Atmospheric aqueous particle live detection system

Abstract

The invention discloses a live detection system and method for atmospheric water particles, the detection system is composed of a detection device, a balloon and a ground receiving system, and the detection device comprises a cylindrical shell, an inner cylinder, a Faraday ring, a photoelectric detection assembly, a signal processing unit, a data transmission unit and a power supply. Aqueous particles in the atmosphere penetrate through the inner cylinder, the Faraday ring detects the electric charge quantity, and the particle speed is inverted through the double rings. The photoelectric detection assembly captures particle images and obtains image textures. The signal processing unit integrates a cross-correlation algorithm, and a charge integration window dynamically adjusts an integration window delta t = L / v. A classification method based on fuzzy logic constructs a multi-parameter classification system through the speed (v), the particle size (d) and the scattering intensity (Z value), and establishes a particle classification model. And the data transmission unit transmits the collected charge signal waveform, particle image and environment temperature and humidity parameters by adopting a wireless data transmission transmitter and transmits the data back to a ground receiving system.

Description

Technical Field

[0001] The present invention relates to the technical field of atmospheric science detection, and particularly relates to a detection device and method for real-time monitoring of the charged characteristics, movement speed, particle size and type classification of atmospheric hydrometeor particles. Background Art

[0002] Under the atmospheric environment, hydrometeor particles (such as raindrops, ice crystals, snowflakes, graupel, hailstones, etc.) become charged due to electron transfer during collisions caused by differences in surface microstructure, surface induction polarization by the background electric field in the cloud, charge distribution according to surface area when large water droplets break in high-speed airflows, and charge separation caused by temperature differences when there are temperature gradients in different parts of ice crystals. Detecting the charged characteristics of atmospheric hydrometeor particles has important scientific value and application potential. In scientific research, it can reveal the formation mechanism of the charge structure of thunderstorm clouds, improve the coupled model of cloud microphysics and electrical processes, and study the global atmospheric electrical balance and climate change. In the application field, it can enhance lightning warning and protection capabilities, optimize the safety design of aviation and spacecraft, support artificial weather modification technology, and protect power and communication infrastructure. In terms of technology and engineering, it can promote the research and development of high-precision sensors, collaborative innovation in fields such as atmospheric electricity, microfluidics, and materials science, promote the application of intelligent algorithms (such as deep learning charge classification models) in the meteorological field, and promote multi-disciplinary cross-integration. Detecting the charge of atmospheric hydrometeors is not only the basis for understanding natural lightning phenomena, but also an important technical support for optimizing disaster warnings, ensuring critical infrastructure, and promoting climate research. With the improvement of detection accuracy and real-time performance, its application scenarios will be further extended to emerging fields such as environmental monitoring and national defense security, becoming the cross-cutting frontier of meteorological science and engineering technology.

[0003] For the detection of the charge of existing atmospheric hydrometeor particles, mainly sounding balloons, rockets, airplanes and other carrier tools are used to carry electric field sounding instruments for detection, with sounding balloons being the main one, and the charged charges of particles in the atmosphere are inferred from the detected electric field profiles. However, the detection devices of the existing technology mainly detect particle charges through Faraday rings and calculate the charge quantity by integrating current, which can only measure static charges, cannot capture dynamic particle characteristics, and cannot determine particle types. There are also some detections of thunderstorms using various detection devices such as weather radars. Dual-polarization Doppler weather radar echo data is used to analyze the charge sensitivity of different types of hydrometeor particles, obtain the charge sensitivity factors of hydrometeor particles, and invert the particle charge quantity. This method is mainly applicable to the detection of the charged characteristics of thunderstorm clouds, cannot reveal the microscopic characteristics of hydrometeor particles, and cannot classify hydrometeor particles in the atmospheric environment. Summary of the Invention

[0004] The object of the present invention is to provide an atmospheric hydrometeor particle charge detection system and method. The detection system consists of a detection device, a balloon, and a ground receiving system. The detection device includes: a cylindrical housing, an inner cylinder, a Faraday ring, a photoelectric detection component, a signal processing unit, a data transmission unit, and a power supply. The detection device is carried into the air by a balloon. Atmospheric hydrometeor particles pass through the inner cylinder, and the Faraday ring detects the charge quantity, and the particle velocity is inversed through a double ring. The photoelectric detection component captures the particle imaging and obtains the image texture. The signal processing unit integrates a cross-correlation algorithm, and the charge integration window dynamically adjusts the integration window Δ t = L / v 。Based on the fuzzy logic classification method, a multi-parameter classification system is constructed through velocity (v), particle size (d), and scattering intensity (Z value), and a particle classification model is established. The signal processing unit outputs the charge quantity, velocity, particle size, and type classification. The data transmission unit transmits the collected charge signal waveform, particle image, and environmental temperature and humidity parameters by using a wireless data transmitter and returns them to the ground receiving system. The ground receiving system uses a wireless data receiver to receive and store the data.

[0005] The present invention is realized through the following technical solutions: An atmospheric hydrometeor particle charge detection system consists of a detection device 100, a balloon 200, and a ground receiving system 300. The detection device 100 includes: a cylindrical housing 1, an inner cylinder 2, a Faraday ring 3, a photoelectric detection component 4, a signal processing unit 5, a data transmission unit 6, and a power supply 7. The inner cylinder 2 is concentrically installed in the housing 1. Particle inlets 21 and outlets 22 are provided at both axial ends of the inner cylinder 2. Anti-static coatings are provided at the particle inlets 21 and outlets 22 to avoid charge accumulation from interfering with the signal, and a tapered opening 23 is provided at the inlet 21 to prevent particles from contacting the inner wall of the inner cylinder 2.

[0006] The Faraday ring 3 includes a first Faraday ring 3a, a second Faraday ring 3b, and a third Faraday ring 3c. The three rings are arranged at intervals along the axial direction of the inner cylinder 2.

[0007] The photoelectric detection component 4 includes a light source module 41, a high-speed imaging module 42, and a reflection prism 43. The optical path is turned by 90° through the reflection prism 43 to form an orthogonal light curtain with the high-speed imaging module 42 to capture particle imaging.

[0008] The signal processing unit 5 integrates a cross-correlation algorithm, and the charge integration window dynamically adjusts the integration window Δ t = L / v ,A particle classification model is established, and the charge quantity, velocity, particle size, and type classification are output.

[0009] The described data transmission unit 6 transmits the collected charge signal waveforms, particle images, and environmental temperature and humidity parameters using a wireless data transmitter and sends them back to the ground receiving system 300. The ground receiving system 300 receives the data using a wireless data receiver and stores it. The ground receiving system 300 can synchronously generate a charge density heat map.

[0010] The described detection device 100 is carried into the air by a balloon 200.

[0011] The described housing 1 is made of a polymer composite material. The lower annular space between the housing 1 and the inner cylinder 2 is the equipment compartment 11, which internally houses an optoelectronic detection component 4, a signal processing module 5, a data transmission unit 6, and a power supply 7. The upper annular space is the parachute compartment 12. The top of the parachute compartment 12 is provided with a parachute compartment cover 13, and four suspension rings 14 are provided on the parachute compartment cover 13 for the balloon 200 to suspend the detection device 100. A parachute 15 is provided in the parachute compartment 12, and the parachute compartment cover 13 is controlled by a parachute opener 16.

[0012] The described inner cylinder 2 adopts a double-layer shielding structure. The inner layer is made of Teflon insulating material, and the outer layer is a composite coating conductive shielding layer for isolating external electromagnetic interference. The inner diameter of the inner cylinder 2 is 50 - 100 mm, and the length-to-diameter ratio is 3:1 - 5:1 to ensure the linearity of the particle passing path. The inner cylinder 2 consists of four sections, which are the first inner cylinder 2a, the second inner cylinder 2b, the third inner cylinder 2c, and the fourth inner cylinder 2d from top to bottom. The second inner cylinder 2b is provided with a transparent cylinder 24, and the transparent cylinder 24 is made of PMMA material for the optical path transmission of the optoelectronic detection component 4.

[0013] The described Faraday ring 3 is made of high-purity copper or gold-plated copper. The Faraday ring 3 is embedded in the cylinder wall at the joint of the inner cylinder 2 sections. The thickness of the ring body is 0.5 - 1.0 mm. The distance between the first Faraday ring 3a and the second Faraday ring 3b L 1 = 100 - 200 mm, and the distance between the second Faraday ring 3b and the third Faraday ring 3c L 2 = 50 - 100 mm. When calculating the particle velocity v, there are L 1, L 2, L 1 + L 2 three distances. The first Faraday ring 3a is installed on the upper side of the optoelectronic detection component 4, and the second Faraday ring 3b and the third Faraday ring 3c are installed on the lower side of the optoelectronic detection component 4. Each ring is connected to a low-noise charge amplifier. The gain of the first Faraday ring 3a is 3 - 5 times that of the second Faraday ring 3b, and the gain of the third Faraday ring 3c is 5 - 10 times that of the second Faraday ring 3b.

[0014] The light source module 41 adopts a white parallel backlight source, the high-speed imaging module 42 is equipped with a global shutter CMOS camera, and the reflective prism 43 is a prism, which is arranged between the light path of the light source module 41 and the high-speed imaging module 42. The photoelectric detection component 4 also includes an orthogonal APD photodetector 44 and a rear APD photodetector 45. The orthogonal APD photodetector 44 and the light path of the light source module 41 are 90°, and the rear APD photodetector 45 and the light path of the light source module 41 have an angle α=30~60°. The photoelectric detection component 4 is combined with backlight illumination, and an orthogonal light curtain is formed by a reflective prism to trigger particle imaging, capture the particle morphology and motion trajectory shadow image, trigger the particle passing event and synchronize the imaging clock. The parallel light of the light source module 41 and the CMOS camera of the high-speed imaging module 42 are synchronously triggered through the first Faraday ring 3a, and the trigger signal is isolated from the charge amplifier by an optical coupler.

[0015] The signal processing unit 5 includes a GNSS satellite navigation receiving device 51, a charge amplifier 52, an ADC analog-to-digital converter 53, and an operation module 54. The GNSS satellite navigation receiving device 51 is used for positioning the detection device 100, detecting signal synchronization, and measuring the acceleration of the detection device 100. The charge amplifier 52 amplifies the charge detected by the Faraday ring 3. The ADC analog-to-digital converter 53 performs analog-to-digital conversion on the output data of the charge amplifier 52, the detection data of the orthogonal APD photodetector 44 and the rear APD photodetector 45, and the output digital signal is sent to the operation module 54. The operation module 54 calculates the charge amount Q and the charge density. ρ Calculation, particle speed v calculate.

[0016] The ascent speed of the balloon 200 is controlled by net lifting force feedback. The helium filling amount is adjusted according to the real-time air pressure and temperature. The ascent speed of the balloon is determined by the net lifting force, air density and resistance. The helium filling amount of the balloon is adjusted according to the expected ascent speed, the volume of the balloon is changed, and the cross-sectional area is controlled according to the shape of the balloon. The calculation formula is: , in, A is the cross-sectional area of the balloon, F is the net lifting force (total buoyancy minus the weight of the balloon and attachments), γ is the air density, C d is the drag coefficient, v p is the expected balloon ascent speed. As the pressure of balloon 200 decreases, its volume gradually expands, and its cross-sectional area AAs it expands, the balloon 200 is provided with a deflation valve 201. According to the required ascending speed and detection height requirements, the signal processing unit 5 controls the opening or closing of the deflation valve 201 based on the ascending speed detected by the GNSS satellite navigation receiving device 51, controls the cross-sectional area of the balloon 200 to control the ascending speed. When the detection device 100 reaches the predetermined end detection height, the deflation valve 201 is opened for deflation, and the net lifting force F is negative (the total buoyancy is less than the weight of the balloon itself and the additional objects), the balloon 200 gradually descends. When the GNSS satellite navigation receiving device 51 detects that the detection device 100 lands at the preset parachute opening height, the signal processing unit 5 controls the parachute opener 16 to act, the parachute compartment cover 13 is separated from the housing 1, the detection device 100 falls by its own weight, the parachute 15 is opened, and the detection device 100 slowly descends to the ground.

[0017] An atmospheric hydrometeor particle charge detection method of the present invention includes charge quantity calculation, image texture acquisition, and particle classification model establishment. The charge quantity calculation steps are as follows: S1. Signal acquisition and preprocessing; S11. Signal acquisition and amplification. Use a double Faraday ring 3a / 3b or 3b / 2c or 3a / 3c to collect the charge signal of the particle passing through the detection area, and realize the two-channel signal synchronous trigger through the GNSS second pulse, with a time error <50 ns and a sampling rate ≥1 MHz. The collected weak electrical signal is amplified by a charge amplifier circuit; S12. Denoising and filtering. Use a sliding window to perform time-domain normalization on the signal to suppress high-energy noise interference, and filter out non-target frequency band noise through a Butterworth low-pass filter to improve the signal-to-noise ratio. The transfer function of the Butterworth filter is: , Where: s is the complex frequency variable, ω c is the cut-off frequency, n is the order of the filter, which determines the steepness of the filter.

[0018] S2. Signal synchronization. The light source module 41, the high-speed imaging module 42, the orthogonal / rear APD detectors 44 / 45, and the three Faraday rings 3a / 3b / 3c are synchronized by the GNSS second pulse; S21. Time reference alignment. The exposure time window of the global shutter CMOS camera is synchronized with the pulse signal of the APD detector to avoid motion blur or signal misalignment; S22. Charge signal synchronization. The charge integration window of the Faraday ring 3 needs to match the time when the particle passes through the detection area. Use the GNSS second pulse to trigger all devices to ensure that the time stamp alignment error <1 μ s; S23. Data fusion: Align the pixel data of high-speed imaging with the signal time dimension of the APD to achieve multi-modal data correlation analysis.

[0019] S3. Cross-correlation calculation: Determine the time delay (Δt) between the signals of the Faraday ring 3 and the signal correlation between the orthogonal APD photodetector 44 and the rear APD photodetector 45 and the camera of the high-speed imaging module 42. S31. Time-domain cross-correlation calculation: , where x n and y n are two signals, k is the number of sampling points of the time delay; S32. FFT-accelerated frequency-domain optimization: Perform fast Fourier transform (FFT) on the two time-domain signals x n and y n respectively to obtain their frequency-domain representations: , , where: X ( f ) and Y ( f ) are the frequency-domain representations of the signals x n and y n respectively, f is the frequency variable. Cross-power spectrum calculation: The cross-power spectrum S xy ( f ) is the complex conjugate product of X ( f ) and Y ( f ) in the frequency domain: , where: Y ∗ ( f ) is the complex conjugate of Y ( f ). Perform inverse FFT (IFFT) on the cross-power spectrum S xy ( f ) to obtain the cross-correlation function R xy τ in the time domain: ​​​​​​​ , Among them, R xy τ is the cross-correlation function, representing the correlation between two signals, τ is the time-delay variable.

[0020] S4. Particle velocity calculation; S41. Calculate the relative velocity of the particle with respect to the detection device 100 based on the time delay between two Faraday rings and the distance between the two Faraday rings v = L / Δ t Among them, L is the distance between the two Faraday rings ( L 1 or L 2 or L 1 + L 2), Δ t is the time delay of the signals of the two Faraday rings. There are three Δ t corresponding to the three distances, which are Δ t 1 、 Δ t 2 、 Δ t 1+2 . The three calculated relative velocities v 1 、v 2 、v 3. Take the arithmetic mean or the median as the particle relative velocity v ; S42. For the rising speed of the detection device 100, use GNSS Doppler frequency shift or carrier phase change rate to solve the three-dimensional velocity of the detection device 100, and the vertical direction is the rising speed of the detection device 100; S43. Correction of the particle velocity. When the rising direction of the detection device 100 is opposite to the particle movement direction, the actual velocity of the particle needs to be corrected by vector superposition: , Among them, v c is the corrected particle velocity, v is the calculated particle velocity, v b is the rising speed of the detection device 100.

[0021] S5. Charge integration window dynamic adjustment calculation. Through the corrected particle velocity v c and the distance between the two Faraday rings L , the integration time window is dynamically adjusted according to the particle velocity and the rising speed. The adjusted time delay Δ t int ​The formula is as follows: .

[0022] S6. Charge density inversion, the amount of charge within the integration window Q Passing current I ( t ) The integration calculation in the time interval t 1, t 2]: , where: Q is the amount of charge within the integration window, I ( t ) is the current varying with time, t 1, t 2 is the integration time window; Charge density ρ Passing the integrated charge Q , the corrected particle velocity v c and the cross-sectional area S Calculate: , where: ρ is the charge density, Q is the amount of charge within the integration window, v c is the corrected particle velocity, S is the area of the Faraday ring.

[0023] The steps for obtaining the image texture are as follows: S1. System synchronization and optical path triggering; S11. Synchronization of the light source and imaging. The white parallel backlight of the light source module 41 and the global shutter CMOS camera of the high-speed imaging module 42 are hardware-synchronized through the GNSS clock to ensure that the timing error between the optical pulse and the camera exposure < 1 μ s, achieving spatio-temporal alignment of the particle motion trajectory and the charge signal; S12. Trigger the parallel backlight light curtain. When the particle passes through the parallel backlight source, an orthogonal light curtain (90°) and an inclined light curtain (30 - 60°) are formed by reflection. The orthogonal APD photodetector 44 triggers the imaging clock, and the rear APD photodetector 45 verifies the particle passing event, forming a dual-light-curtain redundant trigger mechanism to reduce the false detection rate.

[0024] S2. Shadow image capture and preprocessing; S21. High-frame-rate imaging: The global shutter CMOS camera of the high-speed imaging module 42 captures the particle shadow image at a frame rate of ≥100fps, and the resolution supports the morphological details of particles with a size of 2 μ m. S22. Dynamic denoising: By matching the light source pulse width ( μ s level) with the CMOS exposure time, motion blur is suppressed. Median filtering and histogram equalization are performed in real time through the operation module 54 to enhance texture contrast; S23. Super-resolution reconstruction: For low signal-to-noise ratio images, an improved Real-ESRGAN network is used. By integrating multi-channel gradient information through the structure tensor (ST) branch, the microscopic texture details on the particle surface are restored.

[0025] S3. Texture feature extraction; S31. Using the gray-level co-occurrence matrix (GLCM), for the gray-level distribution of the particle shadow image, contrast, energy, and entropy statistics are calculated to quantify the surface uniformity; S32. Through local binary pattern (LBP), analyze the neighborhood coding histogram of the particle edge to distinguish organic particles (random texture) from crystal particles (periodic texture); S33. Using multi-scale Gabor filtering, extract frequency domain features through a multi-directional and multi-scale Gabor filter bank to identify microscopic fluctuations in the deposition path of charged particles (such as the spiral trajectory of biological particles).

[0026] S4. Spatiotemporal feature fusion; S41. Dynamic trajectory modeling: Superimpose the charge signal Q on the consecutive frames of high-speed imaging. Calculate the particle motion speed v through the optical flow method, and combine the particle size d to construct a four-dimensional feature vector (Q, v, d, texture entropy); S42. Optoelectronic signal alignment: The charge amplifier and the APD trigger signal are isolated by an optocoupler to avoid electromagnetic interference. Use timestamps to match the charge pulse waveform with the shadow image sequence, and correlate the charge density ρ with the texture parameters.

[0027] For the establishment of the particle classification model, a classification method based on fuzzy logic constructs a multi-parameter classification system through velocity (v), particle size (d), and scattering intensity (Z value). The steps are as follows: S1. Definition of membership function; S11. Divide the fuzzy sets according to the particle motion characteristics. Use trapezoidal or triangular membership functions to divide them into "low speed" 0 - 5 m / s (the peak membership degree is at 2 m / s), "medium speed" medium speed: 3 - 15 m / s (the peak membership degree is at 8 m / s), and "high speed" 10 - 30 m / s (the peak membership degree is at 20 m / s); S12. Based on the optical imaging of a global shutter CMOS camera, it is divided into "small particles" (<0.1 mm), "medium particles" (0.1 - 2 mm), and "large particles" (>2 mm) according to the particle size (d), and a Gaussian membership function is used for smooth transition; S13. Particles in different phases (such as rain, snow, hail) have different light scattering abilities. The scattering intensities of the orthogonal APD photodetector 44 and the rear APD photodetector 45 are divided into weak, medium, and strong. The weak scattering is a trapezoidal membership function, corresponding to clouds or small ice crystals; the medium scattering is a triangular membership function, corresponding to raindrops or wet snow; the strong scattering is a trapezoidal membership function, corresponding to hail or dry sleet;

[0028] S2. Construction of the fuzzy rule base. According to physical observations and experimental data, if - then rules are designed. IF speed = high speed AND particle size = large AND reflectivity = strong, THEN particle type = hail. IF speed = medium speed AND particle size = medium AND reflectivity = medium, THEN particle type = raindrop. IF speed = low speed AND particle size = small AND reflectivity = weak, THEN particle type = aerosol or cloud particle (AERO).

[0029] S3. Fuzzy inference and classification. Based on the membership function in fuzzy logic, physical quantities are converted into membership degree values from 0 to 1 through function mapping; The membership degree of the triangular function is calculated as follows: , where a, b, and c are the left endpoint, vertex, and right endpoint of the triangular membership function respectively; The membership degree of the trapezoidal function is calculated as follows: , where a, b, c, and d are the left boundary, left vertex, right vertex, and right boundary of the trapezoidal membership function respectively.

[0030] S4. Optimization of logical operations; S41. Adopt MIN - MAX logic. Take the minimum membership degree of multi - condition rules as the rule trigger intensity, and finally take the maximum value of the outputs of each rule; S42. Introduce weight assignment. The weight of scattering intensity for hail classification is higher than that of speed, which is achieved by adjusting the shape of the membership function or the rule priority; S43. Calibrate parameters according to experimental data to achieve the best classification effect and provide type library data for the detection of electrically charged particles.

[0031] The beneficial effects of the present invention are as follows: By using three Faraday rings 3a / 3b / 3c with different gains for segmented detection and combining with optimized spacing (L1 = 100 - 200 mm, L2 = 50 - 100 mm), the particle velocity ( v ), and the electric charge amount ( Q ) are calculated through the cross-correlation algorithm, reducing the velocity measurement error and improving the charge sensitivity. According to the particle velocity ( v c ), the charge integration time (Δ tint = L / v c ) is dynamically adjusted to avoid the charge signal leakage of high-speed particles or the noise interference of low-speed particles. The double-light curtain triggering mechanism of the orthogonal APD and the rear APD, combined with the GNSS second pulse synchronization, ensures the spatio-temporal alignment of the charge signal and the image data. The global shutter CMOS camera combines with the Real-ESRGAN super-resolution reconstruction to extract the microscopic texture on the particle surface. Based on the velocity (v), particle size (d), and reflectivity (Z), a multi-parameter membership function is constructed, and the weight allocation is optimized by combining the MIN-MAX rule to accurately classify hail, wet snow, raindrops, etc. By correlating the charge density ( ρ ), particle size (d), velocity ( v ), and texture entropy through the optical flow method, a dynamic trajectory model is constructed to support the accurate identification of complex particles. Through multi-sensor fusion, dynamic signal processing, and intelligent classification algorithms, the present invention solves the core problems such as insufficient dynamic response, low classification accuracy, and poor environmental adaptability in the prior art, and realizes the high-precision, multi-parameter, and real-time detection of the charged characteristics of atmospheric hydrometeor particles. Its technical advantages have significant application values in the fields of lightning warning, artificial weather modification, aviation safety, etc., and at the same time provide reliable data support for atmospheric electricity and climate research. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 Schematic diagram of the system of the present invention; Figure 2 Longitudinal sectional view of the photoelectric detection component and the power supply position of the detection device ( Figure 4 A - A section of Figure 3 ); Figure 4 Longitudinal sectional view of the signal processing unit and the data transmission unit position of the detection device ( Figure 4 B - B section of Figure 5 ); Figure 6 Transverse sectional view of the equipment bin position of the detection device; Figure 7 Transverse sectional view of the umbrella bin position of the detection device;

[0033] In the figure: 1 - housing, 11 - equipment compartment, 12 - parachute compartment, 13 - parachute compartment cover, 14 - suspension ring, 15 - parachute, 16 - parachute opener, 17 - positioning groove, 18 - slider, 2 - inner cylinder, 2a - first inner cylinder, 2b - second inner cylinder, 2c - third inner cylinder, 2d - fourth inner cylinder, 21 - incident port, 22 - exit port, 23 - tapered port, 3 - Faraday ring, 3a - first Faraday ring, 3b - second Faraday ring, 3c - third Faraday ring, 4 - photoelectric detection component, 41 - light source module, 42 - high-speed imaging module, 43 - reflecting prism, 44 - orthogonal APD photodetector, 45 - rear APD photodetector, 5 - signal processing unit, 51 - GNSS satellite navigation receiving device, 52 - charge amplifier, 53 - ADC analog-to-digital converter, 54 - operation module, 6 - data transmission unit, 7 - power supply, 71 - thermal insulation layer, 72 - heating compartment, 100 - detection device, 200 - balloon, 201 - deflation valve, 300 - ground receiving system. Detailed implementation manner

[0034] For better understanding of the present invention by those skilled in the art, in combination with Figures 1 to 7 the present application is further described. In the description of this specification, the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship commonly understood by those skilled in the art. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the components or parts referred to must have a specific orientation, be constructed and operated in a specific orientation. The content mentioned in the implementation manner is not a limitation to the present invention. The terms "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0035] The charged detection system for atmospheric hydrometeor particles of the present invention is composed of a detection device 100, a balloon 200, and a ground receiving system 300. The system schematic diagram is shown in Figure 1 . The detection device 100 includes: a cylindrical housing 1, an inner cylinder 2, a Faraday ring 3, a photoelectric detection component 4, a signal processing unit 5, a data transmission unit 6, and a power supply 7. The housing 1 is made of a polymer material. The inner cylinder 2 is concentrically installed inside the housing 1. The particle incident port 21 and the exit port 22 are provided at both axial ends of the inner cylinder 2. The particle incident port 21 and the exit port 22 are provided with an anti-static coating to avoid charge accumulation interfering with the signal. The incident port 21 is provided with a tapered port 23 to avoid particles contacting the inner wall of the inner cylinder 2. See Figure 2 , Figure 3 . The lower annular space between the housing 1 and the inner cylinder 2 is the equipment compartment 11, which internally houses the photoelectric detection component 4, the signal processing module 5, the data transmission unit 6, and the power supply 7. See Figure 4。The upper annular space is the parachute compartment 12. A parachute compartment cover 13 is provided at the top of the parachute compartment 12. Four suspension rings 14 are provided on the parachute compartment cover 13 for the balloon 200 to suspend the detection device 100. A parachute 15 is arranged in the parachute compartment 12. The parachute compartment cover 13 is controlled by a parachute opener 16, see Figure 2 、 Figure 5 。Two positioning grooves 17 are provided at the inner edge of the housing 1 of the parachute compartment 12. Sliders 18 are provided at the positions of the parachute compartment cover 13 corresponding to the positioning grooves 17. The sliders 18 are embedded in the positioning grooves 17. The parachute compartment cover 13 is locked by combining with the parachute opener 16, and the parachute compartment cover 13 is firmly combined with the housing 1. Metal contacts are provided on the contact surfaces of the positioning grooves 17 and the sliders 18. The contacts of the two sliders 18 are respectively connected to the control lines of the air release valve 201, and the contact of the positioning groove 17 is connected to the signal processing unit 5.

[0036] The detection device 100 is carried into the air by the balloon 200. The lifting speed of the balloon 200 is controlled by net lift feedback. The helium filling amount is adjusted according to the real-time air pressure and temperature. The rising speed of the balloon is determined by the net lift, air density and resistance. The helium filling amount of the balloon is adjusted according to the expected required rising speed, the volume of the balloon is changed, and the cross-sectional area is controlled according to the shape of the balloon. The calculation formula is: , where, A is the cross-sectional area of the balloon, F is the net lift (total buoyancy minus the weight of the balloon itself and additional objects), γ is the air density, C d is the drag coefficient, v p is the expected rising speed of the balloon. As the balloon 200 rises, the volume gradually expands with the decrease of air pressure, and the cross-sectional area A expands accordingly. An air release valve 201 is provided on the balloon 200, see Figure 1 。According to the required rising speed and detection height requirements, the signal processing unit 5 controls the opening or closing of the air release valve 201 according to the rising speed detected by the GNSS satellite navigation receiving device 51. The air release valve 201 is in a closed state under normal conditions. The air release valve 201 controls the cross-sectional area of the balloon 200 to control the rising speed. When the rising speed exceeds the expected rising speed of the balloon, the air release valve 201 opens to release air, and the volume of the balloon 200 shrinks and the net lift F decreases. When the rising speed meets the expected rising speed of the balloon, the air release valve 201 closes. When the detection device 100 reaches the predetermined end detection height, the air release valve 201 opens to release air, and the net lift FWhen it is negative (the total buoyancy is less than the self-weight of the balloon and the weight of the additional objects), the balloon 200 gradually descends. During the descent process, the descent speed of the balloon 200 can still be controlled by the air release valve 201, and particle charge detection can still be carried out during the descent. When the GNSS satellite navigation receiving device 51 detects that the detection device 100 has descended to the preset parachute opening altitude (generally 300 - 500 from the ground), the signal processing unit 5 controls the parachute opener 16 to act. The parachute compartment cover 13 separates from the housing 1, and the detection device 100 falls freely by its own weight, opening the parachute 15. The detection device 100 slowly descends to the ground without causing danger to the ground due to high-speed falling. If the ground receiving system 300 can still receive the coordinate position transmitted back by the data transmission unit 6, the detection device 100 can be recovered if conditions permit. At this time, the helium gas in the balloon 200 has been basically emptied. After the detection device 100 separates, the balloon 200 slowly descends under the action of air resistance.

[0037] The inner cylinder 2 adopts a double-layer shielding structure. The inner layer is made of Teflon insulating material, and the outer layer is a composite coating conductive shielding layer, which is used to isolate external electromagnetic interference. The inner diameter of the inner cylinder 2 is 50 - 100 mm, and the length-diameter ratio is 3:1 - 5:1 to ensure the linearity of the particle passing path. The inner cylinder 2 is composed of four sections, which are the first inner cylinder 2a, the second inner cylinder 2b, the third inner cylinder 2c, and the fourth inner cylinder 2d from top to bottom. The second inner cylinder 2b is provided with a transparent cylinder 24, and the transparent cylinder 24 is made of PMMA material and is used for the optical path transmission of the photoelectric detection component 4.

[0038] The Faraday ring 3 is composed of three rings, including the first Faraday ring 3a, the second Faraday ring 3b, and the third Faraday ring 3c. The three rings are arranged at intervals along the axial direction of the inner cylinder 2. The Faraday ring 3 is made of high-purity copper or gold-plated copper ring. The thickness of the ring body of the Faraday ring 3 is 0.5 - 1.0 mm, and it can be bent from high-purity copper or gold-plated copper with a thickness of 0.5 mm and a width of 1 mm. The butt joint is welded and led out by the inner core wire of the shielded wire, and the outer lead wire of the shielded wire is connected to the composite coating conductive shielding layer of the inner cylinder 2. The Faraday ring 3 is embedded in the cylinder wall at the joint of the inner cylinder 2 sections. The Faraday ring 3 is enclosed inside the inner cylinder 2 and will not contact the particles passing through the inner cylinder 2. The distance between the first Faraday ring 3a and the second Faraday ring 3b L 1 = 100 - 200 mm, and the distance between the second Faraday ring 3b and the third Faraday ring 3c L 2 = 50 - 100 mm. When calculating the particle velocity v, there is L 1. L 2. L 1 + L 2 three distances. The optimized distance setting can adapt to different particle velocities. The first Faraday ring 3a is installed on the upper side of the photoelectric detection component 4, and the second Faraday ring 3b and the third Faraday ring 3c are installed on the lower side of the photoelectric detection component 4.

[0039] The optoelectronic detection component 4 includes a light source module 41, a high-speed imaging module 42, and a reflecting prism 43. The light path is folded by 90° through the reflecting prism 43 to form an orthogonal light curtain with the high-speed imaging module 42 to capture particle imaging. The light source module 41 uses a white parallel backlight. The high-speed imaging module 42 is equipped with a global shutter CMOS camera. The reflecting prism 43 is a triangular prism, and the triangular prism is arranged between the light paths of the light source module 41 and the high-speed imaging module 42, see Figure 2 . The light source emitted by the light source module 41 passes through the transparent cylinder 24 and passes through the inner cylinder 2, and is reflected by 90° through the reflecting prism 43 to reach the high-speed imaging module 42, increasing the light path to adapt to the focal length of the global shutter CMOS camera. The global shutter CMOS camera is an imaging device that uses synchronous exposure technology. Each pixel of the global shutter CMOS camera is equipped with an independent storage unit. Its core feature is that all pixels start and end exposure at the same time, and transfer the charge to the storage area after the exposure ends, and then read out the signal line by line, effectively solving the problem of image deformation of traditional rolling shutters when shooting high-speed moving objects. The global shutter CMOS camera preferably selects Sony IMX925, and its high frame rate and low noise characteristics can accurately analyze the particle motion trajectory.

[0040] The optoelectronic detection component 4 further includes an orthogonal APD photodetector 44 and a rear APD photodetector 45. The orthogonal APD photodetector 44 is at a 90° angle to the light path of the light source module 41, and the angle α between the rear APD photodetector 45 and the light path of the light source module 41 is 30° - 60°, see Figure 4 . The orthogonal APD photodetector 44 and the rear APD photodetector 45 use avalanche photodiodes. During the detection process, the parallel backlight of the light source module 41 is always on. The optoelectronic detection component 4 combines backlight illumination. When the particle passes through the parallel backlight source, the parallel light source is blocked, and an orthogonal light curtain is formed through the reflecting prism to trigger particle imaging, capturing the particle morphology and the shadow image of the motion trajectory, see Figure 6 the shadow, triggering the particle to pass through the event and synchronizing the imaging clock. The parallel light of the light source module 41 and the CMOS camera of the high-speed imaging module 42 are synchronously triggered through the first Faraday ring 3a, and the trigger signal and the charge amplifier are isolated by an optocoupler. The position of the particle passing through the inner cylinder 2 is uncertain and the shape is irregular ( Figure 6 drawn as a dot only for illustration), and there may be a blind area in the angle of reflecting the parallel backlight. In order to reduce the false detection rate, the particle forms an orthogonal light curtain (90°) and an inclined light curtain (30° - 60°) in the backlight illumination reflection. The orthogonal APD photodetector 44 triggers the imaging clock, and the rear APD photodetector 45 verifies the particle passing through the event, forming a dual-light-curtain redundant trigger mechanism, see Figure 6 , Figure 6 The dashed rectangular box in

[0041] The signal processing unit 5 includes a GNSS satellite navigation receiving device 51, a charge amplifier 52, an ADC analog-to-digital converter 53, and an operation module 54. The GNSS satellite navigation receiving device 51 is used to locate the detection device 100, detect signal synchronization, and determine the acceleration of the detection device 100. The charge amplifier 52 amplifies the charge detected by the Faraday ring 3. The three Faraday rings 3 correspond to three charge amplifiers 52 respectively. The Faraday rings 3 are connected to the reverse input end of the charge amplifier 52, and the outer shielding layer of the Faraday ring 3 is connected to the same direction input end of the charge amplifier 52. The shielding layer is grounded. Figure 7 The charge amplifier 52 uses the ADA4530-1 charge amplifier chip, supports charge and voltage dual input modes, has a gain range of 0.01~1000 mV / pC, and has built-in adjustable high-pass (0.3~100Hz) and low-pass (0.3~100kHz) filters, optimizes the signal-to-noise ratio, and has low noise (≤5μV), which is suitable for weak signal amplification. The three Faraday rings 3 are respectively connected to the three charge amplifiers 52, the gain of the first Faraday ring 3a is 3~5 times that of the second Faraday ring 3b, and the gain of the third Faraday ring 3c is 5~10 times that of the second Faraday ring 3b.

[0042] The charge amplifier 52, the orthogonal APD photodetector 44 and the rear APD photodetector 45 output analog signals, which need to be converted into digital signals. The ADC analog-to-digital converter 53 performs analog-to-digital conversion on the output data of the charge amplifier 52, the detection data of the orthogonal APD photodetector 44 and the rear APD photodetector 45, and the output digital signals are sent to the operation module 54. The three Faraday rings 3 are amplified by the three charge amplifiers 52 and output three analog signals. The orthogonal APD photodetector 44 and the rear APD photodetector 45 output two signals. The ADC analog-to-digital converter 53 can use the AD7606C conversion chip, which has 8 analog input channels, 16-bit resolution, and a sampling rate of 1MSPS. It has good performance and can quickly sample and convert multiple signals in sequence to achieve synchronous sampling of multiple channels.

[0043] Signal processing unit 5, integrated cross-correlation algorithm, calculation module 54 performs charge Q Calculation, charge density ρ Calculation, particle speed v Calculation. Charge integration window dynamically adjusts the integration window Δ t = L / v , the particle classification model is established, and the charge, velocity, particle size and type classification are output.

[0044] The data transmission unit 6 uses frequency shift keying (FSK) modulation to modulate signals, converting binary data into FSK signals. The modulated signals are transmitted by the transmitter and use the dedicated meteorological sounding frequency band of 403 MHz for real-time wireless communication to transmit sounding data. After being received on the ground, the received FSK signals are demodulated back into binary data. The data transmission unit 6 transmits the collected charge signal waveforms, particle images, and environmental temperature and humidity parameters using a wireless data transmitter and sends them back to the ground receiving system 300. The ground receiving system 300 uses a wireless data receiver to receive and store the data, and the ground receiving system 300 can synchronously generate a charge density thermogram.

[0045] The power supply 7 is powered by a lithium battery. Considering that lithium batteries cannot work properly in low-temperature environments and have low self-heating during discharge, and the designed operating environment temperature of the detection system can reach as low as -50°C at the lowest, therefore, a thermal insulation layer 71 is installed outside the battery compartment, and a heating chamber 72 for placing an auxiliary heat source is left outside the thermal insulation layer 71 for placing a heating agent, as shown in Figure 2 、 Figure 4 。Before releasing the detection device 100, ferric oxide material is added to the heating chamber 72 as an auxiliary heat source. Ferric oxide can react chemically with oxygen in the air to become ferric tetroxide and release heat. This process is slow and persistent, and can meet the need to maintain the temperature inside the battery compartment. This heating device is light in weight and does not require additional power consumption, making it suitable for use in a radiosonde.

[0046] The method for detecting charged atmospheric hydrometeor particles of the present invention includes charge quantity calculation, image texture acquisition, and particle classification model establishment. The steps for calculating the charge quantity are as follows: S1. Signal acquisition and preprocessing; S11. Signal acquisition and amplification. Use a double Faraday ring 3a / 3b or 3b / 2c or 3a / 3c to collect the charge signals of particles passing through the detection area, and use the GNSS second pulse to achieve two-way signal synchronous triggering with a time error < 1 ns and a sampling rate ≥ 1 MHz (matching the particle speed range of 0.1 - 100 m / s). The collected weak electrical signals are amplified using a charge amplifier circuit. S12. Denoising and filtering. Use a sliding window to perform time-domain normalization on the signal to suppress high-energy noise interference, and filter out non-target frequency band noise through a Butterworth low-pass filter to improve the signal-to-noise ratio. The transfer function of the Butterworth filter is: , Where: s is the complex frequency variable, ω c is the cut-off frequency, n is the order of the filter, which determines the steepness of the filter.

[0047] S2. Signal synchronization. The light source module 41, the high-speed imaging module 42, the orthogonal / rear APD detectors 44 / 45, and the three Faraday rings 3a / 3b / 3c are synchronized using the GNSS second pulse. S21. Time reference alignment. The exposure time window of the global shutter CMOS camera is synchronized with the pulse signal of the APD detector to avoid motion blur or signal misalignment. S22. Charge signal synchronization. The charge integration window of the Faraday ring 3 needs to match the time when the particle passes through the detection area. All devices are triggered using the GNSS second pulse to ensure that the timestamp alignment error is <1 μ s. S23. Data fusion. The pixel data of the high-speed imaging is aligned with the signal time dimension of the APD to achieve multimodal data correlation analysis.

[0048] S3. Cross-correlation calculation. Determine the time delay (Δt) between the signals of the Faraday ring 3 and the signal correlation between the orthogonal APD photodetector 44 and the rear APD photodetector 45 and the high-speed imaging module 42 camera. S31. Time-domain cross-correlation calculation: , where x n and y n are two signals, k is the number of sampling points of the time delay; S32. FFT-accelerated frequency-domain optimization. Perform fast Fourier transform (FFT) on the two time-domain signals x n and y n respectively to obtain their frequency-domain representations: , , where: X ( f ) and Y ( f ) are the frequency-domain representations of the signals x n and y n respectively, f is the frequency variable. Cross-power spectrum calculation: The cross-power spectrum S xy ( f ) is in the frequency domain X ( f ) and Y ( f ​​​​​​Complex conjugate product of ( , where: Y ∗ ( f ) is Y ( f )'s complex conjugate. For the cross-power spectrum S xy ( f ), perform inverse FFT (IFFT) to obtain the cross-correlation function R xy τ : , where R xy τ is the cross-correlation function, representing the correlation between two signals, τ is the time-delay variable.

[0049] S4. Particle velocity calculation; S41. Calculate the relative velocity of the particle with respect to the detection device 100 through the time delay between two Faraday rings and the distance between the two Faraday rings v = L / Δ t , where L is the distance between the two Faraday rings ( L 1 or L 2 or L 1 + L 2), and Δ t is the time delay of the two Faraday ring signals. The three distances correspond to three Δ t , which are Δ t 1 、 Δ t 2 、 Δ t 1+2 . Calculate the three obtained relative velocities v 1 、v 2 、v 3, and use the arithmetic mean or the median as the particle relative velocity v ; S42. For the rising speed of the detection device 100, use GNSS Doppler frequency shift or carrier phase change rate to solve the three-dimensional velocity of the detection device 100, and the vertical direction is the rising speed of the detection device 100; S43. Correction of the particle velocity. If the rising direction of the detection device 100 is opposite to the particle movement direction, the actual velocity of the particle needs to be corrected by vector superposition: ,​​ Among them, v c is the corrected particle velocity, v is the calculated particle velocity, v b is the rising velocity of the detection device 100.

[0050] S5. Dynamically adjust and calculate the charge integration window, through the corrected particle velocity v c and the distance between the two Faraday rings L , the integration time window is dynamically adjusted according to the particle velocity and the rising speed, and the adjustment time delay Δ t int The formula is: .

[0051] S6. Charge density inversion, the charge quantity within the integration window Q is obtained through the current I ( t ) within the time interval t 1, t 2] by integration calculation: , where: Q is the charge quantity within the integration window, I ( t ) is the current varying with time, t 1, t 2 are the integration time windows; The charge density ρ is calculated through the integrated charge quantity Q , the corrected particle velocity v c and the cross-sectional area S : , where: ρ is the charge density, Q is the charge quantity within the integration window, v c is the corrected particle velocity, S is the area of the Faraday ring.

[0052] The steps for obtaining the image texture are as follows: S1. System synchronization and optical path triggering; S11. Synchronize the light source and imaging. The white parallel backlight of the light source module 41 and the global shutter CMOS camera of the high-speed imaging module 42 are hardware-synchronized through the GNSS clock to ensure that the timing error between the light pulse and the camera exposure < 1 μs to achieve spatio-temporal alignment of the particle motion trajectory and the charge signal; S12. Parallel backlight triggers the light screen. When the particle passes through the parallel backlight source, orthogonal light screens (90°) and inclined light screens (30 - 60°) are formed by reflection. The orthogonal APD photodetector 44 triggers the imaging clock, and the rear APD photodetector 45 verifies the particle passing event, forming a dual light screen redundant trigger mechanism to reduce the false detection rate.

[0053] S2. Shadow image capture and preprocessing; S21. High-frame-rate imaging: The global shutter CMOS camera of the high-speed imaging module 42 captures the particle shadow image at a frame rate of ≥100fps, and the resolution supports the morphological details of particles with a diameter of 2 μ m; S22. Dynamic denoising: Combine the light source pulse width ( μ s level) with the CMOS exposure time to match, suppress motion blur, and perform median filtering and histogram equalization in real time through the operation module 54 to enhance the texture contrast; S23. Super-resolution reconstruction: For low signal-to-noise ratio images, use the improved Real-ESRGAN network, integrate multi-channel gradient information through the structure tensor (ST) branch, and restore the microscopic texture details of the particle surface.

[0054] S3. Texture feature extraction; S31. Use the gray-level co-occurrence matrix (GLCM), calculate the contrast, energy, and entropy statistics for the gray-level distribution of the particle shadow image, and quantify the surface uniformity; S32. Through local binary pattern (LBP), analyze the neighborhood coding histogram of the particle edge to distinguish organic particles (random texture) from crystal particles (periodic texture); S33. Use multi-scale Gabor filtering, extract frequency domain features through a multi-directional and multi-scale Gabor filter bank, and identify the microscopic fluctuations of the charged particle deposition path (such as the spiral trajectory of biological particles).

[0055] S4. Spatio-temporal feature fusion; S41. Dynamic trajectory modeling: Superimpose the charge signal Q on the consecutive frames of high-speed imaging, calculate the particle motion speed v by the optical flow method, and construct a four-dimensional feature vector (Q, v, d, texture entropy) in combination with the particle size d; S42. Optoelectronic signal alignment: The charge amplifier and the APD trigger signal are isolated by an optocoupler to avoid electromagnetic interference. Use the time stamp to match the charge pulse waveform with the shadow image sequence, and correlate the charge density ρ with the texture parameters.

[0056] Particle classification model establishment. The falling speed of hydrometeor particles is closely related to the particle type. Different types of particles have different falling speeds due to their different physical characteristics (such as shape, density, size, etc.). The classification method based on fuzzy logic constructs a multi-parameter classification system through velocity (v), particle size (d), and scattering intensity (Z value). The steps are as follows: S1. Definition of membership function: S11. Divide the fuzzy sets according to the particle motion characteristics. Use trapezoidal or triangular membership functions to divide into "low speed" 0 - 5 m / s (the peak membership degree is at 2 m / s), "medium speed" 3 - 15 m / s (the peak membership degree is at 8 m / s), and "high speed" 10 - 30 m / s (the peak membership degree is at 20 m / s); S12. Based on the optical imaging of the global shutter CMOS camera, divide according to the particle size (d) into "small particles" (<0.1mm), "medium particles" (0.1 - 2 mm), and "large particles" (>2 mm), and use Gaussian membership functions for smooth transition; S13. Particles in different phases (such as rain, snow, hail) have different light scattering abilities. Through the orthogonal APD photodetector 44 and the rear APD photodetector 45, the scattering intensity is divided into weak, medium, and strong. The weak scattering is a trapezoidal membership function, corresponding to clouds or small ice crystals; the medium scattering is a triangular membership function, corresponding to raindrops or wet snow; the strong scattering is a trapezoidal membership function, corresponding to hail or dry sleet.

[0057] S2. Construction of fuzzy rule base: According to physical observations and experimental data, design if - then rules. IF velocity = high speed AND particle size = large AND reflectivity = strong, THEN particle type = hail; IF velocity = medium speed AND particle size = medium AND reflectivity = medium, THEN particle type = raindrop; IF velocity = low speed AND particle size = small AND reflectivity = weak, THEN particle type = aerosol or cloud particle (AERO).

[0058] S3. Fuzzy reasoning and classification. Based on the membership function in fuzzy logic, convert physical quantities into membership degree values from 0 to 1 through function mapping; The membership degree of the triangular function is calculated as follows: , where a, b, and c are the left endpoint, vertex, and right endpoint of the triangular membership function respectively; The membership degree of the trapezoidal function is calculated as follows: , where a, b, c, and d are the left boundary, left vertex, right vertex, and right boundary of the trapezoidal membership function respectively.

[0059] S4, Logic operation optimization; S41. Adopt MIN-MAX logic, take the minimum membership degree of multi-condition rules as the rule trigger intensity, and finally take the maximum value of the outputs of each rule; S42. Introduce weight assignment, the weight of scattering intensity for hail classification is higher than that of velocity, which is achieved by adjusting the shape of the membership function or the rule priority; S43. Calibrate parameters according to experimental data to achieve the best classification effect and provide type library data for the detection of electrically charged particles.

[0060] The description and drawings of this application are only a specific embodiment, rather than restrictive. Those skilled in the art, under the inspiration of this application and without departing from the purpose of this application and the scope protected by the claims, can also make many forms, all of which are within the protection scope of this application.

Claims

1. An atmospheric hydrometeor particle charge detection system, characterized in that: It consists of a detection device (100), a balloon (200), and a ground receiving system (300). The detection device (100) includes: a cylindrical housing (1), an inner cylinder (2), a Faraday ring (3), a photoelectric detection component (4), a signal processing unit (5), a data transmission unit (6), and a power supply (7). The inner cylinder (2) is concentrically installed inside the housing (1). Particle inlets (21) and outlets (22) are provided at both axial ends of the inner cylinder (2). Anti-static coatings are provided at the particle inlets (21) and outlets (22) to avoid charge accumulation from interfering with the signal. A conical opening (23) is provided at the inlet (21) to prevent particles from contacting the inner wall of the inner cylinder (2). The Faraday ring (3) includes a first Faraday ring (3a), a second Faraday ring (3b), and a third Faraday ring (3c). The three rings are arranged at intervals along the axial direction of the inner cylinder (2). The photoelectric detection component (4) includes a light source module (41), a high-speed imaging module (42), and a reflecting prism (43). The optical path is deflected by 90° through the reflecting prism (43) to form an orthogonal light curtain with the high-speed imaging module (42) to capture particle images. The described signal processing unit (5) integrates the cross-correlation algorithm, and the charge integration window dynamically adjusts the integration window Δ t = L / v , establishes a particle classification model, and outputs the classified charge quantity, velocity, particle size, and type; The data transmission unit (6) transmits the collected charge signal waveforms, particle images, and environmental temperature and humidity parameters using a wireless data transmitter and sends them back to the ground receiving system (300). The ground receiving system (300) receives and stores the data using a wireless data receiver. The ground receiving system (300) can synchronously generate a charge density heat map. The detection device (100) is carried into the air by the balloon (200).

2. The atmospheric hydrometeor particle charge detection system according to claim 1, characterized in that: The housing (1) is made of a polymer composite material. The lower annular space between the housing (1) and the inner cylinder (2) is an equipment compartment (11), which internally houses the photoelectric detection component (4), the signal processing module (5), the data transmission unit (6), and the power supply (7). The upper annular space is a parachute compartment (12). A parachute compartment cover (13) is provided at the top of the parachute compartment (12). Four suspension rings (14) are provided on the parachute compartment cover (13) for the balloon (200) to suspend the detection device (100). A parachute (15) is provided inside the parachute compartment (12). The parachute compartment cover (13) is controlled by a parachute opener (16).

3. The atmospheric hydrometeor particle charge detection system according to claim 1, wherein: The inner cylinder (2) adopts a double-layer shielding structure. The inner layer is made of Teflon insulating material, and the outer layer is a composite coating conductive shielding layer for isolating external electromagnetic interference. The inner diameter of the inner cylinder (2) is 50 - 100 mm, and the length-diameter ratio is 3:1 - 5:1 to ensure the linearity of the particle passing path. The inner cylinder (2) consists of four sections, which are the first inner cylinder (2a), the second inner cylinder (2b), the third inner cylinder (2c), and the fourth inner cylinder (2d) from top to bottom. A transparent cylinder (24) is provided on the second inner cylinder (2b). The transparent cylinder (24) is made of PMMA material and is used for the optical path transmission of the photoelectric detection component (4).

4. The atmospheric hydrometeor particle charge detection system according to claim 1, characterized in that: The described Faraday ring (3) is made of high-purity copper or gold-plated copper ring. The Faraday ring (3) is embedded in the cylinder wall at the joint of the inner cylinders (2). The thickness of the ring body is 0.5 to 1.0 mm. The distance between the first Faraday ring (3a) and the second Faraday ring (3b) L 1 = 100 to 200 mm, and the distance between the second Faraday ring (3b) and the third Faraday ring (3c) L 2 = 50 to 100 mm. When calculating the particle velocity v, there are L 1. L 2. L 1 + L 2 three distances. The first Faraday ring (3a) is installed on the upper side of the photoelectric detection component (4), and the second Faraday ring (3b) and the third Faraday ring (3c) are installed on the lower side of the photoelectric detection component (4). Each ring is connected to a low-noise charge amplifier. The gain of the first Faraday ring (3a) is 3 to 5 times that of the second Faraday ring (3b), and the gain of the third Faraday ring (3c) is 5 to 10 times that of the second Faraday ring (3b).

5. The atmospheric hydrometeor particle charge detection system according to claim 1, characterized in that: The described light source module (41) uses a white parallel backlight source. The high-speed imaging module (42) is equipped with a global shutter CMOS camera. The reflection prism (43) is a triangular prism, which is arranged between the light path of the light source module (41) and the high-speed imaging module (42). The photoelectric detection component (4) further includes an orthogonal APD photodetector (44) and a rear APD photodetector (45). The orthogonal APD photodetector (44) forms a 90° angle with the light path of the light source module (41), and the included angle α between the rear APD photodetector (45) and the light path of the light source module (41) is 30° - 60°. The photoelectric detection component (4) combines backlight illumination, forms an orthogonal light curtain through the reflection prism to trigger particle imaging, captures the particle morphology and the shadow image of the movement track, triggers the particle passing event and synchronizes the imaging clock. The parallel light of the light source module (41) and the CMOS camera of the high-speed imaging module (42) are synchronously triggered by the first Faraday ring (3a), and the trigger signal and the charge amplifier are isolated by an optocoupler.

6. The atmospheric hydrometeor particle charge detection system according to claim 1, wherein: The signal processing unit (5) comprises a GNSS satellite navigation receiving device (51), a charge amplifier (52), an ADC analog-to-digital converter (53), and an operation module (54). The GNSS satellite navigation receiving device (51) is used for positioning the detection device (100), detecting signal synchronization, and measuring the acceleration of the detection device (100). The charge amplifier (52) amplifies the charge amount detected by the Faraday ring (3). The ADC analog-to-digital converter (53) performs analog-to-digital conversion on the output data of the charge amplifier (52), the detection data of the orthogonal APD photoelectric detector (44) and the rear APD photoelectric detector (45), and the output digital signal is sent to the operation module (54). The operation module (54) calculates the charge amount Q and the charge density ρ Calculation, particle speed v calculate.

7. The atmospheric hydrometeor particle charge detection system according to claim 1, characterized in that: The lifting speed of the described balloon (200) adopts net lift force feedback control. The helium filling amount is adjusted according to the real-time air pressure and temperature. The balloon rising speed is determined by the net lift force, air density and resistance. The helium filling amount of the balloon is adjusted according to the expected required rising speed, the volume of the balloon is changed, and the cross-sectional area is controlled according to the shape of the balloon. The calculation formula is: It is characterized in that: The lifting speed of the described balloon (200) adopts net lift force feedback control. The helium filling amount is adjusted according to the real-time air pressure and temperature. The balloon rising speed is determined by the net lift force, air density and resistance. The helium filling amount of the balloon is adjusted according to the expected required rising speed, the volume of the balloon is changed, and the cross-sectional area is controlled according to the shape of the balloon. The calculation formula is: ， Among them, A is the cross-sectional area of the balloon, F is the net lifting force (total buoyancy minus the weight of the balloon itself and the weight of additional objects), γ is the air density, C d is the drag coefficient, v p is the expected ascending speed of the balloon; As the balloon (200) rises, with the decrease of air pressure, its volume gradually expands, and the cross-sectional area A also expands accordingly. A deflation valve (201) is provided on the balloon (200). According to the required ascent speed and detection altitude requirements, the signal processing unit (5) controls the opening or closing of the deflation valve (201) based on the ascent speed detected by the GNSS satellite navigation receiving device (51), controls the cross-sectional area of the balloon (200) to control the ascent speed. When the detection device (100) reaches the predetermined end detection altitude, the deflation valve (201) opens to deflate, and when the net lifting force F is negative (the total buoyancy is less than the weight of the balloon itself and the additional objects), the balloon (200) gradually descends. When the GNSS satellite navigation receiving device (51) detects that the detection device (100) has descended to the preset parachute opening altitude, the signal processing unit (5) controls the parachute opener (16) to act, the parachute compartment cover (13) separates from the housing (1), the detection device (100) falls freely under its own weight, and the parachute (15) is opened, and the detection device (100) slowly descends to the ground.

8. The method for detecting charged atmospheric hydrometeor particles according to any one of claims 1 to 7, characterized in that: Including charge quantity calculation, image texture acquisition, and particle classification model establishment; The steps of the described charge quantity calculation are as follows: S1. Signal acquisition and preprocessing: S11. Signal acquisition and amplification: Use a double Faraday ring (3a / 3b) or (3b / 2c) or (3a / 3c) to collect the charge signal of the particle passing through the detection area, and realize the synchronous triggering of the two signals through the GNSS second pulse. The time error < 50ns, and the sampling rate ≥ 1MHz. The collected weak electrical signal is amplified by a charge amplifier circuit. S12. Denoising and filtering: Use a sliding window to perform time-domain normalization on the signal to suppress high-energy noise interference. Filter the non-target frequency band noise through a Butterworth low-pass filter to improve the signal-to-noise ratio. The transfer function of the Butterworth filter is: ， Wherein: s is a complex frequency variable, ω c is the cut-off frequency, n is the order of the filter, which determines the steepness of the filter; S2. Signal synchronization: The light source module (41), the high-speed imaging module (42), the orthogonal / rear APD detector (44 / 45), and the three Faraday rings (3a / 3b / 3c) are synchronized by the GNSS second pulse. S21. Time reference alignment: The exposure time window of the global shutter CMOS camera is synchronized with the pulse signal of the APD detector to avoid motion blur or signal misalignment. S22. Charge signal synchronization. The charge integration window of the Faraday ring (3) needs to match the time when the particle passes through the detection area. All devices are triggered by the GNSS second pulse to ensure that the alignment error of the timestamps is <1 μ s; S23. Data fusion: Align the pixel data of high-speed imaging with the signal time dimension of the APD to achieve multi-modal data correlation analysis; S3. Cross-correlation calculation: Determine the time delay (Δt) between the signals of the Faraday ring (3) and the signal correlation between the orthogonal APD photodetector (44) and the rear APD photodetector (45) and the camera of the high-speed imaging module (42); S31. Time-domain cross-correlation calculation: ， Among them, x n and y n are two-way signals, k is the number of sampling points for time delay;​​ S32. FFT-accelerated frequency domain optimization, for two time-domain signals x n and y n respectively perform fast Fourier transform (FFT) to obtain their frequency domain representations:​​ ， ， Wherein: X ( f ) and Y ( f ) are respectively the frequency domain representations of signals x n and y n , where f is the frequency variable; Cross-power spectrum calculation: The cross-power spectrum S xy ( f ) is the complex conjugate product of X ( f ) and Y ( f ) in the frequency domain:​​ ， Wherein: Y ∗ ( f ) is Y ( f )'s complex conjugate; for the cross-power spectrum S xy ( f ) perform an inverse FFT (IFFT) to obtain the cross-correlation function in the time domain R xy τ :​ ， Among them, R xy τ is the cross-correlation function, representing the correlation between two signals, τ is the time-delay variable;​ S4. Particle velocity calculation: S41. Calculate the relative velocity of the particle with respect to the detection device (100) based on the time delay between two Faraday rings and the distance between the two Faraday rings. v = L / Δ t , where L is the distance between the two Faraday rings ( L 1 or L 2 or L 1 + L 2), and Δ t is the time delay of the signals of the two Faraday rings; there are three Δ t corresponding to the three distances, which are Δ t 1 、 Δ t 2 、 Δ t 1+2 . Calculate the three obtained relative velocities v 1 、v 2 、v 3, and use the arithmetic mean or the median as the relative velocity of the particle v ; S42. Ascent velocity of the detection device (100): Use GNSS Doppler frequency shift or carrier phase change rate to calculate the three-dimensional velocity of the detection device (100), and the vertical direction is the ascent velocity of the detection device (100); S43. Correction of particle velocity: When the ascent direction of the detection device (100) is opposite to the particle movement direction, the actual particle velocity needs to be corrected by vector superposition: ， Among them, v c is the corrected particle velocity, v is the calculated particle velocity, v b is the rising velocity of the detection device (100); S5. Charge integration window dynamic adjustment calculation, based on the corrected particle velocity v c and the distance between the two Faraday rings L , the integration time window is dynamically adjusted according to the particle velocity and acceleration, and the adjustment time delay Δ t int The formula is as follows: ； S6. Charge density inversion, the amount of charge within the integration window Q Passing current I ( t ) Integration calculation over the time interval t 1, t 2]: ， Wherein: Q is the electric charge within the integration window, I ( t ) is the current varying with time, t 1, t 2 is the time window of integration; Charge density ρ By integrating the charge quantity Q , the corrected particle velocity v c and the cross-sectional area S Calculate: ， Wherein: ρ is the charge density, Q is the amount of charge within the integration window, v c is the corrected particle velocity, S is the area of the Faraday ring.

9. The method for detecting charged atmospheric hydrometeor particles according to claim 8, wherein: The steps for obtaining the image texture are as follows: S1. System synchronization and optical path triggering: S11. The light source is synchronized with imaging. The white parallel backlight source of the light source module (41) and the global shutter CMOS camera of the high-speed imaging module (42) are hardware-synchronized through the GNSS clock to ensure that the timing error between the light pulse and the camera exposure is < 1 μ s, achieving the spatio-temporal alignment of the particle motion trajectory and the charge signal; S12. Parallel backlight triggers the light curtain. When the particle passes through the parallel backlight source, orthogonal light curtains (90°) and inclined light curtains (30~60°) are formed by reflection. The orthogonal APD photodetector (44) triggers the imaging clock, and the rear APD photodetector (45) verifies the particle passing event, forming a dual-light curtain redundant trigger mechanism to reduce the false detection rate; S2. Shadow image capture and preprocessing; S21. High frame rate imaging: The global shutter CMOS camera of the high-speed imaging module (42) captures particle shadow images at a frame rate of ≥ 100 fps, and the resolution supports morphological details of particles with a size of 2 μ m particle size; S22. Dynamic denoising: By matching the light source pulse width ( μ s level) with the CMOS exposure time, motion blur is suppressed, and median filtering and histogram equalization are performed in real time through the operation module (54) to enhance texture contrast; S23. Super-resolution reconstruction: Use the improved Real-ESRGAN network for low signal-to-noise ratio images, integrate multi-channel gradient information through the structure tensor (ST) branch, and restore the microscopic texture details of the particle surface; S3. Texture feature extraction: S31. Use the gray-level co-occurrence matrix (GLCM) to calculate contrast, energy, and entropy statistics for the gray-level distribution of the particle shadow image to quantify surface uniformity; S32. Analyze the neighborhood coding histogram of the particle edge through local binary pattern (LBP) to distinguish organic particles (random texture) from crystal particles (periodic texture); S33. Use multi-scale Gabor filtering to extract frequency domain features through a multi-directional and multi-scale Gabor filter bank to identify microscopic fluctuations in the deposition path of charged particles (such as the spiral trajectory of biological particles); S4. Spatiotemporal feature fusion: S41. Dynamic trajectory modeling: Superimpose the charge signal Q on the consecutive frames of high-speed imaging, calculate the particle movement speed v by the optical flow method, and construct a four-dimensional feature vector (Q, v, d, texture entropy) in combination with the particle size d; S42. Photoelectric signal alignment: The charge amplifier and the APD trigger signal are isolated by an optocoupler to avoid electromagnetic interference; the charge pulse waveform and the shadow image sequence are matched using timestamps to correlate the charge density ρ with the texture parameters.

10. The method for detecting charged atmospheric hydrometeor particles according to claim 8, wherein: The establishment of the particle classification model: Based on the fuzzy logic classification method, construct a multi-parameter classification system through velocity (v), particle size (d), and scattering intensity (Z value). The steps are as follows: S1. Definition of membership function: S11. Divide the fuzzy sets according to the particle movement characteristics, use trapezoidal or triangular membership functions, and divide them into "low speed" 0~5 m / s (membership peak at 2 m / s), "medium speed" medium speed: 3~15 m / s (membership peak at 8 m / s), "high speed" 10~30m / s (membership peak at 20 m / s); S12. For the optical imaging based on a global shutter CMOS camera, according to the particle size (d), it is divided into "small particles" (<0.1 mm), "medium particles" (0.1 - 2 mm), and "large particles" (>2 mm), and a Gaussian membership function is used for smooth transition; S13. Particles in different phases (such as rain, snow, hail) have different light scattering abilities. The scattering intensities are divided into weak, medium, and strong by the orthogonal APD photodetector (44) and the rear APD photodetector (45). The weak scattering is a trapezoidal membership function, corresponding to clouds or small ice crystals; the medium scattering is a triangular membership function, corresponding to raindrops or wet snow; the strong scattering is a trapezoidal membership function, corresponding to hail or dry sleet; S2. Fuzzy rule base construction: According to physical observations and experimental data, if-then rules are designed. IF velocity = high speed AND particle size = large AND reflectivity = strong, THEN particle type = hail. IF velocity = medium speed AND particle size = medium AND reflectivity = medium, THEN particle type = raindrop. IF velocity = low speed AND particle size = small AND reflectivity = weak, THEN particle type = aerosol or cloud particle (AERO); S3. Fuzzy reasoning and classification, based on the membership function in fuzzy logic, convert physical quantities into membership degree values from 0 to 1 through function mapping; The membership degree of the triangular function is calculated as follows: ， where a, b, and c are the left endpoint, vertex, and right endpoint of the triangular membership function respectively; The membership degree of the trapezoidal function is calculated as follows: ， where a, b, c, and d are the left boundary, left vertex, right vertex, and right boundary of the trapezoidal membership function respectively; S4. Logic operation optimization: S41. Adopt MIN - MAX logic, take the minimum membership degree of multi - condition rules as the rule trigger intensity, and finally take the maximum value of the outputs of each rule; S42. Introduce weight assignment. The weight of scattering for hail classification is higher than that of velocity, which is achieved by adjusting the shape of the membership function or the rule priority; S43. Calibrate parameters according to experimental data to achieve the best classification effect and provide type library data for the detection of electrically charged particles.

Images