Cloud component identification method and cloud component detection laser radar system

Abstract

The invention provides a cloud component identification method and a cloud component detection laser radar system. The method comprises the following steps: acquiring and processing a mie scattering polarization signal and a fluorescence spectrum signal of a target cloud body in the atmosphere; carrying out quantitative inversion on the processed signal to obtain observation characteristic parameters required by cloud component identification, wherein the observation characteristic parameters comprise fluorescence spectrum characteristics, a depolarization ratio and an attenuation backscattering coefficient; inputting the observation characteristic parameters into a trained machine learning algorithm, and outputting a cloud component identification result of the target cloud body; the training method of the machine learning algorithm comprises the following steps: based on a pre-constructed fluorescence spectrum database, adopting a random forest classifier at the same time, and performing model training by taking an attenuation backscattering coefficient, a depolarization ratio and a fluorescence spectrum as feature vectors. According to the method, the new dimension of the fluorescence spectrum is introduced and is combined with machine learning, so that the problem of cloud component identification is effectively solved, and important technical support can be provided for the fields of cloud microphysical research, artificial influence on weather, disaster prevention and reduction and the like.

Description

Technical Field

[0001] This invention relates to the field of atmospheric remote sensing technology, specifically providing a cloud component identification method and a cloud component detection lidar system. Background Technology

[0002] As a key regulator of the Earth's climate system, clouds exert a profound influence on climate change by modulating global radiation budget and water cycle processes. Cloud components include cloud phases (liquid water, ice, and their mixtures), cloud condensation nuclei (CCN), ice nuclei (IN), and various aerosol particles. Their spatial distribution and temporal evolution directly determine the macroscopic characteristics and microscopic physical processes of clouds. Therefore, achieving high-precision detection of cloud components is of great scientific value and practical significance for improving numerical weather prediction models, optimizing weather modification operations, enhancing disaster prevention and mitigation capabilities, and deepening our understanding of cloud physics.

[0003] Currently, cloud component detection technology mainly faces the following challenges: Passive satellite remote sensing (such as the Fengyun series satellites) has wide-area coverage capabilities, but its vertical resolution is limited, making it difficult to detect the fine structure inside clouds, and its ability to identify key components such as cloud condensation nuclei and ice nuclei is insufficient. Active lidar detection (such as the CALIPSO spaceborne radar) can effectively distinguish between water clouds and ice clouds by analyzing the depolarization ratio, but it has ambiguity in identifying the coexistence of supercooled water and ice crystals in mixed-phase clouds, and it cannot effectively distinguish cloud condensation nuclei. Traditional ground-based observations can provide high spatiotemporal resolution data, but the detection methods are limited, mainly relying on optical parameter inversion, and lacking the ability to directly detect the intrinsic physicochemical properties of particles.

[0004] The main limitations of existing technologies are as follows: First, the detection dimension is limited, relying too much on a few optical parameters such as depolarization ratio, which cannot fully reflect the complex composition of cloud particles; second, the ability to identify key cloud components (such as ice nuclei and cloud condensation nuclei) is insufficient, which restricts in-depth research on the physical processes of cloud precipitation; third, it is difficult to balance spatiotemporal resolution and detection accuracy, which cannot meet the needs of rapid and accurate detection for applications such as weather modification; and fourth, there is a lack of effective means to identify special components such as biological ice nuclei, which limits the research on related ice formation mechanisms. Summary of the Invention

[0005] To address the aforementioned problems, in a first aspect, the present invention provides a cloud component identification method, comprising: acquiring and processing the Mie scattering polarization signal and fluorescence spectral signal of a target cloud in the atmosphere; quantitatively inverting the processed signal to obtain the observation feature parameters required for cloud component identification, wherein the observation feature parameters include: fluorescence spectral features, depolarization ratio, and attenuated backscattering coefficient; inputting the observation feature parameters into a trained machine learning algorithm, and outputting the cloud component identification result of the target cloud; the training method of the machine learning algorithm includes: based on a pre-constructed fluorescence spectral database, and simultaneously employing a random forest classifier, using the attenuated backscattering coefficient, depolarization ratio, and fluorescence spectrum as feature vectors for model training.

[0006] In one technical solution of the above-mentioned cloud component identification method, the pre-constructed fluorescence spectral database is constructed by collecting fluorescence spectral data of cloud condensation nuclei, ice nuclei, liquid water clouds, ice clouds, mixed-phase clouds and typical aerosols through field observation, laboratory simulation measurement and data collection from publicly available literature.

[0007] In one technical solution of the above-mentioned cloud component identification method, the Mie scattering polarization signal includes a parallel polarization component signal and a vertical polarization component signal, and the fluorescence spectral signal is a spectral signal in the 370-550nm band.

[0008] In one technical solution of the above-mentioned cloud component identification method, the acquisition and processing of the Mie scattering polarization signal and fluorescence spectrum signal of the target cloud in the atmosphere includes: background noise correction, distance correction, geometric overlap factor correction, polarization correction, spectrometer crosstalk correction, signal-to-noise ratio control, extreme value processing, and smoothing and denoising.

[0009] Secondly, the present invention provides a cloud component detection lidar system, comprising: a laser emitting unit for emitting a high-energy excitation laser in the ultraviolet band into the atmosphere to excite a target cloud to generate Mie scattering signals and fluorescence signals; a signal receiving unit for receiving and separating the Mie scattering polarization signals and fluorescence spectral signals in the atmospheric echo; an acquisition and storage unit for acquiring and storing the Mie scattering polarization signals and fluorescence spectral signals; and a data inversion unit, which incorporates a machine learning algorithm model for processing the acquired and stored signals, inverting feature parameters, identifying cloud components, and performing multi-source verification, and outputting the final cloud component detection results.

[0010] In one technical solution of the aforementioned cloud component detection lidar system, the laser emitting unit includes a laser and a beam expander and reflector assembly; the laser outputs a 500mJ ultraviolet laser with a wavelength of 355nm, and the beam expander and reflector assembly includes a dichroic beam splitter, a beam expander and collimator, and a 355nm high-reflection mirror.

[0011] In one technical solution of the aforementioned cloud component detection lidar system, the signal receiving unit includes a telescope, a beam splitter, and a signal detector; the beam splitter includes an aperture, a dichroic mirror, a filter, a polarizing beam splitter prism, and a focusing lens; and the signal detector includes two photomultiplier tubes and a 32-channel spectrometer.

[0012] In one technical solution of the aforementioned cloud component detection lidar system, the two photomultiplier tubes respectively receive the parallel polarization component signal and the vertical polarization component signal of Mie scattering, and the 32-channel spectrometer receives fluorescence spectral signals in the range of 370-550nm; the acquisition and storage unit includes a data acquisition card and an industrial control computer, the data acquisition card is used for signal acquisition, and the industrial control computer is used for storing the acquired signals and providing data support for the data inversion unit.

[0013] The beneficial effects of this invention are: by applying laser-induced fluorescence technology to cloud component identification, while maintaining the advantage of high spatiotemporal resolution, it achieves high-precision identification of cloud components, providing more reliable observational support for cloud physics research and related application fields. Attached Figure Description

[0014] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein: Figure 1 This is a schematic diagram of the structure and inversion process of a cloud component detection lidar system according to an embodiment of the present invention; Figure 2 This is a comparison diagram of the normalized fluorescence spectral characteristics of different cloud components and standard biological ice nuclei and organic matter according to an embodiment of the present invention; Figure 3 This is a concurrent FY-4B satellite cloud product image based on the identification results of an embodiment of the present invention. Detailed Implementation

[0015] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0016] Example 1 like Figure 1-3 As shown, the present invention provides a cloud component identification method, comprising the following steps: Step S1: Acquire and process the Mie scattering polarization signal and fluorescence spectrum signal of the target cloud in the atmosphere; Step S2: Quantitatively invert the processed signal to obtain the observational characteristic parameters required for cloud component identification. The observational characteristic parameters include: fluorescence spectral characteristics, depolarization ratio, and attenuation backscattering coefficient. Step S3: Input the observed feature parameters into the trained machine learning algorithm and output the cloud component identification result of the target cloud body; The training methods for machine learning algorithms include: using a pre-built fluorescence spectrum database and a random forest classifier, with attenuated backscattering coefficient, depolarization ratio, and fluorescence spectrum as feature vectors for model training.

[0017] Finally, the results were verified based on multi-source data, including spatial matching and temporal synchronization comparison with satellite cloud component products and reanalysis data.

[0018] In this embodiment, in step S1, the Mie scattering polarization signal is acquired by a polarization lidar, and the fluorescence spectral signal is acquired by a high-resolution fluorescence spectrometer. In step S2, the fluorescence spectral features include core indicators such as characteristic peak wavelength and peak intensity, which reflect the material composition and energy level characteristics of cloud particles. The depolarization ratio is calculated by the ratio of the vertical to the parallel components of the polarization signal, used to characterize the non-sphericity of cloud particles (e.g., the depolarization ratio of ice crystal clouds is significantly higher than that of water droplet clouds). The attenuated backscattering coefficient is obtained by combining the lidar echo signal intensity and transmission path loss, which can quantify the concentration and scattering ability of cloud particles. In step S3, the training stage of the machine learning algorithm involves a pre-constructed fluorescence spectral database covering fluorescence spectral data of common atmospheric cloud components under different environmental conditions. The training parameters of the random forest classifier include the number of decision trees, feature sampling ratio, and node splitting criteria. The training process uses 5-fold cross-validation to optimize the parameters, forming a multi-dimensional feature vector composed of the attenuated backscattering coefficient, depolarization ratio, and fluorescence spectral features. After inputting into the classifier, iterative learning establishes a mapping relationship between features and cloud components. The trained model extracts and classifies the input observational features, and outputs the cloud component category and recognition confidence level.

[0019] Of course, the specific implementation methods mentioned above are not limited to the situations listed above: the fluorescence spectral features can be changed according to the recognition requirements; the random forest classifier can be replaced by other machine learning algorithms such as gradient boosting trees and support vector machines; and the training parameters of the classifier can be flexibly adjusted according to the database size and recognition accuracy.

[0020] In one embodiment, the pre-constructed fluorescence spectral database is constructed by collecting fluorescence spectral data of cloud condensation nuclei, ice nuclei, liquid water clouds, ice clouds, mixed-phase clouds, and typical aerosols through field observations, laboratory simulation measurements, and publicly available literature data collection.

[0021] In this embodiment, field observations are conducted using drones, radiosondes, or ground-based observation stations equipped with fluorescence spectroscopy detection devices. The observation area covers different climate zones and altitudes, collecting in-situ fluorescence spectral data of various clouds and aerosols under different weather conditions. Environmental parameters (such as temperature, humidity, and air pressure) are simultaneously labeled during data recording. Laboratory simulation measurements are based on a cloud chamber simulation system, precisely controlling parameters such as temperature, humidity, and supersaturation to artificially generate liquid water clouds, ice clouds, mixed-phase clouds, and different types of cloud condensation nuclei / ice nuclei. High-precision fluorescence spectrometers are used to collect spectral data under controlled conditions, eliminating external interference factors. Public literature data collection involves screening peer-reviewed research results on cloud and aerosol fluorescence spectroscopy published in the past 10 years, extracting standardized spectral characteristic data (such as characteristic peak wavelength and peak intensity), and unifying the data format and dimensions.

[0022] The database covers cloud condensation nuclei including typical types such as sulfate, sea salt, and organic carbon; ice nuclei including categories such as mineral dust and biological particles (e.g., bacteria, pollen); and typical aerosols including dust, black carbon, and sulfate aerosols. Each data set includes a complete spectral wavelength range, characteristic peak information, and corresponding cloud / aerosol type labels. Of course, the specific implementation methods are not limited to the scenarios listed above: tethered balloons can be added to the equipment carriers for field observations; the parameters of the cloud chamber in laboratory simulations can be adjusted according to the atmospheric characteristics of the study area; the scope of publicly available literature data collection can be expanded to include industry reports, standard datasets, etc.; and those skilled in the art can flexibly adjust and set the parameters according to the application scenarios of cloud component identification.

[0023] In one embodiment, the Mie scattering polarization signal includes a parallel polarization component signal and a vertical polarization component signal, and the fluorescence spectral signal is a spectral signal in the 370-550 nm band.

[0024] In this embodiment, the parallel polarization component of the Mie scattering polarization signal refers to the scattered echo signal parallel to the polarization direction of the incident laser, and the perpendicular polarization component refers to the scattered echo signal perpendicular to the polarization direction of the incident laser. These two types of signals are collected separately using a polarization beam splitter. During the acquisition process, the accuracy of polarization state detection must be ensured to provide precise data support for subsequent depolarization ratio calculations. The acquisition frequency of the parallel and perpendicular components is synchronized with the lidar emission frequency to ensure signal matching in the time dimension. The fluorescence spectrum signal is selected in the 370-550 nm band, which covers the characteristic fluorescence peaks of liquid water clouds, ice clouds, and typical aerosols.

[0025] Of course, the specific implementation methods mentioned above are not limited to the situations listed above: the acquisition of Mie scattering polarization signals can be supplemented with other dimensions of data such as polarization components at different angles and different excitation wavelengths; the band range of fluorescence spectral signals can be expanded according to the cloud shape of the target identification (such as extending to 550-700nm to cover more aerosol characteristic peaks); and the resolution of the spectrometer can be flexibly set in combination with the detection scenario.

[0026] In one embodiment, acquiring and processing the Mie scattering polarization signal and fluorescence spectrum signal of a target cloud in the atmosphere includes: background noise correction, distance correction, geometric overlap factor correction, polarization correction, spectrometer crosstalk correction, signal-to-noise ratio control, extremum processing, and smoothing and denoising.

[0027] In this embodiment, background noise correction is performed separately for Mie scattering polarization signals and fluorescence spectral signals: For polarization signals, atmospheric background echo signals from cloudless areas are collected and averaged, and noise such as lidar dark current and ambient stray light is removed by the difference method; For fluorescence spectral signals, the spectrum under clear nighttime atmospheric conditions is collected as background, and baseline fitting subtraction is used to eliminate detector dark noise and light source stray light interference, ensuring that the signal only reflects the characteristic response of the target cloud.

[0028] Range correction is used to calibrate the range attenuation effect of Mie scattering polarization signals. Based on the radar equation (power is inversely proportional to the square of the range), the intensity of echo signals at different detection distances is corrected to make the signals of clouds at different altitudes comparable.

[0029] Geometric overlap factor correction addresses the near-field detection blind zone of lidar. By observing and calibrating the overlap factor curve, it compensates for the polarization signal intensity of low-altitude clouds, correcting signal attenuation caused by incomplete overlap between the laser beam and the receiving field of view. Polarization correction corrects measurement deviations of parallel / perpendicular polarization components by calibrating the polarization transmittance of the polarization beam splitter and the difference in the polarization response of the detector, ensuring the accuracy of the depolarization ratio calculation. Spectrometer crosstalk correction addresses signal crosstalk between different bands of the fluorescence spectrometer. It eliminates cross-interference between adjacent bands using a matrix correction method, ensuring the purity of signals at each wavelength within the 370-550nm band.

[0030] Signal-to-noise ratio (SNR) control filters valid data by setting a signal strength threshold (e.g., SNR ≥ 1) and removes invalid signal segments with low SNR. Extremum processing uses the 3σ criterion to identify and mark abnormal extreme points in the signal, and corrects these points using neighborhood mean replacement or interpolation to avoid single outliers affecting overall data quality. Smoothing and denoising involves filtering the preprocessed signal. A moving average filter is used for polarization signals, and a Savitzky-Golay filter is used for fluorescence spectral signals to reduce random noise while preserving signal characteristics.

[0031] Of course, the specific implementation methods mentioned above are not limited to the situations listed above: background noise correction can use the multiple acquisition averaging method to improve stability; smoothing and denoising can also use wavelet transform, Gaussian filtering and other algorithms; the thresholds and parameters of each correction stage can be flexibly adjusted according to the performance of the detection equipment and the cloud observation environment.

[0032] Example 2 like Figure 1 As shown, this invention provides a cloud component detection lidar system, comprising: a laser emitting unit for emitting a high-energy excitation laser in the ultraviolet band into the atmosphere to excite a target cloud to generate Mie scattering and fluorescence signals; a signal receiving unit for receiving and separating the Mie scattering polarization signal and fluorescence spectral signal from the atmospheric echo; an acquisition and storage unit for acquiring and storing the Mie scattering polarization signal and fluorescence spectral signal; and a data inversion unit with a built-in machine learning algorithm model for processing the acquired and stored signals, inverting feature parameters, identifying cloud components, and performing multi-source verification, outputting the final cloud component detection result.

[0033] In one embodiment, the laser emitting unit includes a laser and a beam expander and reflector; the laser outputs a 500mJ ultraviolet laser with a wavelength of 355nm, and the beam expander and reflector includes a dichroic beam splitter, a beam expander and collimator, and a 355nm high-reflectivity mirror.

[0034] In this embodiment, the 355nm ultraviolet laser pulse width output by the laser is 7-9ns, and the repetition frequency is 10Hz, ensuring that the laser energy is sufficient to excite clouds of different phases to generate stable Mie scattering and fluorescence signals. In the beam expander and reflector assembly, a dichroic beam splitter is used to separate the pure 355nm laser and avoid interference from 532nm green light; a beam expander and collimator expands the laser beam to a diameter of 3-10mm to reduce the divergence angle; and a 355nm high-reflectivity mirror with a reflectivity ≥99% is used to adjust the laser emission direction to adapt to different detection elevation angle requirements.

[0035] In one embodiment, the signal receiving unit includes a telescope, a beam splitter, and a signal detector; the beam splitter includes an aperture, a dichroic mirror, a filter, a polarizing beam splitter prism, and a focusing lens; and the signal detector includes two photomultiplier tubes and a 32-channel spectrometer.

[0036] In this embodiment, a 400mm Cassegrain telescope is selected as the telescope, with the receiving field of view matched to the laser emission field of view to ensure efficient capture of atmospheric echo signals. In the beam splitter assembly, an aperture stop with a diameter of 2mm is used to filter stray light. A dichroic mirror separates the Mie scattering signal (355nm) from the fluorescence spectral signal (370-550nm), and a narrow-band filter (center wavelength 355nm, bandwidth ±1nm) is used to further purify the Mie scattering signal. A polarizing beam splitter separates the Mie scattering signal into parallel and perpendicular polarization components, and a focusing lens focuses the separated signal onto the detector's photosensitive surface. In the signal detector, the photomultiplier tube's response band covers 355nm, and its gain can be adjusted via a sensitive voltage to adapt to polarization signals of different intensities. The 32-channel spectrometer has a spectral resolution of 5.8nm, covering the 370-550nm fluorescence band.

[0037] In one embodiment, the two photomultiplier tubes respectively receive the parallel polarization component signal and the vertical polarization component signal of Mie scattering, and the 32-channel spectrometer receives fluorescence spectral signals in the range of 370-550nm; the acquisition and storage unit includes a data acquisition card and an industrial control computer, the data acquisition card is used for signal acquisition, and the industrial control computer is used for storing the acquired signals and providing data support for the data inversion unit.

[0038] In this embodiment, photomultiplier tubes and polarization beam splitters are connected one-to-one via shielded cables to ensure no crosstalk between polarization signals. The 32-channel spectrometer has a sampling frequency of 230MHz, enabling real-time capture of the time-domain variations of fluorescence signals. The data acquisition card has a sampling frequency of 80MHz and a resolution of 16bit, supporting simultaneous acquisition of two polarization signals and one spectral signal. The industrial control computer is equipped with a large-capacity solid-state drive, capable of storing raw signal data in real time, and is also configured with a multi-core processor to provide computing power support for the real-time calculations of the data inversion unit.

[0039] Of course, the specific implementation methods mentioned above are not limited to the situations listed above: the output parameters of the laser can be adjusted according to the detection scenario, including increasing the excitation wavelength, increasing the excitation energy, and increasing the laser frequency; the aperture of the signal receiving telescope, the special coatings of the main / sub-lens, and the tilt angle of the telescope can also be further improved according to the detection scenario to achieve target cloud detection at long distances and in different directions; the beam splitting component can increase the diffraction grating line density to further subdivide the spectral signal, and multiple spectrometers can be configured in parallel to expand the detection spectral range; the sampling rate of the data acquisition card and the configuration of the industrial control computer can be upgraded according to the amount of data and computing needs to improve the spatiotemporal resolution of the data.

[0040] Taking the observations conducted from July to August 2024 at the Semi-Arid Climate and Environment Observation Station (SACOL Station) of Lanzhou University as an example.

[0041] Reference Figure 1A cloud component detection lidar system performs vertical detection, simultaneously acquiring 355nm Mie scattering polarization signals and fluorescence spectral signals. Standardized preprocessing and quality control procedures are performed on the received signals. Among the key feature parameters, the attenuated backscattering coefficient is inverted using the Raman method, the depolarization ratio is calculated from the Mie scattering polarization components, and the fluorescence spectrum is normalized using nitrogen Raman signals. A pre-built fluorescence spectral database is used, and a random forest classifier is employed to train the model using the attenuated backscattering coefficient, depolarization ratio, and fluorescence spectrum as feature vectors.

[0042] Reference Figure 2 As shown, the comparison between the normalized fluorescence spectra of water clouds, ice clouds, mixed-phase clouds, and aerosols and the spectra of standard samples clearly demonstrates that different cloud components exhibit significantly different fluorescence spectral characteristics under 355 nm laser excitation. By analyzing characteristic parameters such as spectral shape, peak position, and intensity ratio of specific bands, various cloud components can be effectively distinguished. Importantly, the fluorescence spectrum of mixed-phase clouds highly matches the fluorescence spectrum of the standard biological ice core sample, *Pseudomonas syringae*, in terms of main peak positions and spectral shape. Simultaneously, the fluorescence spectra of aerosol components show good consistency with the characteristic spectra of polycyclic aromatic hydrocarbon standards, further verifying the ability and reliability of the method of this invention to qualitatively identify different components in clouds.

[0043] Reference Figure 3 Three different phase cloud results identified by the inversion method of this invention were selected and spatially matched and temporally synchronized with cloud component products from the FY-4B satellite. The results showed good consistency in both spatial distribution and temporal variation, further verifying the reliability of the invention in practical applications.

[0044] The technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles and objectives of the present invention, those skilled in the art can make equivalent changes or substitutions to the original technical features, specific structures, materials, connection methods, arrangement methods, etc., and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for identifying cloud components, characterized in that, include: Acquire and process the Mie scattering polarization signal and fluorescence spectrum signal of the target cloud in the atmosphere; The processed signal is quantitatively inverted to obtain the observational characteristic parameters required for cloud component identification. The observational characteristic parameters include: fluorescence spectral characteristics, depolarization ratio, and attenuation backscattering coefficient. The observed feature parameters are input into a trained machine learning algorithm, which outputs the cloud component identification results of the target cloud body. The training method of the machine learning algorithm includes: based on a pre-built fluorescence spectrum database, a random forest classifier is used to train the model with attenuated backscattering coefficient, depolarization ratio, and fluorescence spectrum as feature vectors.

2. The method according to claim 1, characterized in that, The pre-constructed fluorescence spectral database is built by collecting data from field observations, laboratory simulations, and publicly available literature, and includes fluorescence spectral data of cloud condensation nuclei, ice nuclei, liquid water clouds, ice clouds, mixed-phase clouds, and typical aerosols.

3. The method according to claim 1, characterized in that, The Mie scattering polarization signal includes a parallel polarization component signal and a vertical polarization component signal, and the fluorescence spectrum signal is a spectral signal in the 370-550nm band.

4. The method according to claim 1, characterized in that, The acquisition and processing of Mie scattering polarization and fluorescence spectral signals of target clouds in the atmosphere includes: background noise correction, distance correction, geometric overlap factor correction, polarization correction, spectrometer crosstalk correction, signal-to-noise ratio control, extremum processing, and smoothing and denoising.

5. The method according to claim 3, characterized in that, The fluorescence spectral characteristics include: normalized fluorescence spectrum and total fluorescence intensity in the 370-550 nm range; the total fluorescence intensity is obtained by integrating the fluorescence spectral signal in the 420-520 nm band, and deducting the leakage components of the Raman scattering and Mie scattering signals in this band.

6. A cloud component detection lidar system, characterized in that, include: The laser emitting unit is used to emit high-energy excitation lasers in the ultraviolet band into the atmosphere to excite the target cloud to generate Mie scattering and fluorescence signals; The signal receiving unit is used to receive and separate the Mie scattering polarization signal and fluorescence spectrum signal in the atmospheric echo; The acquisition and storage unit is used to acquire and store the Mie scattering polarization signal and fluorescence spectrum signal; The data inversion unit has a built-in machine learning algorithm model for processing the acquired and stored signals, inverting feature parameters, identifying cloud phases, and performing multi-source verification, outputting the final cloud component detection results.

7. The system according to claim 6, characterized in that, The laser emitting unit includes a laser and a beam expander and reflector assembly; the laser outputs a 500mJ ultraviolet laser with a wavelength of 355nm, and the beam expander and reflector assembly includes a dichroic beam splitter, a beam expander and collimator, and a 355nm high-reflection mirror.

8. The system according to claim 7, characterized in that, The signal receiving unit includes a telescope, a beam splitter, and a signal detector; the beam splitter includes an aperture, a dichroic mirror, a filter, a polarizing beam splitter prism, and a focusing lens; the signal detector includes two photomultiplier tubes and a 32-channel spectrometer.

9. The system according to claim 8, characterized in that, The two photomultiplier tubes respectively receive the parallel polarization component signal and the vertical polarization component signal of Mie scattering, and the 32-channel spectrometer receives fluorescence spectral signals in the range of 370-550nm; the acquisition and storage unit includes a data acquisition card and an industrial control computer. The data acquisition card is used for signal acquisition, and the industrial control computer is used for storing the acquired signals and providing data support for the data inversion unit.

Images