology, and discloses a method, system, and medium for lightning location error correction using optical and VLF data collaboration. The method determines the observed point and deploys optical equipment and a VLF location network around it. When the observed point is struck by lightning, the time of the lightning strike recorded by the optical equipment, as well as the VLF three-dimensional location coordinates and detection time detected by the VLF location network, are acquired. The detection time and the lightning strike time are filtered and matched through a set time window, and the VLF three-dimensional location coordinates are calculated with the known three-dimensional spatial coordinates of the observed point to obtain the corresponding error vector. A training sample set is constructed. Based on the training sample set, a linear mapping model between the VLF three-dimensional location coordinates and the error correction vector is established. When the VLF location network detects a new lightning event, the error correction vector is calculated through the linear mapping model to correct the VLF three-dimensional location coordinates. This invention improves the positioning accuracy of existing VLF location networks.

Description

Technical Field

[0001] This invention relates to the field of lightning location technology, specifically to a lightning location error correction method that combines optical and VLF data, as well as a system and storage medium that applies this method. Background Technology

[0002] Lightning location systems are core equipment for monitoring lightning activity, providing early warning of lightning disasters, and analyzing lightning strike faults. Among them, very low frequency (VLF) three-dimensional location technology based on the time difference of arrival (TDOA) method has become the mainstream lightning detection method due to its high detection sensitivity to electromagnetic radiation sources of cloud-to-ground lightning.

[0003] However, this technology faces inherent limitations in practical deployment: First, the propagation delay error caused by electromagnetic waves propagating in a non-ideal homogeneous atmosphere is unavoidable; second, measurement errors in the coordinates of the VLF positioning network's detection stations, slight deviations in instrument clock synchronization, and model approximations in the solution algorithm collectively introduce significant systematic positioning errors. These errors result in positioning results with uncertainties of hundreds of meters, making it difficult to meet the high-precision application requirements such as accurate location of lightning strike faults in power systems and safety monitoring of aerospace launch sites. In summary, how to perform real-time accuracy correction on the operating VLF positioning network to improve the accuracy of lightning positioning results is a pressing technical problem to be solved in current lightning positioning engineering. Summary of the Invention

[0004] To address the technical problems existing in the prior art, this invention provides a lightning positioning error correction method, system, and medium that integrates optical and VLF data. This method enables deep fusion of optical and VLF data and completes the calibration from error data processing to real-time positioning compensation closed loop, thereby improving the positioning accuracy of existing VLF positioning networks.

[0005] To achieve the above objectives, the present invention provides the following technical solution: This invention discloses a lightning location error correction method based on the coordination of optical and VLF data, comprising the following steps: S1. Determine the observed point with known three-dimensional spatial coordinates, and deploy optical equipment and VLF positioning network around the observed point; S2. When the observed point is struck by lightning, the time of the lightning event recorded by the optical equipment and the VLF three-dimensional positioning coordinates and detection time detected by the VLF positioning network are obtained respectively; the detection time and the time of the lightning event are filtered and matched through a set time window. After the match is successful and it is determined that the two are obtained from the same lightning event, the VLF three-dimensional positioning coordinates and the known three-dimensional spatial coordinates of the observed point are calculated to obtain the corresponding error vector. S3. Collect VLF three-dimensional positioning coordinates and corresponding error vectors from multiple historical lightning strike events, and construct a training sample set accordingly; perform calculations based on the training sample set to obtain the weight matrix used to characterize the spatial distribution law of VLF positioning network errors, and then establish a linear mapping model between VLF three-dimensional positioning coordinates and error correction vectors based on the weight matrix. S4. When the VLF positioning network detects a new lightning event, it obtains the current VLF three-dimensional positioning coordinates and substitutes them as input parameters into the linear mapping model for calculation to obtain an error correction vector. The error correction vector is then added to the current VLF three-dimensional positioning coordinates to obtain the final corrected positioning coordinates.

[0006] As a further improvement to the above scheme, in step S3, the specific calculation process of the weight matrix and the linear mapping model includes: Based on the N training samples in the training sample set, construct the original positioning matrix containing the VLF three-dimensional positioning coordinates. and the target matrix containing the corresponding error vector ; Using mean squared error as the loss function, the expression is: ; In the formula, Represents the weight matrix The loss function; Let Frobenius norm be the matrix. Apply the loss function to the weight matrix The gradient matrix is ​​obtained by taking the derivative. Setting the gradient matrix to zero, the optimal weight matrix is ​​then obtained. ; In the formula, T represents the matrix transpose operation; , Represents a 4x3 real number matrix; Based on the weight matrix obtained by the solution Establish a linear mapping model: ; In the formula, This is the error correction vector; VLF three-dimensional positioning coordinates Augmented feature vectors.

[0007] As a further improvement to the above scheme, after step S4, the following steps are also included: Following the method in step S2, new lightning strike event data occurring at the observed point are continuously acquired, and the corresponding new error vector is calculated and expanded into the training sample set; When the number of new samples added reaches a preset threshold, or when the time since the last weight matrix update reaches a preset period, the updated training sample set is used to resolve and replace the current weight matrix.

[0008] As a further improvement to the above scheme, in step S2, the specific process of the optical device recording the time of the lightning strike includes: Set a fixed time in microseconds and a threshold for brightness changes in consecutive image frames; The brightness change of consecutive image frames within the fixed time period is monitored. When the brightness change exceeds the brightness change threshold, it is determined that the observed point has been struck by lightning, and the timestamp of the frame is extracted as the time of the lightning strike event.

[0009] As a further improvement to the above scheme, in step S2, the specific process of the optical device recording the time of the lightning strike includes: Set a fixed time in microseconds and a threshold for brightness changes in consecutive image frames; Monitor the brightness change of consecutive image frames within the fixed time period. When the brightness change exceeds the brightness change threshold, extract the current observed point image block. The observed point image patch is used as input parameter and fed into a pre-trained convolutional neural network. The convolutional neural network extracts hierarchical features step by step through a feature extraction network containing convolutional layers and pooling layers, and outputs a visual feature vector. The visual feature vector is then input into a fully connected layer and linearly transformed and mapped through a sigmoid activation function to output the confidence probability of the lightning strike event. The convolutional neural network is pre-trained using a positive sample set containing real lightning strikes on the observed point image frames and a negative sample set containing light interference or lightning image patches that are not struck. When the confidence probability of the lightning strike event is greater than the set probability threshold, it is finally determined that the observed point has actually been struck by lightning, and the timestamp of the frame is extracted as the time of the lightning strike event.

[0010] As a further improvement to the above scheme, in step S1, the observed point is a structure with a lightning rod tip on top, and the three-dimensional spatial coordinates of the observed point are determined by pre-measuring the latitude, longitude and elevation data of the observed point; the optical equipment includes n high-speed cameras deployed within a range of 500m to 2000m from the observed point, and the azimuth angle between adjacent high-speed cameras is 360 / n°; the deployment altitude of each high-speed camera is lower than the altitude of the observed point, and the observation elevation angle is between 25° and 45°.

[0011] As a further improvement to the above scheme, in step S1, multiple observation points that meet the conditions and have known three-dimensional spatial reference coordinates are selected within the same geographical area. For each observed point, optical devices are deployed collaboratively to obtain the corresponding error vector. Then, the error vector data obtained from multiple observed points are summarized to jointly construct and expand the training sample set, thereby shortening the data collection cycle required to solve the weight matrix.

[0012] As a further improvement to the above scheme, each detection substation in the VLF positioning network collects lightning radiation signals and performs preliminary processing on the lightning radiation signals. Each detection substation transmits the pre-processed data to the central server, which then performs comprehensive positioning calculations on the multiple data streams to obtain the VLF three-dimensional positioning coordinates.

[0013] This invention also discloses a lightning location error correction system based on optical and VLF data coordination, used to implement the steps of the aforementioned lightning location error correction method based on optical and VLF data coordination. The system includes: The data acquisition module includes optical equipment and a VLF positioning network deployed around the observed point; the optical equipment is used to acquire optical image data of the observed point and record the time of the lightning strike when it is determined that the observed point has been struck by lightning; the VLF positioning network is used to acquire the VLF three-dimensional positioning coordinates and detection time of the lightning event within its own detection range. The error calculation and sample construction module is used to match the same lightning strike event through a time window and calculate the error vector of the VLF positioning coordinates relative to the true coordinates of the observed point in order to construct a training sample set. The matrix solving and modeling module is used to solve for the weight matrix representing the distribution law of system error based on the training sample set, and to establish a linear mapping model accordingly. The positioning compensation processing module, deployed on the central server, is used to call the linear mapping model to calculate the error correction vector when the VLF three-dimensional positioning coordinates of a new lightning event are obtained, thereby correcting the current VLF three-dimensional positioning coordinates.

[0014] The present invention also discloses a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of the above-described method for correcting lightning positioning errors by coordinating optical and VLF data are implemented.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. The lightning positioning error correction method disclosed in this invention, which combines optical and VLF data, calibrates the error of the VLF positioning network using optical equipment. Verification has shown a significant improvement in positioning accuracy. Furthermore, it transforms optical observation from an offline, verification tool into part of an online calibration system, forming a closed-loop control of "perception-modeling-decision-execution," thus achieving a leap from offline to online operation.

[0016] 2. The weight matrix of this invention is not fixed. As optical observations continue, the error vector dataset will gradually expand, and the weight matrix can be updated and made more reliable. This enables the positioning network to adapt to error drift caused by seasonal changes, equipment aging, etc., and has dynamic adaptive capabilities.

[0017] 3. The present invention can also employ cascaded determination for the acquisition method of optical data. Compared with the traditional optical data acquisition method, this method greatly improves the effectiveness and reliability of data samples, thereby further improving positioning accuracy. Attached Figure Description

[0018] Figure 1 This is a flowchart of the lightning positioning error correction method based on optical and VLF data coordination in Embodiment 1 of the present invention.

[0019] Figure 2 This is a schematic diagram of the deployment of the optical high-speed camera in Embodiment 1 of the present invention.

[0020] Figure 3 This is a VLF positioning network deployment diagram in Embodiment 1 of the present invention.

[0021] Figure 4 This is a schematic diagram of the error vector in Embodiment 1 of the present invention.

[0022] Figure 5 This is a comparison diagram of the positioning results before and after error correction for a certain lightning strike event in Embodiment 1 of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1 Please see Figure 1 This embodiment provides a lightning positioning error correction method based on the coordination of optical and VLF data, including the following steps: S1. Determine the observation point with known three-dimensional spatial coordinates, and deploy optical equipment and VLF positioning network around the observation point.

[0025] The first step is the selection of the observation points. To ensure data sufficiency and the reliability of subsequent localization algorithms, the selection of observation points is crucial. They must meet the following conditions: ① This point is located in an area where natural lightning events occur frequently throughout the year, and it may be a building or structure with a lightning rod on top.

[0026] ②The precise latitude and elevation of this point must be obtained through high-precision measuring equipment.

[0027] ③ The selection of this location should fully consider the feasibility of equipment deployment and maintenance, and should not be located in a hazardous environment.

[0028] In this embodiment, as Figure 2 As shown, two high-speed optical cameras are deployed symmetrically on both sides of the tower at a distance of approximately 620m. The elevation angle of the cameras is adjusted to 30°, ensuring that the cameras can clearly observe the top of the tower at this elevation angle.

[0029] Next is the layout and installation of the optical equipment. To prevent interference from clouds and fog, ensure the accuracy of optical observation results, and reduce errors, two or more high-speed cameras are required. These cameras should have all-weather protection, accurate time synchronization, and backend communication capabilities. Furthermore, to withstand natural wind forces and prevent image shake and camera damage, the cameras must be mounted on extremely stable platforms. In addition, the following issues need to be considered in the camera layout and installation: ① To ensure that the camera can clearly see the structural details of the tower top (such as the lightning rod and its base), it needs to be installed within a distance of 500m-2000m from the observation point. Too close may result in an excessively high elevation angle, which is not conducive to observation, while too far away may be easily affected by rain and fog, affecting the observation results.

[0030] ② The azimuth angle between multiple cameras should be as large as possible. For example, in a dual-camera setup, the angle can be 180°, and in a three-camera setup, they should be evenly spaced at 120° intervals.

[0031] ③ The altitude at which the camera is installed should be lower than the altitude of the observed point, and the elevation angle of the camera should be between 25° and 45° for optimal results.

[0032] Finally, there is the equipment layout and installation in the VLF positioning network. To ensure minimal error in the equipment positioning results, different deployment methods must be adopted based on the positioning principle and specific circumstances of the VLF positioning equipment to be installed. Special cases require special handling, and the deployment plan must be centered on the observed point. The following issues need to be considered when installing a single VLF positioning device: ① It should be installed in an open location with clean electromagnetic fields, as far away as possible from electromagnetic interference sources such as power line carrier communication, transformers, and variable frequency motors.

[0033] ② It should be installed on a solid flat surface to prevent the influence of natural factors such as wind and loose soil.

[0034] ③ The difficulty of installation and maintenance should be fully considered, and locations with lower risk factors should be prioritized for installation.

[0035] In this embodiment, based on factors such as the site's electromagnetic environment, natural environment, and communication conditions, five sites were selected as the VLF positioning network. Their deployment is as follows: Figure 3 As shown. By selecting the central station from the five stations as the reference station for the time difference of arrival method, the positioning result can be obtained using a conventional five-station three-dimensional positioning algorithm. The five-station three-dimensional positioning algorithm can be found in the following literature: Cai Li. Research on three-dimensional full lightning positioning technology of ground-based VLF / LF [D]. Wuhan University, 2013. In some embodiments, multiple observation points that meet the conditions and have known three-dimensional spatial reference coordinates can be selected within the same geographical area; for each observation point, optical devices are deployed collaboratively to obtain the corresponding error vector, and then the error vector data obtained from multiple observation points are summarized to jointly construct and expand the training sample set, so as to shorten the data collection cycle required to solve the weight matrix.

[0036] S2. When the observed point is struck by lightning, the time of the lightning strike event recorded by the optical equipment and the VLF three-dimensional positioning coordinates and detection time detected by the VLF positioning network are obtained respectively. The detection time and the time of the lightning strike event are filtered and matched through a set time window. After the match is successful and it is determined that the two are obtained from the same lightning strike event, the VLF three-dimensional positioning coordinates and the known three-dimensional spatial coordinates of the observed point are calculated to obtain the corresponding error vector.

[0037] Since the three-dimensional spatial position of the observed point is fixed, the data processing for the optical high-speed camera only needs to enable it to identify whether a lightning event has occurred at the observed point.

[0038] When a high-speed camera observes a designated location, if no lightning event occurs, the brightness change of consecutive image frames is slow; however, if a lightning event occurs, the brightness change of consecutive image frames within a fixed time period will be significant. Therefore, let the fixed time period be... (Given the instantaneous nature of lightning strikes, this time should be on the order of microseconds), fixed time The change within is Then, simply set the brightness change threshold for consecutive image frames (set to...). ),like If so, it is assumed that the observed point was struck by lightning. If so, it is assumed that the observed point was not struck by lightning. Specifically... The value needs to be determined by the sensitivity of the corresponding high-speed camera to changes in brightness.

[0039] when When a lightning event occurs, the high-speed camera records the time of the event and stores it in its log. The point of lightning strike is then recorded as the three-dimensional spatial coordinates of the observed point.

[0040] Processing VLF data relies on a pre-deployed VLF positioning network. This network should include substations, a central server, and a host computer. The central server should possess comprehensive positioning calculation, communication, data storage, and query functions. The general data processing flow is as follows: substations in the VLF positioning network collect lightning radiation signals and perform preliminary processing; the substations transmit the pre-processed data to the central server, which then performs comprehensive positioning calculations to obtain the final 3D positioning result; the positioning result and related information are stored in a database.

[0041] Finally, by filtering and matching the time window and optical data, it was determined that both data originated from the same lightning event. Finally, the error vector of the two data sets was calculated. Figure 4 As shown.

[0042] S3. Collect VLF three-dimensional positioning coordinates and corresponding error vectors from multiple historical lightning strike events, and construct a training sample set accordingly; perform calculations based on the training sample set to obtain the weight matrix used to characterize the spatial distribution law of VLF positioning network errors, and then establish a linear mapping model between VLF three-dimensional positioning coordinates and error correction vectors based on the weight matrix.

[0043] Over a given period, after collecting a sufficient number of error vectors from different lightning events, these are integrated into an error vector set. The calculated error vector dataset is then used to train a linear mapping model. In this embodiment, to ensure the reliability of the real-time positioning compensation algorithm and improve the positioning accuracy of the VLF positioning network, an error vector set consisting of 200-300 error vector samples is required. Ideally, for a single observed point, obtaining error vector samples from 200-300 lightning events would require continuous observation for approximately 3-5 years. Therefore, multiple observed points can be selected to shorten the period for collecting error vector samples.

[0044] In step S3, the specific calculation process of the weight matrix and the linear mapping model includes: Let the coordinates given by the VLF positioning network be... The corresponding error vector is Establish the following linear mapping relationship: ; in To augment the feature vector, This is the weight matrix; This represents a 4x3 real matrix; the superscript T is the transpose symbol.

[0045] Collect N training samples Construct the original positioning matrix and target matrix Original positioning matrix As shown below: ; Target matrix as follows: ; To ensure the optimal weight matrix is ​​obtained, given the known design and target matrices, this matrix model training uses mean squared error as the loss function to measure the weight matrix. The advantages and disadvantages of the loss function are determined by the following expression: ; in, Let Frobenius be the Frobenius norm of the matrix, which is the sum of squares of all elements in the matrix, mathematically represented as the sum of squared residuals. Applying the loss function of the above equation to the matrix... Taking the derivative yields the gradient matrix of the given matrix. Setting the gradient matrix to zero, the derivation is as follows: , as well as The normal equation between them: ; Then solve for the optimal weight matrix. : ; After calculating a specific weight matrix using data samples, the weight matrix is ​​compiled and tested. When the VLF positioning network detects a lightning event, the feature vector... This will be used as input, and the error vector can be obtained from the formula of the linear mapping model. Then, the result is compared with the actual error vector of the sample to determine whether the deviation is large. If the deviation is small, the weight matrix is ​​considered to be mature and usable.

[0046] S4. When the VLF positioning network detects a new lightning event, obtain its current VLF three-dimensional positioning coordinates. The error correction vector is obtained by substituting it as an input parameter into the linear mapping model for calculation. The error correction vector is added to the current VLF 3D positioning coordinates to obtain the final corrected positioning coordinates: .

[0047] Once the mature real-time positioning compensation algorithm, consisting of a weighted matrix and a positioning compensation module, is completed, it can be embedded into the central server. Its matrix operations can be completed within microseconds, making the latency of the entire error compensation process negligible. This high-efficiency computing power ensures seamless processing of lightning positioning data streams. On the Ubuntu operating system of the VLF positioning network central server, the successfully implemented algorithm source code is packaged into a callable, independent function. Then, after comprehensively calculating the original VLF positioning coordinates, the real-time positioning compensation algorithm function is called, with everything else remaining unchanged.

[0048] When the tower spire is struck by lightning, each VLF positioning substation transmits the pre-processed data back to the central server for comprehensive calculation. After the central server completes the calculation of the original VLF positioning coordinates, it calls the real-time positioning compensation algorithm function to quickly calculate the correction result, thereby realizing real-time correction of the VLF positioning result.

[0049] In some embodiments, the weight matrix is ​​not fixed. As optical observations continue, new lightning strike event data occurring at the observed point can be continuously acquired, as in step S2, and the corresponding new error vector can be calculated and expanded into the training sample set. Whenever the number of expanded new samples reaches a preset threshold, or the time since the last weight matrix update reaches a preset period, the updated training sample set is used to resolve and replace the current weight matrix. The weight matrix can then be updated more reliably, enabling the VLF positioning network to adapt to error drift caused by seasonal changes, equipment aging, etc., and possessing dynamic adaptive capabilities.

[0050] To verify the effectiveness of the method proposed in this invention, the following experiments are also provided in this embodiment: Based on the existing VLF positioning network (the laboratory has the necessary facilities), two high-speed optical cameras were scientifically installed at the observation point. The datasets collected in the experiment were the original positioning dataset collected by the VLF positioning network and the real dataset collected by the completed high-speed optical cameras. Based on the former two datasets, one-to-one pairing operations were performed through time windows to output an error vector dataset. Then, a weight matrix containing the error characteristics of the positioning network was obtained by training with this dataset. Finally, a positioning compensation algorithm was designed using this weight matrix, and the algorithm was burned to the server to give the final positioning correction result.

[0051] The deployment information of each site in the positioning network is shown in Table 1. Table 1: Deployment Information of Each Site in the Location Network

[0052] After incorporating the compensation algorithm, the experimental results of a certain lightning strike event are as follows: Figure 5 As shown, the lightning positioning error correction method of the present invention significantly improves positioning accuracy.

[0053] Example 2 This invention provides a lightning location error correction method that combines optical and VLF data. The difference between this method and the correction method in Embodiment 1 lies in the specific process of the optical device recording the time of the lightning strike in step S2. This embodiment uses a dual determination to determine whether a lightning event has occurred: First, a brightness change threshold is set within a fixed time t. If the brightness change of the high-speed camera exceeds the threshold within this time period, it is considered that a lightning event has initially occurred. Second, based on the former, a trained lightweight neural network (convolutional neural network) is used to further determine whether a lightning event has occurred.

[0054] First, if no lightning event occurs in the observed area, the brightness change of consecutive image frames is slow. However, if a lightning event occurs, the brightness change of consecutive image frames within a fixed time period will be significant. Therefore, let the fixed time period be... (Given the instantaneous nature of lightning strikes, this time should be on the order of microseconds), fixed time The change within is Then, simply set the brightness change threshold for consecutive image frames (set to...). ),like If so, it is preliminarily concluded that the observed point was struck by lightning, and the next step of judgment is initiated. If If so, it is assumed that the observed point was not struck by lightning. Specifically... The value needs to be determined by the sensitivity of the corresponding high-speed camera to changes in brightness.

[0055] when At this point, the high-speed camera's determination of the occurrence of a lightning event moves to the next step: lightweight neural network determination.

[0056] The aforementioned neural network is a pre-trained feature model. Its input is an image patch of the observed point at the current time, and its output is the confidence probability of a lightning event for that image patch. When the confidence probability is greater than 0.9, it is considered that the current lightning event has actually occurred, and the high-speed camera will record the time of the event and store it in the log. The lightning strike point is the three-dimensional spatial coordinates of the observed point.

[0057] The model training process is as follows: ① Collect positive sample set Image frames that have been manually confirmed as being struck by lightning from historical observation data are extracted and integrated into a positive sample set.

[0058] ② Negative sample set Collect various interfering images, such as lightning that did not hit the observation point nearby, searchlight interference, vehicle lights, and mirror reflections, and integrate them into a negative sample set.

[0059] ③Build network architecture Construct a lightweight neural network whose structure includes an input layer, several convolutional layers, pooling layers, fully connected layers, and an output layer.

[0060] ④ Feature extraction After integrating the positive and negative sample sets, use them as input to begin model training, input image patches. Hierarchical features are extracted step by step through multiple convolutional and pooling layers: ; in The feature vector contains key visual features indicating whether the image contains a lightning event. This represents a feature extraction network consisting of convolutional layers, pooling layers, and activation functions.

[0061] ⑤ Confidence Probability Output eigenvectors The input to the fully connected layer is mapped to the confidence probability of a lightning strike event through the Sigmoid activation function: ; FC stands for Fully Connected Layer, which plays a role in mapping from features to the decision space through a linear transformation from the input node to the output node; The sigmoid activation function is defined as follows: ; Its function is to map any real number to the interval (0,1), giving it a probabilistic meaning.

[0062] Example 3 This embodiment provides a lightning positioning error correction system that combines optical and VLF data, used to implement the steps of the lightning positioning error correction method that combines optical and VLF data as described in Embodiment 1 or Embodiment 2. The system includes: The data acquisition module includes optical equipment and a VLF positioning network deployed around the observed point; the optical equipment is used to acquire optical image data of the observed point and record the time of the lightning strike when it is determined that the observed point has been struck by lightning; the VLF positioning network is used to acquire the VLF three-dimensional positioning coordinates and detection time of the lightning event within its own detection range. The error calculation and sample construction module is used to match the same lightning strike event through a time window and calculate the error vector of the VLF positioning coordinates relative to the true coordinates of the observed point in order to construct a training sample set. The matrix solving and modeling module is used to solve for the weight matrix representing the distribution law of system error based on the training sample set, and to establish a linear mapping model accordingly. The positioning compensation processing module, deployed on the central server, is used to call the linear mapping model to calculate the error correction vector when the VLF three-dimensional positioning coordinates of a new lightning event are obtained, thereby correcting the current VLF three-dimensional positioning coordinates.

[0063] Example 4 This embodiment provides a computer-readable storage medium storing a computer program thereon. When the program is executed by a processor, it implements the steps of the lightning positioning error correction method based on optical and VLF data coordination as described in Embodiment 1 or Embodiment 2.

[0064] The computer-readable storage medium may include flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the storage medium may be an internal storage unit of a computer device, such as the hard disk or memory of the computer device. In other embodiments, the storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc., provided on the computer device. Of course, the storage medium may include both internal storage units and external storage devices of the computer device. In this embodiment, the memory is typically used to store the operating system and various application software installed on the computer device. In addition, the memory can also be used to temporarily store various types of data that have been output or will be output.

[0065] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for correcting lightning location errors using optical and VLF data coordination, characterized in that, Includes the following steps: S1. Determine the observed point with known three-dimensional spatial coordinates, and deploy optical equipment and VLF positioning network around the observed point; S2. When the observed point is struck by lightning, the time of the lightning event recorded by the optical equipment, and the three-dimensional positioning coordinates and detection time of the VLF detected by the VLF positioning network are obtained respectively. The detection time and the occurrence time of the lightning strike are filtered and matched by a set time window. After a successful match and it is determined that the two are from the same lightning strike event, the VLF three-dimensional positioning coordinates and the known three-dimensional spatial coordinates of the observed point are calculated to obtain the corresponding error vector. S3. Collect the VLF three-dimensional positioning coordinates and corresponding error vectors of multiple historical lightning strike events, and construct a training sample set accordingly; The weight matrix is ​​obtained by performing calculations based on the training sample set to characterize the spatial distribution of errors in the VLF positioning network. Then, a linear mapping model between the VLF three-dimensional positioning coordinates and the error correction vector is established based on the weight matrix. S4. When the VLF positioning network detects a new lightning event, it obtains the current VLF three-dimensional positioning coordinates and substitutes them as input parameters into the linear mapping model for calculation to obtain an error correction vector. The error correction vector is then added to the current VLF three-dimensional positioning coordinates to obtain the final corrected positioning coordinates.

2. The lightning positioning error correction method based on optical and VLF data coordination according to claim 1, characterized in that, In step S3, the specific calculation process of the weight matrix and the linear mapping model includes: Based on the N training samples in the training sample set, construct the original positioning matrix containing the VLF three-dimensional positioning coordinates. and the target matrix containing the corresponding error vector ; Using mean squared error as the loss function, the expression is: In the formula, Represents the weight matrix The loss function; Let Frobenius norm be the matrix. Apply the loss function to the weight matrix The gradient matrix is ​​obtained by taking the derivative. Setting the gradient matrix to zero, the optimal weight matrix is ​​then obtained. In the formula, T represents the matrix transpose operation; , Represents a 4x3 real number matrix; Based on the weight matrix obtained by the solution Establish a linear mapping model: In the formula, This is the error correction vector; VLF three-dimensional positioning coordinates Augmented feature vectors.

3. The lightning positioning error correction method based on optical and VLF data coordination according to claim 1, characterized in that, Following step S4, the following steps are also included: Following the method in step S2, new lightning strike event data occurring at the observed point are continuously acquired, and the corresponding new error vector is calculated and expanded into the training sample set; When the number of new samples added reaches a preset threshold, or when the time since the last weight matrix update reaches a preset period, the updated training sample set is used to resolve and replace the current weight matrix.

4. The lightning positioning error correction method based on optical and VLF data coordination according to claim 1, characterized in that, In step S2, the specific process of the optical device recording the time of the lightning strike includes: Set a fixed time in microseconds and a threshold for brightness changes in consecutive image frames; The brightness change of consecutive image frames within the fixed time period is monitored. When the brightness change exceeds the brightness change threshold, it is determined that the observed point has been struck by lightning, and the timestamp of the frame is extracted as the time of the lightning strike event.

5. The lightning positioning error correction method based on optical and VLF data coordination according to claim 1, characterized in that, In step S2, the specific process of the optical device recording the time of the lightning strike includes: Set a fixed time in microseconds and a threshold for brightness changes in consecutive image frames; Monitor the brightness change of consecutive image frames within the fixed time period. When the brightness change exceeds the brightness change threshold, extract the current observed point image block. The observed point image patch is used as input parameter and fed into a pre-trained convolutional neural network. The convolutional neural network extracts hierarchical features step by step through a feature extraction network containing convolutional layers and pooling layers, and outputs a visual feature vector. The visual feature vector is then input into a fully connected layer and linearly transformed and mapped through a sigmoid activation function to output the confidence probability of the lightning strike event. The convolutional neural network is pre-trained using a positive sample set containing real lightning strikes on the observed point image frames and a negative sample set containing light interference or lightning image patches that are not struck. When the confidence probability of the lightning strike event is greater than the set probability threshold, it is finally determined that the observed point has actually been struck by lightning, and the timestamp of the frame is extracted as the time of the lightning strike event.

6. The lightning positioning error correction method based on optical and VLF data coordination according to claim 1, characterized in that, In step S1, the observed point is a structure with a lightning rod tip on top, and the three-dimensional spatial coordinates of the observed point are determined by pre-measuring the latitude, longitude and elevation data of the observed point; the optical equipment includes n high-speed cameras deployed within a range of 500m to 2000m from the observed point, and the azimuth angle between adjacent high-speed cameras is 360 / n°; the deployment altitude of each high-speed camera is lower than the altitude of the observed point, and the observation elevation angle is between 25° and 45°.

7. The lightning positioning error correction method based on optical and VLF data coordination according to claim 6, characterized in that, In step S1, multiple observation points that meet the conditions and have known three-dimensional spatial reference coordinates are selected within the same geographical area; For each observed point, optical devices are deployed collaboratively to obtain the corresponding error vector. Then, the error vector data obtained from multiple observed points are summarized to jointly construct and expand the training sample set, thereby shortening the data collection cycle required to solve the weight matrix.

8. The lightning positioning error correction method based on optical and VLF data coordination according to claim 1, characterized in that, Each detection substation in the VLF positioning network collects lightning radiation signals and performs preliminary processing on the lightning radiation signals. Each detection substation transmits the pre-processed data to the central server, which then performs comprehensive positioning calculations on the multiple data streams to obtain the VLF three-dimensional positioning coordinates.

9. A lightning location error correction system based on optical and VLF data coordination, used to implement the steps of the lightning location error correction method based on optical and VLF data coordination as described in any one of claims 1 to 8, the system comprising: The data acquisition module includes optical equipment and a VLF positioning network deployed around the observed point; Optical equipment is used to acquire optical image data of the observed point and to record the time of the lightning strike when it is determined that the observed point has been struck by lightning. The VLF positioning network is used to collect the VLF three-dimensional positioning coordinates and detection time of lightning events within its own detection range; The error calculation and sample construction module is used to match the same lightning strike event through a time window and calculate the error vector of the VLF positioning coordinates relative to the true coordinates of the observed point in order to construct a training sample set. The matrix solving and modeling module is used to solve for the weight matrix representing the distribution law of system error based on the training sample set, and to establish a linear mapping model accordingly. The positioning compensation processing module, deployed on the central server, is used to call the linear mapping model to calculate the error correction vector when the VLF three-dimensional positioning coordinates of a new lightning event are obtained, thereby correcting the current VLF three-dimensional positioning coordinates.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the lightning positioning error correction method based on optical and VLF data coordination as described in any one of claims 1 to 8.

Images