Real-time monitoring method for atmospheric visibility based on micro-pulse polarization laser radar

Abstract

The present application belongs to the technical field of laser radar atmospheric detection, and provides a real-time monitoring method for atmospheric visibility based on a micro-pulse polarization laser radar, comprising: obtaining parallel channel and vertical channel original echo signals output by the micro-pulse polarization laser radar in a plurality of continuous detection periods, extracting an aerosol-free tail data segment to obtain a double-channel background noise sequence, and calculating a background light real-time polarization degree time sequence; comparing the real-time polarization degree with a natural light polarization threshold value, identifying polarization background light pollution, when the pollution exists, analyzing dynamic correlation between the double-channel background noise sequence and the real-time polarization degree time sequence, constructing an asymmetric polarization degree restoration model, generating a baseline for background deduction of each channel, performing asymmetric baseline deduction on the original echo to obtain a denoised double-channel signal, and finally calculating an atmospheric depolarization ratio based on the denoised signal, and combining a radar ratio inversion to obtain atmospheric visibility.

Description

Technical Field

[0001] This invention belongs to the field of lidar atmospheric detection technology, specifically a method for real-time atmospheric visibility monitoring based on micropulse polarization lidar. Background Technology

[0002] When deploying micropulse polarized lidar in core urban business districts and major transportation hubs to conduct real-time atmospheric visibility monitoring, during the low solar altitude angle periods in the early morning and evening, sunlight is easily projected onto surrounding glass curtain walls, open water surfaces, and polished facades, resulting in directional specular reflection. The reflected sunlight is then converted into high-intensity linearly polarized light and enters the radar's receiving field of view.

[0003] The background noise correction logic of existing conventional micropulse polarization lidar assumes that ambient natural light is unpolarized light, and the background offset of the parallel polarization channel and the vertical polarization channel is approximately constant. Background subtraction is completed by fixing the DC offset.

[0004] However, the polarized background light generated by specular reflection will only cause a short-term, sudden increase in the background noise of one of the polarization receiving channels, resulting in asymmetrical fluctuations in the noise of the two channels. Existing algorithms cannot distinguish whether the channel signal fluctuations originate from the polarization scattering characteristics of atmospheric aerosol particles or from external specular reflection polarization interference. Ultimately, the aerosol depolarization ratio calculation is distorted, and the atmospheric visibility value obtained by real-time inversion experiences irregular, abrupt changes.

[0005] Furthermore, micropulse lidar has low laser pulse energy and a weak original echo signal-to-noise ratio, making it more susceptible to polarization background light pollution compared to high-power lidar. Relying on shading structures and adjusting the radar installation angle cannot cope with the random reflection interference caused by the continuous changes in the sun's position. Currently, there is a lack of correction schemes adapted to the abrupt polarization background light pollution in complex urban environments. Therefore, this invention provides a method for real-time atmospheric visibility monitoring based on micropulse polarization lidar. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is:

[0008] To address the aforementioned technical problems, this invention provides a method for real-time atmospheric visibility monitoring based on micropulse polarization lidar, comprising:

[0009] Acquire the parallel channel raw echo signal and vertical channel raw echo signal output by the micropulse polarization lidar in multiple consecutive detection cycles;

[0010] Extract the aerosol-free tail data segment within the far-field distance range of the original echo signal in each detection cycle, calculate the parallel channel background noise sequence and the vertical channel background noise sequence, and calculate the real-time polarization degree time sequence of the background light in the current field of view based on the parallel channel background noise sequence and the vertical channel background noise sequence.

[0011] The real-time polarization degree at the current moment is compared with the natural light polarization threshold to determine whether there is polarization background light pollution. If so, the dynamic correlation between the parallel channel background noise sequence and the real-time polarization degree time series, as well as the vertical channel background noise sequence and the real-time polarization degree time series, is analyzed. Based on the correlation fluctuation analysis results, a parallel channel-polarization degree coupling baseline model and a vertical channel-polarization degree coupling baseline model are established.

[0012] The parallel channel-polarization degree coupled baseline model and the vertical channel-polarization degree coupled baseline model are merged into an asymmetric polarization degree restoration model. Using the asymmetric polarization degree restoration model, the parallel channel background subtraction baseline and the vertical channel background subtraction baseline under the current detection period are generated. The original echo signals of the parallel channel and the original echo signals of the vertical channel are then asymmetrically subtracted and restored to obtain denoised dual-channel signals. The atmospheric depolarization ratio is calculated based on the denoised dual-channel signals, and the atmospheric visibility is obtained by combining the radar ratio inversion.

[0013] As a further improvement of the present invention, the specific process for obtaining the original parallel channel echo signal and the original vertical channel echo signal output by the micropulse polarization lidar in multiple consecutive detection cycles is as follows:

[0014] Micropulse polarization lidar is used to emit linearly polarized laser pulses into the atmosphere and receive backscattered light generated by atmospheric aerosols and atmospheric molecules.

[0015] By utilizing the polarization beam splitter inside the micropulse polarization lidar, backscattered light is separated into parallel polarized light that is parallel to the emitted polarization state, and perpendicular polarized light that is orthogonal to the emitted polarization state.

[0016] The photoelectric detection module performs photoelectric conversion and spatial distance gating on the parallel polarized light and the vertical polarized light respectively to obtain single-pulse parallel channel discrete signal and single-pulse vertical channel discrete signal.

[0017] Within a preset single detection cycle time window, the discrete signals of the single-pulse parallel channel and the discrete signals of the single-pulse vertical channel generated by multiple transmissions are accumulated and averaged to generate the original echo signals of the parallel channel and the original echo signals of the vertical channel under the current detection cycle.

[0018] A preset number of detection cycles are used as a sliding time window to acquire the parallel channel raw echo signal and the vertical channel raw echo signal output by multiple consecutive detection cycles within the sliding time window.

[0019] As a further improvement of the present invention, the specific process of calculating the parallel channel background noise sequence and the vertical channel background noise sequence is as follows:

[0020] Obtain the spatial distance sequence corresponding to the original echo signal of the parallel channel and the original echo signal of the vertical channel. Determine the interval in the spatial distance sequence that is greater than the far-field start threshold and less than the far-field end threshold as the far-field distance range. According to the preset far-field distance range, extract the data corresponding to the distance gate from the original echo signal of the parallel channel and the original echo signal of the vertical channel respectively as the aerosol-free tail data segment.

[0021] For each detection cycle, the arithmetic mean of the parallel channel data and the arithmetic mean of the vertical channel data within the aerosol-free tail data segment are calculated to obtain the mean of the parallel channel background noise and the mean of the vertical channel background noise corresponding to the current detection cycle.

[0022] Following the chronological order of multiple consecutive detection cycles, the mean values ​​of the background noise of the parallel channels corresponding to each detection cycle are spliced ​​together in the time dimension to generate a parallel channel background noise sequence that varies with time. Similarly, following the chronological order of multiple consecutive detection cycles, the mean values ​​of the background noise of the vertical channels corresponding to each detection cycle are spliced ​​together in the time dimension to generate a vertical channel background noise sequence that varies with time.

[0023] As a further improvement of the present invention, the specific process for calculating the real-time polarization degree time series of the background light in the current field of view is as follows:

[0024] For each detection cycle, the mean values ​​of the parallel channel background noise and the vertical channel background noise are calculated; the mean value of the vertical channel background noise is multiplied by the system polarization gain constant to obtain the corrected mean value of the vertical channel background noise.

[0025] Calculate the absolute value of the difference between the mean background noise of the parallel channel and the mean background noise of the corrected vertical channel, and use it as the background light polarization difference; calculate the sum of the mean background noise of the parallel channel and the mean background noise of the corrected vertical channel, and use it as the total background light intensity.

[0026] Divide the background light polarization difference by the total background light intensity to calculate the real-time polarization degree corresponding to the current detection cycle. Then, according to the time sequence of multiple consecutive detection cycles, the real-time polarization degrees corresponding to each detection cycle are spliced ​​together in the time dimension to generate a real-time polarization degree time series.

[0027] As a further improvement of the present invention, the specific process for determining whether polarization background light pollution exists is as follows:

[0028] Obtain the value corresponding to the latest detection period in the real-time polarization degree time series, and use it as the real-time polarization degree at the current moment;

[0029] If the real-time polarization degree at the current moment is greater than the natural light polarization threshold, then it is determined that there is polarized background light pollution in the current field of view;

[0030] If the real-time polarization degree at the current moment is less than or equal to the natural light polarization threshold, it is determined that there is no polarized background light pollution in the current field of view.

[0031] As a further improvement of the present invention, the specific analysis process of the dynamic correlation is as follows:

[0032] Obtain the parallel channel background noise sequence, the vertical channel background noise sequence, and the real-time polarization time sequence within the current sliding time window;

[0033] The Pearson correlation coefficient algorithm is used to calculate the first covariance between the parallel channel background noise sequence and the real-time polarization time series. The first covariance is then divided by the product of the standard deviation of the parallel channel background noise sequence and the standard deviation of the real-time polarization time series to obtain the parallel channel dynamic correlation coefficient.

[0034] The Pearson correlation coefficient algorithm is used to calculate the second covariance between the vertical channel background noise sequence and the real-time polarization time series. The second covariance is then divided by the product of the standard deviation of the vertical channel background noise sequence and the standard deviation of the real-time polarization time series to obtain the vertical channel dynamic correlation coefficient.

[0035] If the dynamic correlation coefficient of the parallel channel is greater than or equal to the threshold of the dynamic correlation coefficient of the parallel channel, then the background noise sequence of the parallel channel and the real-time polarization time series show a linear relationship.

[0036] If the dynamic correlation coefficient of the parallel channel is less than the threshold of the dynamic correlation coefficient of the parallel channel, then the background noise sequence of the parallel channel and the real-time polarization time series exhibit a nonlinear relationship.

[0037] If the vertical channel dynamic correlation coefficient is greater than or equal to the vertical channel dynamic correlation coefficient threshold, then the vertical channel background noise sequence and the real-time polarization time series show a linear relationship.

[0038] If the dynamic correlation coefficient of the vertical channel is less than the threshold of the dynamic correlation coefficient of the vertical channel, then the background noise sequence of the vertical channel and the real-time polarization time series exhibit a nonlinear relationship.

[0039] As a further improvement of the present invention, the specific process of establishing the parallel channel-polarization degree coupling baseline model and the vertical channel-polarization degree coupling baseline model is as follows:

[0040] If the parallel channel background noise sequence and the real-time polarization degree time series show a linear relationship, then:

[0041] Using the real-time polarization time series as the independent variable sample and the parallel channel background noise series as the dependent variable sample, a linear regression analysis was performed. The parallel channel dynamic correlation coefficient was used as a scaling weight factor and multiplied by the ratio of the standard deviation of the parallel channel background noise series to the real-time polarization time series to calculate the parallel channel baseline offset slope. The parallel channel baseline intercept was then calculated using the least squares method.

[0042] Based on the parallel channel baseline offset slope and the parallel channel baseline intercept, a linear equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the linear equation is defined as the parallel channel background subtraction baseline, thus obtaining the parallel channel-polarization degree coupled baseline model.

[0043] If the parallel channel background noise sequence and the real-time polarization degree time sequence exhibit a nonlinear relationship, then:

[0044] The real-time polarization degree time series was used as the independent variable sample, and the parallel channel background noise series was used as the dependent variable sample.

[0045] A polynomial regression algorithm of a preset order is used to perform nonlinear fitting on the independent variable samples and the dependent variable samples, and the parallel channel polynomial coefficient matrix is ​​calculated based on the least squares method.

[0046] Based on the parallel channel polynomial coefficient matrix, a nonlinear polynomial equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the nonlinear polynomial equation is defined as the parallel channel background subtraction baseline, thus obtaining the parallel channel-polarization degree coupled baseline model.

[0047] If the vertical channel background noise sequence and the real-time polarization degree time series show a linear relationship, then:

[0048] Using the real-time polarization time series as the independent variable sample and the vertical channel background noise series as the dependent variable sample, a linear regression analysis was performed. The vertical channel dynamic correlation coefficient was used as a scaling weight factor and multiplied by the ratio of the standard deviation of the vertical channel background noise series to the real-time polarization time series to calculate the vertical channel baseline offset slope. The vertical channel baseline intercept was then calculated using the least squares method.

[0049] Based on the vertical channel baseline offset slope and the vertical channel baseline intercept, a linear equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the linear equation is defined as the vertical channel background subtraction baseline, thus obtaining the vertical channel-polarization degree coupled baseline model.

[0050] If the vertical channel background noise sequence and the real-time polarization degree time series exhibit a nonlinear relationship, then:

[0051] The real-time polarization degree time series was used as the independent variable sample, and the vertical channel background noise series was used as the dependent variable sample.

[0052] A polynomial regression algorithm of a preset order is used to perform nonlinear fitting on the independent variable samples and the dependent variable samples, and the vertical channel polynomial coefficient matrix is ​​calculated based on the least squares method.

[0053] Based on the vertical channel polynomial coefficient matrix, a nonlinear polynomial equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the nonlinear polynomial equation is defined as the vertical channel background subtraction baseline, thus obtaining the vertical channel-polarization degree coupled baseline model.

[0054] As a further improvement of the present invention, the specific process of merging the parallel channel-polarization degree coupling baseline model and the vertical channel-polarization degree coupling baseline model into an asymmetric polarization degree restoration model is as follows:

[0055] Configure the parallel channel-polarization degree coupled baseline model as the first prediction branch of the model, and configure the vertical channel-polarization degree coupled baseline model as the second prediction branch of the model;

[0056] Using real-time polarization degree as a unified input variable, the first prediction branch and the second prediction branch are encapsulated into a joint mathematical equation system with dual-channel independent output characteristics, and the joint mathematical equation system is used as an asymmetric polarization degree restoration model.

[0057] As a further improvement of the present invention, the specific process of obtaining the denoised dual-channel signal is as follows:

[0058] The real-time polarization degree at the current moment is input into the asymmetric polarization degree restoration model. The parallel channel background subtraction baseline under the current detection period is calculated and output through the first prediction branch, and the vertical channel background subtraction baseline under the current detection period is calculated and output through the second prediction branch.

[0059] The parallel channel background subtraction baseline is subtracted from each spatial distance gate data in the original parallel channel echo signal to obtain the denoised parallel channel signal; the vertical channel background subtraction baseline is subtracted from each spatial distance gate data in the original vertical channel echo signal to obtain the denoised vertical channel signal.

[0060] The denoised parallel channel signal and the denoised vertical channel signal are synchronously combined to obtain a denoised dual-channel signal.

[0061] As a further improvement of the present invention, the specific calculation process for atmospheric visibility is as follows:

[0062] The atmospheric depolarization ratio is calculated by dividing the denoised vertical channel signal in the denoised dual-channel signal by the denoised parallel channel signal and multiplying it by the system polarization gain constant.

[0063] Match the corresponding radar ratio based on the atmospheric depolarization ratio;

[0064] The total echo signal is obtained by adding the denoised parallel channel signal and the denoised vertical channel signal. The total echo signal and the radar ratio are substituted into the lidar equation, and the atmospheric extinction coefficient is obtained by using the Fernald inversion algorithm or the Klett inversion algorithm.

[0065] The atmospheric extinction coefficient of the near-surface target altitude layer is obtained, and the atmospheric visibility is calculated using the Cauchy-Mead's law formula.

[0066] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0067] This invention constructs a dual-channel background noise sequence and a real-time polarization degree time series using far-field tail data. It relies on thresholding to identify polarized background light pollution, overcoming the limitation of traditional fixed-constant baseline subtraction in handling dual-channel asymmetric baseline drift. Pearson correlation coefficients are used to analyze the dynamic correlation between background noise and polarization degree, distinguishing between linear and nonlinear polarization interference. Independent dual-channel polarization degree-baseline coupling models are established using linear regression and multinomial regression, respectively, adaptable to varying polarized background light in complex scenarios such as buildings and water surfaces. An asymmetric polarization degree restoration model is constructed, outputting dedicated background subtraction baselines for parallel and perpendicular channels, achieving differentiated baseline subtraction, reducing the impact of baseline residual error on depolarization ratio, and improving aerosol discrimination accuracy. Based on the denoised dual-channel signals, extinction coefficient inversion and visibility calculation are performed, effectively suppressing inversion bias introduced by polarized background light and improving visibility monitoring accuracy in complex urban environments. Attached Figure Description

[0068] The invention will now be further described with reference to the accompanying drawings.

[0069] Figure 1This is a flowchart of the steps of the real-time atmospheric visibility monitoring method based on micropulse polarization lidar of the present invention;

[0070] Figure 2 This is a logic diagram of the real-time atmospheric visibility monitoring method based on micropulse polarization lidar of the present invention. Detailed Implementation

[0071] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0072] Example 1

[0073] like Figures 1-2 As shown in the embodiment of the present invention, the method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar includes:

[0074] Step 1: Acquire the original echo signals of the parallel channel and the original echo signals of the vertical channel output by the micropulse polarization lidar in multiple consecutive detection cycles;

[0075] In some embodiments, the specific process of obtaining the original parallel channel echo signal and the original vertical channel echo signal output by the micropulse polarization lidar in multiple consecutive detection cycles in step one is as follows:

[0076] The micropulse polarization lidar emits linearly polarized laser pulses into the atmosphere and receives backscattered light generated by atmospheric aerosols and atmospheric molecules.

[0077] The backscattered light is separated into parallel polarized light that is parallel to the emitted polarization state and perpendicular polarized light that is orthogonal to the emitted polarization state by using the polarization beam splitter inside the micropulse polarization lidar.

[0078] The parallel polarized light and the vertical polarized light are respectively photoelectrically converted and spatially gated by the photoelectric detection module to obtain single-pulse parallel channel discrete signal and single-pulse vertical channel discrete signal.

[0079] Within a preset single detection cycle time window, the discrete signals of the single-pulse parallel channel and the discrete signals of the single-pulse vertical channel generated by multiple transmissions are accumulated and averaged to generate the original echo signal of the parallel channel and the original echo signal of the vertical channel under the current detection cycle.

[0080] The above steps are executed continuously according to the time sequence. A preset number of detection cycles are used as a sliding time window to obtain the original echo signals of the parallel channel and the original echo signals of the vertical channel output in multiple consecutive detection cycles within the sliding time window.

[0081] It should be noted that, in order to ensure the accuracy and real-time performance of the subsequent background light polarization fluctuation analysis, the single detection cycle time window is usually set to 1 to 5 minutes.

[0082] If the time window of a single detection cycle is set too short, the amount of photons accumulated within a single detection cycle will be insufficient, resulting in a low signal-to-noise ratio and thus masking the true background light fluctuation characteristics.

[0083] If the time window for a single detection cycle is set too long, it will smooth out the transient changes in the polarization state of the background light in complex urban environments (such as sunlight reflected from glass curtain walls), resulting in the inability to trigger the subsequent asymmetric subtraction model in a timely manner.

[0084] It should be understood that the spatial resolution of the spatial distance gate sampling is usually set to 15 to 30 meters to ensure sufficient data density in the vertical space; the number of detection cycles included in the sliding time window is usually set to 10 to 30, so as to provide sufficient data sample support for subsequent calculation of the dynamic correlation of the time series by maintaining a certain length of historical data queue.

[0085] Step 2: Extract the aerosol-free tail data segment within the preset far-field distance range of the original echo signal in each detection cycle, calculate the parallel channel background noise sequence and the vertical channel background noise sequence that change with time, and calculate the real-time polarization degree time sequence of the background light in the current field of view based on the parallel channel background noise sequence and the vertical channel background noise sequence.

[0086] In some embodiments, the specific process of extracting the aerosol-free tail data segment within a preset far-field distance range from the original echo signal in each detection period in step two is as follows:

[0087] Obtain the spatial distance sequence corresponding to the original echo signal of the parallel channel and the original echo signal of the vertical channel;

[0088] The interval in the spatial distance sequence that is greater than a preset far-field start threshold and less than a preset far-field end threshold is determined as the preset far-field distance range;

[0089] According to the preset far-field distance range, data corresponding to the distance gate are extracted from the original echo signal of the parallel channel and the original echo signal of the vertical channel, respectively, and used as the aerosol-free tail data segment.

[0090] It should be noted that the preset far-field start threshold is usually set to 15 to 20 kilometers, and the preset far-field end threshold is usually set to 30 to 40 kilometers (depending on the maximum detection range of the lidar).

[0091] In a physical sense, the height of the tropopause is usually around 10 to 12 kilometers. Above 15 kilometers, there are almost no aerosol particles in the atmosphere, and the Rayleigh scattering signal of atmospheric molecules has also been attenuated to an extremely weak level.

[0092] Therefore, by extracting the signal within the far-field range, interference from laser echoes can be eliminated to the greatest extent, ensuring that the acquired data segment contains only pure solar background light noise and dark current noise within the system, providing a clean data source for subsequent accurate evaluation of the background light polarization state.

[0093] In some embodiments, the specific process of calculating the time-varying parallel channel background noise sequence and vertical channel background noise sequence in step two is as follows:

[0094] For each detection cycle, the arithmetic mean of the parallel channel data and the arithmetic mean of the vertical channel data within the aerosol-free tail data segment are calculated to obtain the mean of the parallel channel background noise and the mean of the vertical channel background noise corresponding to the current detection cycle.

[0095] According to the time sequence of the multiple consecutive detection cycles, the mean values ​​of the parallel channel background noise corresponding to each detection cycle are spliced ​​together in the time dimension to generate the parallel channel background noise sequence that changes with time.

[0096] According to the time sequence of the multiple consecutive detection cycles, the mean values ​​of the vertical channel background noise corresponding to each detection cycle are spliced ​​together in the time dimension to generate the vertical channel background noise sequence that changes with time.

[0097] In some embodiments, in step two, the specific process of calculating the real-time polarization time series of the background light in the current field of view based on the parallel channel background noise sequence and the vertical channel background noise sequence is as follows:

[0098] For each detection cycle, the mean value of the background noise in the parallel channel and the mean value of the background noise in the vertical channel are obtained.

[0099] Multiply the mean value of the vertical channel background noise by the system polarization gain constant to obtain the corrected mean value of the vertical channel background noise.

[0100] The absolute value of the difference between the mean background noise of the parallel channel and the mean background noise of the corrected vertical channel is calculated as the background light polarization difference value.

[0101] The sum of the mean background noise of the parallel channels and the mean background noise of the corrected vertical channels is calculated as the total background light intensity.

[0102] Divide the background light polarization difference by the total intensity of the background light to calculate the real-time polarization degree corresponding to the current detection period;

[0103] According to the chronological order of the multiple consecutive detection cycles, the real-time polarization degrees corresponding to each detection cycle are spliced ​​together in the time dimension to generate the real-time polarization degree time series.

[0104] It should be noted that the system polarization gain constant is an inherent hardware parameter used to calibrate the difference in responsivity between the parallel channel and the vertical channel photodetectors. In physical optics, the standard definition of the degree of linear polarization (DOLP) is the ratio of the difference in intensity of orthogonal polarization components to the sum of their intensities.

[0105] By substituting the pure background noise in the far field into the polarization degree formula through the above steps, the calculated real-time polarization degree is a dimensionless physical quantity with a value strictly between 0 and 1.

[0106] The closer the value of a dimensionless physical quantity is to 0, the closer the background light is to natural light (unpolarized); the closer the value of a dimensionless physical quantity is to 1, the more severe the polarization contamination of the background light.

[0107] By stitching together the real-time polarization degrees of multiple consecutive periods into a time series, the most crucial mathematical input features are provided for determining the pollution threshold and for constructing a coupled baseline model as an independent variable in subsequent steps.

[0108] It should be understood that when calculating the mean background noise of a single detection cycle, the arithmetic mean is used to effectively filter out the high-frequency random fluctuations of shot noise and thermal noise generated in the far field by photodetectors (such as photomultiplier tubes PMT or avalanche photodiodes APD), thereby extracting a stable DC baseline representing the background light intensity at the current moment.

[0109] By stitching together the mean values ​​of multiple consecutive detection cycles in chronological order, the far-field tail data, which originally belonged to the spatial dimension, was reduced in dimensionality and transformed into a time series reflecting the evolution of background light intensity over time, laying a mathematical foundation for subsequent dynamic correlation analysis.

[0110] Step 3: Compare the real-time polarization degree at the current moment with the preset natural light polarization threshold to determine whether there is polarization background light pollution. If there is polarization background light pollution, analyze the dynamic correlation between the parallel channel background noise sequence and the real-time polarization degree time series, as well as the vertical channel background noise sequence and the real-time polarization degree time series. Based on the correlation fluctuation analysis results, establish the parallel channel-polarization degree coupling baseline model and the vertical channel-polarization degree coupling baseline model.

[0111] In some embodiments, in step three, the specific process of comparing the real-time polarization degree at the current moment with a preset natural light polarization threshold to determine whether polarization background light pollution exists is as follows:

[0112] Obtain the value corresponding to the latest detection period in the real-time polarization time series, and use it as the real-time polarization at the current moment;

[0113] The real-time polarization degree at the current moment is compared with the preset natural light polarization threshold.

[0114] If the real-time polarization degree at the current moment is greater than the preset natural light polarization threshold, then it is determined that there is polarized background light pollution in the current field of view.

[0115] If the real-time polarization degree at the current moment is less than or equal to the preset natural light polarization threshold, it is determined that there is no polarized background light pollution in the current field of view.

[0116] It should be noted that the preset natural light polarization threshold is typically set to 0.05 to 0.10. Physically, natural solar background light is approximately unpolarized, with its polarization degree theoretically close to 0.

[0117] Considering the weak polarization effect caused by multiple scattering of atmospheric molecules, and the inherent polarization crosstalk inside the lidar receiving optical system, the background light polarization degree calculated under normal, pollution-free conditions will fluctuate slightly around 0.

[0118] Therefore, setting a tolerance threshold of 0.05 to 0.10 can effectively prevent false triggering. When the real-time polarization degree at the current moment exceeds the preset natural light polarization threshold, it means that the radar has received highly polarized light generated by directional reflection from urban glass curtain walls, calm water surfaces, or the exterior of specific metal buildings, i.e., polarized background light pollution has occurred. At this time, the traditional symmetrical background subtraction method must be abandoned.

[0119] In some embodiments, in step three, the specific process of analyzing the dynamic correlation between the parallel channel background noise sequence and the real-time polarization degree time series, and between the vertical channel background noise sequence and the real-time polarization degree time series, if polarization background light pollution exists, is as follows:

[0120] Obtain the parallel channel background noise sequence, the vertical channel background noise sequence, and the real-time polarization time sequence within the current sliding time window;

[0121] The Pearson correlation coefficient algorithm is used to calculate the first covariance between the parallel channel background noise sequence and the real-time polarization time series. The first covariance is then divided by the product of the standard deviation of the parallel channel background noise sequence and the standard deviation of the real-time polarization time series to obtain the parallel channel dynamic correlation coefficient.

[0122] The Pearson correlation coefficient algorithm is used to calculate the second covariance between the vertical channel background noise sequence and the real-time polarization time series. The second covariance is then divided by the product of the standard deviation of the vertical channel background noise sequence and the standard deviation of the real-time polarization time series to obtain the vertical channel dynamic correlation coefficient.

[0123] Obtain the dynamic correlation coefficient of the parallel channel and the dynamic correlation coefficient of the vertical channel, and compare them with the threshold values ​​of the dynamic correlation coefficient of the parallel channel and the dynamic correlation coefficient of the vertical channel, respectively;

[0124] Compare the dynamic correlation coefficient of parallel channels with the threshold of the dynamic correlation coefficient of parallel channels;

[0125] If the dynamic correlation coefficient of the parallel channel is greater than or equal to the threshold of the dynamic correlation coefficient of the parallel channel, it indicates that the background noise sequence of the parallel channel and the real-time polarization time series have a linear relationship.

[0126] If the dynamic correlation coefficient of the parallel channel is less than the threshold of the dynamic correlation coefficient of the parallel channel, it indicates that the background noise sequence of the parallel channel and the real-time polarization time series have a nonlinear relationship.

[0127] Compare the dynamic correlation coefficient of the vertical channel with the threshold value of the dynamic correlation coefficient of the vertical channel;

[0128] If the vertical channel dynamic correlation coefficient is greater than or equal to the vertical channel dynamic correlation coefficient threshold, it indicates that the vertical channel background noise sequence and the real-time polarization time series have a linear relationship.

[0129] If the vertical channel dynamic correlation coefficient is less than the vertical channel dynamic correlation coefficient threshold, it indicates that the vertical channel background noise sequence and the real-time polarization time series have a non-linear relationship.

[0130] This completes the dynamic correlation analysis.

[0131] It should be noted that the core physical significance of introducing dynamic correlation calculation lies in tracing the source of pollution. When the polarization background light changes abruptly, due to the random spatial angle of the urban reflective surface, the polluted light may mainly exhibit a parallel polarization state or a vertical polarization state.

[0132] By calculating the Pearson correlation coefficient of the time series (ranging from -1 to 1), it is possible to quantitatively assess which channel's noise fluctuations dominate the anomalous change in polarization. For example, if the dynamic correlation coefficient of the parallel channel is close to 1, while the dynamic correlation coefficient of the vertical channel is close to 0, it fully demonstrates that the current polarization background light pollution is mostly superimposed on the parallel channel. This correlation quantitative analysis based on time series fluctuations provides a solid mathematical basis for the subsequent establishment of an asymmetric coupled baseline model.

[0133] In some embodiments, in step three, the specific process of establishing the parallel channel-polarization degree coupling baseline model and the vertical channel-polarization degree coupling baseline model based on the correlation fluctuation analysis results is as follows:

[0134] Obtain the results of correlation fluctuation analysis;

[0135] If the parallel channel background noise sequence and the real-time polarization degree time series show a linear relationship, then:

[0136] Linear regression analysis was performed using the real-time polarization time series as the independent variable sample and the parallel channel background noise series as the dependent variable sample.

[0137] When calculating the slope of the linear regression, the parallel channel dynamic correlation coefficient is used as a scaling weight factor and multiplied by the ratio of the standard deviation of the parallel channel background noise sequence to the real-time polarization time series to calculate the parallel channel baseline offset slope, and the parallel channel baseline intercept is calculated according to the least squares method.

[0138] Based on the parallel channel baseline offset slope and the parallel channel baseline intercept, a linear equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the linear equation is defined as the parallel channel background subtraction baseline, thereby obtaining the parallel channel-polarization degree coupled baseline model.

[0139] If the parallel channel background noise sequence and the real-time polarization degree time sequence exhibit a nonlinear relationship, then:

[0140] Using the real-time polarization degree time series as the independent variable sample and the parallel channel background noise series as the dependent variable sample;

[0141] A multinomial regression algorithm of a preset order is used to perform nonlinear fitting on the independent variable samples and the dependent variable samples, and the parallel channel polynomial coefficient matrix is ​​calculated according to the least squares method.

[0142] Based on the parallel channel polynomial coefficient matrix, a nonlinear polynomial equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the nonlinear polynomial equation is defined as the parallel channel background subtraction baseline, thereby obtaining the parallel channel-polarization degree coupled baseline model.

[0143] If the vertical channel background noise sequence and the real-time polarization degree time series show a linear relationship, then:

[0144] Linear regression analysis was performed using the real-time polarization time series as the independent variable sample and the vertical channel background noise series as the dependent variable sample.

[0145] When calculating the slope of the linear regression, the dynamic correlation coefficient of the vertical channel is used as a scaling weight factor and multiplied by the ratio of the standard deviation of the vertical channel background noise sequence to the real-time polarization time series to calculate the vertical channel baseline offset slope, and the vertical channel baseline intercept is calculated according to the least squares method.

[0146] Based on the vertical channel baseline offset slope and the vertical channel baseline intercept, a linear equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the linear equation is defined as the vertical channel background subtraction baseline, thereby obtaining the vertical channel-polarization degree coupled baseline model.

[0147] If the vertical channel background noise sequence and the real-time polarization degree time series exhibit a nonlinear relationship, then:

[0148] Using the real-time polarization time series as the independent variable sample and the vertical channel background noise series as the dependent variable sample;

[0149] A multinomial regression algorithm of a preset order is used to perform nonlinear fitting on the independent variable samples and the dependent variable samples, and the vertical channel multinomial coefficient matrix is ​​calculated according to the least squares method.

[0150] Based on the vertical channel polynomial coefficient matrix, a nonlinear polynomial equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the nonlinear polynomial equation is defined as the vertical channel background subtraction baseline, thereby obtaining the vertical channel-polarization degree coupled baseline model.

[0151] It should be understood that the preset order is usually preferred to be second or third order. If the order is too low (such as first order), it degenerates into linear regression and cannot represent complex nonlinear fluctuations; if the order is too high (such as fourth order and above), it is easy to overfit when fitting historical data, resulting in severe oscillations and distortions in the prediction of the background subtraction baseline at the current moment.

[0152] It should be noted that the threshold values ​​for the dynamic correlation coefficients of the parallel channel and the vertical channel are typically set to 0.6 to 0.8 (preferably 0.7). In statistical and physical observations, when the Pearson correlation coefficient is greater than or equal to 0.7, it indicates a strong linear positive or negative correlation between background noise and polarization. In this case, linear regression analysis is used, and the correlation coefficient is used as a scaling weight factor, which not only has high computational efficiency but also accurately tracks the linear shift of noise. When the correlation coefficient is less than 0.7, it indicates that the reflecting surface may have curvature changes (such as irregular building surfaces) or dynamic deformations (such as water ripples), resulting in complex nonlinear fluctuations in polarization pollution. In this case, a multinomial regression algorithm can be switched to.

[0153] The physical significance of establishing the above coupled baseline model lies in realizing adaptive asymmetric prediction of background noise. In traditional lidar data processing, the background subtraction baseline is a fixed constant.

[0154] In this invention, the mathematical law of background noise shifting with abrupt changes in polarization degree can be learned based on historical time series. This ensures that in complex light pollution environments, the true background noise value to be subtracted at the current moment (i.e., background subtraction baseline) can be dynamically and accurately predicted for different channels based on their pollution level and pollution form (linear or nonlinear), thereby maximizing the preservation of the original echo signal.

[0155] Step 4: Merge the parallel channel-polarization degree coupled baseline model and the vertical channel-polarization degree coupled baseline model into an asymmetric polarization degree restoration model. Using the asymmetric polarization degree restoration model, generate the parallel channel background subtraction baseline and the vertical channel background subtraction baseline under the current detection period. Perform asymmetric subtraction restoration on the original echo signals of the parallel channel and the vertical channel to obtain denoised dual-channel signals. Calculate the atmospheric depolarization ratio based on the denoised dual-channel signals and obtain the atmospheric visibility by combining it with the preset radar ratio inversion.

[0156] In some embodiments, the specific process of merging the parallel channel-polarization degree coupling baseline model and the vertical channel-polarization degree coupling baseline model into an asymmetric polarization degree restoration model in step four is as follows:

[0157] Obtain the established parallel channel-polarization degree coupling baseline model and the vertical channel-polarization degree coupling baseline model;

[0158] The parallel channel-polarization degree coupled baseline model is configured as the first prediction branch of the model, and the vertical channel-polarization degree coupled baseline model is configured as the second prediction branch of the model;

[0159] Using real-time polarization degree as a unified input variable, the first prediction branch and the second prediction branch are encapsulated into a joint mathematical equation system with dual-channel independent output characteristics, and the joint mathematical equation system is used as the asymmetric polarization degree restoration model.

[0160] In some embodiments, in step four, the specific process of generating the parallel channel background subtraction baseline and the vertical channel background subtraction baseline under the current detection period using the asymmetric polarization degree restoration model, and performing asymmetric subtraction restoration on the original parallel channel echo signal and the original vertical channel echo signal to obtain the denoised dual-channel signal is as follows:

[0161] Obtain the real-time polarization degree corresponding to the current moment in the current detection cycle;

[0162] The real-time polarization degree at the current moment is input into the asymmetric polarization degree restoration model. The parallel channel background subtraction baseline under the current detection period is calculated and output through the first prediction branch, and the vertical channel background subtraction baseline under the current detection period is calculated and output through the second prediction branch.

[0163] The parallel channel background subtraction baseline is subtracted from each spatial distance gate data in the original parallel channel echo signal to obtain the denoised parallel channel signal.

[0164] The background subtraction baseline of the vertical channel is subtracted from each spatial distance gate data in the original vertical channel echo signal to obtain the denoised vertical channel signal.

[0165] The denoised parallel channel signal and the denoised vertical channel signal are synchronously combined to obtain the denoised dual-channel signal.

[0166] It should be noted that the traditional background subtraction method subtracts the same fixed DC bias from both channels (i.e., symmetrical subtraction); however, in this invention, due to the different degrees of interference from polarized background light pollution on the two channels, the generated background subtraction baselines for the parallel channels and the vertical channel are not numerically equal.

[0167] This asymmetric subtraction and restoration technique can remove the distortion noise superimposed on each channel without damaging the real atmospheric aerosol echo signal, thus preserving the polarization characteristics of the original signal to the greatest extent.

[0168] In some embodiments, the specific process of calculating the atmospheric depolarization ratio based on the denoised dual-channel signal and obtaining the atmospheric visibility by combining it with a preset radar ratio inversion in step four is as follows:

[0169] The atmospheric depolarization ratio is calculated by dividing the denoised vertical channel signal in the denoised dual-channel signal by the denoised parallel channel signal and multiplying it by the system polarization gain constant.

[0170] Based on the value of the atmospheric depolarization ratio, the corresponding preset radar ratio is matched in the preset aerosol classification mapping table;

[0171] The total echo signal is obtained by adding the denoised parallel channel signal and the denoised vertical channel signal. The total echo signal and the preset radar ratio are substituted into the lidar equation, and the atmospheric extinction coefficient is obtained by using the Fernald inversion algorithm or the Klett inversion algorithm.

[0172] The atmospheric extinction coefficient of the near-surface target altitude layer is obtained, and the atmospheric visibility is calculated using the Cauchy-Mead's law formula.

[0173] It should be understood that the formula for Koschmieder's law is as follows: ,in Atmospheric visibility, The atmospheric extinction coefficient, with a constant of 3.912, is based on the human visual contrast threshold (usually taken as...). The physical constants derived from this.

[0174] It should be noted that the preset aerosol classification mapping table is an empirical lookup table pre-constructed based on the non-spherical characteristics and optical scattering properties of atmospheric aerosol particles. In a specific embodiment, the preset aerosol classification mapping table specifically includes the following mapping relationships:

[0175] When the atmospheric depolarization ratio is less than 0.05, the current aerosol type is determined to be marine aerosol, and the preset radar ratio is 20sr to 30sr (preferably 25sr).

[0176] When the atmospheric depolarization ratio is greater than or equal to 0.05 and less than 0.15, the current aerosol type is determined to be urban / industrial aerosol, and the preset radar ratio is 60sr to 70sr (preferably 65sr).

[0177] When the atmospheric depolarization ratio is greater than or equal to 0.15 and less than 0.30, the current aerosol type is determined to be a mixed aerosol, and the preset radar ratio is 45 sr to 55 sr (preferably 50 sr).

[0178] When the atmospheric depolarization ratio is greater than or equal to 0.30, the current aerosol type is determined to be sand and dust aerosol, and the preset radar ratio is 40sr to 50sr (preferably 45sr).

[0179] It should be understood that the depolarization ratio is a core physical quantity for measuring the degree of non-sphericality of aerosol particles. Spherical particles (such as ocean droplets and urban fine particulate matter) have extremely low depolarization ratios; while non-spherical particles (such as dust and ice crystals) change the polarization state of incident light during backscattering, resulting in a significantly higher depolarization ratio. This invention, through a pre-applied asymmetric denoising model, eliminates abnormally high depolarization ratios caused by polarized background light pollution (e.g., misclassifying urban aerosols as dust), thereby ensuring that the system can accurately match the true radar ratio from the aforementioned mapping table and avoiding systematic deviations in subsequent extinction coefficient and visibility inversion results.

[0180] This invention ensures the absolute accuracy of atmospheric depolarization ratio calculation through pre-processed asymmetric denoising, thereby guaranteeing the correct selection of the preset radar ratio. Ultimately, the atmospheric visibility derived by substituting Cauchy's law still has extremely high reliability in complex urban light pollution environments, completely solving the industry pain point of abrupt changes in visibility values ​​in existing technologies.

[0181] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar, characterized in that: include: Acquire the parallel channel raw echo signal and vertical channel raw echo signal output by the micropulse polarization lidar in multiple consecutive detection cycles; Extract the aerosol-free tail data segment within the far-field distance range of the original echo signal in each detection cycle, calculate the parallel channel background noise sequence and the vertical channel background noise sequence, and calculate the real-time polarization degree time sequence of the background light in the current field of view based on the parallel channel background noise sequence and the vertical channel background noise sequence. The real-time polarization degree at the current moment is compared with the natural light polarization threshold to determine whether there is polarization background light pollution. If so, the dynamic correlation between the parallel channel background noise sequence and the real-time polarization degree time series, as well as the vertical channel background noise sequence and the real-time polarization degree time series, is analyzed. Based on the correlation fluctuation analysis results, a parallel channel-polarization degree coupling baseline model and a vertical channel-polarization degree coupling baseline model are established. The parallel channel-polarization degree coupled baseline model and the vertical channel-polarization degree coupled baseline model are merged into an asymmetric polarization degree restoration model. Using the asymmetric polarization degree restoration model, the parallel channel background subtraction baseline and the vertical channel background subtraction baseline under the current detection period are generated. The original echo signals of the parallel channel and the original echo signals of the vertical channel are then asymmetrically subtracted and restored to obtain denoised dual-channel signals. The atmospheric depolarization ratio is calculated based on the denoised dual-channel signals, and the atmospheric visibility is obtained by combining the radar ratio inversion.

2. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific process for obtaining the original parallel channel echo signal and the original vertical channel echo signal output by the micro-pulse polarization lidar within multiple consecutive detection cycles is as follows: Micropulse polarization lidar is used to emit linearly polarized laser pulses into the atmosphere and receive backscattered light generated by atmospheric aerosols and atmospheric molecules. By utilizing the polarization beam splitter inside the micropulse polarization lidar, backscattered light is separated into parallel polarized light that is parallel to the emitted polarization state, and perpendicular polarized light that is orthogonal to the emitted polarization state. The photoelectric detection module performs photoelectric conversion and spatial distance gating on the parallel polarized light and the vertical polarized light respectively to obtain single-pulse parallel channel discrete signal and single-pulse vertical channel discrete signal. Within a preset single detection cycle time window, the discrete signals of the single-pulse parallel channel and the discrete signals of the single-pulse vertical channel generated by multiple transmissions are accumulated and averaged to generate the original echo signals of the parallel channel and the original echo signals of the vertical channel under the current detection cycle. A preset number of detection cycles are used as a sliding time window to acquire the parallel channel raw echo signal and the vertical channel raw echo signal output by multiple consecutive detection cycles within the sliding time window.

3. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific process for calculating the parallel channel background noise sequence and the vertical channel background noise sequence is as follows: Obtain the spatial distance sequence corresponding to the original echo signal of the parallel channel and the original echo signal of the vertical channel. Determine the interval in the spatial distance sequence that is greater than the far-field start threshold and less than the far-field end threshold as the far-field distance range. According to the preset far-field distance range, extract the data corresponding to the distance gate from the original echo signal of the parallel channel and the original echo signal of the vertical channel respectively as the aerosol-free tail data segment. For each detection cycle, the arithmetic mean of the parallel channel data and the arithmetic mean of the vertical channel data within the aerosol-free tail data segment are calculated to obtain the mean of the parallel channel background noise and the mean of the vertical channel background noise corresponding to the current detection cycle. Following the chronological order of multiple consecutive detection cycles, the mean values ​​of the background noise of the parallel channels corresponding to each detection cycle are spliced ​​together in the time dimension to generate a parallel channel background noise sequence that varies with time. Similarly, following the chronological order of multiple consecutive detection cycles, the mean values ​​of the background noise of the vertical channels corresponding to each detection cycle are spliced ​​together in the time dimension to generate a vertical channel background noise sequence that varies with time.

4. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific process for calculating the real-time polarization degree time series of the background light within the current field of view is as follows: For each detection cycle, the mean values ​​of the parallel channel background noise and the vertical channel background noise are calculated; the mean value of the vertical channel background noise is multiplied by the system polarization gain constant to obtain the corrected mean value of the vertical channel background noise. The absolute value of the difference between the mean background noise of the parallel channel and the mean background noise of the corrected vertical channel is calculated as the background light polarization difference. The sum of the mean background noise of the parallel channels and the mean background noise of the corrected vertical channels is calculated as the total background light intensity. Divide the background light polarization difference by the total background light intensity to calculate the real-time polarization degree corresponding to the current detection cycle. Then, according to the time sequence of multiple consecutive detection cycles, the real-time polarization degrees corresponding to each detection cycle are spliced ​​together in the time dimension to generate a real-time polarization degree time series.

5. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific process for determining whether polarization background light pollution exists is as follows: Obtain the value corresponding to the latest detection period in the real-time polarization degree time series as the real-time polarization degree at the current moment; If the real-time polarization degree at the current moment is greater than the natural light polarization threshold, then it is determined that there is polarized background light pollution in the current field of view; If the real-time polarization degree at the current moment is less than or equal to the natural light polarization threshold, it is determined that there is no polarized background light pollution in the current field of view.

6. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific analysis process of the dynamic correlation is as follows: Obtain the parallel channel background noise sequence, the vertical channel background noise sequence, and the real-time polarization time sequence within the current sliding time window; The Pearson correlation coefficient algorithm is used to calculate the first covariance between the parallel channel background noise sequence and the real-time polarization time series. The first covariance is then divided by the product of the standard deviation of the parallel channel background noise sequence and the standard deviation of the real-time polarization time series to obtain the parallel channel dynamic correlation coefficient. The Pearson correlation coefficient algorithm is used to calculate the second covariance between the vertical channel background noise sequence and the real-time polarization time series. The second covariance is then divided by the product of the standard deviation of the vertical channel background noise sequence and the standard deviation of the real-time polarization time series to obtain the vertical channel dynamic correlation coefficient. If the dynamic correlation coefficient of the parallel channel is greater than or equal to the threshold of the dynamic correlation coefficient of the parallel channel, then the background noise sequence of the parallel channel and the real-time polarization time series show a linear relationship. If the dynamic correlation coefficient of the parallel channel is less than the threshold of the dynamic correlation coefficient of the parallel channel, then the background noise sequence of the parallel channel and the real-time polarization time series exhibit a nonlinear relationship. If the vertical channel dynamic correlation coefficient is greater than or equal to the vertical channel dynamic correlation coefficient threshold, then the vertical channel background noise sequence and the real-time polarization time series show a linear relationship. If the dynamic correlation coefficient of the vertical channel is less than the threshold of the dynamic correlation coefficient of the vertical channel, then the background noise sequence of the vertical channel and the real-time polarization time series exhibit a nonlinear relationship.

7. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific process for establishing the parallel channel-polarity degree coupling baseline model and the vertical channel-polarity degree coupling baseline model is as follows: If the parallel channel background noise sequence and the real-time polarization degree time series show a linear relationship, then: Using the real-time polarization time series as the independent variable sample and the parallel channel background noise series as the dependent variable sample, a linear regression analysis was performed. The parallel channel dynamic correlation coefficient was used as a scaling weight factor and multiplied by the ratio of the standard deviation of the parallel channel background noise series to the real-time polarization time series to calculate the parallel channel baseline offset slope. The parallel channel baseline intercept was then calculated using the least squares method. Based on the parallel channel baseline offset slope and the parallel channel baseline intercept, a linear equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the linear equation is defined as the parallel channel background subtraction baseline, thus obtaining the parallel channel-polarization degree coupled baseline model. If the parallel channel background noise sequence and the real-time polarization degree time sequence exhibit a nonlinear relationship, then: The real-time polarization degree time series was used as the independent variable sample, and the parallel channel background noise series was used as the dependent variable sample. A polynomial regression algorithm of a preset order is used to perform nonlinear fitting on the independent variable samples and the dependent variable samples, and the parallel channel polynomial coefficient matrix is ​​calculated based on the least squares method. Based on the parallel channel polynomial coefficient matrix, a nonlinear polynomial equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the nonlinear polynomial equation is defined as the parallel channel background subtraction baseline, thus obtaining the parallel channel-polarization degree coupled baseline model. If the vertical channel background noise sequence and the real-time polarization degree time series show a linear relationship, then: Using the real-time polarization time series as the independent variable sample and the vertical channel background noise series as the dependent variable sample, a linear regression analysis was performed. The vertical channel dynamic correlation coefficient was used as a scaling weight factor and multiplied by the ratio of the standard deviation of the vertical channel background noise series to the real-time polarization time series to calculate the vertical channel baseline offset slope. The vertical channel baseline intercept was then calculated using the least squares method. Based on the vertical channel baseline offset slope and the vertical channel baseline intercept, a linear equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the linear equation is defined as the vertical channel background subtraction baseline, thus obtaining the vertical channel-polarization degree coupled baseline model. If the vertical channel background noise sequence and the real-time polarization degree time series exhibit a nonlinear relationship, then: The real-time polarization degree time series was used as the independent variable sample, and the vertical channel background noise series was used as the dependent variable sample. A polynomial regression algorithm of a preset order is used to perform nonlinear fitting on the independent variable samples and the dependent variable samples, and the vertical channel polynomial coefficient matrix is ​​calculated based on the least squares method. Based on the vertical channel polynomial coefficient matrix, a nonlinear polynomial equation is constructed with polarization degree as the input independent variable and the predicted background noise value as the output dependent variable. The predicted background noise value output by the nonlinear polynomial equation is defined as the vertical channel background subtraction baseline, thus obtaining the vertical channel-polarization degree coupled baseline model.

8. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific process of merging the parallel channel-polarity degree coupled baseline model and the vertical channel-polarity degree coupled baseline model into an asymmetric polarity degree restoration model is as follows: Configure the parallel channel-polarization degree coupled baseline model as the first prediction branch of the model, and configure the vertical channel-polarization degree coupled baseline model as the second prediction branch of the model; Using real-time polarization degree as a unified input variable, the first prediction branch and the second prediction branch are encapsulated into a joint mathematical equation system with dual-channel independent output characteristics, and the joint mathematical equation system is used as an asymmetric polarization degree restoration model.

9. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific process for obtaining the denoised dual-channel signal is as follows: The real-time polarization degree at the current moment is input into the asymmetric polarization degree restoration model. The parallel channel background subtraction baseline under the current detection period is calculated and output through the first prediction branch, and the vertical channel background subtraction baseline under the current detection period is calculated and output through the second prediction branch. The parallel channel background subtraction baseline is subtracted from each spatial distance gate data in the original parallel channel echo signal to obtain the denoised parallel channel signal; the vertical channel background subtraction baseline is subtracted from each spatial distance gate data in the original vertical channel echo signal to obtain the denoised vertical channel signal. The denoised parallel channel signal and the denoised vertical channel signal are synchronously combined to obtain a denoised dual-channel signal.

10. The method for real-time monitoring of atmospheric visibility based on micropulse polarization lidar according to claim 1, characterized in that, The specific calculation process for atmospheric visibility is as follows: The atmospheric depolarization ratio is calculated by dividing the denoised vertical channel signal in the denoised dual-channel signal by the denoised parallel channel signal and multiplying it by the system polarization gain constant. Match the corresponding radar ratio based on the atmospheric depolarization ratio; The total echo signal is obtained by adding the denoised parallel channel signal and the denoised vertical channel signal. The total echo signal and the radar ratio are substituted into the lidar equation, and the atmospheric extinction coefficient is obtained by using the Fernald inversion algorithm or the Klett inversion algorithm. The atmospheric extinction coefficient of the near-surface target altitude layer is obtained, and the atmospheric visibility is calculated using the Cauchy-Mead's law formula.

Images