Lightning protection detection system applied to unmanned aerial vehicle

Abstract

The invention discloses a lightning protection detection system applied to unmanned aerial vehicles, and relates to the technical field of unmanned aerial vehicles, and the system comprises an unmanned aerial vehicle cooperative detection unit which comprises at least two unmanned aerial vehicles, each unmanned aerial vehicle carries a lightning protection detection sensor, and the lightning protection detection sensor is used for recording original data when a lightning stroke event occurs; and processing the original data to obtain detection data. According to the lightning protection detection system applied to the unmanned aerial vehicle, through the lightning protection detection sensor carried on the unmanned aerial vehicle, various original data when a lightning stroke event occurs can be recorded at the same time; compared with a single data source, the multi-source data is integrated for analysis, so that the characteristics and intensity of lightning activities can be more accurately judged, the conditions of misjudgment and missed judgment are reduced, and a more reliable basis is provided for a lightning protection decision; the data verification and comparative analysis module carries out a series of accurate processing on original data collected by each unmanned aerial vehicle; and the interference of abnormal data on subsequent analysis is avoided.

Description

Technical Field

[0001] The invention relates to the technical field of unmanned aerial vehicles, and in particular to a lightning protection detection system applied to unmanned aerial vehicles. Background Art

[0002] In today's era of rapid technological development, drones have been widely used in many fields due to their flexibility, maneuverability and ease of operation, such as meteorological research, power inspection, forestry monitoring, geological exploration, etc. These application scenarios often expose drones to complex and changeable natural environments, among which lightning activities pose a serious threat to the safety of drones. Therefore, the importance of drone lightning protection detection technology has become increasingly prominent. However, in complex environments, drone lightning protection detection faces many challenges. In the existing technology, there are still many defects in the UAV lightning protection detection system; most systems tend to rely on a single data source for detection, and only rely on the data collected by a certain type of sensor carried by the UAV to judge the lightning activity; this single data source approach has great risks, because any sensor may fail or be interfered by specific environmental factors and produce erroneous data; once the single data source has problems, the entire detection system will lose its reliable detection basis and will not be able to accurately judge lightning activity, which will seriously affect the scientificity and timeliness of lightning protection decisions. Summary of the invention

[0003] In order to solve the above technical problems, the present invention is implemented through the following technical solutions: A lightning protection detection system applied to unmanned aerial vehicles, comprising a unmanned aerial vehicle cooperative detection unit: comprising at least two unmanned aerial vehicles, each of which is equipped with a lightning protection detection sensor, the lightning protection detection sensor is used to record the original data when a lightning strike event occurs, and process the original data to obtain detection data; The detection data includes lightning parameters, lightning occurrence time, and lightning duration; among which, lightning parameters include the change in electric field intensity generated by lightning, the change in magnetic field intensity caused by lightning, and lightning pulse signals; Data verification and comparative analysis module: used to receive the detection data sent by each drone, and to conduct mutual verification and comparative analysis on each detection data, to identify and eliminate data anomalies caused by environmental factors; Data sharing and interaction module, which realizes real-time data transmission and integration between drones through wireless communication technology; The ground control center is used to receive the detection data transmitted by the drone and generate a real-time map of lightning activity based on the detection data.

[0004] Preferably, the data sharing and interaction module supports Wi-Fi, Bluetooth, ZigBee, LTE, LoRaWAN, XBee, MAVLink and OcuSync series communication protocols to meet the data transmission requirements between different drones and ground control centers.

[0005] Preferably, the data verification and comparative analysis module is used to perform preliminary processing on the raw data collected by the lightning protection detection sensor on each UAV, and the process is as follows: For the raw data collected by the lightning protection detection sensor on each drone, data cleaning is used to remove data points with obvious errors or missing data, including outliers caused by sensor failure or communication interference; moving average filtering and Kalman filtering are used to denoise and smooth the raw data to reduce the impact of environmental noise on the data, so that the data can better reflect the real characteristics of lightning signals; then time-space synchronization operations are performed: since different drones collect data at different locations and times, the data needs to be processed in time and space synchronization; the timestamps and position coordinates obtained by the GPS module on the drone are used to align the detection data of each drone in time and space; including: for time alignment, the time synchronization algorithm is used to ensure that the time reference of the data of each drone is consistent, so as to compare the data at the same time; for space alignment, the detection data is mapped to the standard geographic coordinate system according to the position coordinates of the drone, so as to facilitate the analysis of data correlation at different locations.

[0006] Preferably, the process of ensuring the consistency of the time reference of each drone data through the time synchronization algorithm is as follows: Install a high-precision crystal oscillator or clock chip on the circuit board of the drone, and synchronize the clock source through hardware wiring; in the production or initialization stage, calibrate the crystal oscillator or clock chip to keep the frequency of the crystal oscillator or clock chip consistent with the initial time; Or a clock distribution module is used to randomly determine a drone, install a master clock generator inside it, and then install clock receiving modules inside the remaining drones to distribute a high-precision master clock signal to the subsystems of each drone: a high-precision clock pulse is generated by the master clock generator to form a master clock signal, and then the master clock signal is transmitted to the clock receiving modules of each drone through a coaxial cable or an optical fiber. After the clock receiving module receives the master clock signal, it removes high-frequency noise through a low-pass filter, and amplifies the weak signal to the corresponding level of the correct recognition range of the subsequent digital circuit through an amplifier to obtain an adaptation signal, wherein the adaptation signal includes a start signal and a frequency signal; then the adaptation signal is input into the local clock circuit of the drone, and the counter and register in the local clock circuit are initialized according to the received start signal, and their own time counting starts from the start time; at the same time, the voltage-controlled oscillator VCO in the clock circuit is locked according to the received frequency signal, and its own oscillation frequency is adjusted to be consistent with the frequency signal of the master clock generator; ensuring that the clock circuit of each drone receives the same start signal and frequency signal.

[0007] Preferably, for spatial alignment, the detection data is mapped to a standard geographic coordinate system according to the position coordinates of the drone, and the process is as follows: The WGS-84 coordinate system is selected as the standard geographic coordinate system. Each drone obtains its own real-time position coordinates through the GPS module it carries. The real-time position coordinates include longitude, latitude and altitude information. The GPS module is used to receive signals from satellites and determine the three-dimensional position of the drone on the earth through calculation. According to the selected standard geographic coordinate system and the real-time position coordinates of the UAV, a mapping relationship between the detection data and the geographic spatial position is established; this involves associating the data collected by the sensor with the corresponding longitude, latitude and altitude information; adding geographic coordinate tags to the electric field strength, magnetic field strength and lightning pulse signal respectively, so as to clarify the spatial position corresponding to the data; including binding the electric field strength data detected at a specific longitude, latitude and altitude at a certain moment with the location information; converting and projecting the real-time position coordinates obtained by the UAV to adapt to specific map projection methods or data processing requirements; including: for real-time position coordinate conversion, converting the longitude and latitude coordinates in the WGS-84 coordinate system into coordinates in the plane rectangular coordinate system through Gauss projection, and for real-time position coordinate projection, projecting the three-dimensional earth surface data onto a two-dimensional plane: that is, based on the display of lightning activity distribution on a plane map, converting the longitude and latitude coordinates in the real-time position coordinates into plane rectangular coordinates for display and analysis on the map; The test data mapped to the standard geographic coordinate system is stored and a data management mechanism is established to facilitate subsequent data query, retrieval and analysis; including: using a geographic information database to store data, storing the test data and geographic coordinate information as record fields; at the same time, establishing indexes and data structures to improve data access and processing efficiency, so as to quickly obtain corresponding test data based on geographic location.

[0008] Preferably, the process of mutual verification and comparative analysis of each test data to identify and eliminate data anomalies caused by environmental factors is as follows: Feature extraction: Extract characteristic parameters related to lightning from the detection data after data preprocessing and time-space synchronization; including: for electric field strength data, extract the peak value, change rate, and duration characteristics of the electric field strength; for magnetic field strength data, extract the amplitude and frequency characteristics of the magnetic field strength; for lightning pulse signals, extract the rise time, fall time, pulse width, and peak voltage characteristics of the pulse; these characteristic parameters will serve as the basis for subsequent comparative analysis and can more effectively reflect the characteristics and change laws of lightning; Correlation analysis: Calculate the correlation coefficient between different drone detection data to determine the similarity between the data; including: for electric field strength data, compare whether the change trend of electric field strength detected by drones at different locations is consistent; for magnetic field strength data, analyze whether the fluctuation of magnetic field strength is correlated; for lightning pulse signals, compare the similarity of pulse characteristic parameters to obtain correlation analysis results; among them, if the correlation coefficient is high, it means that there is good consistency between the detection data, reflecting the actual lightning activity; if the correlation coefficient is low, it means that there is data anomaly or it is affected by local environmental factors; Data consistency test: Based on statistical principles, the detection data of multiple drones are tested for consistency; including: using hypothesis testing methods, assuming that multiple groups of data come from the same population and reflect the same lightning activity, calculating test statistics and comparing them with corresponding critical values ​​to obtain consistency test results; statistics include chi-square test statistics and t-test statistics; if the test statistic is less than the critical value, the null hypothesis is accepted and the data is considered to be consistent; if the test statistic is greater than the critical value, the null hypothesis is rejected, indicating that there are significant differences in the data and abnormal data exists; Abnormal data identification: The detection results are formed based on the correlation analysis results and the consistency test results. Based on the detection results, the detection data with low correlation with the drone data or that fails the consistency test are marked as abnormal data. Identify abnormal data; at the same time, combine the flight trajectory of the drone and the surrounding environment information to analyze the causes of abnormal data; Data correction and verification: Correct or supplement the identified abnormal data; including: if the abnormal data is a slight deviation caused by local environmental interference, use interpolation method and average method to correct it according to the trend and characteristics of normal data; after correction, perform correlation analysis and data consistency check again to verify whether the corrected data is consistent with the remaining data to ensure the accuracy and reliability of the data; then generate an analysis report based on the results of mutual verification and comparative analysis, which includes data preprocessing process, correlation analysis results, consistency test results, abnormal data identification and processing; it also includes a comprehensive assessment of lightning activity, such as the intensity, location, and activity range of lightning, as well as evaluation and suggestions on the performance of the detection system, to provide a reference for subsequent lightning protection decisions and drone flight safety.

[0009] Preferably, the ground control center generates a real-time map of lightning activity based on the detection data: The ground control center continuously receives the detection data transmitted by each drone through wireless communication links. The detection data includes electric field strength, magnetic field strength and lightning pulse signals with geographic coordinate tags. Then the received detection data is integrated in real time, sorted and stored according to the time sequence and geographic location information of the data to ensure the integrity and consistency of the data, providing an accurate data basis for subsequent map generation. Build a GIS-based map processing platform that has the functions of geospatial data processing, visualization and analysis; obtain basic geographic information data from the geographic information database, and import the basic geographic information data into the map processing platform to provide geographic background information for the overlay and display of lightning activity data; among them, the basic geographic information data includes topographic and geomorphic layers, administrative division layers, and building distribution layers; convert the received detection data into GeoJSON format to facilitate GIS platform recognition and processing, set the display properties of the data according to the requirements of the GIS platform to ensure that the data can be accurately displayed and analyzed on the map, and the display properties include symbol systems and color grading to intuitively represent lightning activities of different intensities; on the GIS platform, draw the integrated lightning detection data on the map according to its geographic coordinate information; A real-time update mechanism is established to ensure that the map can promptly reflect the latest lightning activity. As the drone continuously transmits new detection data, the ground control center promptly integrates the new detection data into the map and updates the lightning activity display information at the corresponding location. The map is updated every three seconds through the set update frequency to ensure that the lightning activity status displayed on the map is basically synchronized with the actual situation.

[0010] Preferably, the ground control center also includes: in the process of generating real-time maps, further analyzing the lightning activity data in combination with the data analysis algorithm to identify the hot spots of lightning activity and predict the movement path; wherein the process of identifying the hot spots is as follows: first, selecting the lightning detection data within the time window of the past 5-10 minutes, including the electric field strength, magnetic field strength and lightning pulse signal; using the clustering algorithm to perform cluster analysis on the detection data, classifying the data points with similar geographical positions and similar lightning activity characteristics into one category, and obtaining the clustering results; according to the clustering results, counting the number of data points in each cluster or calculating the comprehensive index of lightning activity intensity within the cluster; the comprehensive index of lightning activity intensity includes the sum of electric field strength and the total number of lightning pulses; determining the area where the cluster with a large number of data points or a high comprehensive index is located as the hot spot of lightning activity; highlighting it with purple or lightning symbol on the map for intuitive display.

[0011] Preferably, the process of moving path prediction is as follows: Collect continuous lightning detection data sequences within the past 15-30 minutes to ensure that the data contains sufficient time and space information; extract lightning-related mobile feature data from the detection data sequence, the mobile feature data includes the longitude and latitude position coordinates of the lightning activity center at different times, the changing trend of the electric field strength and magnetic field strength, and the propagation direction characteristics of the lightning pulse signal; select a prediction model combining a convolutional neural network and a recurrent neural network based on the extracted mobile feature data; use historical lightning data to train the selected model and adjust the model parameters to accurately capture the movement patterns of lightning activities; use the latest mobile feature data of lightning activities acquired in real time as input data The prediction model outputs the prediction results according to the input data and calculates the future movement direction and speed of lightning activities. The predicted positions of different time steps are connected on the real-time map in the form of lines or arrows to form a predicted movement path. The predicted movement path is displayed on the real-time map to intuitively present the possible movement direction of lightning activities. As new detection data is continuously transmitted, the above steps are repeated to update the prediction results of the movement path in real time to improve the accuracy and timeliness of the prediction. At the same time, the prediction results are evaluated and adjusted in combination with the actual situation. When sudden meteorological changes or terrain influences occur, the model prediction results are corrected in time.

[0012] Preferably, during the identification of the hotspot area by the ground control center, the steps of using a clustering algorithm to perform cluster analysis on the detection data are as follows: S1. Data preparation: Obtain the selected time window from the geographic information database: lightning detection data within the past 5-10 minutes, including the longitude and latitude geographic coordinates of each data point and lightning characteristic parameters, which include electric field intensity, magnetic field intensity and lightning pulse signal; standardize the detection data so that different characteristic parameters have the same dimension and importance, and the detection data has zero mean and unit variance; S2. Determine the clustering algorithm and parameters: According to the data characteristics and analysis requirements, select the K-Means algorithm or the DBSCAN algorithm; if the K-Means algorithm is selected, the number of clusters K is pre-determined by the elbow method, and the clustering error square sum SSE curve under different K values ​​is plotted, and the K value at the inflection point of the curve is selected as the initial number of clusters; at the same time, set the algorithm's iteration termination conditions: the maximum number of iterations or the convergence threshold, where the maximum number of iterations is set to 100-300 times; convergence threshold: when the moving distance of the cluster center between two iterations is less than the threshold, the algorithm is considered to have converged; if the DBSCAN algorithm is selected, determine the key parameters: neighborhood radius Epsilon and minimum number of points MinPts; where the neighborhood radius defines the neighborhood range of the data point, which is determined by preliminary analysis of the data distribution or empirical value; the minimum number of points specifies the minimum number of data points required to form a cluster, which is set according to the density characteristics of the data and the expected clustering quality; S3. Initialize cluster centers, only applicable to K-Means algorithm: For K-Means algorithm, randomly select K data points as initial cluster centers; the selection of these initial cluster centers will affect the final clustering results, so you can randomly initialize and run the algorithm multiple times to select the optimal clustering result or the result that minimizes SSE; S4. Calculate the distance matrix: For the K-Means algorithm, calculate the distance between each data point and the cluster center; for the DBSCAN algorithm, calculate the distance between each data point; for geographic coordinate data, since the earth is an approximate sphere, use the Haversine formula as a distance metric to calculate the spherical distance between two points; construct a distance matrix in which each element represents the distance between two data points; for the K-Means algorithm, calculate the distance between each data point and the K cluster centers; for the DBSCAN algorithm, calculate the distance between each data point and its neighboring data points; S5. Assign data points to clusters: According to the distance matrix, for the K-Means algorithm, each data point is assigned to the cluster to which the nearest cluster center belongs, and for the DBSCAN algorithm, each data point is assigned to the cluster that meets the density condition; in the K-Means algorithm, data points are assigned to the cluster to which the nearest cluster center belongs; in the DBSCAN algorithm, if the number of neighboring data points of a data point within its neighborhood radius Epsilon is greater than or equal to the minimum number of points MinPts, then the data point is regarded as a core point and forms a new cluster, and the data points in its neighborhood are also assigned to the cluster; if a data point is not a core point but is in the neighborhood of a core point, then the data point is assigned to the cluster to which the core point belongs, and is called a boundary point; if a data point is neither a core point nor in the neighborhood of any core point, then the data point is regarded as a noise point and does not participate in clustering; S6. Update cluster centers, only applicable to the K-Means algorithm: In the K-Means algorithm, recalculate the new cluster center of each cluster. The calculation method of the new cluster center is to take the average value of all data points in the cluster in each feature dimension; S7. Iterative optimization: repeat steps S4-S6 until the termination condition of the clustering algorithm is met; including: for the K-Means algorithm, the iteration is stopped when the maximum number of iterations is reached or the moving distance of the cluster center is less than the convergence threshold; for the DBSCAN algorithm, the iteration is stopped when all data points are assigned to clusters and no noise points are included or no new clusters are formed; S8. Evaluate clustering results: calculate clustering evaluation indicators to evaluate the quality of clustering results; clustering evaluation indicators include: Silhouette Coefficient, Calinski-Harabasz index; judge whether the clustering results are reasonable based on the evaluation indicators. If the evaluation indicators are not ideal, adjust the parameters of the clustering algorithm, such as K value, Epsilon and MinPts, or change the clustering algorithm, and re-perform clustering analysis until a satisfactory clustering result is obtained.

[0013] The present invention provides a lightning protection detection system applied to unmanned aerial vehicles, which has the following beneficial effects: The lightning protection detection system applied to UAVs can simultaneously record a variety of raw data when a lightning strike occurs, including changes in the electric field intensity, magnetic field intensity and lightning pulse signals generated by lightning, through the lightning protection detection sensors carried on the UAVs, and process these data to obtain comprehensive detection data; comprehensive multi-source data is analyzed, and compared with a single data source, it can more accurately judge the characteristics and intensity of lightning activities, reduce misjudgments and missed judgments, and provide a more reliable basis for lightning protection decisions; the data verification and comparative analysis module performs a series of precise processing on the raw data collected by each UAV; avoids the interference of abnormal data on subsequent analysis; moving average filtering and Kalman filtering denoising and smoothing processing effectively reduce the impact of environmental noise on the data, so that the data can better reflect the real lightning signal characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flowchart of a lightning protection detection system applied to an unmanned aerial vehicle according to the present invention; Figure 2 It is a flowchart of the data verification and comparative analysis module of the present invention; Figure 3 This is a flowchart of the preliminary processing of raw data in the present invention. DETAILED DESCRIPTION

[0015] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0016] like Figures 1 to 3 As shown, the present invention provides a technical solution: a lightning protection detection system applied to a UAV, comprising at least two UAVs, each of which is equipped with a lightning protection detection sensor, the lightning protection detection sensor is used to record the original data when a lightning strike event occurs, and process the original data to obtain detection data; The detection data includes lightning parameters, lightning occurrence time, and lightning duration; among which, lightning parameters include the change in electric field intensity generated by lightning, the change in magnetic field intensity caused by lightning, and lightning pulse signals; Data verification and comparative analysis module: used to receive the detection data sent by each drone, and to conduct mutual verification and comparative analysis on each detection data, to identify and eliminate data anomalies caused by environmental factors; Data sharing and interaction module, which realizes real-time data transmission and integration between drones through wireless communication technology; The ground control center is used to receive the detection data transmitted by the drone and generate a real-time map of lightning activity based on the detection data.

[0017] The data sharing and interaction module supports Wi-Fi, Bluetooth, ZigBee, LTE, LoRaWAN, XBee, MAVLink and OcuSync series communication protocols to meet the data transmission needs between different drones and ground control centers; it can adapt to the data transmission needs between different types of drones and ground control centers to ensure the efficiency and stability of data transmission; this wide range of communication protocol compatibility enables the system to work smoothly in different application scenarios and device configurations, enhancing the versatility and scalability of the system.

[0018] The data verification and comparative analysis module is used to perform preliminary processing on the raw data collected by the lightning protection detection sensor on each drone. The process is as follows: For the raw data collected by the lightning protection detection sensor on each drone, data cleaning is used to remove obviously erroneous data points, including outliers caused by sensor failure; moving average filtering and Kalman filtering are used to denoise and smooth the raw data to reduce the impact of environmental noise on the data, so that the data can better reflect the real characteristics of lightning signals; then time-space synchronization operations are performed: since different drones collect data at different locations and times, the data needs to be processed in time and space synchronization; the timestamps and position coordinates obtained by the GPS module on the drone are used to align the detection data of each drone in time and space; including: for time alignment, the time synchronization algorithm is used to ensure that the time reference of the data of each drone is consistent, so as to compare the data at the same time; for space alignment, the detection data is mapped to the standard geographic coordinate system according to the position coordinates of the drone, so as to facilitate the analysis of data correlation at different locations.

[0019] The process of ensuring the consistency of the time base of each drone data through the time synchronization algorithm is as follows: A high-precision crystal oscillator is installed on the circuit board of the drone, and the clock source is synchronized through hardware wiring; during the production stage, the crystal oscillator is calibrated to keep the frequency of the crystal oscillator consistent with the initial time; A clock distribution module is used to randomly determine a drone, install a master clock generator inside it, and then install clock receiving modules inside the remaining drones to distribute a high-precision master clock signal to the subsystems of each drone: a high-precision clock pulse is generated by the master clock generator to form a master clock signal, and then the master clock signal is transmitted to the clock receiving modules of each drone through a coaxial cable. After the clock receiving module receives the master clock signal, it removes high-frequency noise through a low-pass filter, and amplifies the weak signal to the corresponding level of the correct recognition range of the subsequent digital circuit through an amplifier to obtain an adaptation signal, wherein the adaptation signal includes a start signal and a frequency signal; then the adaptation signal is input into the local clock circuit of the drone, and the counter and register in the local clock circuit are initialized according to the received start signal, and their own time counting starts from the start time; at the same time, the voltage-controlled oscillator VCO in the clock circuit is locked according to the received frequency signal, and its own oscillation frequency is adjusted to be consistent with the frequency signal of the master clock generator; ensure that the clock circuit of each drone receives the same start signal and frequency signal; It should be further explained that, in the specific implementation process, the master clock generator uses a miniaturized chip-level atomic clock based on the atomic clock principle as the master clock source, which can provide accurate frequency signals for a long time; Clock receiving module: A dedicated clock receiving module is installed on the circuit board of each drone. The clock receiving module has high-precision signal receiving and processing capabilities, and can accurately capture and process signals from the main clock generator; including the use of high-speed differential signal receivers, which can effectively reduce interference during signal transmission and ensure signal integrity; Setting of the master clock generator: When the system starts, the master clock generator is initialized and set up first. The master clock generator is calibrated with a precisely calibrated laboratory atomic clock signal so that its output frequency signal is accurate to the specified frequency value: 10MHz, 100MHz, and the start time is determined; After receiving the main clock signal, the clock receiving module performs signal preprocessing, including removing high-frequency noise through a low-pass filter and amplifying the weak signal to a suitable level through an amplifier to ensure that the subsequent circuit can accurately process the signal; because if the level is too high, it may damage the subsequent circuit components; if the level is too low, the subsequent circuit may not recognize it as a valid signal; Local clock circuit synchronization: After the clock receiving module processes the signal, it inputs the signal into the local clock circuit of the drone; the counters and registers in the local clock circuit will be initialized according to the received start signal, and their own time counting will start from the start time; at the same time, the voltage-controlled oscillator VCO in the clock circuit will be locked according to the received frequency signal, and adjust its own oscillation frequency to make it consistent with the frequency signal of the main clock generator; including: locking the output frequency of the local VCO with the received main clock frequency signal through the phase-locked loop PLL circuit for phase and frequency, ensuring that the local clock circuit can stably output a signal with the same frequency as the main clock.

[0020] For spatial alignment, the detection data is mapped to a standard geographic coordinate system based on the position coordinates of the drone. The process is as follows: The WGS-84 coordinate system is selected as the standard geographic coordinate system. Each drone obtains its own real-time position coordinates through the GPS module it carries. The real-time position coordinates include longitude, latitude and altitude information. The GPS module is used to receive signals from satellites and determine the three-dimensional position of the drone on the earth through calculation. According to the selected standard geographic coordinate system and the real-time position coordinates of the UAV, a mapping relationship between the detection data and the geographic spatial position is established; this involves associating the data collected by the sensor with the corresponding longitude, latitude and altitude information; adding geographic coordinate tags to the electric field strength, magnetic field strength and lightning pulse signal respectively, so as to clarify the spatial position corresponding to the data; including binding the electric field strength data detected at a specific longitude, latitude and altitude at a certain moment with the location information; converting and projecting the real-time position coordinates obtained by the UAV; including: for real-time position coordinate conversion, converting the longitude and latitude coordinates in the WGS-84 coordinate system into coordinates in the plane rectangular coordinate system through Gauss projection, and for real-time position coordinate projection, projecting the three-dimensional earth surface data onto a two-dimensional plane: that is, based on the display of lightning activity distribution on a plane map, converting the longitude and latitude coordinates in the real-time position coordinates into plane rectangular coordinates for display and analysis on the map; The test data mapped to the standard geographic coordinate system is stored and a data management mechanism is established to facilitate subsequent data query, retrieval and analysis; including: using a geographic information database to store data, storing the test data and geographic coordinate information as record fields; at the same time, establishing indexes and data structures to improve data access and processing efficiency, so as to quickly obtain corresponding test data based on geographic location.

[0021] The process of mutual verification and comparative analysis of each test data and identification and elimination of data anomalies caused by environmental factors is as follows: Feature extraction: Extract characteristic parameters related to lightning from the detection data after data preprocessing and time-space synchronization; including: for electric field strength data, extract the peak value, change rate, and duration characteristics of the electric field strength; for magnetic field strength data, extract the amplitude and frequency characteristics of the magnetic field strength; for lightning pulse signals, extract the rise time, fall time, pulse width, and peak voltage characteristics of the pulse; these characteristic parameters will serve as the basis for subsequent comparative analysis and can more effectively reflect the characteristics and change laws of lightning; Correlation analysis: Calculate the correlation coefficient between different drone detection data to determine the similarity between the data; including: for electric field strength data, compare whether the change trend of electric field strength detected by drones at different locations is consistent; for magnetic field strength data, analyze whether the fluctuation of magnetic field strength is correlated; for lightning pulse signals, compare the similarity of pulse characteristic parameters to obtain correlation analysis results; among them, if the correlation coefficient is high, it means that there is good consistency between the detection data, reflecting the actual lightning activity; if the correlation coefficient is low, it means that there is data anomaly; Data consistency test: Based on statistical principles, the detection data of multiple drones are tested for consistency; including: using hypothesis testing methods, assuming that multiple groups of data come from the same population and reflect the same lightning activity, calculating test statistics and comparing them with corresponding critical values ​​to obtain consistency test results; statistics include chi-square test statistics and t-test statistics; if the test statistic is less than the critical value, the null hypothesis is accepted and the data is considered to be consistent; if the test statistic is greater than the critical value, the null hypothesis is rejected, indicating that there are significant differences in the data and abnormal data exists; Abnormal data identification: The detection results are formed based on the correlation analysis results and the consistency test results. Based on the detection results, the detection data with low correlation with the drone data and failing to pass are marked as abnormal data. Identify abnormal data; at the same time, combine the flight trajectory of the drone and the surrounding environment information to analyze the causes of abnormal data, such as the drone approaching large metal objects causing magnetic field interference and local meteorological anomalies affecting electric field measurement. The surrounding environment information includes terrain, buildings, and meteorological conditions; Data correction and verification: Correct the identified abnormal data; including: if the abnormal data is a slight deviation caused by local environmental interference, then use interpolation and average methods to correct it according to the trend and characteristics of normal data; after correction, perform correlation analysis and data consistency check again to verify whether the corrected data is consistent with the remaining data to ensure the accuracy and reliability of the data; then generate an analysis report based on the results of mutual verification and comparative analysis. It should be further explained that in the specific implementation process, the analysis report includes data preprocessing, correlation analysis results, consistency test results, abnormal data identification and processing; it also includes a comprehensive assessment of lightning activity, such as lightning intensity, location, activity range information, and evaluation and suggestions on the performance of the detection system, which is used to provide a reference for subsequent lightning protection decisions and drone flight safety.

[0022] The process by which the ground control center generates a real-time map of lightning activity based on detection data: The ground control center continuously receives the detection data transmitted by each drone through wireless communication links. The detection data includes electric field strength, magnetic field strength and lightning pulse signals with geographic coordinate tags. Then the received detection data is integrated in real time, sorted and stored according to the time sequence and geographic location information of the data to ensure the integrity and consistency of the data, providing an accurate data basis for subsequent map generation. Build a GIS-based map processing platform that has the functions of geospatial data processing, visualization and analysis; obtain basic geographic information data from the geographic information database and import the basic geographic information data into the map processing platform to provide geographic background information for the overlay and display of lightning activity data; among them, the basic geographic information data includes topographic layer, administrative division layer, and building distribution layer; convert the received detection data into GeoJSON format to facilitate GIS platform recognition and processing; set the display attributes of the data according to the requirements of the GIS platform to ensure that the data can be accurately displayed and analyzed on the map. The display attributes include symbol system and color grading so that Intuitively represent lightning activities of different intensities; on the GIS platform, the integrated lightning detection data is plotted on the map according to its geographic coordinate information; it should be further explained that in the specific implementation process, the electric field strength data is visualized using different colors according to its numerical value, including: red represents high-intensity electric field areas, blue represents low-intensity electric field areas, and the distribution of colors shows the changing trend of electric field strength in geographic space; similarly, the magnetic field strength and lightning pulse signal data are also visualized in the same way, including using symbols of different shapes and sizes to represent the intensity and frequency of lightning pulse signals, so that operators can intuitively observe the characteristics of lightning activities at different locations; A real-time update mechanism is established to ensure that the map can promptly reflect the latest lightning activity. As the drone continuously transmits new detection data, the ground control center promptly integrates the new detection data into the map and updates the lightning activity display information at the corresponding location. The map is updated every three seconds through the set update frequency to ensure that the lightning activity status displayed on the map is basically synchronized with the actual situation.

[0023] Based on the detection data transmitted by the drone, the ground control center can generate a map of lightning activity in real time; by integrating the electric field strength, magnetic field strength and lightning pulse signal data with geographic coordinate tags into the map, the location, intensity and distribution range of lightning activities can be presented in an intuitive and visual way; operators can understand the lightning dynamics in the detection area in real time, discover potential danger areas in time, and provide timely warning information for related activities, which helps to take lightning protection measures in advance and ensure the safety of personnel and equipment.

[0024] The ground control center also includes: in the process of generating real-time maps, combining data analysis algorithms to further analyze lightning activity data, identify hot spots of lightning activity, and predict movement paths; the process of identifying hot spots is as follows: first, select lightning detection data within a time window of the past 7.5 minutes, including electric field strength, magnetic field strength, and lightning pulse signals; use a clustering algorithm to perform cluster analysis on the detection data, and classify data points with similar geographical locations and similar lightning activity characteristics into one category to obtain clustering results. For example, the electric field strength and lightning pulse frequency are used as clustering feature vectors to calculate the similarity between data points; based on the clustering results, count the number of data points in each cluster; determine the area where the cluster with a large number of data points is located as a hot spot for lightning activity; highlight it on the map with purple and lightning symbols for intuitive display.

[0025] The process of moving path prediction is as follows: Collect continuous lightning detection data sequences within the past 22.5 minutes to ensure that the data contains sufficient time and space information; extract lightning-related mobile feature data from the detection data sequence, the mobile feature data includes the longitude and latitude position coordinates of the lightning activity center at different times, the changing trend of the electric field strength and magnetic field strength, and the propagation direction characteristics of the lightning pulse signal; select a prediction model combining a convolutional neural network and a recurrent neural network based on the extracted mobile feature data; use historical lightning data to train the selected model and adjust the model parameters to accurately capture the movement law of lightning activity; use the latest mobile feature data of lightning activity obtained in real time as Input data and input it into the trained prediction model; the prediction model outputs prediction results based on the input data and calculates the future movement direction and speed of lightning activity; connects the predicted positions of different time steps on the real-time map in the form of lines to form a predicted movement path, and displays the predicted movement path on the real-time map to intuitively present the possible movement direction of lightning activity; as new detection data is continuously transmitted, repeat the above steps, update the prediction results of the movement path in real time, and improve the accuracy and timeliness of the prediction; at the same time, evaluate and adjust the prediction results based on actual conditions; when sudden meteorological changes occur, promptly correct the model prediction results.

[0026] In the process of generating real-time maps, the lightning activity data is further analyzed in combination with the data analysis algorithm, which can identify the hot spots of lightning activity and predict its movement path; the detection data within the past 7.5 minutes is clustered and analyzed by the clustering algorithm to determine the hot spots with frequent lightning activity and highlight them on the map for easy focus and prevention; at the same time, the prediction model combining convolutional neural network and recurrent neural network is used to predict the future movement direction and speed of lightning activity based on the continuous detection data within the past 22.5 minutes, plan the flight path in advance, and take corresponding protective measures, thereby improving the initiative and timeliness of responding to lightning disasters.

[0027] During the identification of hotspots by the ground control center, the steps of clustering analysis of detection data using clustering algorithms are as follows: S1. Data preparation: Obtain the selected time window from the geographic information database: lightning detection data within the past 7.5 minutes, including the longitude and latitude geographic coordinates of each data point and lightning characteristic parameters, which include electric field intensity, magnetic field intensity and lightning pulse signal; standardize the detection data so that different characteristic parameters have the same dimension and importance, and the detection data has zero mean and unit variance; S2. Determine the clustering algorithm and parameters: According to the data characteristics and analysis requirements, select the K-Means algorithm. If the K-Means algorithm is selected, the number of clusters K value is pre-determined by the elbow method, and the clustering error square sum SSE curve under different K values ​​is plotted. The K value at the inflection point of the curve is selected as the initial number of clusters; at the same time, set the algorithm's iteration termination condition: the maximum number of iterations, where the maximum number of iterations is set to 100-300 times; convergence threshold: when the moving distance of the cluster center between two iterations is less than the threshold, the algorithm is considered to have converged; S3. Initialize cluster centers: For the K-Means algorithm, randomly select K data points as initial cluster centers. The selection of these initial cluster centers will affect the final clustering results, so you can randomly initialize and run the algorithm multiple times to select the optimal clustering result. S4. Calculate the distance matrix: For the K-Means algorithm, calculate the distance between each data point and the cluster center; construct a distance matrix in which each element represents the distance between two data points; for the K-Means algorithm, calculate the distance between each data point and the K cluster centers; S5. Assign data points to clusters: According to the distance matrix, for the K-Means algorithm, each data point is assigned to the cluster to which the nearest cluster center belongs. In the K-Means algorithm, data points are assigned to the cluster to which the nearest cluster center belongs. If a data point is not a core point but is in the neighborhood of a core point, the data point is assigned to the cluster to which the core point belongs, and is called a boundary point. If a data point is neither a core point nor in the neighborhood of any core point, the data point is regarded as a noise point and does not participate in clustering. S6. Update cluster centers: In the K-Means algorithm, recalculate the new cluster center of each cluster. The calculation method of the new cluster center is to take the average value of all data points in the cluster on each feature dimension. For example, for a cluster containing n data points, the value of its new cluster center on a certain feature dimension is equal to the sum of the values ​​of the n data points in the cluster on this dimension divided by n. S7. Iterative optimization: repeating steps S4-S6 until the termination condition of the clustering algorithm is met; including: for the K-Means algorithm, stopping the iteration when the maximum number of iterations is reached; S8. Evaluate clustering results: calculate clustering evaluation indicators to evaluate the quality of clustering results; clustering evaluation indicators include: Silhouette Coefficient, Calinski-Harabasz index; it should be further explained that in the specific implementation process, the silhouette coefficient measures the ratio of the average distance of each data point to other data points in its cluster to which it belongs and the average distance to the data points in the nearest neighbor cluster, and its value range is between [-1,1]. The closer the value is to 1, the better the clustering quality; the Calinski-Harabasz index evaluates the clustering quality by calculating the ratio of the inter-cluster dispersion to the intra-cluster dispersion. The larger the ratio, the better the clustering quality; judge whether the clustering results are reasonable based on the evaluation indicators. If the evaluation indicators are not ideal, adjust the parameters of the clustering algorithm, such as K value, Epsilon and MinPts, and re-perform clustering analysis until a satisfactory clustering result is obtained.

[0028] It should be further explained that, in the specific implementation process, the lightning protection detection sensors mounted on the UAV can simultaneously record a variety of raw data when a lightning strike occurs, including changes in the electric field strength, magnetic field strength and lightning pulse signals generated by lightning, and process these data to obtain comprehensive detection data; comprehensive multi-source data analysis can more accurately judge the characteristics and intensity of lightning activities compared to a single data source, reduce misjudgment and missed judgments, and provide a more reliable basis for lightning protection decision-making; the data verification and comparative analysis module performs a series of precise processing on the raw data collected by each UAV; avoids the interference of abnormal data on subsequent analysis; moving average filtering and Kalman filtering denoising and smoothing processing effectively reduce the impact of environmental noise on the data, so that the data can better reflect the real characteristics of lightning signals; The use of at least two drones working together increases the coverage of detection and data redundancy; when the data of a drone is abnormal, the data of other drones can be used as supplement and verification, which improves the reliability of the detection results of the entire system; through data sharing and interaction modules, real-time data transmission and integration between drones can be achieved, which can quickly obtain comprehensive detection information, and ensure the stable operation of the system even in complex environments, effectively avoiding the problem of detection interruption or error caused by single point failure; The GIS-based map processing platform built by the ground control center integrates geographic spatial data processing, visualization and analysis functions; by importing basic geographic information data such as topography, administrative divisions, and building distribution, it provides rich geographic background information for the overlay and display of lightning activity data, allowing operators to more intuitively analyze the relationship between lightning activity and the geographical environment; lightning detection data is displayed on the map in different colors, symbols, and layers, which facilitates data mining and spatial analysis, helps to discover the laws and trends of lightning activities, and provides an effective tool for long-term lightning protection planning and research.

[0029] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. A lightning protection detection system for unmanned aerial vehicles, characterized in that: include: UAV collaborative detection unit: includes at least two UAVs, each of which is equipped with a lightning protection detection sensor, which is used to record the original data when the lightning strike occurs, and process the original data to obtain detection data; wherein the detection data includes lightning parameters, lightning occurrence time, and lightning duration; wherein the lightning parameters include the change in electric field intensity generated by lightning, the change in magnetic field intensity caused by lightning, and the lightning pulse signal; Data verification and comparative analysis module: used to receive the detection data sent by each drone, and to conduct mutual verification and comparative analysis on each detection data, to identify and eliminate data anomalies caused by environmental factors; Data sharing and interaction module, which realizes real-time data transmission and integration between drones through wireless communication technology; The ground control center is used to receive the detection data transmitted by the drone and generate a real-time map of lightning activity based on the detection data.

2. According to claim 1, a lightning protection detection system for unmanned aerial vehicles is characterized in that: The data sharing and interaction module supports Wi-Fi, Bluetooth, ZigBee, LTE, LoRaWAN, XBee, MAVLink and OcuSync series communication protocols.

3. The lightning protection detection system for unmanned aerial vehicles according to claim 2 is characterized in that: The data verification and comparison analysis module is used to perform preliminary processing on the raw data collected by the lightning protection detection sensor on each UAV. The process is as follows: For the raw data collected by the lightning protection detection sensor on each drone, data cleaning is used to remove data points with obvious errors or missing data points; moving average filtering and Kalman filtering are used to denoise and smooth the raw data; Then, the time-space synchronization operation is performed: the timestamp and position coordinates obtained by the GPS module on the drone are used to align the detection data of each drone in time and space; including: for time alignment, the time synchronization algorithm is used to ensure that the time base of the data of each drone is consistent; For spatial alignment, the detection data is mapped to a standard geographic coordinate system based on the position coordinates of the drone.

4. The lightning protection detection system for unmanned aerial vehicles according to claim 3 is characterized in that: The process of ensuring the consistency of the time base of each drone data through the time synchronization algorithm is as follows: Install a crystal oscillator or clock chip on the circuit board of the drone, and synchronize the clock source through hardware wiring; wherein, during the production or initialization stage, calibrate the crystal oscillator or clock chip to keep the frequency of the crystal oscillator or clock chip consistent with the initial time; Or use a clock distribution module, randomly determine a drone, install a master clock generator inside it, and then install clock receiving modules inside the remaining drones, form a master clock signal through the clock pulse generated by the master clock generator, and then transmit the master clock signal to the clock receiving modules of each drone through a coaxial cable or optical fiber. After the clock receiving module receives the master clock signal, it removes high-frequency noise through a low-pass filter, and amplifies the weak signal to the corresponding level of the correct recognition range of the subsequent digital circuit through an amplifier to obtain an adaptation signal, wherein the adaptation signal includes a start signal and a frequency signal; then the adaptation signal is input into the local clock circuit of the drone, and the counter and register in the local clock circuit are initialized according to the received start signal, and their own time counting starts from the start time; at the same time, the voltage-controlled oscillator VCO in the clock circuit is locked according to the received frequency signal, and its own oscillation frequency is adjusted to be consistent with the frequency signal of the master clock generator.

5. The lightning protection detection system for unmanned aerial vehicles according to claim 4 is characterized in that: For spatial alignment, the detection data is mapped to a standard geographic coordinate system based on the position coordinates of the drone. The process is as follows: The WGS-84 coordinate system is selected as the standard geographic coordinate system. Each drone obtains its own real-time position coordinates through the GPS module it carries. The real-time position coordinates include longitude, latitude and altitude information. According to the selected standard geographic coordinate system and the real-time position coordinates of the UAV, a mapping relationship between the detection data and the geographic space position is established; geographic coordinate tags are added to the electric field strength, magnetic field strength and lightning pulse signal respectively; the real-time position coordinates obtained by the UAV are converted and projected; including: for real-time position coordinate conversion, the longitude and latitude coordinates in the WGS-84 coordinate system are converted into coordinates in the plane rectangular coordinate system through Gauss projection; for real-time position coordinate projection, the three-dimensional earth surface data is projected onto a two-dimensional plane: that is, based on the display of lightning activity distribution on a plane map, the longitude and latitude coordinates in the real-time position coordinates are converted into plane rectangular coordinates; The detection data mapped to the standard geographic coordinate system is stored and a data management mechanism is established; including: using a geographic information database to store data, storing the detection data and geographic coordinate information as record fields; at the same time, establishing indexes and data structures to obtain corresponding detection data based on geographic location.

6. The lightning protection detection system for unmanned aerial vehicles according to claim 5, characterized in that: The process of mutual verification and comparative analysis of each test data and identification and elimination of data anomalies caused by environmental factors is as follows: Feature extraction: Extract characteristic parameters related to lightning from the detection data after data preprocessing and time-space synchronization; including: for electric field strength data, extract the peak value, change rate, and duration characteristics of the electric field strength; for magnetic field strength data, extract the amplitude and frequency characteristics of the magnetic field strength; for lightning pulse signals, extract the pulse rise time, fall time, pulse width, and peak voltage characteristics; Correlation analysis: Calculate the correlation coefficient between different drone detection data to determine the similarity between the data; including: for electric field strength data, compare whether the change trend of electric field strength detected by drones at different locations is consistent; for magnetic field strength data, analyze whether the fluctuation of magnetic field strength is correlated; for lightning pulse signals, compare the similarity of pulse characteristic parameters to obtain correlation analysis results; among them, if the correlation coefficient is high, it means that the detection data are consistent; if the correlation coefficient is low, it means that there are data anomalies or they are affected by local environmental factors; Data consistency test: Based on statistical principles, the detection data of multiple drones are tested for consistency; including: using hypothesis testing methods, assuming that multiple groups of data come from the same population and reflect the same lightning activity, calculating the test statistic and comparing it with the corresponding critical value to obtain the consistency test result; among them, if the test statistic is less than the critical value, the original hypothesis is accepted and the data is considered to be consistent; if the test statistic is greater than the critical value, the original hypothesis is rejected, indicating that there are significant differences in the data and abnormal data exists; Abnormal data identification: The detection results are composed of the correlation analysis results and the consistency test results. Based on the detection results, the detection data with low correlation with the drone data or that fails the consistency test are marked as abnormal data. Data correction and verification: Correct or supplement the identified abnormal data; including: if the abnormal data is a slight deviation caused by local environmental interference, use interpolation and average method to correct it according to the trend and characteristics of normal data; after correction, perform correlation analysis and data consistency check again to verify whether the corrected data is consistent with the remaining data; then generate an analysis report based on the results of mutual verification and comparative analysis.

7. The lightning protection detection system for unmanned aerial vehicles according to claim 6 is characterized in that: The process by which the ground control center generates a real-time map of lightning activity based on detection data: The ground control center continuously receives the detection data transmitted by each drone through wireless communication links. The detection data includes electric field strength, magnetic field strength and lightning pulse signals with geographic coordinate tags. Then the received detection data is integrated in real time, sorted and classified according to the time sequence and geographic location information of the data. Build a GIS-based map processing platform that has the functions of geospatial data processing, visualization and analysis; obtain basic geographic information data from the geographic information database, and import the basic geographic information data into the map processing platform to provide geographic background information for the overlay and display of lightning activity data; among them, the basic geographic information data includes topographic and geomorphic layers, administrative division layers, and building distribution layers; convert the received detection data into GeoJSON format, and set the display properties of the data according to the requirements of the GIS platform, including symbol system and color grading; on the GIS platform, draw the integrated lightning detection data on the map according to its geographic coordinate information; A real-time update mechanism is established. As the drone continuously transmits new detection data, the ground control center promptly integrates the new detection data into the map and updates the lightning activity display information at the corresponding location. The map is updated every three seconds through the set update frequency to ensure that the lightning activity status displayed on the map is basically synchronized with the actual situation.

8. The lightning protection detection system for unmanned aerial vehicles according to claim 7, characterized in that: The ground control center also includes: in the process of generating real-time maps, combining data analysis algorithms to further analyze lightning activity data, identify hot spots of lightning activity, and predict movement paths; the process of identifying hot spots is as follows: first, select the time window of lightning detection data within the past 5-10 minutes, including electric field strength, magnetic field strength and lightning pulse signals; use clustering algorithms to cluster the detection data, classify data points with similar geographical locations and similar lightning activity characteristics into one category, and obtain clustering results. According to the clustering results, count the number of data points in each cluster or calculate the comprehensive index of lightning activity intensity within the cluster; the comprehensive index of lightning activity intensity includes the sum of electric field strength and the total number of lightning pulses; determine the area where the cluster with a large number of data points or a high comprehensive index is located as a hot spot of lightning activity; highlight it with purple or lightning symbols on the map.

9. The lightning protection detection system for unmanned aerial vehicles according to claim 8, characterized in that: The process of moving path prediction is as follows: Collect continuous lightning detection data sequences within the past 15-30 minutes, and extract lightning-related mobile feature data from the detection data sequences. The mobile feature data include the longitude and latitude coordinates of the lightning activity center at different times, the changing trends of the electric and magnetic field strengths, and the propagation direction characteristics of lightning pulse signals; Based on the extracted mobile feature data, a prediction model combining a convolutional neural network and a recurrent neural network is selected; the selected model is trained using historical lightning data, and the model parameters are adjusted to accurately capture the movement patterns of lightning activities; the latest mobile feature data of lightning activities acquired in real time is used as input data and input into the trained prediction model; the prediction model outputs the prediction results based on the input data and calculates the future movement direction and speed of lightning activities; the predicted positions of different time steps are connected on the real-time map in the form of lines or arrows to form a predicted movement path.

10. The lightning protection detection system for unmanned aerial vehicles according to claim 9, characterized in that: During the identification of hotspots by the ground control center, the steps of clustering analysis of detection data using clustering algorithms are as follows: S1: Obtain the selected time window from the geographic information database: lightning detection data within the past 5-10 minutes, including the longitude and latitude geographic coordinates of each data point and lightning characteristic parameters; standardize the detection data; S2: Select K-Means algorithm or DBSCAN algorithm; if K-Means algorithm is selected, the number of clusters K is determined in advance by the elbow method, the clustering error sum of squares SSE curves under different K values ​​are plotted, and the K value at the inflection point of the curve is selected as the initial number of clusters; at the same time, the iteration termination condition of the algorithm is set: the maximum number of iterations or the convergence threshold; if DBSCAN algorithm is selected, the key parameters are determined: the neighborhood radius Epsilon and the minimum number of points MinPts; the neighborhood radius defines the neighborhood range of the data point, which is determined by preliminary analysis of the data distribution or empirical value; the minimum number of points specifies the minimum number of data points required to form a cluster, which is set according to the density characteristics of the data and the expected clustering quality; S3: For the K-Means algorithm, randomly select K data points as the initial cluster centers; randomly initialize and run the algorithm multiple times, and select the optimal clustering result or the result that minimizes SSE; S4: For the K-Means algorithm, calculate the distance between each data point and the cluster center; for the DBSCAN algorithm, calculate the distance between each data point; for geographic coordinate data, use the Haversine formula as a distance metric to calculate the spherical distance between two points; construct a distance matrix in which each element represents the distance between two data points; for the K-Means algorithm, calculate the distance between each data point and the K cluster centers; for the DBSCAN algorithm, calculate the distance between each data point and its neighboring data points; S5: According to the distance matrix, for the K-Means algorithm, each data point is assigned to the cluster to which the nearest cluster center belongs, and for the DBSCAN algorithm, each data point is assigned to the cluster that meets the density condition; in the K-Means algorithm, data points are assigned to the cluster to which the nearest cluster center belongs; in the DBSCAN algorithm, if the number of neighboring data points of a data point within its neighborhood radius Epsilon is greater than or equal to the minimum number of points MinPts, then the data point is regarded as a core point and forms a new cluster, and the data points in its neighborhood are also assigned to the cluster; if a data point is not a core point but is in the neighborhood of a core point, then the data point is assigned to the cluster to which the core point belongs, and is called a boundary point; if a data point is neither a core point nor in the neighborhood of any core point, then the data point is regarded as a noise point and does not participate in clustering; S6: In the K-Means algorithm, the new cluster center of each cluster is recalculated. The calculation method of the new cluster center is to take the average value of all data points in the cluster in each feature dimension; S7: repeating steps S4-S6 until the termination condition of the clustering algorithm is met; including: for the K-Means algorithm, stopping the iteration when the maximum number of iterations is reached or the moving distance of the cluster center is less than the convergence threshold; for the DBSCAN algorithm, stopping the iteration when all data points are assigned to clusters and do not include noise points or no new clusters are formed; S8: Calculate clustering evaluation indicators to evaluate the quality of clustering results; clustering evaluation indicators include: Silhouette Coefficient, Calinski-Harabasz index; judge whether the clustering results are reasonable based on the evaluation indicators. If the evaluation indicators are not ideal, adjust the parameters of the clustering algorithm or replace the clustering algorithm, and re-perform clustering analysis until a satisfactory clustering result is obtained.

Images