Meteorological data comprehensive analysis platform based on multi-sensor fusion

Abstract

The invention discloses a multi-sensor fusion meteorological data comprehensive analysis platform, and relates to the technical field of meteorological monitoring. The platform comprises a three-level sensing layer, a dynamic collaborative architecture and a meteorological data analysis layer. The third-level sensing layer works cooperatively through a ground total-factor meteorological station, an unmanned aerial vehicle cluster and a FY-4 satellite and is integrated with a bimodal ultrasonic wind measurement sensor, a miniature meteorological sounding cabin and an HRIT / LRIT data receiving module, so that multi-dimensional data acquisition from the ground to a 5km high altitude is realized; the dynamic collaborative architecture is based on an edge computing node and a cloud fusion module, adopts a lightweight LSTM (Long Short Term Memory) model to detect abnormity in real time, and dynamically optimizes the weight of multi-source data through Kalman filtering; and the meteorological data analysis layer is combined with the element coincidence index, the layering stability index and the risk assessment model to generate a graded early warning signal. According to the scheme, the problems of vertical data fault, low multi-source data fusion efficiency, early warning lag and the like in the traditional technology are solved, and the severe convective weather prediction precision is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological monitoring, and in particular to a multi-sensor fusion meteorological data comprehensive analysis platform. Background Art

[0002] Traditional meteorological monitoring technology faces technical bottlenecks in multiple dimensions. In terms of vertical data collection, ground observation stations and satellite remote sensing systems are limited by fixed detection altitudes, making it difficult to obtain meteorological parameters such as temperature gradient and stratification stability data for continuous vertical profiles from 0 to 5 kilometers, resulting in a prediction error rate of up to 30% for severe convective weather. At the data fusion level, the sampling density of conventional drone routes is insufficient, and ground, drone and satellite data lack a dynamic calibration mechanism due to equipment attitude deviation and environmental interference. For example, when the drone elevation error exceeds 30 degrees or the ultrasonic wind measurement signal attenuation exceeds 3 decibels due to snow accumulation, the data consistency is less than 85%. In terms of complex scene response capabilities, traditional drone systems have significant defects in strong convective environments with wind speeds exceeding 15 meters per second, with obstacle avoidance response delays exceeding 500 milliseconds and positioning errors reaching ±5 meters. At the same time, the transmission delay introduced by traditional encryption algorithms such as AES exceeds 100 milliseconds, which cannot meet the real-time requirements of graded warnings, and extreme weather warnings lag by more than 30 minutes. These problems together lead to limited vertical resolution of meteorological monitoring, inefficient multi-source data fusion and insufficient emergency response capabilities, which seriously restrict the accurate prediction and rapid disposal of disaster weather. Summary of the invention

[0003] The present invention provides a multi-sensor fusion meteorological data comprehensive analysis platform to solve the problems of limited vertical resolution of meteorological monitoring, low efficiency of multi-source data fusion and insufficient emergency response capability in the prior art.

[0004] The present invention provides a multi-sensor fusion meteorological data comprehensive analysis platform, including: a three-level perception layer including a ground observation layer, a drone cluster layer, and a satellite coordination layer, wherein the satellite coordination layer receives the HRIT / LRIT data stream of the Fengyun-4 satellite through a 3.7m parabolic antenna; a dynamic coordination architecture including an edge computing node and a cloud fusion module; a meteorological data analysis layer including a factor analysis module, a level analysis module, and a meteorological assessment module;

[0005] The dynamic collaborative framework uses the ground station as a carrier and has a built-in NPU acceleration module, runs a lightweight LSTM model with a parameter volume of <1MB, detects data anomalies in real time, and outputs fault codes to obtain error detection results E. error , the abnormal judgment condition is: |X obs -X pred |>3σ, where X obs is the observed value, X predis the predicted value, σ is the standard deviation of historical data; based on the spatiotemporal interpolation algorithm, multi-source data are integrated, and the weight coefficient is dynamically optimized through Kalman filtering;

[0006] The meteorological data analysis layer calculates the element conformity index I based on the trend similarity corr and numerical similarity sim between real-time data and historical data. element , when I element <0.6 triggers a data anomaly warning; combines the vertical profile data of the drone with the satellite cloud map to calculate the atmospheric stratification stability index I layer , when I layer >5℃ / km triggers convective instability warning; comprehensive factor compliance index, stratification stability index and error detection results generate meteorological risk assessment value R risk , R risk ≥0.7, triggers a red alert, 0.5≤R risk <0.7, triggering an orange warning.

[0007] Preferably, the ground observation layer deploys a full-element meteorological station integrating a dual-mode ultrasonic wind sensor and a self-calibration temperature, humidity and pressure composite sensor. The ultrasonic wind sensor adopts a concave reflection ellipsoid structure, measures wind speed and direction in real time by the time difference method, and dynamically calibrates the drone and satellite data to satisfy the equation: Among them, L = 300mm ± 5% is the horizontal distance between the two foci of the ultrasonic transmitter, and H = 150mm ± 5% is the height of the concave reflector corresponding to the minor semi-axis of the ellipsoid; the ultrasonic operating frequency is 200kHz, and the signal attenuation is <3dB. The ultrasonic wind sensor has a built-in heating circuit module, and the heating is started when the temperature is <0℃ to eliminate the measurement error caused by snow cover.

[0008] Preferably, the drone cluster layer is equipped with a micro-meteorological sounding cabin and a three-axis stabilized optoelectronic pod, and the meteorological sounding cabin specifically includes: a Pt100 platinum resistance temperature sensor: using a four-wire measurement circuit ±0.2°C, with a temperature measurement range of -50°C to +50°C;

[0009] Humidity-sensitive capacitance humidity sensor: Through the capacitance-voltage conversion circuit, it outputs 0-1V linear voltage to achieve 0-100%RH measurement, ±3%RH@≤80%RH, and the humidity data is corrected for temperature drift through the Kalman filter algorithm;

[0010] Three-axis stabilized optoelectronic pod: equipped with a 2-megapixel CMOS sensor, a frame rate of 30fps, combined with a gyroscope zero bias stability of 0.5° / h to achieve a stabilization accuracy of ±0.1°.

[0011] Preferably, the drones in the drone cluster layer execute a three-dimensional spiral scanning path: x=r(θ)cosθ, y=r(θ)sinθ, z=kθ, θ is the pitch angle deviation angle of the drone, wherein the three-dimensional spiral scanning path satisfies the dynamic adjustment formula: Among them, R0 = 200m is the baseline radius, ΔP = 50m / hPa is the pressure gradient response coefficient, is the temperature gradient modulus; the vertical step coefficient k=10m / rad, the UAV uses Beidou / GPS dual-mode navigation, the positioning accuracy is ±1.5m, the flight trajectory is adjusted in real time, and the altitude profile of 0-5km is covered.

[0012] Preferably, the drone data collection calibration realizes attitude error correction through quaternion transformation and three-dimensional rotation matrix:

[0013] Step 1. Input parameters: roll angle roll: -30°≤roll≤+30°, pitch angle pitch: -15°≤pitch≤+15°, horizontal speed U drone , vertical speed V drone ;

[0014] Step 2: Calculate the normal vector of the inclined surface h = (h x ,h y ,h z ),in:

[0015] h x =-tan -1 (pitch),h y =-tan -1 (roll),h z =1

[0016] Step 3: Construct the three-dimensional rotation matrix R:

[0017]

[0018] Step 4: Correct the wind speed component:

[0019]

[0020] Step 5: Output corrected wind speed and wind direction WD corrected =arctan 2(f y ,f x ), eliminating the 30° elevation error.

[0021] Preferably, the platform further autonomously operates the system: Polarization coded QKD protocol: Based on the BB84 protocol variant, Alice and Bob's basis vector selection probability P H =0.5,P +=0.5, 25km air-to-ground channel key rate>1kbps, bit error rate<3.8×10 -5 ;

[0022] Data packet structure: 32bit quantum check code: Where D i , is the i-th binary, P i ∈{|H>,|V>,|+>,|->}; 64-bit BeiDou-3 nanosecond timestamp, timing accuracy ±3ns; LZ4 compressed meteorological data, compression ratio 4:1, end-to-end transmission delay <45ms.

[0023] A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 6, characterized in that the autonomous operation system further includes: 77GHz millimeter wave radar, detection distance ≥150m and ORB-SLAM algorithm, matching accuracy ≤0.1px coordination, obstacle avoidance response time <200ms; using polarization coding BB84 protocol variant, data contains 32bit quantum check code and Beidou-3 timestamp, field of view angle ±45°, speed resolution 0.1m / s;

[0024] Doppler shift formula: v r The radial velocity of the target is obtained by fusing the ORB-SLAM algorithm with the IMU data, λ =

[0025] 3.89mm millimeter wave radar operating wavelength;

[0026] ORB-SLAM algorithm: extract ORB feature points, ≥1000 points per frame, matching accuracy ≤0.1 pixel, fuse IMU data, and output six-degree-of-freedom pose at 100Hz;

[0027] DS Evidence Theory Fusion: Conflict Factors Obstacle avoidance response time <200ms@5m / s wind speed.

[0028] Preferably, the meteorological data analysis layer specifically includes: calculating the element conformity index I based on the trend similarity corr and the numerical similarity sim of the real-time data and the historical data element , When I element <0.6 triggers a data anomaly warning, N is the number of sampling points, corr i is the correlation coefficient of the ith sampling point, sim i is the similarity of the i-th sampling point; combining the vertical profile data of the UAV and the satellite cloud map, the atmospheric stratification stability index I is calculated layer , When I layer >5℃ / km triggers convective instability warning, temperature gradient The unit is ℃ / km. It is obtained by the Kriging interpolation algorithm of the vertical profile data of the UAV, and the interpolation resolution is ≤50m. The meteorological risk assessment value R is generated by integrating the element conformity index, stratification stability index and error detection results. risk , R risk =0.4I element +0.4I layer -0.2E error , when R risk ≥0.7 triggers a red warning, and 0.5≤R risk <0.7 triggers an orange alert.

[0029] Preferably, the platform further includes a data receiving subsystem, which includes: an antenna configured to receive satellite signals; a low noise amplifier connected to the antenna and configured to amplify and filter the received signal; a down converter connected to the low noise amplifier and configured to convert the signal into an intermediate frequency signal; an indoor data receiver connected to the down converter and configured to demodulate, descramble and channel decode the intermediate frequency signal to generate demodulated and decoded data of IP group packets; a data receiving computer configured to run data receiving software, perform quick visual processing and save the demodulated and decoded data, generate LO level data and transmit it to the server;

[0030] The data processing and product generation and display subsystem is configured as follows: according to the preset time scheduler, poll the data temporary storage area to obtain the files to be processed; perform quality inspection on the files, and if the inspection passes, trigger the product generation process to generate multi-channel synthetic data and vertical profiles; perform projection processing based on the area and projection type selected by the user, generate quick-view images and overlay geographic information; store the generated images and data products in the temporary storage area for the data archiving program to call.

[0031] Preferably, the platform further includes a data storage and retrieval subsystem configured to: schedule the storage program according to a preset time plan; detect the temporary storage area of ​​files and images to obtain the files to be archived; perform quality checks on the files to be archived, extract metadata and store them in the database; create a directory according to the cataloging scheme and complete the archiving based on the detector type, product type, resolution, projection mode and time information;

[0032] The system status monitoring and management subsystem is configured to provide scheduled data reception, product generation, data archiving and time calibration services.

[0033] The multi-sensor fusion meteorological data comprehensive analysis platform of the present invention realizes high-precision meteorological data collection of 0-5km continuous vertical profiles through the combination of the three-dimensional spiral scanning path of the unmanned aerial vehicle and the satellite collaborative layer. The resolution can reach ≤50m, which effectively solves the problem of discontinuous vertical data collection in traditional meteorological monitoring. The platform adopts a dynamic collaborative architecture, integrates edge computing and cloud-based Kalman filtering technology, and optimizes the spatiotemporal interpolation algorithm, which significantly improves the efficiency of multi-source data fusion, and the data consistency reaches >85%, solving the problem of low efficiency of multi-source data fusion. In addition, the platform introduces a lightweight LSTM model with a parameter volume of <1MB for real-time anomaly detection, and the end-to-end transmission delay is controlled at <45ms. The warning timeliness is improved to seconds, which significantly shortens the warning response time and solves the problem of delayed warning response. In terms of data transmission, the platform adopts the polarization coded QKD protocol with a bit error rate of <3.8×10 -5 Combined with Beidou timestamp, the data's tamper-proof and anti-interference capabilities are ensured, effectively improving the security of data transmission. The platform of the present invention has a significant effect in improving the accuracy, efficiency and security of meteorological monitoring, and can better meet the needs of modern meteorological monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a general architecture diagram of a multi-sensor fusion meteorological data comprehensive analysis platform of the present invention;

[0035] Figure 2 It is a schematic diagram of the concave reflection structure;

[0036] Figure 3 This is a principle block diagram of the circuit system of the ultrasonic wind sensor;

[0037] Figure 4 This is a schematic diagram of the waveform received by the ultrasonic wind sensor;

[0038] Figure 5 This is a schematic diagram of the ultrasonic wind measurement principle;

[0039] Figure 6 This is the system information flow chart. DETAILED DESCRIPTION

[0040] The present application provides a multi-sensor fusion meteorological data comprehensive analysis platform to solve the problems of low data calibration efficiency, insufficient vertical profile coverage, and delayed abnormal warning in traditional meteorological monitoring.

[0041] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. 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. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0042] Embodiment 1

[0043] like Figure 1 As shown: The three-level perception layer includes the ground observation layer: deploying a full-element meteorological station, integrating a dual-mode ultrasonic wind sensor and a self-calibrated temperature, humidity and pressure composite sensor. The ultrasonic wind sensor adopts a concave reflection ellipsoid structure, measures wind speed and direction in real time through the time difference method, and dynamically calibrates drone and satellite data;

[0044] UAV cluster layer: equipped with a miniature meteorological sounding cabin and a three-axis stabilized optoelectronic pod. The meteorological sounding cabin integrates a Pt100 platinum resistance temperature sensor and a humidity-sensitive capacitor humidity sensor.

[0045] Satellite coordination layer: Receive the HRIT / LRIT data stream of the Fengyun-4 satellite through a 3.7m parabolic antenna, and use polarization multiplexing technology to achieve dual-channel concurrent transmission;

[0046] Specifically, Figure 2 The ultrasonic wind sensor of this scheme adopts a concave reflection ellipsoid structure. The concave reflection structure is based on the plane reflection structure, and the reflection surface is designed to be concave. It not only changes the direction of the reflected beam and increases the measurement range of the measurement system, but also makes the reflected beam more concentrated, effectively enhancing the strength of the reflected signal. The ultrasonic beams emitted by sensors T1 and T2 can be received by each other, and the propagation distance is the same. According to geometric knowledge, the ellipse with T1 and T2 as the focus, O as the origin, and H as the minor semi-axis meets the requirements, that is: In order to simultaneously meet the measurement requirements of another pair of sensors in the orthogonal direction, the concave surface should be an ellipsoidal surface formed by rotating the ellipse described by the above formula around the y-axis, that is: Among them, L = 300mm ± 5% is the horizontal distance between the two emitter foci, H = 150mm ± 5% is the height of the concave reflector corresponding to the minor semi-axis of the ellipsoid; the ultrasonic working frequency is 200kHz, and the signal attenuation is <3dB. In the concave reflection structure, the wind speed measurement range is not limited by the sensor beam angle, but is determined by the size of the ellipsoid surface. The advantages of this structure are mainly reflected in the following aspects:

[0047] 1) The reflected beam is more concentrated, and the beams in the directions of T1A and T1B can be received by the T2 sensor, which greatly enhances the strength of the received signal and can effectively overcome the difficulty of severe signal attenuation under strong wind conditions.

[0048] 2) The sensor beam angle can be designed to be smaller. Under the same signal strength conditions, not only is the voltage required to drive the sensor lower, but the sensitivity requirement for circuit detection is also lower, reducing the difficulty of circuit design.

[0049] 3) It can effectively reduce interference and improve the accuracy of wind speed and direction measurement.

[0050] like Figure 3 The circuit system of the ultrasonic wind sensor is composed of a control unit, a transmitting drive circuit, a receiving circuit, an amplifying and filtering circuit, a heating circuit, a monitoring module, etc.

[0051] 1) Control unit

[0052] The control unit is mainly responsible for generating pulse square waves and driving ultrasonic emission; at the same time, it controls the ultrasonic receiving circuit on the other side to process the received echo signal, find the best receiving time point, and thus calculate the time difference, and finally calculate the wind speed and direction values; the data is displayed and transmitted through the display screen or communication interface.

[0053] 2) Transmitter drive circuit

[0054] The driving signal of the ultrasonic sensor usually adopts a square wave pulse driving method, because it is relatively easy to implement with a single-chip microcomputer. The driving circuit mainly controls the ultrasonic probe to emit ultrasonic waves.

[0055] 3) Receiving circuit

[0056] During the driving pulse emission process, the ultrasonic sensor chip will experience three states: forced vibration, balanced vibration and attenuated vibration. In the process of receiving the echo, because the piezoelectric crystal has a certain vibration inertia, after receiving the echo, the amplitude increases according to the exponential curve, and it takes several cycles to saturate. Moreover, when the emission signal ends, the chip will maintain several or more cycles of residual vibration. Therefore, accurately judging the arrival time of the ultrasonic wave is a key factor in improving the accuracy of wind speed measurement. Figure 4 As shown, high-speed pulse acquisition is used to realize ultrasonic echo acquisition, and the correct receiving position is found from the echo signal, so as to accurately calculate the wind speed and wind direction values.

[0057] 4) Heating circuit

[0058] Because ultrasonic wind sensors need to adapt to the usage requirements of different regions across the country, they have a built-in heating circuit module. When the air temperature is below 0°C, the heating circuit module starts to prevent the sensor surface from being affected by ice, frost or snow accumulation, which may affect the accuracy of wind direction and speed measurements.

[0059] 5) Monitoring module

[0060] The wind direction and speed sensor has a built-in monitoring circuit, which monitors the ultrasonic sensor calibration period, data measurement range and power supply voltage in real time. If an abnormality occurs, it will output the corresponding abnormal code to prompt the user the cause of the fault so that it can be replaced or repaired in time.

[0061] Working principle: ultrasonic time difference method is used to measure wind.

[0062] Ultrasonic sensors realize the mutual conversion between high-frequency sound energy and electrical energy through positive and negative piezoelectric effects, thereby realizing the emission and reception of ultrasonic waves.

[0063] Figure 5 Schematic diagram of ultrasonic wind measurement principle

[0064] Two ultrasonic sensor probes with transmitting and receiving functions are placed orthogonally to each other, and ultrasonic waves are transmitted and received alternately. Affected by the movement of airflow, the propagation time of ultrasonic waves in two directions is different, and the time difference is a function of wind speed. After calculation and calibration, the airflow velocity between the two probes can be measured, and then the wind direction and wind speed values ​​can be obtained by vector synthesis.

[0065] Assume that the distance between the ultrasonic transmitting and receiving sensors is L, the wind speed during measurement is V, the speed of sound is C, and when the ultrasonic signal propagates downwind, the time required for the ultrasonic signal to propagate from the transmitting sensor to the receiving sensor is t w ,

[0066] Then: t w =L / (C+V) Since each ultrasonic sensor has the function of transmitting and receiving, when the two ultrasonic sensors transmit and receive in reverse direction, the time required for the ultrasonic signal to propagate against the wind is t a , then: t a =L / (CV)From the formula:

[0067] From the above formula, we can know that the wind speed is proportional to the inverse of the time difference required for ultrasound to propagate along the sound path. w and t a The wind speed value can be calculated by the distance L between the transmitting and receiving sensors. Due to the attenuation of ultrasonic signal propagation, the distance L should not be too large. At the same time, a smaller distance L will make t w and t a High precision becomes difficult.

[0068] The meteorological sensors carried by the UAV platform conduct real-time sampling to collect high-precision meteorological data such as temperature, humidity, air pressure, wind direction, wind speed, etc. in the detection area. The mounted optoelectronic pod uses a visible light camera to collect clear and stable video image data in the detection area. Meteorological data and video image data rely on a wireless data transmission system to transmit the collected meteorological data back to the ground measurement and control platform in real time. The ground measurement and control platform communicates with the aerial UAV in real time through a data link to realize remote control, telemetry, tracking, monitoring information transmission, data transmission quality control box data processing, analysis and display of the UAV system. It can receive and display meteorological detection data, image transmission information, data location information, data time information, etc. from the aircraft platform in real time. The system information flow chart is as follows: Figure 6 shown.

[0069] The drones in the drone cluster layer execute a three-dimensional spiral scanning path: x = r(θ)cosθ, y = r(θ)sinθ, z = kθ, θ is the pitch angle deviation angle of the drone, and the three-dimensional spiral scanning path satisfies the dynamic adjustment formula: Among them, R0 = 200m is the baseline radius, ΔP = 50m / hPa is the pressure gradient response coefficient, is the temperature gradient modulus; the vertical step coefficient k=10m / rad, the UAV uses Beidou / GPS dual-mode navigation, the positioning accuracy is ±1.5m, the flight trajectory is adjusted in real time, and the altitude profile of 0-5km is covered.

[0070] The calibration of drone data acquisition realizes attitude error correction through quaternion transformation and three-dimensional rotation matrix:

[0071] Step 1. Input parameters: roll angle roll: -30°≤roll≤+30°, pitch angle pitch: -15°≤pitch≤+15°, horizontal speed U drone , vertical speed V drone ;

[0072] Step 2: Calculate the normal vector of the inclined surface h = (h x ,h y ,h z ),in:

[0073] h x =-tan -1 (pitch),h y =-tan -1 (roll),h z =1

[0074] Step 3: Construct the three-dimensional rotation matrix R:

[0075]

[0076] Step 4: Correct the wind speed component:

[0077]

[0078] Step 5: Output corrected wind speed and wind direction WD corrected =arctan 2(f y ,f x ), eliminating the 30° elevation error.

[0079] Satellite coordination layer: The data receiving subsystem consists of a 3.7m antenna device, a dual-path feed, a low-noise amplifier, a satellite communication lightning arrester, a HRIT-1 data receiver, a HRIT-2 data receiver, a HRIT-1 data entry station, and a HRIT-2 data entry station. Polarization coded QKD protocol: Based on a variant of the BB84 protocol, the basis vector selection probability P of Alice and Bob is H =0.5,P + =0.5, 25km air-to-ground channel key rate>1kbps, bit error rate<3.8×10-5; data packet structure: 32bit quantum check code: Where D i , is the i-th binary, P i ∈{|H>,|V>,|+>,|->}; 64-bit BeiDou-3 timestamp, timing accuracy ±3ns; LZ4 compressed meteorological data, compression ratio 4:1, end-to-end transmission delay <45ms. Further includes: 77GHz millimeter wave radar: detection distance ≥150m, field of view ±45°, velocity resolution 0.1m / s, Doppler frequency shift formula: v r Target radial velocity, obtained by fusing ORB-SLAM algorithm with IMU data, λ = 3.89mm millimeter-wave radar working wavelength; RB-SLAM algorithm: extract ORB feature points, ≥ 1000 points per frame, matching accuracy ≤ 0.1 pixel, fuse IMU data, output six-degree-of-freedom pose at 100Hz; DS evidence theory fusion: conflict factor Obstacle avoidance response time <200ms@5m / s wind speed.

[0080] The main functions of the data receiving subsystem are as follows: (1) Receive the L-band horizontally polarized and vertically polarized signals transmitted by the Fengyun-4 satellite; (2) Perform network input and processing on the received signals; (3) Have data recovery capabilities such as signal demodulation, serial-to-parallel conversion, and descrambling; (4) Have self-check and test functions.

[0081] Working principle: The receiving antenna is aimed at the satellite, and the satellite signal is focused to the feed source by the parabolic antenna and sent to the high-frequency extension via the feeder; the low-noise amplifier amplifies and filters the signal and sends it to the down-converter; the down-converter outputs the intermediate frequency signal and transmits it to the indoor data receiver; the receiver amplifies, filters, demodulates, descrambles and decodes the channel of the signal, outputs the demodulated and decoded data of the IP package, and sends it to the data receiving computer; the data receiving software of the data receiving computer quickly views and saves the received data to generate L0 level data and sends it to the server.

[0082] Furthermore, the dynamic collaborative architecture includes edge computing nodes: a ground station with a built-in NPU acceleration module, a lightweight LSTM model with a running parameter volume of <1MB, real-time detection of data anomalies and output of fault codes to obtain error detection results E error , the abnormal judgment condition is: |X obs -X pred |>3σ, where σ is the standard deviation of historical data;

[0083] Cloud fusion module: Fusion of multi-source data based on spatiotemporal interpolation algorithm, weight coefficients α, β, γ are dynamically optimized through Kalman filtering: X(t,h) = α·X ground (t)+β·X drone (t,h)+γ·X satellite (t) Kalman filtering dynamically adjusts the weight coefficients through prediction and update steps to ensure the accuracy of data fusion.

[0084] The meteorological data analysis layer includes the element analysis module: the element conformity index I is calculated based on the trend similarity corr and numerical similarity sim between real-time data and historical data. element , when I element When <0.6, a data anomaly warning is triggered;

[0085] Layer analysis module: Combines the vertical profile data of drones with satellite cloud images to calculate the atmospheric stratification stability index I layer , when I layer >5℃ / km triggers convective instability warning;

[0086] Meteorological assessment module: Comprehensive element compliance index, stratification stability index and error detection results to generate a meteorological risk assessment value R risk , R risk ≥0.7, triggers a red alert, 0.5≤R risk <0.7, triggering an orange warning.

[0087] Specifically, the meteorological data analysis layer includes: calculating the element conformity index I based on the trend similarity corr and numerical similarity sim between real-time data and historical data element , Where Xhist is the historical data series, corr is the Pearson correlation coefficient, When I element When <0.6, a data anomaly warning is triggered;

[0088] Furthermore, the atmospheric stratification stability index I was calculated by combining the vertical profile data of the UAV with the satellite cloud image. layer , When I layer >5℃ / km triggers convective instability warning, temperature gradient The unit is ℃ / km. It is obtained by the Kriging interpolation algorithm of the vertical profile data of the UAV, and the interpolation resolution is ≤50m;

[0089] Furthermore, the meteorological risk assessment value R is generated by integrating the factor compliance index, stratification stability index and error detection results. risk , R risk =0.4I element +0.4I layer -0.2E error , when R risk ≥0.7 triggers a red warning, and 0.5≤R risk <0.7 triggers an orange alert.

[0090] Further, autonomous operation system: Polarization coded QKD protocol: Based on BB84 protocol variant, Alice and Bob's basis vector selection probability P H =0.5,P + =0.5, 25km air-to-ground channel key rate>1kbps, bit error rate<3.8×10 -5 ;

[0091] Data packet structure: 32bit quantum check code: Where D i , is the i-th binary, P i ∈{|H>,|V>,|+>,|->}; 64-bit BeiDou-3 nanosecond timestamp, timing accuracy ±3ns; LZ4 compressed meteorological data, compression ratio 4:1, end-to-end transmission delay <45ms.

[0092] The autonomous operation system further includes: 77GHz millimeter wave radar, detection distance ≥150m and ORB-SLAM algorithm, matching accuracy ≤0.1px coordination, obstacle avoidance response time <200ms; adopts polarization coding BB84 protocol variant, data includes 32bit quantum check code and Beidou-3 timestamp, field of view angle ±45°, speed resolution 0.1m / s;

[0093] Doppler shift formula: v rThe radial velocity of the target is obtained by fusing the ORB-SLAM algorithm with the IMU data, λ =

[0094] 3.89mm millimeter wave radar operating wavelength;

[0095] ORB-SLAM algorithm: extract ORB feature points, ≥1000 points per frame, matching accuracy ≤0.1 pixel, fuse IMU data, and output six-degree-of-freedom pose at 100Hz;

[0096] DS Evidence Theory Fusion: Conflict Factors Obstacle avoidance response time <200ms@5m / s wind speed.

[0097] Further, a data receiving subsystem includes: an antenna configured to receive satellite signals; a low noise amplifier connected to the antenna, configured to amplify and filter the received signal; a down converter connected to the low noise amplifier, configured to convert the signal into an intermediate frequency signal; an indoor data receiver connected to the down converter, configured to demodulate, descramble and channel decode the intermediate frequency signal, and generate demodulated and decoded data of IP group packets; a data receiving computer configured to run data receiving software, perform quick visual processing and save the demodulated and decoded data, generate LO level data and transmit it to the server;

[0098] The data processing and product generation and display subsystem is configured as follows: according to the preset time scheduler, poll the data temporary storage area to obtain the files to be processed; perform quality inspection on the files, and if the inspection passes, trigger the product generation process to generate multi-channel synthetic data and vertical profiles; perform projection processing based on the area and projection type selected by the user, generate quick-view images and overlay geographic information; store the generated images and data products in the temporary storage area for the data archiving program to call.

[0099] Furthermore, the data storage and retrieval subsystem is configured to: schedule the storage program according to a preset time plan; detect the temporary storage area of ​​files and images to obtain the files to be archived; perform quality checks on the files to be archived, extract metadata and store them in the database; create a directory according to the cataloging plan and complete the archiving based on the detector type, product type, resolution, projection mode and time information;

[0100] The system status monitoring and management subsystem is configured to provide scheduled data reception, product generation, data archiving and time calibration services.

[0101] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A multi-sensor fusion meteorological data comprehensive analysis platform, characterized in that: include: The three-level perception layer includes ground observation layer, drone cluster layer, and satellite coordination layer. The satellite coordination layer receives the HRIT / LRIT data stream of Fengyun-4 satellite through a 3.7m parabolic antenna; the dynamic coordination architecture includes edge computing nodes and cloud fusion modules; the meteorological data analysis layer includes factor analysis module, level analysis module, and meteorological assessment module; The dynamic collaborative framework uses the ground station as a carrier and has a built-in NPU acceleration module, runs a lightweight LSTM model with a parameter volume of <1MB, detects data anomalies in real time, and outputs fault codes to obtain error detection results E. error , the abnormal judgment condition is: |X obs -X pred |>3σ, where X obs is the observed value, X pred is the predicted value, σ is the standard deviation of historical data; based on the spatiotemporal interpolation algorithm, multi-source data are integrated, and the weight coefficient is dynamically optimized through Kalman filtering; The meteorological data analysis layer calculates the element conformity index I based on the trend similarity corr and numerical similarity sim between real-time data and historical data. element , when I element <0.6 triggers a data anomaly warning; combines the vertical profile data of the drone with the satellite cloud map to calculate the atmospheric stratification stability index I layer , when I layer >5℃ / km triggers convective instability warning; comprehensive factor compliance index, stratification stability index and error detection results generate meteorological risk assessment value R risk , R risk ≥0.7, triggers a red alert, 0.5≤R risk <0.7, triggering an orange warning.

2. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 1, characterized in that: The ground observation layer deploys a full-element meteorological station integrating a dual-mode ultrasonic wind sensor and a self-calibrated temperature, humidity and pressure composite sensor. The ultrasonic wind sensor adopts a concave reflection ellipsoid structure, measures wind speed and direction in real time through the time difference method, and dynamically calibrates the drone and satellite data to satisfy the equation: Among them, L = 300mm ± 5% is the horizontal distance between the two foci of the ultrasonic transmitter, and H = 150mm ± 5% is the height of the concave reflector corresponding to the minor semi-axis of the ellipsoid; the ultrasonic operating frequency is 200kHz, and the signal attenuation is <3dB. The ultrasonic wind sensor has a built-in heating circuit module, and the heating is started when the temperature is <0℃ to eliminate the measurement error caused by snow cover.

3. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 1, characterized in that: The drone cluster layer is equipped with a micro-weather sounding cabin and a three-axis stabilized optoelectronic pod. The meteorological sounding cabin specifically includes: a Pt100 platinum resistance temperature sensor: a four-wire measurement circuit of ±0.2°C is used, and the temperature measurement range is -50°C to +50°C; Humidity-sensitive capacitance humidity sensor: Through the capacitance-voltage conversion circuit, it outputs 0-1V linear voltage to achieve 0-100%RH measurement, ±3%RH@≤80%RH, and the humidity data is corrected for temperature drift through the Kalman filter algorithm; Three-axis stabilized optoelectronic pod: equipped with a 2-megapixel CMOS sensor, a frame rate of 30fps, combined with a gyroscope with zero bias stability of 0.5° / achieving a stabilization accuracy of ±0.1°.

4. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 1, characterized in that: The drones in the drone cluster layer execute a three-dimensional spiral scanning path: x=r(θ)cosθ, y=r(θ)sinθ, z=kθ, θ is the pitch angle deviation angle of the drone, and the three-dimensional spiral scanning path satisfies the dynamic adjustment formula: Among them, R0 = 200m is the baseline radius, ΔP = 50m / Pa is the pressure gradient response coefficient, is the temperature gradient modulus; the vertical step coefficient k=10m / rad, the UAV uses Beidou / GPS dual-mode navigation, the positioning accuracy is ±1.5m, the flight trajectory is adjusted in real time, and the altitude profile of 0-5km is covered.

5. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 1, characterized in that: The drone data collection calibration of the drone cluster layer realizes attitude error correction through quaternion transformation and three-dimensional rotation matrix: Step 1. Input parameters: roll angle roll: -30°≤roll≤+30°, pitch angle pitc: -15°≤pitc≤+15°, horizontal speed U drone , vertical speed V drone ; Step 2: Calculate the normal vector of the inclined surface = ( x , y , z ),in: x =-so -1 (pitc), y =-so -1 (roll), z =1 Step 3: Construct the three-dimensional rotation matrix R: Step 4: Correct the wind speed component: Step 5: Output corrected wind speed and wind direction WD corrected =arctan 2(f y ,f x ), eliminating the 30° elevation error.

6. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 1, characterized in that: The platform further includes an autonomous operating system: Polarization coded QKD protocol: Based on a BB84 protocol variant, Alice and Bob’s basis vector selection probability P H =0.5,P + =0.5, 25km air-to-ground channel key rate>1kbps, bit error rate<3.8×10 -5 ; Data packet structure: 32bit quantum check code: Where D i , is the i-th binary, P i ∈{|H>,|V>,|+>,|->}; 64-bit BeiDou-3 nanosecond timestamp, timing accuracy ±3ns; LZ4 compressed meteorological data, compression ratio 4:1, end-to-end transmission delay <45ms.

7. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 6, characterized in that: The autonomous operation system further includes: 77GHz millimeter wave radar, detection distance ≥150m and ORB-SLAM algorithm, matching accuracy ≤0.1px coordination, obstacle avoidance response time <200ms; adopts polarization coding BB84 protocol variant, data contains 32bit quantum check code and Beidou-3 timestamp, field of view angle ±45°, speed resolution 0.1m / s; Doppler shift formula: v r The target radial velocity is obtained by fusing the ORB-SLAM algorithm with the IMU data, λ = 3.89 mm millimeter-wave radar operating wavelength; ORB-SLAM algorithm: extract ORB feature points, ≥1000 points per frame, matching accuracy ≤0.1 pixel, fuse IMU data, and output six-degree-of-freedom pose at 100Hz; DS Evidence Theory Fusion: Conflict Factors Obstacle avoidance response time <200ms@5m / s wind speed.

8. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 1, characterized in that: The meteorological data analysis layer specifically includes: calculating the element conformity index I based on the trend similarity corr and numerical similarity sim of real-time data and historical data element , When I element <0.6 triggers a data anomaly warning, N is the number of sampling points, corr i is the correlation coefficient of the i-th sampling point, sim i is the similarity of the i-th sampling point; Combine the vertical profile data of UAV with the satellite cloud image to calculate the atmospheric stratification stability index I layer , When I layer >5℃ / km triggers convective instability warning, temperature gradient The unit is ℃ / km. It is obtained by the Kriging interpolation algorithm of the vertical profile data of the UAV, and the interpolation resolution is ≤50m; The meteorological risk assessment value R is generated by integrating the element compliance index, stratification stability index and error detection results. risk , R risk =0.4I element +0.4I layer -0.2E error , when R risk ≥0.7 triggers a red warning, and 0.5≤R risk <0.7 triggers an orange alert.

9. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 1, characterized in that: The platform further includes a data receiving subsystem, which includes: an antenna configured to receive satellite signals; a low noise amplifier connected to the antenna and configured to amplify and filter the received signal; a down converter connected to the low noise amplifier and configured to convert the signal into an intermediate frequency signal; an indoor data receiver connected to the down converter and configured to demodulate, descramble and channel decode the intermediate frequency signal to generate demodulated and decoded data of IP group packets; a data receiving computer configured to run data receiving software, perform quick visual processing and save the demodulated and decoded data, generate LO level data and transmit it to the server; The data processing and product generation and display subsystem is configured as follows: according to the preset time scheduler, poll the data temporary storage area to obtain the files to be processed; perform quality inspection on the files, and if the inspection passes, trigger the product generation process to generate multi-channel synthetic data and vertical profiles; perform projection processing based on the area and projection type selected by the user, generate quick-view images and overlay geographic information; store the generated images and data products in the temporary storage area for the data archiving program to call.

10. A multi-sensor fusion meteorological data comprehensive analysis platform according to claim 1, characterized in that: The platform further includes a data storage and retrieval subsystem configured to: schedule storage programs according to a preset time plan; detect temporary storage areas for files and images to obtain files to be archived; perform quality checks on the files to be archived, extract metadata and store them in a database; Create a catalog and complete the archive according to the cataloging scheme based on the detector type, product type, resolution, projection method and time information; The system status monitoring and management subsystem is configured to provide scheduled data reception, product generation, data archiving and time calibration services.

Images