Multi-dimensional visual analysis method for meteorological observation detection data

Abstract

The invention provides a multi-dimensional visual analysis method for meteorological observation and detection data, and relates to the field of visualization. According to the method, an unmanned aerial vehicle platform carries a meteorological sensor to collect data, and primary processing, abnormal value detection and intelligent data compression are carried out; transmitting the processed data to a ground measurement and control terminal; the ground terminal fuses the multi-source data to generate a three-dimensional meteorological data set; constructing an interactive three-dimensional visual scene by utilizing self-adaptive interaction logic and a dynamic data display technology; a user can check and perform multi-view linkage operation through the interactive operation interface; a severe weather and weather raster data set can be constructed based on the three-dimensional weather data set and historical severe weather records, and simulation and display are carried out; and finally exporting the data, verifying the precision and integrity, and generating a visual analysis report.

Description

Technical Field

[0001] The present invention relates to the field of visualization, and in particular to a multi-dimensional visualization analysis method for meteorological observation and detection data. Background Art

[0002] The demand for accurate and timely meteorological data in the field of meteorological observation is growing; meteorological data is crucial for weather forecasting, climate research and many industry applications; with the development of science and technology, meteorological observation technology continues to improve, from traditional ground observation stations to today's satellite remote sensing and other methods; but the acquisition and analysis of meteorological data still faces many challenges, such as multi-dimensional visualization analysis of data, in order to better meet the needs of various fields for in-depth mining and utilization of meteorological information.

[0003] Among the common meteorological observation data processing solutions currently, one is to collect data through fixed ground meteorological stations and then transmit it to the analysis center for processing; these ground stations are relatively fixed in distribution, and data collection is stable, but the coverage is limited; the other is to use satellite remote sensing technology to obtain large-scale meteorological data, but its data resolution is relatively low and is affected by many factors; in terms of data fusion and visualization, more basic algorithms and display methods are mostly used, which makes it difficult to achieve multi-dimensional accurate visualization analysis.

[0004] The shortcomings of the existing technology are that, although the data collection of existing ground meteorological stations is stable, the spatial resolution of the data is low due to the limited number of stations and their uneven distribution, and it is impossible to obtain comprehensive and detailed meteorological information; although satellite remote sensing data has a wide coverage, the resolution is difficult to meet the needs of local fine meteorological analysis; in the data fusion process, the existing algorithm has insufficient alignment accuracy for data with different sampling frequencies; the visualization display technology is also relatively simple and cannot dynamically adjust the display method according to user behavior and needs; and in terms of severe weather simulation and display, the functions are not perfect enough. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the shortcomings of the prior art, the present invention provides a multi-dimensional visualization analysis method for meteorological observation and detection data to solve the problems that the existing ground meteorological stations have low spatial resolution of data and cannot obtain comprehensive and detailed meteorological information due to the limited number of stations and uneven distribution; although satellite remote sensing data has a wide coverage, the resolution is difficult to meet the needs of local fine meteorological analysis; in the data fusion process, the existing algorithms have insufficient alignment accuracy for data with different sampling frequencies; the visualization display technology is relatively single and cannot dynamically adjust the display method according to user behavior and needs; and the functions in terms of severe weather simulation and display are not perfect.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multi-dimensional visualization analysis method for meteorological observation and detection data, comprising the following steps:

[0009] S1: Use the meteorological sensor on the UAV platform to detect the route and collect meteorological data, which is transmitted through the wireless transmission link;

[0010] S2: performing preliminary data processing, outlier detection and intelligent data compression on the meteorological data at the UAV end to reduce the amount of transmitted data;

[0011] S3: Transmit the processed data to the ground measurement and control terminal via a wireless transmission link;

[0012] S4: After receiving the data, the ground terminal combines the satellite remote sensing data and the ground meteorological station data for data fusion, generates a grid field through the Kriging interpolation algorithm, and uses the dynamic time warping algorithm to align data with different sampling frequencies to generate a three-dimensional meteorological data set containing spatial and temporal information;

[0013] S5: Through adaptive interactive logic and dynamic data display technology, the data display method is dynamically adjusted according to user behavior and needs. The interactive 3D visualization scene is constructed using a 3D scene construction engine that supports WebGL rendering and dynamic particle streamline visualization. Real-time updates ensure that users obtain the latest meteorological data.

[0014] S6: The user views the three-dimensional chart and the interactive three-dimensional visualization scene simultaneously through the interactive operation interface, which supports rotation, zooming and multi-view linkage operations;

[0015] S7: Based on the three-dimensional meteorological data set and the historical measured severe weather records, a high-resolution numerical weather forecast model is called to construct a severe weather meteorological grid data set, and simulate and display it;

[0016] S8: Export the final data into a preset format and verify the data accuracy and completeness, and generate a visual analysis report.

[0017] Preferably, the UAV platform is equipped with a temperature sensor, a humidity sensor, an air pressure sensor, a wind speed sensor, a wind direction sensor and a visibility sensor, wherein the temperature sensor adopts a platinum resistance temperature measuring element that complies with the IEC 60751 standard, and eliminates the influence of the wire resistance through a four-wire measurement method, with a measurement range of -50°C to 60°C and an accuracy of ±0.1°C; the humidity sensor is based on the principle of capacitive polymer film, adopts a dual-channel differential measurement circuit, and has a measurement range of 0%RH to 100%RH and an accuracy of ±2%RH; the air pressure sensor integrates a piezoresistive MEMS chip, with a measurement range of 300hPa to 1100hPa and an accuracy of ±0.5hPa; the wind speed and wind direction sensors are measured by an ultrasonic time difference method Real-time measurement, wind speed measurement range is 0m / s to 30m / s, accuracy is ±0.1m / s, wind direction measurement range is 0° to 360°, accuracy is ±1°; visibility sensor adopts forward scattering optical measurement technology, the transmitting end and the receiving end form an angle of 35°, visibility value is inverted through Mie scattering theory, measurement range is 10m to 20000m, accuracy is ±5%m; UAV platform generates three-dimensional waypoint sequence according to latitude and longitude coordinates and altitude parameters of preset route, and adjusts flight through improved PID control algorithm attitude, ensuring that the sensor can stably collect data under complex meteorological conditions. In airspace below 3,000 meters above sea level, the horizontal flight speed is dynamically adjusted from 1m / s to 15m / s according to detection requirements to meet fixed-point observation requirements. The data acquisition synchronously records the output signals of each sensor at a sampling frequency of 10Hz, adopts hardware-level timestamp alignment technology, and achieves nanosecond synchronization accuracy through FPGA to ensure the temporal and spatial consistency of multi-source data. The collected raw data is encapsulated into data frames after verification by CRC-32 algorithm. Each frame contains a frame header, sensor ID, timestamp, data body and check code. The wireless transmission link adopts adaptive modulation and coding technology, dynamically selects QPSK, 16QAM or 64QAM modulation mode according to channel quality, and the transmission power is automatically compensated with link attenuation. The ground measurement and control terminal receives signals through a MIMO antenna array, and uses the maximum ratio combining algorithm to improve the signal-to-noise ratio. The data transmission delay is controlled within 200 milliseconds, realizing efficient and stable data transmission between the drone and the ground, laying the foundation for subsequent data processing and analysis.

[0018] Preferably, at the drone end, the collected raw meteorological data is first processed preliminarily; the temperature data is subjected to a sliding average filter with a window size of 1 second to eliminate high-frequency noise, and the middle data points are replaced by the average value of the data in the window to effectively smooth the temperature curve; the humidity data is subjected to a median filter to suppress pulse interference, and 5 adjacent data points are selected, sorted by size, and the median is taken to replace the original data to remove abnormal pulses; the air pressure data is smoothed using a second-order Butterworth low-pass filter with a cutoff frequency of 0.1Hz, which can effectively filter out high-frequency interference signals and retain the trend of air pressure changes; the wind speed and wind direction data use a Kalman filter algorithm to separate the real signal from the turbulent disturbance component, and by establishing a state space model, combining the prior estimate with the observed value, recursively calculating the optimal estimate, and accurately extracting the real information of wind speed and direction; then the outlier detection is performed, and based on the dynamic threshold method, the maximum hourly change rate of the temperature data is set to 5℃ / min. If the data exceeds this change range, it is marked as abnormal; the humidity mutation threshold is set to 10%RH / s, and data exceeding this value is considered abnormal; The interquartile range method is used to detect outliers in the air pressure data. If the standard deviation of a data point and the adjacent 10-second data exceeds 3 times, it is determined to be an outlier and removed to ensure the accuracy and reliability of the data. Subsequently, the data compression engine adopts a hybrid compression strategy, and the key parameters use the Zstandard lossless compression algorithm. The key parameters include temperature, air pressure, wind speed and wind direction. The compression level is set to 5 to balance speed and compression rate. The auxiliary parameters use lossy compression. After the data is converted to the frequency domain through discrete cosine transform, the first 8 coefficients are retained. The compression ratio is dynamically adjusted in the range of 1:3 to 1:15, which reduces the amount of data while retaining important information as much as possible. Finally, metadata tags are added to the compressed data packets, including data collection time, geographic location hash value and data quality score. The quality score is calculated based on the sensor calibration status, signal-to-noise ratio and the number of outliers. The quality score algorithm is: quality score = 0.4×calibration coefficient + 0.3×signal-to-noise ratio (dB) / 30 + 0.3×(1-outlier ratio). Data packets with scores below 80 are downgraded to background transmission to optimize data transmission efficiency.

[0019] Preferably, the wireless transmission adopts a time division multiple access technology based on TDMA, divides the transmission time slot into a 10ms period, each period includes an uplink control channel and a downlink confirmation channel, and sends data in sequence according to the time slot number assigned by the ground meteorological station to avoid channel conflicts and ensure orderly data transmission; the ground receiving end deploys a software-defined radio module to support real-time spectrum sensing function, and automatically switches to the backup frequency band when interference in the same frequency band is detected. The frequency band switching response time is less than 50ms. During the switching process, a double buffer mechanism is enabled to temporarily store data in the buffer area to avoid data loss and ensure the continuity of data transmission; a parallel processing architecture is adopted to realize CRC check, data decapsulation and timestamp synchronization functions through FPGA hardware acceleration, and the decoded data is stored in a circular buffer with a buffer capacity of 1GB. When the data accumulation exceeds 80%, a flow control signal is triggered, and the drone end automatically reduces the sampling frequency to 5Hz to prevent data overflow, ensuring that the ground receiving end can stably receive and process the data transmitted by the drone, providing a complete data source for subsequent data fusion and analysis.

[0020] Preferably, after receiving the data, the ground terminal performs multi-source data alignment; the satellite remote sensing data matches the drone data coordinate system through geographic coordinate projection transformation, and the data in different coordinate systems are unified to the same benchmark; the ground meteorological station uses the NTP protocol to synchronize the data, and the data with a time deviation of more than 1 second triggers the interpolation compensation algorithm, and the linear interpolation method is used to estimate the accurate value of the deviation data based on adjacent data points to ensure the consistency of the time series data; the spatial data uses the Kriging interpolation algorithm to generate a 1km×1km resolution grid field, and the sparse observation area of ​​the drone is supplemented with dense data from the ground station, and the SRTM is introduced The elevation correction factor calculated by 30-meter terrain data is: ΔT = 0.0065 × ΔH, where ΔH is the relative height difference. The influence of terrain undulation on the distribution of meteorological elements is taken into account, and the influence of mountain terrain on the vertical gradient of temperature is eliminated, so that the generated grid field is closer to the actual meteorological distribution. The time series data uses a dynamic time warping algorithm to align data sources with different sampling frequencies, and the 1-minute interval data of the ground station and the 10Hz data of the drone are uniformly resampled to a 1-second time base. The resampling process uses a linear interpolation method to retain the statistical characteristics of the original data, and the variance change rate is controlled within ±5% to ensure the integrity and consistency of the time series data. The generated three-dimensional meteorological data set performs a three-level inspection. The first The first level test eliminates data that exceeds the historical climate extreme range, such as temperature below -50℃ or above 60℃, and excludes obviously erroneous data; the second level test marks the data that differs from adjacent stations by more than 3 standard deviations through spatial consistency analysis to identify local abnormal data; the third level test verifies whether the physical relationship between air pressure and altitude conforms to the fluid static equilibrium equation specified by the International Standard Atmospheric Model ISA. The physical relationship formula is ΔP=ρgΔH, where the air density ρ is calculated according to the ISA standard atmospheric model. When the deviation between the measured value and the theoretical value exceeds 5%, an alarm is triggered to ensure the physical rationality of the data. The strictly tested data set provides an accurate and reliable data basis for subsequent visualization analysis.

[0021] Preferably, the user behavior analysis engine is used to capture interactive events in real time, including view zoom ratio, element selection combination, timeline drag speed, and click hot zone distribution, and the behavior data is stored in the Redis cache database for the machine learning model to call, so that the system can automatically optimize the display effect according to user habits; WebGL technology is used to build a three-dimensional visualization scene, in which the temperature field rendering adopts HSV color temperature mapping, which is represented by a red-blue gradient surface, red represents a high temperature area, blue represents a low temperature area, and the overall hue saturation is proportional to the temperature gradient. The specific operation is to linearly map the entire temperature range to the hue range. Surrounded; assuming the lowest temperature is T_min and the highest temperature is T_max, then the hue H corresponding to any temperature T can be calculated according to the formula H=(T-T_min)÷(T_max-T_min)*(240°-0°), the low temperature area will be close to 240°, and the high temperature area will be close to 0°, which can intuitively show the difference in temperature distribution; the humidity field is displayed as a superposition of semi-transparent isosurfaces, and different humidity values ​​correspond to different transparencies, which is convenient for observing the changes in humidity levels; the wind speed vector field is visualized using dynamic particle streamlines, and the particle density is automatically adjusted according to the user's zoom level. When observed at a close distance, the particle density The particle density is adjusted according to the user's zoom level, taking into account both display effect and performance. It supports users to customize data point size, grid density and color mapping rules. The data point size is 1 to 20 pixels, the grid density is 1 to 10 levels, and the color mapping rules are HSV, RGB, and gradient mapping. The configuration parameters are synchronized to the interactive 3D visualization scene and 3D chart in real time to meet the personalized needs of users. Users can select different meteorological elements for display through interactive operations, or draw vertical profiles to view the comparison curves of multiple meteorological elements and gain a deeper understanding of the relationship between meteorological elements. When a user selects a single meteorological element, the 3D chart automatically focuses on the time change curve of the element, and the transparency of other elements is adjusted to 30%. The data extreme points are highlighted to highlight key information and help users quickly grasp the changing characteristics of meteorological elements.

[0022] Preferably, the multi-view linkage operation supports gesture recognition function, and the user can pinch and zoom the three-dimensional model perspective by two fingers to conveniently adjust the observation range; the rotation operation is driven by a gyroscope sensor to change the observation angle naturally and smoothly; when the device is tilted more than 15°, it automatically switches to a vertical profile view to intuitively display the distribution of meteorological parameters at different altitudes. When the user performs a zoom operation, the display granularity and detail level of the relevant meteorological data are automatically adjusted, and closer observations display more detailed data, and longer-distance observations display macro data trends; after receiving the latest meteorological data, the corresponding area of ​​the interactive three-dimensional visualization scene is refreshed with a pulse halo effect to attract the user's attention to the latest data changes; The three-dimensional chart automatically scrolls the time axis when adding new data points to ensure that the latest data is located in the center of the visible area, thereby ensuring that users can obtain the latest meteorological information in real time; when the temperature change rate exceeds 2°C / min or the wind speed mutation exceeds 5m / s, the relevant area is displayed as a red flashing border in the interactive three-dimensional visualization scene, and the duration of the border is positively correlated with the temperature change rate or the wind speed mutation value, highlighting the abnormal meteorological conditions to alert users to pay attention; after the user draws the vertical profile line, the system generates wind speed and humidity comparison curves for different altitude layers with an interval of 50m, and marks the position of the maximum gradient change, clearly showing the change pattern of meteorological elements in the vertical direction, providing users with a comprehensive meteorological data perspective.

[0023] Preferably, severe weather events that meet the conditions are retrieved from the relational database. The screening conditions include sustained wind speed ≥ 17.2 m / s, 1-hour precipitation ≥ 50 mm, and visibility ≤ 500 m. The search results are sorted by frequency to generate a candidate list, which is convenient for selecting typical severe weather cases for simulation. The WRF model kernel is called through the numerical simulation engine, the horizontal grid resolution is set to 500 m, the vertical layer is 50 layers, and the terrain data is based on SRTM. 30-meter DEM, underlying surface type data fused with MODIS land cover classification products, dynamic adjustment of simulation time step, shortening the step to 2 seconds in the severe convection stage, improving simulation accuracy, and accurately reproducing the occurrence and development of severe weather; historical severe weather raster data and real-time monitoring data are superimposed, and a difference heat map is generated through the Mahalanobis distance algorithm. Areas with differences exceeding the threshold of ±20% are marked with purple contour lines, which intuitively show the difference between real-time data and historical data and provide a reference for weather forecasting; spatial coordinates are mapped to GIS three-dimensional terrain base map, and the time axis is associated with meteorological data timestamps to generate spatiotemporal dynamic heat maps. The length of the wind speed vector arrow is linearly related to the wind speed value, that is, 1cm / m / s, the spacing between isobaric surfaces is set to 5hPa, and the transparency of the visibility fog effect is inversely correlated with the measured value, vividly showing the spatiotemporal evolution characteristics of severe weather; it supports intercepting interactive three-dimensional visualization scene key frames and saving them in PNG or TIFF format, exporting time series data as CSV or JSON structured files, and automatically attaching data quality statistics and abnormal event marker lists during report generation, which is convenient for users to further analyze and share simulation results.

[0024] Preferably, the visual analysis report supports export in text format, XLS format and NetCDF standard format to meet the data usage needs of different users in different scenarios; the data integrity is verified before export, and if there is a missing data, an alarm is triggered and a log is recorded to ensure the integrity of the exported data; the consistency of the exported data and the original data is compared by cross-validation to ensure that the error range is less than 1%, thereby ensuring the accuracy and reliability of the data; the visual analysis report includes a data quality score, abnormal event statistics and simulation accuracy indicators calculated based on the sensor calibration status, signal-to-noise ratio and number of outliers, and the report template supports Markdown and PDF format export, which is convenient for users to generate professional and standardized analysis reports.

[0025] (III) Beneficial effects

[0026] The present invention provides a multi-dimensional visualization analysis method for meteorological observation and detection data.

[0027] Beneficial effects:

[0028] 1. The present invention uses a drone platform equipped with a variety of meteorological sensors to conduct route detection and collect meteorological data, and performs preliminary data processing, outlier detection and intelligent data compression on the drone side to reduce the amount of transmitted data and effectively improve the accuracy of data collection and transmission efficiency; at the same time, it combines satellite remote sensing data and ground meteorological station data for data fusion to generate a three-dimensional meteorological data set containing spatial and temporal information, providing a more comprehensive and accurate data basis for subsequent visualization analysis.

[0029] 2. The present invention utilizes adaptive interactive logic and dynamic data display technology to construct an interactive three-dimensional visualization scene. Users can view three-dimensional charts and scenes through an interactive operation interface, and supports multi-view linkage operations. It can also construct a severe weather meteorological raster dataset based on a three-dimensional meteorological dataset and historical severe weather records for simulation and display, so that users can intuitively view and analyze meteorological data, facilitate a better understanding of meteorological phenomena and trends, and provide strong support for meteorological research, weather forecasting, and related industry applications. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] The embodiment of the present invention provides a multi-dimensional visualization analysis method for meteorological observation and detection data. In practical applications, it is first necessary to prepare an unmanned aerial vehicle platform and carry a variety of meteorological sensors, including temperature sensors, humidity sensors, air pressure sensors, wind speed sensors, wind direction sensors and visibility sensors; these sensors must meet specific measurement range and accuracy requirements; for example, the temperature sensor adopts a platinum resistance temperature measuring element that complies with the IEC 60751 standard, and eliminates the influence of wire resistance through a four-wire measurement method. The measurement range is -50°C to 60°C, and the accuracy is ±0.1°C; the humidity sensor is based on the principle of capacitive polymer film, adopts a dual-channel differential measurement circuit, and the measurement range is 0%RH to 100%RH, and the accuracy is ±2%RH; the air pressure sensor integrates a piezoresistive MEMS chip, and the measurement range is 300hPa to 1100hPa, and the accuracy is ±0.5hPa; the wind speed and wind direction sensors use ultrasonic time difference The wind speed measurement range is 0m / s to 30m / s with an accuracy of ±0.1m / s, and the wind direction measurement range is 0° to 360° with an accuracy of ±1°. The visibility sensor adopts forward scattering optical measurement technology, with the transmitting end and the receiving end forming an angle of 35°. The visibility value is inverted through Mie scattering theory, with a measurement range of 10m to 20000m and an accuracy of ±5%m. The UAV platform generates a three-dimensional waypoint sequence based on the latitude and longitude coordinates and altitude parameters of the preset route, and adjusts the waypoint sequence through an improved PID control algorithm. The aircraft adjusts its flight attitude to ensure that the sensor can collect data stably under complex meteorological conditions. In airspace below 3,000 meters above sea level, the horizontal flight speed is dynamically adjusted from 1m / s to 15m / s according to detection requirements to meet fixed-point observation requirements. Data acquisition synchronously records the output signals of each sensor at a sampling frequency of 10Hz, adopts hardware-level timestamp alignment technology, and achieves nanosecond synchronization accuracy through FPGA to ensure the temporal and spatial consistency of multi-source data. The collected raw data is encapsulated into data frames after verification by the CRC-32 algorithm. Each frame contains a frame header, sensor ID, timestamp, data body and check code. The wireless transmission link adopts adaptive modulation and coding technology, dynamically selects QPSK, 16QAM or 64QAM modulation mode according to channel quality, and the transmission power is automatically compensated with link attenuation. The ground measurement and control terminal receives signals through the MIMO antenna array, uses the maximum ratio combining algorithm to improve the signal-to-noise ratio, and controls the data transmission delay within 200 milliseconds, realizing efficient and stable data transmission between the drone and the ground, laying the foundation for subsequent data processing and analysis.

[0032] On the drone side, the collected raw meteorological data is first processed preliminarily; the temperature data is filtered with a sliding average filter with a window size of 1 second to eliminate high-frequency noise, and the middle data points are replaced by the average value of the data in the window to effectively smooth the temperature curve; the humidity data is filtered to suppress pulse interference, and 5 adjacent data points are selected, sorted by size, and the median is taken to replace the original data to remove abnormal pulses; the air pressure data is smoothed using a second-order Butterworth low-pass filter with a cutoff frequency of 0.1Hz, which can effectively filter out high-frequency interference signals and retain the trend of air pressure changes; the wind speed and wind direction data use the Kalman filter algorithm to separate the real signal from the turbulent disturbance component, and by establishing a state space model, combining the prior estimate with the observed value, recursively calculating the optimal estimate, and accurately extracting the real information of wind speed and direction; then outlier detection is performed, and based on the dynamic threshold method, the maximum hourly change rate of the temperature data is set to 5℃ / min. If the data exceeds this range, it is marked as abnormal; the humidity mutation threshold is set to 10%RH / s, and data exceeding this value is considered abnormal; the air pressure data The interquartile range method is used to detect outliers. If the standard deviation of a data point with the adjacent 10-second data exceeds 3 times, it is determined to be an outlier and removed to ensure the accuracy and reliability of the data. Subsequently, the data compression engine adopts a hybrid compression strategy. The key parameters use the Zstandard lossless compression algorithm. The key parameters include temperature, air pressure, wind speed and wind direction. The compression level is set to 5 to balance speed and compression rate. The auxiliary parameters use lossy compression. After the data is converted to the frequency domain through discrete cosine transform, the first 8 coefficients are retained. The compression ratio is dynamically adjusted from 1:3 to 1:15, which reduces the amount of data while retaining important information as much as possible. Finally, metadata tags are added to the compressed data packets, including data collection time, geographic location hash value and data quality score. The quality score is calculated based on the sensor calibration status, signal-to-noise ratio and the number of outliers. The quality score algorithm is: quality score = 0.4×calibration coefficient + 0.3×signal-to-noise ratio (dB) / 30 + 0.3×(1-outlier ratio). Data packets with scores below 80 are downgraded to background transmission to optimize data transmission efficiency.

[0033] The wireless transmission adopts TDMA-based time division multiple access technology, which divides the transmission time slot into 10ms cycles. Each cycle contains an uplink control channel and a downlink confirmation channel. Data is sent in sequence according to the time slot number assigned by the ground meteorological station to avoid channel conflicts and ensure orderly data transmission. The ground receiving end deploys a software-defined radio module to support real-time spectrum sensing function. When interference in the same frequency band is detected, it automatically switches to the backup frequency band. The frequency band switching response time is less than 50ms. During the switching process, the double buffer mechanism is enabled to temporarily store the data in the buffer area to avoid data loss and ensure the continuity of data transmission. The parallel processing architecture is adopted to realize CRC check, data decapsulation and timestamp synchronization functions through FPGA hardware acceleration. The decoded data is stored in a circular buffer with a buffer capacity of 1GB. When the data accumulation exceeds 80%, the flow control signal is triggered, and the drone end automatically reduces the sampling frequency to 5Hz to prevent data overflow, ensuring that the ground receiving end can stably receive and process the data transmitted by the drone, providing a complete data source for subsequent data fusion and analysis.

[0034] After the ground terminal receives the data, multi-source data alignment is performed; satellite remote sensing data is matched to the drone data coordinate system through geographic coordinate projection transformation, and data in different coordinate systems are unified to the same benchmark; the ground meteorological station uses the NTP protocol for time synchronization data, and the data with a time deviation of more than 1 second triggers the interpolation compensation algorithm. The linear interpolation method is used to estimate the accurate value of the deviation data based on adjacent data points to ensure the consistency of the time series data; the spatial data uses the Kriging interpolation algorithm to generate a 1km×1km resolution grid field, and the sparse observation area of ​​the drone is supplemented with dense data from the ground station, introducing SRTM The 30-meter terrain data is used to calculate the elevation correction factor, taking into account the impact of terrain undulation on the distribution of meteorological elements, eliminating the impact of mountain terrain on the vertical gradient of temperature, and making the generated grid field closer to the actual meteorological distribution. The time series data uses a dynamic time warping algorithm to align data sources with different sampling frequencies, and the 1-minute interval data of the ground station and the 10Hz data of the drone are uniformly resampled to a 1-second time base. The resampling process uses a linear interpolation method to retain the statistical characteristics of the original data, and the variance change rate is controlled within ±5% to ensure the integrity and consistency of the time series data. The generated three-dimensional meteorological data set performs a three-level test. The first level of the test eliminates data that exceeds the historical climate extreme range, such as temperatures below -50℃ or above 60℃, and excludes obviously erroneous data. The second level of the test marks data that differ from adjacent stations by more than 3 times the standard deviation through spatial consistency analysis to identify local abnormal data. The third level of the test verifies whether the physical relationship between air pressure and altitude conforms to the hydrostatic equilibrium equation specified by the International Standard Atmosphere Model ISA. When the deviation between the measured value and the theoretical value exceeds 5%, an alarm is triggered to ensure the physical rationality of the data. The strictly tested data set provides an accurate and reliable data basis for subsequent visualization analysis.

[0035] The user behavior analysis engine captures interactive events in real time, including view zoom ratio, feature selection combination, timeline drag speed, and click hot zone distribution. The behavior data is stored in the Redis cache database for the machine learning model to call, so that the system can automatically optimize the display effect according to user habits. The WebGL technology is used to build a three-dimensional visualization scene, in which the temperature field rendering adopts HSV color temperature mapping, which is represented by a red-blue gradient surface. Red represents the high temperature area, blue represents the low temperature area, and the overall hue saturation is proportional to the temperature gradient. The humidity field is displayed by superimposing a semi-transparent isosurface, and different humidity values ​​correspond to different transparency. The wind speed vector field is visualized by dynamic particle streamlines, and the particle density is automatically adjusted according to the user's zoom level. The visualization mode is dynamically switched according to the user's interactive behavior, and users are supported to customize the data point size, grid density and color mapping rules. Users can select different meteorological elements for display through interactive operations, or draw vertical profiles to view the comparison curves of multiple meteorological elements. When the user selects a single meteorological element, the three-dimensional chart automatically focuses on displaying the time variation curve of the element, the transparency of other elements is adjusted to 30%, and the data extreme points are highlighted.

[0036] The multi-view linkage operation supports gesture recognition function, and users can zoom the 3D model perspective by pinching with two fingers; the rotation operation is driven by the gyroscope sensor; it automatically switches to the vertical profile view when the device is tilted more than 15°; when the user zooms in and out, the display granularity and detail level of the relevant meteorological data are automatically adjusted; after receiving the latest meteorological data, the corresponding area of ​​the interactive 3D visualization scene is refreshed with a pulse halo effect; the timeline is automatically scrolled when new data points are added to the 3D chart; when the temperature change rate exceeds 2℃ / min or the wind speed suddenly changes by more than 5m / s, the relevant area is displayed as a red flashing border in the interactive 3D visualization scene; after the user draws the vertical profile line, the system generates wind speed and humidity comparison curves for different height layers with an interval of 50m.

[0037] Then, severe weather events that meet the conditions are retrieved from the relational database. The screening conditions include sustained wind speed ≥17.2m / s, 1-hour precipitation ≥50mm and visibility ≤500m. The search results are sorted by frequency of occurrence to generate a candidate list; the WRF model kernel is called through the numerical simulation engine to overlay the historical severe weather raster data with the real-time monitoring data to generate a spatiotemporal dynamic heat map; it supports capturing the key frames of the interactive 3D visualization scene and saving them in PNG or TIFF format, exporting time series data as CSV or JSON structured files, and automatically attaching data quality statistics tables and abnormal event marker lists during report generation.

[0038] Finally, the visual analysis report supports export in text format, XLS format and NetCDF standard format; the data integrity is verified before export; the consistency of the exported data and the original data is compared through cross-validation method; the visual analysis report includes data quality scores, abnormal event statistics and simulation accuracy indicators calculated based on sensor calibration status, signal-to-noise ratio and number of outliers; the report template supports export in Markdown and PDF formats.

[0039] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional visualization analysis method for meteorological observation data, characterized in that: The following steps are involved: S1: Use the meteorological sensor on the UAV platform to detect the route and collect meteorological data, which is transmitted through the wireless transmission link; S2: performing preliminary data processing, outlier detection and intelligent data compression on the data of the temperature, humidity, air pressure, wind speed and wind direction sensors in the meteorological data on the drone end to reduce the amount of transmitted data; S3: Transmit the processed data to the ground measurement and control terminal via a wireless transmission link; S4: After receiving the data, the ground terminal combines the satellite remote sensing data and the ground meteorological station data for data fusion, generates a grid field through the Kriging interpolation algorithm, and uses the dynamic time warping algorithm to align data with different sampling frequencies to generate a three-dimensional meteorological data set containing spatial and temporal information; S5: Through adaptive interactive logic and dynamic data display technology, the data display method is dynamically adjusted according to user behavior and needs. The interactive 3D visualization scene is constructed using a 3D scene construction engine that supports WebGL rendering and dynamic particle streamline visualization. Real-time updates ensure that users obtain the latest meteorological data. S6: The user views the three-dimensional chart and the interactive three-dimensional visualization scene simultaneously through the interactive operation interface, which supports rotation, zooming and multi-view linkage operations; S7: Based on the three-dimensional meteorological data set and the historical measured severe weather records, a high-resolution numerical weather forecast model is called to construct a severe weather meteorological grid data set, and simulate and display it; S8: Export the final data into a preset format and verify the data accuracy and completeness, and generate a visual analysis report.

2. The multi-dimensional visualization analysis method of meteorological observation data according to claim 1 is characterized in that: In the step S1, the specific operations include: S11: The UAV platform is equipped with a temperature sensor, a humidity sensor, an air pressure sensor, a wind speed sensor, a wind direction sensor and a visibility sensor, wherein the temperature sensor adopts a platinum resistance temperature measuring element, the humidity sensor is based on the principle of capacitive polymer film, the air pressure sensor integrates a piezoresistive MEMS chip, the wind speed and wind direction sensors are measured in real time by an ultrasonic time difference method, and the visibility sensor adopts a forward scattering optical measurement technology; S12: The UAV platform generates a three-dimensional waypoint sequence according to the latitude and longitude coordinates and altitude parameters of the preset route, and adjusts the flight attitude through the PID control algorithm to ensure that the sensor stably collects data under complex meteorological conditions. When in airspace below 3,000 meters above sea level, the horizontal flight speed is dynamically adjusted in the range of 1 to 15 m / s according to the detection requirements to meet the fixed-point observation requirements; S13: The data acquisition synchronously records the output signals of each sensor at a sampling frequency of 10 Hz, and the time-space consistency of multi-source data is ensured by the timestamp alignment technology. The collected raw data is encapsulated into a data frame after CRC verification. Each frame includes a frame header, a sensor ID, a timestamp, a data body and a check code. The check code is generated by the CRC-32 algorithm; S14: The wireless transmission link adopts adaptive modulation and coding technology, dynamically selects QPSK, 16QAM or 64QAM modulation mode according to channel quality, and automatically compensates for transmission power as link attenuation. The ground measurement and control terminal receives signals through the MIMO antenna array and uses the maximum ratio combining algorithm to improve the signal-to-noise ratio. The data transmission delay is controlled within 200 milliseconds.

3. The multi-dimensional visualization analysis method of meteorological observation data according to claim 1 is characterized in that: In the step S2, the specific operations include: S21: In the initial processing stage of the raw data, the temperature data is subjected to a sliding average filter to eliminate high-frequency noise, the humidity data is subjected to a median filter to suppress pulse interference, the air pressure data is subjected to a second-order Butterworth low-pass filter for smoothing, and the wind speed and direction data are subjected to a Kalman filter algorithm to separate the true signal from the turbulent disturbance component; S22: The outlier detection is based on the dynamic threshold method. The maximum hourly change rate of the temperature data is set to 5°C / min, the humidity mutation threshold is set to 10%RH / s, and the interquartile range method is used to detect outliers for the pressure data. If the standard deviation of a data point with the adjacent 10-second data exceeds 3 times, it is marked as an outlier and removed; S23: The data compression engine adopts a hybrid compression strategy. The key parameters include temperature, air pressure, wind speed and wind direction. The Zstandard lossless compression algorithm is used. The compression level is set to 5 to balance speed and compression rate. The auxiliary parameters adopt lossy compression. After the data is converted to the frequency domain through discrete cosine transform, the first 8 coefficients are retained. The compression ratio is dynamically adjusted in the range of 1:3 to 1:

15. S24: A metadata tag is added to the compressed data packet, including the data collection time, the geographic location hash value and the data quality score. The data quality score result is calculated based on the sensor calibration status, the signal-to-noise ratio and the number of outliers. Data packets with a score lower than 80 points are downgraded to background transmission.

4. The multi-dimensional visualization analysis method of meteorological observation data according to claim 1 is characterized in that: In the step S3, the specific operations include: S31: Using TDMA-based time division multiple access technology, the transmission time slot is divided into 10ms cycles. Each cycle contains an uplink control channel and a downlink confirmation channel. The UAV sends data in sequence according to the time slot number assigned by the ground weather station to avoid channel conflicts. S32: The ground receiving end deploys a software-defined radio module, supports real-time spectrum sensing function, automatically switches to the backup frequency band when co-frequency interference is detected, and the frequency band switching response time is less than 50ms. During the switching process, a double buffer mechanism is enabled to avoid data loss; S33: A parallel processing architecture is used to implement CRC check, data decapsulation and timestamp synchronization functions through FPGA hardware acceleration. The decoded data is stored in a circular buffer with a buffer capacity of 1GB. When the data accumulation exceeds 80%, a flow control signal is triggered and the drone end automatically reduces the sampling frequency to 5Hz.

5. The multi-dimensional visualization analysis method of meteorological observation data according to claim 1 is characterized by: In step S4, the three-dimensional meteorological data set is generated by the following steps: S41: In the multi-source data alignment stage, satellite remote sensing data is transformed through geographic coordinate projection to match the drone data coordinate system. The ground meteorological station uses the NTP protocol to synchronize data. Data with a time deviation of more than 1 second triggers the interpolation compensation algorithm. S42: The spatial data is generated using the Kriging interpolation algorithm to generate a 1km×1km resolution grid field, and the sparsely observed areas of the UAV are supplemented with dense ground station data. The terrain elevation correction factor is introduced in the interpolation process to eliminate the influence of mountain terrain on the vertical gradient of temperature; S43: The time series data uses a dynamic time warping algorithm to align data sources with different sampling frequencies. The 1-minute interval data of the ground station and the 10Hz data of the drone are uniformly resampled to a 1-second time base. The resampling process retains the statistical characteristics of the original data, and the variance change rate is controlled within ±5%; S44: The three-dimensional meteorological dataset is subjected to three levels of inspection. The first level of inspection eliminates data that exceeds the historical climate extreme range. The second level of inspection marks the data that differs from adjacent stations by more than 3 standard deviations through spatial consistency analysis. The third level of inspection verifies whether the physical relationship between air pressure and altitude conforms to the fluid static equilibrium equation specified by the International Standard Atmosphere Model ISA.

6. The multi-dimensional visualization analysis method of meteorological observation data according to claim 1 is characterized by: In step S5, the interactive three-dimensional visualization scene is constructed by the following steps: S51: The user behavior analysis engine is used to capture interactive events in real time, including view zoom ratio, element selection combination, timeline drag speed, and click hot zone distribution. The behavior data is stored in the Redis cache database for machine learning model calls. S52: WebGL technology is used to construct a three-dimensional visualization scene, in which the temperature field is represented by a red-blue gradient surface, the humidity field is displayed as a semi-transparent isosurface overlay, and the wind speed vector field is visualized using dynamic particle streamlines, and the particle density is automatically adjusted according to the user's zoom level; S53: Dynamically switch the visualization mode according to the user's interactive behavior. When the user clicks the same area three times in a row, it automatically switches to the two-dimensional heat map. If the user continues to zoom in and out for more than 5 seconds, the three-dimensional chart is displayed first, and the particle density is adjusted according to the user's zoom level. S54: Supporting user-defined data point size, grid density and color mapping rules, and synchronizing configuration parameters to the interactive 3D visualization scene and 3D chart in real time; S55: Users can select different meteorological elements for display through interactive operations, or draw vertical profiles to view comparison curves of multiple meteorological elements; S56: When the user selects a single meteorological element, the three-dimensional chart automatically focuses on displaying the time variation curve of the element, the transparency of other elements is adjusted to 30%, and the data extreme points are highlighted.

7. The multi-dimensional visualization analysis method of meteorological observation data according to claim 1 is characterized by: In step S6, the specific operation process supported by the user interaction operation includes: S61: The multi-view linkage operation supports gesture recognition function. The user can zoom the 3D model perspective by pinching with two fingers. The rotation operation is driven by the gyroscope sensor. When the device is tilted more than 15°, it automatically switches to the vertical section view to display the distribution of meteorological parameters at different altitudes. When the user performs a zoom operation, the display granularity and detail level of the relevant meteorological data are automatically adjusted. S62: After receiving the latest meteorological data, the corresponding area of ​​the interactive three-dimensional visualization scene is refreshed with a pulse halo effect, and the time axis is automatically scrolled when a new data point is added to the three-dimensional chart to ensure that the latest data is located in the center of the visual area; S63: When the temperature change rate exceeds 2°C / min or the wind speed mutation exceeds 5m / s, the relevant area is displayed as a red flashing border in the interactive 3D visualization scene, and the duration of the red flashing border is positively correlated with the temperature change rate or wind speed mutation value; S64: After the user draws the vertical profile line, the system generates wind speed and humidity comparison curves for different height layers at intervals of 50m, and marks the location of the maximum gradient change.

8. The multi-dimensional visualization analysis method of meteorological observation data according to claim 1 is characterized by: In step S7, the specific operation process of severe weather simulation and display includes: S71: the historical data mining module retrieves the severe weather events that meet the conditions from the relational database, and the screening conditions include continuous wind speed ≥ 17.2m / s, 1-hour precipitation ≥ 50mm or visibility ≤ 500m, and the search results are sorted by occurrence frequency to generate a candidate list; S72: The numerical simulation engine calls the WRF model kernel, the horizontal grid resolution is set to 500m, the vertical layer is 50 layers, the terrain data uses SRTM 30m DEM, the underlying surface type data is fused with MODIS land cover classification products, the simulation time step is dynamically adjusted, and the step length is shortened to 2 seconds in the severe convection stage; S73: Overlay the historical severe weather grid data with the real-time monitoring data, generate a difference heat map through the Mahalanobis distance algorithm, and mark the area where the difference exceeds the threshold with a purple contour line. The threshold is based on the benchmark data that has passed the third-level inspection as a reference value; S74: Map the spatial coordinates to the GIS three-dimensional terrain base map, associate the time axis with the meteorological data timestamp, generate a spatiotemporal dynamic heat map, the wind speed vector arrow length is linearly related to the wind speed value, the isobaric surface spacing is set to 5hPa, and the visibility fog effect transparency is inversely correlated with the measured value; S75: Supports capturing the key frames of the interactive 3D visualization scene and saving them in PNG / TIFF format, exporting time series data as CSV / JSON structured files, and the report generation module automatically attaches a data quality statistics table and an abnormal event marking list.

9. The multi-dimensional visualization analysis method of meteorological observation data according to claim 1, characterized in that: In step S8, the specific operation process of data export and verification includes: S81: Supports exporting in text format, XLS format and NetCDF standard format; S82: Check the data integrity before exporting. If there is any missing data, an alarm will be triggered and a log will be recorded. S83: Compare the consistency of the derived data with the original data by cross-validation method to ensure that the error range is less than 1%; S84: The visual analysis report includes data quality scores, abnormal event statistics, and simulation accuracy indicators calculated based on sensor calibration status, signal-to-noise ratio, and number of outliers. The report template supports Markdown and PDF format exports.