A sea surface tide level real-time monitoring method based on a multi-frequency laser radar

Abstract

The application provides a kind of sea surface tide level real-time monitoring method based on multi-frequency laser radar, belongs to sea surface tide level monitoring technical field, the application is by erecting multi-frequency laser radar equipment to three wavelengths and three incident angles Synchronous acquisition sea surface echo signal, according to echo intensity and background noise adaptive adjustment detector gain and time gate filter, establish sea surface reflection interference matrix and input time folding multilayer perception network identification sea state, query wavelength angle response uniform matrix and get integrated echo signal by weighted fusion, determine optimal wave surface fitting order and peak detection threshold by using double-layer game optimization model, calculate sea surface average height by peak detection and wave surface fitting, combined with tide level mapping vector conversion obtains real-time tide level value, and establish tide level change trend vector for short-term forecast and abnormal early warning, solve the technical problem of unstable laser radar tide level measurement accuracy under complex sea state conditions.

Description

Technical Field

[0001] This invention belongs to the field of sea surface tide monitoring technology, and more specifically, relates to a method for real-time monitoring of sea surface tides based on multi-frequency lidar. Background Technology

[0002] Tide level monitoring is a crucial aspect of marine observation and port management. Traditional tide level monitoring primarily employs techniques such as pressure tide gauges, float tide gauges, and single-frequency lidar. Pressure tide gauges calculate tide levels by measuring the hydrostatic pressure on an underwater pressure sensor. Float tide gauges use mechanical transmission devices to track changes in buoy height to measure tide levels. Single-frequency lidar calculates sea surface distance by emitting a single-wavelength laser pulse and measuring the echo time. However, in current applications of tide level monitoring, the sea surface exhibits dynamic roughness due to wind and waves. The reflection characteristics of the sea surface on lasers vary significantly under different sea conditions. Single-frequency lidar experiences a severe degradation in echo signal quality under rough sea conditions, leading to insufficient ranging accuracy and stability. While pressure and float tide gauges offer better stability, their slow response speed and complex installation and maintenance hinder large-scale, rapid deployment. In other words, existing technologies suffer from the technical problem of unstable lidar tide level measurement accuracy under complex sea conditions. Summary of the Invention

[0003] In view of this, the present invention provides a method for real-time monitoring of sea surface tides based on multi-frequency lidar, which can solve the technical problem of unstable lidar tide measurement accuracy under complex sea conditions in the prior art.

[0004] This invention is implemented as follows: It provides a real-time sea surface tide monitoring method based on multi-frequency lidar. A multi-frequency lidar device is installed above the monitoring sea area. The device emits laser pulses towards the sea surface at an oblique angle and acquires sea surface echo signals. The detector gain coefficient and laser emission power are adjusted based on the average echo intensity. The detector gain factor and time-gated filtering are adjusted based on the sea surface echo signal intensity and background light noise level. After timestamp alignment and amplitude normalization of the sea surface echo signals, a sea surface reflection interference matrix is ​​established. This matrix is ​​then input into a sea state identification model for sea surface state discrimination. The laser pulse emission frequency is adjusted according to the confidence level. The wavelength-angle response uniform matrix is ​​queried and weighted to obtain the comprehensive echo signal. The input is a two-layer game optimization model to output the optimal wavefront fitting order and the optimal peak detection threshold. Peak detection is performed based on the optimal peak detection threshold and the laser round-trip time is calculated. The instantaneous distance is calculated through the time-distance conversion relationship. The instantaneous distance sequence is established and the average sea surface height is obtained by wavefront fitting based on the optimal wavefront fitting order. The real-time tide level is obtained by combining the tide level mapping vector conversion. A tide level change trend vector is established for short-term tide level forecasting and abnormal tide warning.

[0005] The multi-frequency lidar device includes three laser emitting modules with different wavelengths and a matching narrowband filter receiving module, with the three wavelengths being 532nm, 1064nm and 1550nm respectively.

[0006] The narrowband filter receiving module uses a multi-cavity interference filter group to achieve ultra-narrowband filtering function. The multi-cavity interference filter group is composed of 5 layers of Fabry-Perot cavities cascaded together.

[0007] The oblique angle includes three incident angles of 30 degrees, 45 degrees and 60 degrees. Each incident angle corresponds to a laser emission channel. The three laser emission channels work synchronously to acquire sea surface echo signals in three angular directions.

[0008] The mean echo intensity is the average value of the peak power of the sea surface echo signal received in the previous measurement period in the time domain, and the previous measurement period is a complete laser pulse transmission and reception cycle before the current measurement time.

[0009] Specifically, when the sea surface echo signal strength is less than 10mW per square meter and the background light noise level is higher than 200W per square meter, the detector gain factor and start-up time gating filtering are simultaneously enhanced.

[0010] The time-gated filtering process uses the time characteristics of the pulsed laser to suppress background noise. The detector's gating circuit turns on 20ns before the expected echo arrives and turns off after working continuously for 60ns.

[0011] The sea surface reflection interference matrix is ​​a 3x3 matrix, with row indices corresponding to three incident angles and column indices corresponding to three laser wavelengths. The matrix elements represent the echo signal quality assessment values ​​under different incident angles and different wavelength combinations.

[0012] The echo signal quality assessment value is calculated by multiplying the echo peak signal-to-noise ratio by the echo waveform symmetry.

[0013] The structure of the sea state recognition model is a time-folded multilayer perceptron. The input layer receives a 9-dimensional feature vector, and the input layer is connected to a time-folded coding layer. The time-folded coding layer folds the input features of 12 consecutive time moments in the time dimension.

[0014] The temporal folding coding layer consists of 12 parallel feature extraction sub-networks that process the input features at 12 different time points. The outputs of the 12 feature extraction sub-networks are weighted and fused through a temporal attention mechanism.

[0015] The wavelength-angle response uniform matrix is ​​a 5-row, 9-column matrix, with row indices corresponding to 5 sea surface roughness parameter levels, column indices corresponding to 9 wavelength-angle combinations, and matrix elements representing the optimal weight coefficients.

[0016] The two-layer game optimization model includes an upper-layer model that aims to maximize wavefront reconstruction accuracy and a lower-layer model that aims to minimize ranging error. The upper-layer model and the lower-layer model are associated through a distance-wave coupling term.

[0017] The wavefront fitting process employs a least-squares polynomial fitting method, using 100 instantaneous distance data points from the instantaneous distance sequence as observations, and taking the highest degree of the polynomial as the optimal wavefront fitting order.

[0018] The tide mapping vector is a 3-dimensional vector. The first dimension is the vertical height of the installation point of the multi-frequency lidar equipment relative to the local tide reference surface. The second dimension is the cosine of the angle between the optical axis of the multi-frequency lidar equipment and the vertical direction. The third dimension is the statistical deviation correction coefficient of sea surface fluctuation.

[0019] The tidal level change trend vector is a 24-dimensional vector, with each element corresponding to the tidal level change rate over the 12 hours before and after the current moment, totaling 24 hours. The tidal level change rate over the first 12 hours is calculated based on historical measured tidal level data, while the tidal level change rate over the next 12 hours is predicted using a harmonic analysis method.

[0020] This invention solves the technical problem of unstable lidar tide level measurement accuracy under complex sea conditions by establishing a multi-frequency lidar collaborative measurement mechanism, adaptively adjusting the fusion weights of various wavelength and incident angle combinations for different sea conditions, and dynamically optimizing measurement parameters and signal processing strategies in conjunction with a sea state identification model. Specifically, this invention utilizes the differentiated reflection characteristics of three laser wavelengths (532 nm, 1064 nm, and 1550 nm) under different sea conditions, combined with spatial diversity measurements at three incident angles (30°, 45°, and 60°). The echo quality of each channel is quantitatively evaluated using a sea surface reflection interference matrix. The sea state identification model accurately determines the current sea surface roughness based on historical state evolution information encoded by time folding. The wavelength-angle response uniform matrix provides the optimal fusion weights for sea state adaptation, ensuring that the composite echo signal maintains a high signal-to-noise ratio and stable waveform characteristics under both calm and rough sea conditions. In summary, this invention solves the technical problem of unstable lidar tide level measurement accuracy under complex sea conditions mentioned in the background art. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention.

[0022] Figure 2 This is a time sequence comparison diagram of sea surface echo signals at three incident angles in the embodiment.

[0023] Figure 3 This is a time attention weight distribution diagram of the sea state recognition model in the embodiment.

[0024] Figure 4 This is a schematic diagram of the peak detection of the integrated echo signal in the embodiment.

[0025] Figure 5 This is a vector distribution diagram of tidal level variation trends in the embodiment. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0027] like Figure 1 The diagram shown is a flowchart of a real-time sea surface tide level monitoring method based on multi-frequency lidar provided by the present invention. This method includes the following steps:

[0028] S01. A multi-frequency lidar device is installed above the monitored sea area. The multi-frequency lidar device includes three laser emitting modules with different wavelengths and a matching narrowband filter receiving module. The three wavelengths are 532nm, 1064nm and 1550nm respectively. The narrowband filter receiving module uses a multi-cavity interference filter group to achieve ultra-narrowband filtering function.

[0029] S02. The multi-frequency lidar device emits laser pulses towards the sea surface at an oblique angle, which includes three incident angles of 30 degrees, 45 degrees and 60 degrees. Each incident angle corresponds to a laser emission channel. The three laser emission channels work synchronously and acquire sea surface echo signals in three angular directions. After acquiring the sea surface echo signals, the receiving module adjusts the detector gain coefficient and laser emission power for the current period according to the average echo intensity of the previous measurement period.

[0030] S03, When the sea surface echo signal strength is below 10 And the background light noise level is higher than 200. Simultaneously, the detector gain is enhanced and the startup time-gated filtering is applied; when only the sea surface echo signal strength is below 10... When only the detector gain is increased; when only the background light noise level is higher than 200... At that time, only time-gated filtering is activated;

[0031] S04. The sea surface echo signals acquired from the three incident angles are time-stamp aligned and amplitude normalized to establish a sea surface reflection interference matrix. Each element of the sea surface reflection interference matrix represents the echo signal quality assessment value under different incident angles and different wavelength combinations. The normalized sea surface echo signals and the sea surface reflection interference matrix are input into the sea state identification model to determine the sea surface state. The sea state identification model outputs sea surface roughness parameters, wind and wave levels, and confidence levels.

[0032] S05. Adjust subsequent measurement parameters based on the confidence level output by the sea state identification model. When the confidence level is below 75%, increase the laser pulse emission frequency to 150% of the original laser pulse emission frequency; when the confidence level is above 92%, decrease the laser pulse emission frequency to 65% of the original laser pulse emission frequency.

[0033] S06. Query the wavelength-angle response uniform matrix based on the sea surface roughness parameters output by the sea state identification model. The wavelength-angle response uniform matrix provides the optimal weighting coefficients for each combination of wavelength and incident angle under different sea states. Weight the sea surface echo signals at the three incident angles according to the optimal weighting coefficients to obtain the comprehensive echo signal.

[0034] S07. Input the integrated echo signal into the two-layer game optimization model to optimize the signal processing parameters. The two-layer game optimization model includes an upper-layer model with the goal of maximizing the wavefront reconstruction accuracy and a lower-layer model with the goal of minimizing the ranging error. The upper-layer model and the lower-layer model are associated through a distance-wave coupling term. The two-layer game optimization model outputs the optimal wavefront fitting order and the optimal peak detection threshold.

[0035] S08. Detect the peak of the integrated echo signal according to the optimal peak detection threshold and calculate the laser round-trip time. Calculate the instantaneous distance from the sea surface to the multi-frequency lidar device through the time-distance conversion relationship. Collect 100 sets of instantaneous distance data continuously and establish an instantaneous distance sequence. Perform wavefront fitting on the instantaneous distance sequence according to the optimal wavefront fitting order to obtain the average sea surface height value.

[0036] S09. Combining the installation height reference of the multi-frequency lidar equipment and the tide level mapping vector, the average sea surface height value is converted into the real-time tide level value. The tide level mapping vector is used to convert the lidar ranging coordinate system to the local tide level reference coordinate system. The monitoring strategy is adjusted according to the deviation between the real-time tide level value and the historical tide level data. When the deviation is greater than 0.3m, the multi-point cross-verification mode is activated and the measurement frequency is increased to 180% of the original measurement frequency. When the deviation is less than 0.05m, the current monitoring mode is maintained.

[0037] S10. Compare and analyze the real-time tide level with the tide forecast value to establish a tide level change trend vector. The tide level change trend vector includes the tide level change rate and direction 12 hours before and after the current time, and is used for short-term tide level forecast and abnormal tide warning.

[0038] The multi-cavity interference filter array consists of five cascaded Fabry-Perot cavities. The free spectral range and precision parameters of each Fabry-Perot cavity have been optimized, resulting in an overall filter bandwidth narrowed to within 0.6 nm, a peak transmittance of 87% at 532 nm wavelength, and an out-of-band suppression ratio exceeding [missing information]. By precisely matching the center wavelength of the laser, the multi-cavity interference filter group reduces the intensity of background light generated by solar radiation to 0.08% of its original intensity. Combined with the surface treatment of the lens coated with an anti-reflection film, this further reduces stray light interference within the optical system. The free spectral range is the wavelength interval between two adjacent transmission peaks, characterizing the periodic transmission characteristics of the filter. The fineness parameter is the ratio of the free spectral range to the full width at half maximum (FWHM), characterizing the filter's frequency selectivity. The out-of-band suppression ratio (OBS) is the ratio of peak wavelength transmittance to the average transmittance of non-peak wavelengths, characterizing the filter's ability to suppress non-target wavelengths. The design parameters of the multi-cavity interference filter group are optimized using optical thin-film design software, with the optimization goal of achieving the narrowest filter bandwidth and the highest OBS while ensuring high peak transmittance.

[0039] The previous measurement period refers to a complete laser pulse transmission and reception cycle preceding the current measurement moment, and the duration of this complete laser pulse transmission and reception cycle is 100 ms. The average echo intensity is the average value of the peak power of the sea surface echo signal received within the previous measurement period over the time domain, expressed in units of... The detector gain coefficient is the ratio of the detector output electrical signal amplitude to the incident light power, and its adjustment range is ∈ [100, 10000], with units of V / W. The laser emission power is the peak power of a single laser pulse, and its adjustment range is ∈ [50, 500] W. The adjustment rule is to increase the detector gain coefficient and laser emission power when the average echo intensity is lower than a set threshold, and to decrease the detector gain coefficient and laser emission power when the average echo intensity is higher than the set threshold, so that the amplitude of the received sea surface echo signal remains within the linear response range of the detector. The set threshold is 15. The set threshold is determined by experimentally measuring the echo intensity corresponding to the midpoint of the detector's linear response interval.

[0040] The sea surface echo signal strength threshold is 10. And background light noise level threshold 200 The following method was used to determine the threshold values: A 30-day continuous observation experiment was conducted under different time periods and sea conditions. Statistical distribution data of sea surface echo signal intensity and background light noise level were collected. The 15th percentile value of the sea surface echo signal intensity was taken as the threshold value after sorting the data from smallest to largest. The 20th percentile value of the background light noise level was taken as the threshold value after sorting the data from largest to smallest. The 15th percentile value represents 15% of the observed data being lower than the threshold value, and the 20th percentile value represents 20% of the observed data being higher than the threshold value. The enhancement of the detector gain was determined based on the ratio of the sea surface echo signal intensity to the sea surface echo signal intensity threshold. When the ratio was less than 0.5, the detector gain was enhanced to 3 times the original detector gain. When the ratio was ∈ [0.5, 1.0), the detector gain was enhanced to 1.5 times the original detector gain.

[0041] The time-gated filtering process suppresses background noise by utilizing the temporal characteristics of the pulsed laser. The laser pulse width is 8 ns. The time window for the sea surface echo signal to reach the detector is estimated to be ∈ [150, 200] ns based on the sea surface height range. The detector's gating circuit is activated 20 ns before the expected echo arrival and operates continuously for 60 ns before deactivating. During the deactivated period, the detector does not respond to incident photons. The gating activation and deactivation are driven by a high-precision timing control circuit with a time resolution of 2 ns. The time-gated filtering process reduces the proportion of continuous solar background light radiation to 0.04%. This is combined with the differential detection channel to subtract residual background noise in real time. The differential detection channel has the same structure as the main detection channel but does not receive the laser echo signal; it only collects the background light intensity at the same moment. The pure laser echo signal is obtained by subtracting the differential detection channel signal from the main detection channel signal. The laser pulse width of 8 ns is a factory parameter of the laser, determined by the laser's cavity length and Q-switching speed. The estimated echo arrival time is calculated based on the historical statistical range of the installation height of the multi-frequency lidar equipment and the sea surface height. The installation height range is ∈[20, 50]m, and the historical statistical range of the sea surface height is ∈[-3, 3]m. The estimated echo arrival time is the result of subtracting the historical statistical median of the sea surface height from the installation height and dividing by half the speed of light.

[0042] The timestamp alignment process unifies the time reference of the sea surface echo signals acquired by the three laser emission channels to the same reference time, which is the trigger time of the laser pulse emission. This timestamp alignment is achieved by recording the time difference between the arrival time of the sea surface echo signal and the trigger time of the laser pulse emission for each channel. The amplitude normalization process divides the peak amplitude of the sea surface echo signals from the three laser emission channels by the maximum measurable amplitude of each channel, resulting in a normalized amplitude range of [0, 1]. The maximum measurable amplitude is the maximum signal amplitude that the detector can measure within its linear response range, and this maximum measurable amplitude is determined by the detector's saturation threshold.

[0043] The sea surface reflection interference matrix is ​​a 3x3 matrix, with row indices corresponding to three incident angles of 30°, 45°, and 60°, and column indices corresponding to three laser wavelengths of 532nm, 1064nm, and 1550nm. The matrix elements... Indicates the first The incident angle and the first The echo signal quality assessment value for a laser wavelength combination is calculated by multiplying the peak signal-to-noise ratio (PSNR) and the echo waveform symmetry. The PSNR is the ratio of the peak echo power to the standard deviation of the noise power. The echo waveform symmetry is the reciprocal of the absolute value of the difference between the ratio of the echo rise time to the echo fall time and 1. When the sea surface is calm, the matrix element values ​​corresponding to small-angle incident and short-wavelength lasers are larger; when the sea surface is rough, the matrix element values ​​corresponding to large-angle incident and long-wavelength lasers are larger. The distribution characteristics of the matrix elements are analyzed to identify the current sea surface state and provide a basis for subsequent weight allocation. The echo rise time is the time required for the sea surface echo signal to rise from 10% peak amplitude to 90% peak amplitude, and the echo fall time is the time required for the sea surface echo signal to fall from 90% peak amplitude to 10% peak amplitude. The noise power standard deviation is the standard deviation of the signal amplitude within the 100ns time window before the sea surface echo signal arrives.

[0044] The structure of the sea state recognition model is a time-folded multilayer perceptron. The input layer receives a 9-dimensional feature vector, which contains 9 elements of the expanded sea surface reflection interference matrix. The input layer is followed by a time-folding coding layer, which folds the input features from 12 consecutive time points along the time dimension. Specifically, 12 parallel feature extraction sub-networks process the input features from each of the 12 time points. Each feature extraction sub-network contains two fully connected layers. The first fully connected layer maps the 9-dimensional input features to 18-dimensional hidden features, and the second fully connected layer maps the 18-dimensional hidden features to 6-dimensional time features. The outputs of the 12 feature extraction sub-networks are weighted and fused using a time attention mechanism. The weight coefficients of this time attention mechanism are calculated based on the time interval between the current time and historical time points; the closer the time interval, the larger the weight coefficient. The formula for calculating the weight coefficients is as follows: The weight coefficient for each historical moment is equal to the square of the negative time interval of the natural constant divided by the quotient of the time scale parameter raised to the power of the quotient divided by the sum of the weight coefficients for all historical moments. The time-folding coding layer outputs a 72-dimensional fused feature vector, which contains information about the evolution of the sea surface state at different times. The 72-dimensional fused feature vector is input into a 3-layer fully connected network for nonlinear transformation. The first fully connected layer maps the 72-dimensional fused feature vector to a 48-dimensional intermediate feature, the second fully connected layer maps the 48-dimensional intermediate feature to a 24-dimensional intermediate feature, and the third layer... The fully connected network maps 24-dimensional intermediate features to 12-dimensional output features. Finally, the output layer maps the 12-dimensional output features to 2-dimensional output results. The first dimension of the output result is the sea surface roughness parameter, which takes values ​​in the range of [0, 5]. The second dimension of the output result is the wave level, which takes values ​​in the range of [0, 9]. The output layer also generates a confidence level, which is calculated from the probability distribution entropy value of the output result. The smaller the entropy value, the higher the certainty of the sea state identification model's prediction and the higher the confidence level.

[0045] The timescale parameter is determined based on the sea surface fluctuation period and the lidar sampling frequency. The statistical range of the sea surface fluctuation period is ∈ [3, 8] s, the lidar sampling frequency is 10 Hz, the time window length of 12 moments is selected, and the timescale parameter is set to 0.5 s. The physical meaning of the timescale parameter is the decay rate control parameter of the time attention mechanism. The smaller the timescale parameter, the faster the weight coefficient decays with the increase of the time interval. The time folding coding layer assigns higher weight coefficients to recent moments through a weighted fusion mechanism, so that the output result of the sea state recognition model can quickly respond to changes in sea state. At the same time, the information of historical moments provides the periodicity and trend characteristics of sea surface fluctuations, improving the stability and accuracy of sea state recognition. When the sea surface changes from a calm state to a rough state, the elements of the sea surface reflection interference matrix at different moments show a gradual trend. The time folding coding layer captures the gradual trend and encodes the temporal law of sea surface state evolution in the 72-dimensional fusion feature vector. The time scale parameter 0.5s is determined as follows: 1 / 11 of the median of the statistical range of sea surface fluctuation period of 5.5s is selected as the time scale parameter, so that the time window length of 1.2s covers about 2.4 decay time constants, ensuring that the weight coefficient of the near time is significantly higher than that of the far time.

[0046] The steps for establishing the training dataset for the sea state recognition model include collecting measured lidar echo data under different sea areas and seasons. Each data sample contains the sea surface reflection interference matrix for 12 consecutive time moments, as well as the corresponding sea surface roughness parameters and wind and wave level labels. The sea surface roughness parameters are measured by a synchronously installed mechanical wave meter. The wind and wave levels are determined according to the Beaufort scale based on the sea surface wind speed data provided by the meteorological observation station. The data collection time exceeds 2000 hours, covering five typical sea states: calm, light waves, gentle waves, moderate waves, and large waves. The number of samples collected under each sea state is no less than 8000 sets. The collected raw data is cleaned to remove outliers and missing values. The outlier judgment criterion is data points whose echo signal peak power exceeds three times the normal range standard deviation. After data cleaning, the dataset is divided into training set, validation set, and test set in a ratio of 8:1:1.

[0047] The training steps of the sea state recognition model include initializing network parameters using a normal distribution with a mean of 0 and a standard deviation of 0.01, randomly assigning values, training using the batch gradient descent algorithm with a batch size of 64, initial learning rate of 0.001, decaying the learning rate to 90% of the original learning rate after every 50 training rounds, using a loss function that is a weighted sum of the prediction errors of sea surface roughness parameters and wind and wave levels, with a weight ratio of 3:2 for the two prediction errors, evaluating the performance of the sea state recognition model on a validation set every 10 rounds during training, and stopping training and saving the current network parameters when the validation set loss does not decrease for 30 consecutive rounds.

[0048] The confidence level thresholds of 75% and 92% are determined as follows: The relationship between the confidence level and prediction accuracy of the sea state recognition model output is statistically analyzed on the test set. The confidence levels are sorted from smallest to largest. The confidence level corresponding to the first occurrence of 85% prediction accuracy is selected as the low confidence level threshold of 75%, and the confidence level corresponding to the first occurrence of 95% prediction accuracy is selected as the high confidence level threshold of 92%. The original laser pulse emission frequency is 10Hz. When the confidence level is below 75%, the laser pulse emission frequency is increased to 15Hz; when the confidence level is above 92%, the laser pulse emission frequency is decreased to 6.5Hz.

[0049] The wavelength-angle response uniform matrix is ​​a 5x9 matrix, with row indices corresponding to 5 sea surface roughness parameter levels and column indices corresponding to 9 wavelength-angle combinations. Matrix elements... Indicates the first The first of the sea surface roughness parameter levels The optimal weighting coefficients for nine wavelength-angle combinations were determined through offline calibration experiments. These experiments measured the true height of the same sea surface location under different sea conditions, calculating the ranging results using nine wavelength-angle combinations. The reciprocal of the normalized ranging error for each combination was used as the initial value of the weighting coefficient. Statistical analysis was performed on the offline calibration experimental data for all sea conditions, dividing the sea surface roughness parameter into five level intervals. Within each level interval, the average of the initial values ​​of the weighting coefficients for the nine wavelength-angle combinations was used to obtain the matrix element values. The physical significance of the angular response uniform matrix lies in the significant differences in the quality of laser echo signals of different wavelengths and incident angles under different sea conditions. The optimal fusion of multi-channel signals is achieved by adjusting the optimal weight coefficient. Under calm sea conditions, the specular reflection echo intensity of short wavelength and small angle channels is high and the optimal weight coefficient is large. Under rough sea conditions, the diffuse reflection echo of long wavelength and large angle channels has good stability and the optimal weight coefficient is large. The weight allocation strategy is adaptively adjusted according to the sea surface roughness parameters output by the sea condition identification model to ensure that the integrated echo signal has a high signal-to-noise ratio and stable waveform characteristics under various sea conditions.

[0050] The true height of the offline calibration experiment was measured using a high-precision pressure tide gauge with a measurement accuracy of 1 mm. The offline calibration experiment lasted for 60 days, covering the complete tidal cycle and various sea state conditions. The five sea surface roughness parameter level intervals are ∈[0, 1), ∈[1, 2), ∈[2, 3), ∈[3, 4), and ∈[4, 5]. The ranging error normalization method is to divide the ranging error of each wavelength-angle combination by the maximum value of the ranging errors of all nine combinations.

[0051] The upper-level model of the two-layer game optimization model aims to maximize the wavefront reconstruction accuracy. The objective function of the upper-level model is expressed as follows: the objective function value of the upper-level model equals the ratio of the spatial variance of the instantaneous distance sequence divided by the standard spatial variance to the ratio of the temporal variance of the instantaneous distance sequence divided by the standard temporal variance, multiplied by the square of the standard wavefront fitting residual divided by the square of the wavefront fitting residual, multiplied by the distance-wave coupling term divided by the standard time derivative. The input includes the integrated echo signal and the wavefront fitting order, and the output is the optimal wavefront fitting order. The wavefront fitting order ranges from [3, 8] and is an integer. The constraint condition of the upper-level model is that the wavefront fitting order must be an integer and the wavefront fitting residual does not exceed 20% of the instantaneous distance standard deviation. The lower-level model of the two-layer game optimization model aims to minimize the ranging error. The objective function of the lower-level model is expressed as follows: the objective function value of the lower-level model equals the ratio of the ranging repeatability standard deviation divided by the standard time derivative. The sum of the standard range repeatability standard deviation and the absolute value of the range deviation, multiplied by the sum of the absolute values ​​of the standard range deviations, and then multiplied by the distance-wave coupling term divided by the standard time derivative, is used. The input includes the composite echo signal and the peak detection threshold, and the output is the optimal peak detection threshold. The peak detection threshold ranges from 3 to 8 times the noise level. The constraint condition of the lower-level model is that the peak detection threshold is not less than 3 times the noise level. The distance-wave coupling term is defined as the maximum value of the absolute value of the time derivative of the peak position of the composite echo signal. When the wavefront fitting order increases, the distance-wave coupling term decreases; when the peak detection threshold decreases, the distance-wave coupling term increases. The upper-level model, which pursues higher wavefront reconstruction accuracy, tends to choose a higher wavefront fitting order, while the lower-level model, which pursues lower range error, tends to choose a lower peak detection threshold. The two objectives are mutually constrained by the distance-wave coupling term, forming a game relationship.

[0052] The standard spatial variance, standard time variance, standard wavefront fitting residual, standard distance repeatability standard deviation, standard distance deviation absolute value, and standard time derivative value are dimensionless normalized parameters. These normalized parameters are obtained through offline calibration experiments. The standard spatial variance is the statistical median of the variances of instantaneous distance sequences at different spatial locations, with a value of 0.25. The standard time variance is the statistical median of the variances of the instantaneous distance sequences at different times from the same location, and it is set to 0.16. The standard wavefront fitting residual is the statistical median of the wavefront fitting residual, which is 0.08m. The standard ranging repeatability standard deviation is the statistical median of the standard deviation of repeated ranging results under the same conditions, which is 0.03m. The standard ranging deviation absolute value is the statistical median of the absolute value of the deviation between the ranging result and the true distance, which is 0.05m. The standard time derivative value is the statistical median of the absolute value of the time derivative of the peak position of the composite echo signal, which is 0.12m / s.

[0053] The game-theoretic optimization process employs an alternating iterative algorithm. In the first iteration, a peak detection threshold is fixed at 5 times the noise level to solve the upper-level model and obtain the optimal wavefront fitting order. In the second iteration, the optimal wavefront fitting order output by the upper-level model is fixed to solve the lower-level model and obtain the optimal peak detection threshold. In the third iteration, the optimal peak detection threshold output by the lower-level model is fixed, and the upper-level model is solved again. This iterative process is repeated until convergence occurs when the changes in the objective function values ​​of both the upper and lower-level models are less than 0.01. The converged optimal wavefront fitting order and optimal peak detection threshold achieve a balance between wavefront reconstruction accuracy and ranging error. The noise level is the root mean square value of the noise amplitude in the synthesized echo signal.

[0054] The peak detection method involves searching for sampling points in the composite echo signal whose amplitude exceeds the optimal peak detection threshold. The point with the largest amplitude among these sampling points is selected as the echo peak point. The laser round-trip time is obtained by subtracting the laser pulse emission time from the time corresponding to the echo peak point. The time-distance conversion relationship is calculated by multiplying the laser round-trip time by the speed of light and dividing by 2, where the speed of light is 299,792,458 m / s. The instantaneous distance is the distance from the sea surface to the multi-frequency lidar device obtained from a single measurement.

[0055] The wavefront fitting process employs a least-squares polynomial fitting method. This method uses 100 instantaneous distance data points from the instantaneous distance sequence as observations, the observation time as the independent variable, and the optimal wavefront fitting order as the highest degree of the polynomial. The polynomial coefficients are solved using the least-squares method, and the function value of the fitted polynomial at the midpoint of the observation time is taken as the average sea surface height. The wavefront fitting residual is the root mean square of the difference between each instantaneous distance data point in the instantaneous distance sequence and the corresponding function value of the fitted polynomial.

[0056] The tide level mapping vector is a 3-dimensional vector. The first dimension is the vertical height of the multi-frequency lidar equipment installation point relative to the local tide level reference surface. This vertical height is determined through high-precision leveling with an accuracy of 2 mm. The second dimension is the cosine of the angle between the optical axis of the multi-frequency lidar equipment and the vertical direction. This angle is monitored in real time by an tilt sensor to compensate for measurement deviations caused by changes in the attitude of the multi-frequency lidar equipment. The third dimension is the statistical deviation correction coefficient for sea surface fluctuations. This statistical deviation correction coefficient reflects the difference between the instantaneous sea surface height caused by wave motion and the average sea surface height. The systematic differences in degree, the statistical deviation correction coefficient is obtained through statistical analysis of long-term observation data. Under wave motion, the instantaneous distance of the sea surface measured by lidar corresponds to random sampling of wave crests and troughs. The height difference between wave crests and troughs and the mean sea surface leads to statistical deviation in the measurement results. The value range of the statistical deviation correction coefficient is ∈ [0.92, 0.98]. It is determined by looking up the wave height and wave period parameters in a table. The calculation formula of the tide level mapping vector is expressed as follows: the real-time tide level value is equal to the first dimension element minus the mean sea surface height value multiplied by the second dimension element, and then multiplied by the third dimension element.

[0057] The vertical height of the multi-frequency lidar equipment installation point relative to the local tide level reference surface was obtained by leveling points of the national elevation control network after installation. The tilt sensor has a measurement accuracy of 0.01 degrees and a sampling frequency of 1 Hz. The long-term observation data spans more than one year. The statistical deviation correction coefficient lookup table is established based on different combinations of wave height ∈ [0, 5] m and wave period ∈ [3, 15] s. Wave height is divided into 10 levels at 0.5 m intervals, and wave period is divided into 12 levels at 1 s intervals. The table stores the statistical deviation correction coefficient values ​​corresponding to 120 combinations. The statistical deviation correction coefficient value is the statistical average of the ratio of the lidar measurement distance to the pressure tide gauge measurement distance under each combination.

[0058] The historical tide data refers to the measured tide levels of the same monitoring sea area over the past 30 days, with a sampling interval of 1 hour. The deviation value is the absolute value of the difference between the real-time tide level and the average value of the historical tide data at the same tidal time. The same tidal time is the moment corresponding to the tidal periodicity. The method for determining the same tidal time is to take the current moment modulo the tidal period of 12.42 hours to obtain the tidal phase, and then find records in the historical tide data where the tidal phase difference is less than 0.5 hours as the historical tide data for the same tidal time. The deviation thresholds of 0.3m and 0.05m are determined as follows: the standard deviation of the historical tide data is calculated, with 1 standard deviation used as the small deviation threshold of 0.05m and 2 standard deviations used as the large deviation threshold of 0.3m. The statistical time span of the standard deviation is the past year.

[0059] The multi-point cross-validation mode achieves spatial redundancy measurement by deploying multiple multi-frequency lidar devices in the monitored sea area. When the real-time tide level value measured at a single point deviates significantly from historical tide level data, the multi-point cross-validation mode is activated to verify the measurement results. In this mode, the measurement results from each multi-frequency lidar device are fused using a weighted average method. The weighting coefficients are determined based on the measurement accuracy and sea state adaptability of each device; devices with high measurement accuracy and good sea state adaptability have larger weighting coefficients. The weighted average tide level is used as the final output. Good consistency in the measurement results from multiple multi-frequency lidar devices indicates that the single-point deviation is caused by local sea state disturbances. Deviations in the measurement results from multiple multi-frequency lidar devices indicate a regional abnormal tidal event. The measurement accuracy is determined through offline calibration experiments, and the sea state adaptability is determined by statistically analyzing the measurement success rate under different sea state conditions. The original measurement frequency is 10Hz; when the deviation exceeds 0.3m, the measurement frequency is increased to 18Hz.

[0060] The tidal level change trend vector is a 24-dimensional vector. The vector elements correspond to the tidal level change rate over the 12 hours before and after the current moment (24 hours total). The unit of the tidal level change rate is m / h. The tidal level change rate for the first 12 hours is calculated based on historical measured tidal level data, using the difference between adjacent hourly tidal levels divided by the time interval. The tidal level change rate for the next 12 hours is predicted using harmonic analysis. Harmonic analysis decomposes tidal motion into multiple simple harmonic components with different periods. The main components include a semi-diurnal tidal period of 12.42 hours and a full diurnal tidal period of 12.42 hours. The diurnal tidal cycle is 24.84 hours. The amplitude and phase parameters of each simple harmonic component are obtained by least-squares fitting of long-term tidal level observation data. The tidal level value at the predicted time is the sum of the instantaneous values ​​of each simple harmonic component at that time. The establishment of the tidal level change trend vector provides a basis for short-term tidal level forecasting and abnormal tidal identification. When the vector element shows a continuous monotonically increasing or monotonically decreasing trend, it indicates normal tidal change. When the vector element shows a sudden change or the deviation from the predicted value exceeds 0.5 m / h, an abnormal tidal warning is triggered. The warning information includes the time of the abnormality, the magnitude of the deviation, and the scope of the impact.

[0061] The harmonic analysis method uses eight simple harmonic components, including two semi-diurnal tidal components, two diurnal tidal components, two shallow water tidal components, and two long-period tidal components. The time span of the long-term tidal level observation data is no less than one month, and the sampling interval is one hour. The impact range of the abnormal tidal warning is determined based on the hydrogeographic characteristics of the monitored sea area. The impact range along the shoreline is 10 km upstream and downstream of the monitoring point, and the impact range perpendicular to the shoreline is 3 km offshore.

[0062] The present invention also provides a method for implementing a real-time sea surface tide monitoring system based on multi-frequency lidar using a computer. The computer is equipped with a readable storage medium, which stores program instructions. When the program instructions are run in the computer, they execute the above-described method.

[0063] It should be noted that this invention also solves the following technical problem: the insufficient detection capability of lidar echo signals under strong background light interference. Under strong solar radiation during the day, the background light noise power can reach hundreds of watts per square meter, while the intensity of the sea surface laser echo signal is only in the tens of milliwatts per square meter range. The signal is submerged in the strong background noise, leading to detection failure. This invention achieves ultra-narrowband optical filtering by narrowing the filter bandwidth to within 0.6 nanometers using a multi-cavity interference filter group. Combined with a deep suppression capability exceeding 10^4 out-of-band suppression ratio, the intensity of solar background light is reduced to 0.08% of its original intensity. Simultaneously, a short-pulse laser with a pulse width of only 8 nanoseconds and a time-gated circuit with a duration of 60 nanoseconds are used. Utilizing the time concentration characteristics of the pulse laser, the effective interaction time of continuous background light is compressed to 0.04%. The differential detection channel further deducts residual background noise in real time. The synergistic effect of the dual filtering mechanisms in the optical and temporal domains enables the lidar to reliably extract weak echo signals even under strong background light conditions, significantly improving the all-weather continuous monitoring capability.

[0064] Furthermore, this invention also solves the technical problem of systematic deviation between instantaneous distance measurements and the true average tide level caused by sea wave motion. Lidar measurements obtain the instantaneous sea surface height, while tide level monitoring requires the average sea surface height after deducting wave effects. The height difference between wave crests and troughs relative to the average sea surface causes statistical bias in randomly sampled instantaneous distance sequences, which cannot be eliminated by direct averaging. This invention establishes a statistical bias correction coefficient as the third element of the tide level mapping vector. This coefficient is determined by looking up a table based on wave height and wave period parameters. The lookup table for the correction coefficient was established through long-term comparison of lidar measurement distance and pressure-type tide gauge measurement distance. For 120 combinations of wave height and period, the average ratio of the two measurement methods was statistically calculated. A quantitative mapping relationship was established between the instantaneous sampling statistical characteristics of lidar and the time integral smoothing characteristics of the tide gauge. This correction coefficient is introduced into the tide level calculation process for compensation, effectively eliminating the systematic measurement bias caused by wave motion, and ensuring consistency between lidar tide level measurement results and traditional tide gauges.

[0065] Specifically, the principle of this invention is that it can solve the technical problem of unstable accuracy of lidar tide level measurement under complex sea conditions. The fundamental reason is that the multi-frequency and multi-angle collaborative measurement mechanism breaks through the sensitivity limitation of a single measurement channel to changes in sea conditions. As the sea surface transitions from calm to rough, the specular reflection component of short-wavelength lasers weakens while the diffuse reflection component of long-wavelength lasers remains relatively stable. Channels with small incident angles are significantly affected by sea surface tilt, while channels with large incident angles exhibit better geometric robustness. The signal quality differences between channels are quantified in real time using the sea surface reflection interference matrix. The sea state identification model utilizes a time-folding coding layer to capture the temporal patterns of sea state evolution and accurately outputs the current sea surface roughness parameters. The wavelength angle response uniformity matrix, based on the sea state and optimal weight mapping relationship established through offline calibration experiments, assigns fusion weights matching the signal quality of each channel. High-quality channels receive greater weights while low-quality channels receive reduced weights. The weighted fusion of the comprehensive echo signal effectively suppresses the performance degradation of a single channel under unfavorable sea conditions. The two-layer game optimization model further balances the contradictory relationship between wavefront reconstruction accuracy and ranging error, adaptively determining the optimal signal processing parameters. The entire technical solution forms a closed-loop control logic from multi-source information acquisition to adaptive sea state processing and then to optimized parameter output. Therefore, the technical solution of this invention conforms to scientific principles and has inherent consistency.

[0066] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0067] In this embodiment, the specific implementation of step S01 is the same as described above, and will not be repeated in detail here.

[0068] The specific implementation of step S02 is as follows: detector gain coefficient and laser emission power The adjustment formula is expressed as follows:

[0069] ;

[0070] ;

[0071] In the formula, This represents the detector gain coefficient for the current cycle, expressed in V / W. This is the detector gain coefficient for the previous cycle, expressed in V / W. To set the threshold, the value is 15. ; This represents the average echo intensity of the previous period, in units of... ; This represents the laser emission power for the current cycle, expressed in W. This represents the laser emission power of the previous cycle, measured in W. Among these, The initial empirical value is 500V / W. The initial empirical value is 200W. The average value is obtained by sampling the peak power of the sea surface echo signal during the previous measurement period and calculating the time domain average. The duration of the previous measurement period is 100 ms. In the above formula... Item and The terms are all powers of dimensionless ratios, serving as adjustment factors for the gain coefficient and power, respectively. The dimensionless ratios ensure dimensional matching between the left and right sides of the formula, with the exponents 0.5 and 0.3 being empirically optimized parameters. This adjustment formula achieves non-linear adjustment of gain and power through a power function. When the echo intensity is below a set threshold, the gain and power are increased; when the echo intensity is above the set threshold, the gain and power are decreased. The exponents 0.5 and 0.3 control the adjustment sensitivity of gain and power, ensuring that the received signal amplitude remains stable within the detector's linear response range, avoiding saturation distortion or excessively low signal-to-noise ratio.

[0072] The specific implementation of step S03 is as follows: detector gain factor The adjustment formula is expressed as follows:

[0073] ;

[0074] In the formula, The detector gain factor is dimensionless. The current sea surface echo signal strength, in units of ; The threshold value for sea surface echo signal strength is 10. The gate opening time of time-gated filtering. and closing time The calculation formula is expressed as follows:

[0075] ;

[0076] ;

[0077] In the formula, The time of gate opening is expressed in seconds (s). The time of gating closure is expressed in seconds (s). The installation height of the multi-frequency lidar equipment is shown in meters (m). This is the historical median sea level height, in meters (m). The speed of light is 299,792,458 m / s. The value range is 20–50 m. The median is calculated from sea level height observation data over the past year, and is typically taken as 0m. In the formula above... This represents the estimated round-trip time of the laser, in seconds, minus... s as lead time, The speed of light is expressed in m / s. This gating timing design achieves temporal filtering of continuous background light by estimating the echo arrival time window. The gating start time is 20ns in advance to provide timing tolerance, and the continuous operation for 60ns covers the echo signal time, reducing the proportion of background light radiation time to 0.04%. After deducting residual noise with the differential detection channel, a clean laser echo signal is obtained.

[0078] The specific implementation of step S04 is as follows: timestamp alignment is achieved by recording the time difference between the arrival time of the echo signal of each channel and the trigger time of the laser pulse emission. The formula for amplitude normalization is expressed as follows:

[0079] ;

[0080] In the formula, For the first The normalized amplitude of each channel is dimensionless. For the first Peak amplitude of echo signal for each channel, in V; For the first The maximum measurable amplitude of each channel, in V, is determined by the detector saturation threshold, with an empirical value of 10V. This is the channel index, ranging from 1 to 3, corresponding to the three incident angle channels of 30 degrees, 45 degrees, and 60 degrees, respectively. Sea surface reflection interference matrix elements. The calculation formula is expressed as follows:

[0081] ;

[0082] ;

[0083] ;

[0084] In the formula, For the first The incident angle and the first The dimensionless evaluation value of the echo signal quality of a combination of laser wavelengths; The peak signal-to-noise ratio of the echo is dimensionless. The symmetry of the echo waveform is dimensionless. For the first The incident angle and the first Peak power of the echo signal for each wavelength combination, in units of ; For the first The incident angle and the first The standard deviation of noise power for each wavelength combination, in units of ; For the first The incident angle and the first The rise time of the echo for each wavelength combination, in nanoseconds; For the first The incident angle and the first The fall time of the echo for each wavelength combination, in nanoseconds; This is the incident angle index, with values ​​from 1 to 3, corresponding to incident angles of 30 degrees, 45 degrees, and 60 degrees. This is a wavelength index, with values ​​from 1 to 3, corresponding to wavelengths of 532nm, 1064nm, and 1550nm. By collecting the first The incident angle and the first The signal amplitude and standard deviation of the echo signal within the first 100 ns time window before the arrival of each wavelength combination are obtained. and By detecting the first The incident angle and the first The time intervals between the rise and fall of the echo signal from 10% peak amplitude to 90% peak amplitude and from 90% peak amplitude to 10% peak amplitude for each wavelength combination are obtained. This matrix reflects the sea surface state by quantifying the echo quality characteristics of different angles and wavelength combinations. For calm sea surfaces, the matrix elements for small-angle, short-wavelength combinations are larger, while for rough sea surfaces, the matrix elements for large-angle, long-wavelength combinations are larger. The temporal attention weight coefficients of the temporal folding coding layer in the sea state recognition model are also included. The calculation formula is expressed as follows:

[0085] ;

[0086] In the formula, For the first The weighting coefficients for each historical moment are dimensionless. For the current moment and the first The time interval between historical moments, in seconds; This is a time scale parameter, with a value of 0.5s; The index for summing historical moments, with values ​​ranging from 1 to 12; This is a historical time index, with values ​​ranging from 1 to 12; It is a natural exponential function. This weighting coefficient is assigned a higher weight to recent times through a Gaussian decay function, with the weight increasing as the time interval becomes more recent. The time scale parameter controls the rate at which the weight decays with the time interval.

[0087] The specific implementation of step S05 is as follows: laser pulse emission frequency The adjustment formula is expressed as follows:

[0088] ;

[0089] In the formula, The adjusted laser pulse emission frequency is expressed in Hz. The original laser pulse emission frequency is set to 10Hz. The confidence level is dimensionless and ranges from 0 to 1. The calculation formula is expressed as follows:

[0090] ;

[0091] ;

[0092] ;

[0093] In the formula, The entropy value of the probability distribution of the output result, in bits; This is the maximum entropy value, expressed in bits. The output count is 10. For the first The predicted probabilities of each category, dimensionless, ranging from 0 to 1, satisfying the following conditions: ; This is a category index, with values ​​ranging from 1 to 10; This is a logarithmic function to the base 2. The confidence level is measured by the entropy value, which indicates the certainty of the prediction result. The smaller the entropy value, the more certain the prediction, and the higher the confidence level. This adjustment strategy dynamically optimizes the allocation of measurement resources based on the confidence level output by the sea state identification model. When the confidence level is low, the sampling frequency is increased to improve identification accuracy, and when the confidence level is high, the sampling frequency is decreased to save energy.

[0094] The specific implementation of step S06 is as follows: synthesize the echo signal. The weighted fusion formula is expressed as follows:

[0095] ;

[0096] In the formula, To synthesize the echo signal at the sampling time The amplitude, in V; For the first Sea surface roughness level 1 The optimal weighting coefficient for each wavelength-angle combination is dimensionless. For the first The incident angle and the first Each wavelength at the sampling time The amplitude of the echo signal, in V; The time interval for signal sampling is measured in seconds (s). This is a sea surface roughness level index, determined based on the sea surface roughness parameters output by the sea state identification model, with values ​​ranging from 1 to 5. For wavelength-angle combination index, the calculation formula is as follows: The value ranges from 1 to 9; This is the incident angle index, with values ​​ranging from 1 to 3; The wavelength index ranges from 1 to 3. This weighted fusion mechanism achieves the optimal combination of multi-channel signals through adaptive weight allocation. In calm sea conditions, short-wavelength, small-angle channels have higher weights, while in rough sea conditions, long-wavelength, large-angle channels have higher weights, ensuring that the composite echo signal has high signal-to-noise ratio and stable waveform characteristics under various sea conditions.

[0097] The specific implementation of step S07 is as follows: the objective function of the upper-level model. The formula is expressed as follows:

[0098] ;

[0099] In the formula, The objective function value of the upper-level model is dimensionless. The spatial variance of the instantaneous distance sequence is given in units of 1. ; The standard spatial variance is 0.25. ; The time variance of the instantaneous distance sequence, in units of ; This represents the standard time variance, with a value of 0.16. ; The wavefront fitting residual is expressed in meters. The standard wavefront fitting residual is 0.08m. This is the distance-wave coupling term, in m / s; The standard time derivative is 0.12 m / s. The objective function of the lower-level model. The formula is expressed as follows:

[0100] ;

[0101] In the formula, This represents the objective function value of the lower-level model, which is dimensionless. The standard deviation of distance measurement repeatability is expressed in meters. The standard deviation of the standard distance measurement repeatability is 0.03m. This is the absolute value of the distance measurement deviation, in meters (m). This is the absolute value of the standard distance measurement deviation, taken as 0.05m. Distance-fluctuation coupling term. The calculation formula is expressed as follows:

[0102] ;

[0103] In the formula, For the first The peak position time of the composite echo signal corresponding to each instantaneous distance is given in seconds. The sampling time interval is 0.1s. This is the instantaneous distance sequence index, with values ​​ranging from 1 to 99; Indicates all The value is the maximum value. In the formula above... The rate of change of the peak position over time, in dimensionless form, multiplied by the speed of light. Then convert to the unit m / s for matching The dimensions of the quantity, i.e. The game optimization process employs an alternating iterative algorithm, with a fixed peak detection threshold in the first round. Noise level The upper-level model is solved by applying a factor of 5. In the second round, the upper-level output is fixed to solve the lower-level model. This process is repeated iteratively until the objective function converges to a value less than 0.01. The root mean square value of the noise amplitude in the synthesized echo signal, in volts (V), is calculated using the following formula: , For the first The signal amplitude at each noise sampling point, in V. The number of noise sampling points. This serves as an index for noise sampling points. The two-layer game theory model links the upper and lower objectives through a distance-wave coupling term. The upper layer prioritizes wavefront reconstruction accuracy by selecting a higher fitting order, while the lower layer prioritizes minimizing ranging error by selecting a lower detection threshold. The two objectives mutually constrain each other to achieve a balanced optimization, ensuring the accuracy and stability of tide level measurements.

[0104] The specific implementation method of step S08 is as follows: laser round trip time The calculation formula is expressed as follows:

[0105] ;

[0106] In the formula, The round-trip time of the laser is expressed in seconds (s). The detection time corresponding to the echo peak point is expressed in seconds. The moment of laser pulse emission, measured in seconds. Instantaneous distance. The calculation formula is expressed as follows:

[0107] ;

[0108] In the formula, This is the instantaneous distance, in meters (m). The speed of light is taken as 299,792,458 m / s. The wavefront fitting process uses least-squares polynomial fitting, and the fitting polynomial is... The expression is:

[0109] ;

[0110] In the formula, To fit the polynomial at the observation time The function value, in units of m; For the first The polynomial coefficient of the term, when The unit is m. The unit is m / s. The unit of time is Generally expressed as ; The optimal wavefront fitting order is 3 to 8. The observation time is expressed in seconds (s). This is the polynomial degree index, with values ​​ranging from 0 to... Polynomial coefficients Solving using the least squares method, the normal system of equations is expressed as follows:

[0111] ;

[0112] In the formula, The index for summing coefficients ranges from 0 to 1. ; This is an index for instantaneous distance data, with values ​​ranging from 1 to 100; For the first Each observation time is measured in seconds (s). For the first Instantaneous distance data, in meters. Average sea level height. The calculation formula is expressed as follows:

[0113] ;

[0114] ;

[0115] In the formula, This represents the average height of sea level, in meters (m). The midpoint of the observation time is expressed in seconds. The first observation time is expressed in seconds. The timeframe is the 100th observation, measured in seconds. This fitting process eliminates the wave component in the instantaneous distance sequence through polynomial approximation, extracts the average sea surface height information, and the fitting order is determined by a two-layer game model to balance wavefront reconstruction accuracy and ranging error.

[0116] The specific implementation method of step S09 is as follows: real-time tide level value The calculation formula is expressed as follows:

[0117] ;

[0118] In the formula, This is the real-time tide level, in meters (m). The first element of the tide level mapping vector represents the vertical height of the multi-frequency lidar equipment installation point relative to the local tide level reference surface, in meters. This represents the average height of sea level, in meters (m). The second element of the tide level mapping vector represents the cosine of the angle between the optical axis and the vertical direction, which is dimensionless. The third element of the tide level mapping vector represents the statistical deviation correction coefficient, which is dimensionless. The accuracy was determined through high-precision leveling, with a measurement accuracy of 2mm. The value is obtained through real-time monitoring using a tilt sensor, and the calculation formula is: , The angle between the optical axis and the vertical direction is expressed in degrees, and the sensor accuracy is 0.01 degrees. Based on wave height and wave cycle According to the table, the value range is 0.92 to 0.98. The unit is meters. The unit is seconds (s). Deviation value The calculation formula is expressed as follows:

[0119] ;

[0120] ;

[0121] In the formula, This is the deviation value, in meters (m). This represents the average historical tide level data at the same tide time, in meters. For the first Historical tidal level data for the same tidal time, in meters; The number of historical data points for the same tide; This is a historical data index, with values ​​ranging from 1 to... This mapping transformation formula converts the sea surface height in the lidar ranging coordinate system to the local tide level reference coordinate system, taking into account the systematic effects of installation height, optical axis tilt angle, and wave statistical deviation, thus achieving high-precision tide level measurement.

[0122] The specific implementation of step S10 is as follows: the tidal level change trend vector is a 24-dimensional vector, and the tidal level change rate in the previous 12 hours... The calculation formula is expressed as follows:

[0123] ;

[0124] In the formula, The first time before the current moment The rate of change of tidal level per hour, expressed in m / h; The current time is expressed in hours (h). Before the current moment Hourly measured tide level, in meters; This is a time interval, with a value of 1 hour. The index represents past times, with values ​​ranging from 1 to 12. The rate of tidal change over the next 12 hours is predicted using harmonic analysis to determine the predicted tidal value. The calculation formula is expressed as follows:

[0125] ;

[0126] In the formula, For the predicted time The tidal level value, in meters (m); Future time, in hours; For the first The amplitude of each simple harmonic component is expressed in meters (m). For the first The angular frequency of a simple harmonic component, in rad / h; For the first The phase of each simple harmonic component, in rad; This is a simple harmonic component index, with values ​​ranging from 1 to 8, including two semi-diurnal tides, two diurnal tides, two shallow water tides, and two long-period tides. The rate of tidal change in the following 12 hours. The calculation formula is expressed as follows:

[0127] ;

[0128] In the formula, The first time after the current time The rate of change of tidal level per hour, expressed in m / h; The index is set to future times, with values ​​ranging from 1 to 12. This tidal level change trend vector provides short-term tidal level evolution patterns by integrating historical observations and harmonic forecast information. When a sudden change occurs in the vector element or the deviation from the forecast exceeds 0.5 m / h, an abnormal tidal warning is triggered, realizing the forecasting and warning functions of tidal level monitoring.

[0129] It should be noted that this invention utilizes a nonlinear adjustment formula for detector gain and laser power. and Dynamic and stable control of the echo signal amplitude was achieved. The power exponents of 0.5 and 0.3 were determined experimentally to ensure the detector operates within its optimal linear response range. This avoids insufficient signal-to-noise ratio under weak echo conditions and saturation distortion under strong echo conditions, significantly improving measurement reliability in complex sea states. The sea surface reflection interference matrix was determined through... A multidimensional characterization system for sea state characteristics was established by quantifying echo quality assessment values ​​for different combinations of incident angles and wavelengths, including the signal-to-noise ratio term. The waveform symmetry term reflects the echo intensity level. Reflecting the time-domain characteristics of the echo, the product of the two is used to comprehensively evaluate the echo quality, combined with a time-folded multilayer sensing network and time-attention weight coefficients. This enables rapid and accurate identification of sea surface state evolution, providing a scientific basis for subsequent adaptive weight allocation. The two-level game optimization model utilizes distance-wave coupling terms... The objective function to maximize the accuracy of associated wavefront reconstruction and the objective function of minimizing ranging error Two mutually constraining objectives were addressed using an alternating iterative algorithm to determine the optimal wavefront fitting order and peak detection threshold. This approach balanced and optimized tidal level measurement accuracy and stability by controlling random distance measurement errors while ensuring high wavefront detail capture. (Tidal level mapping vector transformation formula) Taking into account three systematic influencing factors—installation height reference, optical axis tilt compensation, and wave statistical deviation correction—the lidar ranging coordinate system is accurately transformed to the local tide level reference coordinate system, eliminating systematic deviations caused by attitude changes and random wave sampling. Combined with harmonic analysis of tide change trend prediction, millimeter-level precision real-time tide level monitoring and abnormal tide early warning capabilities are achieved.

[0130] To better understand and implement this invention, a specific application scenario is provided below as Example 2: A technical team is implementing a real-time tide level monitoring project in a coastal port area. This area has a large tidal range and complex sea conditions, making traditional tide level monitoring methods insufficient to meet real-time and accuracy requirements. The technical team adopted the multi-frequency lidar real-time tide level monitoring method of this invention, installing a multi-frequency lidar device 25m above the port pier. The device is equipped with laser emission modules for three wavelengths: 532nm, 1064nm, and 1550nm, with an initial laser emission power set at 280W. The multi-cavity interference filter group is designed with a 5-layer Fabry-Perot cavity cascade structure, achieving a filter bandwidth of 0.52nm at a wavelength of 532nm. The out-of-band suppression ratio effectively reduces the intensity of solar background light to 0.08% of that before filtering.

[0131] The technical team set the incident angles of the three laser emission channels to 30 degrees, 45 degrees, and 60 degrees, respectively, with each channel having a laser pulse emission frequency of 10 Hz and a pulse width of 8 ns. At the start of monitoring, the system calculated the estimated echo arrival time to be 173 ns based on historical sea surface height statistics ranging from -1.8 m to 2.2 m. The gating circuit was activated at 153 ns and deactivated after 60 ns of continuous operation. The differential detection channel synchronously acquired the background light intensity, and the pure laser echo signal was obtained by subtracting the differential detection channel signal from the main detection channel signal. During one monitoring cycle, the average echo intensity of the previous measurement period was 8.3. 15 below the set threshold The system automatically increased the detector gain from 2500V / W to 3750V / W and the laser emission power from 280W to 320W, while simultaneously achieving a background light noise level of 235. Exceeding the threshold of 200 System startup time gating filtering processing.

[0132] like Figure 2 As shown, after timestamp alignment, the echo arrival times of the sea surface signals acquired from the three incident angles are 174.2 ns for the 30-degree channel, 175.8 ns for the 45-degree channel, and 177.5 ns for the 60-degree channel. Amplitude normalization was performed by dividing the peak amplitude of the echo signals from each of the three channels by their respective maximum measurable amplitude of 4800 mV, resulting in normalized amplitudes of 0.78, 0.82, and 0.75, respectively. The technical team established a sea surface reflection interference matrix, as shown in Table 1.

[0133] Table 1 Sea Surface Reflection Interference Matrix

[0134]

[0135] The matrix elements are calculated by multiplying the peak signal-to-noise ratio (PSNR) of the echo by the echo waveform symmetry. For example, with an incident angle of 45 degrees and a wavelength of 532 nm, the PSNR is 18.5, the rise time is 3.2 ns, the fall time is 3.5 ns, and the echo waveform symmetry is 0.827. Multiplying these two values ​​yields a matrix element value of 15.3. The sea surface reflection interference matrix is ​​expanded into a 9-dimensional feature vector and input into the sea state identification model, such as... Figure 3As shown, the model processes the input features from 12 consecutive time points through a time-folding encoding layer, with the time scale parameter set to 0.5 s. The feature vector from the current monitoring time point is weighted and fused with the feature vectors from the previous 11 historical time points using a time-attention mechanism. The weight coefficient for the current time point is 0.285, the weight coefficient for the time point one cycle ago is 0.193, and the weight coefficient for the time point two cycles ago is 0.124. After the 72-dimensional fused feature vector undergoes a nonlinear transformation through a three-layer fully connected network, the output layer generates a sea surface roughness parameter of 2.7 and a wave level of 4, with a confidence level of 81%.

[0136] Since the confidence level of 81% falls between 75% and 92%, the system maintains a constant laser pulse emission frequency of 10Hz. The technical team consulted the wavelength-angle response uniformity matrix based on the sea surface roughness parameter 2.7, which corresponds to the third sea surface roughness level range, 2 to 3, as shown in Table 2.

[0137] Table 2 Elements in the 3rd row of the wavelength-angle response uniformity matrix

[0138]

[0139] The sea surface echo signals from the three incident angles are weighted and fused using optimal weighting coefficients to obtain a composite echo signal. The absolute value of the time derivative of the peak position of the composite echo signal is 0.087 m / s. This composite echo signal is then input into a two-layer game-theoretic optimization model. The upper-layer model aims to maximize the wavefront reconstruction accuracy, and the spatial variance of the instantaneous distance sequence is 0.28. The time variance is 0.19. The initial wavefront fitting order was set to 5, the wavefront fitting residual was 0.076m, and the distance-wave coupling term was 0.087m / s. The lower-level model aimed to minimize ranging error, with the initial peak detection threshold set to 5 times the noise level (0.65mV), the ranging repeatability standard deviation to 0.028m, and the absolute value of the ranging bias to 0.043m. After four rounds of alternating iterative optimization, the change in the objective function value of the upper-level model was 0.008, and the change in the objective function value of the lower-level model was 0.007. The system converged and output the optimal wavefront fitting order of 6 and the optimal peak detection threshold of 0.78mV.

[0140] like Figure 4As shown, peak detection was performed on the composite echo signal according to the optimal peak detection threshold. The time corresponding to the echo peak point was 175.3 ns. Subtracting the laser pulse emission time (i.e., time zero) yielded the laser round-trip time of 175.3 ns. The instantaneous distance was calculated to be 26.28 m using the time-distance conversion relationship. 100 sets of instantaneous distance data were continuously collected, with the instantaneous distance sequence ranging from 25.82 m to 26.75 m. Least square polynomial fitting was performed using the optimal wavefront fitting order of 6. The function value of the fitted polynomial at the midpoint of the observation time was 26.31 m, which corresponds to the average sea surface height of 26.31 m. The vertical height of the multi-frequency lidar equipment installation point relative to the local tide datum was 25.00 m, and the cosine of the angle between the optical axis and the vertical direction was 0.998. Based on the current wave height of 1.8 m and wave period of 5.5 s, the statistical deviation correction coefficient was obtained from the statistical deviation correction coefficient lookup table, resulting in a statistical deviation correction coefficient of 0.95. The real-time tide level was calculated to be -1.24 m.

[0141] The technical team compared the real-time tide level of -1.24m with historical tide data. The current tidal phase is 8.3 hours. In the past 30 days of historical tide data, they searched for 28 records with a tidal phase difference of less than 0.5 hours. The average tide level of these records was -1.18m, and the absolute value of the deviation was 0.06m. Since the deviation of 0.06m is greater than the small deviation threshold of 0.05m but less than the large deviation threshold of 0.3m, the system maintains the current monitoring mode. Figure 5 As shown, the technical team established a tidal level change trend vector. The rate of tidal level change for the first 12 hours was calculated based on historical measured tidal level data. For example, the rate of tidal level change for the first hour at the current moment was -0.32 m / h, for the first two hours it was -0.28 m / h, and for the first three hours it was -0.21 m / h. The rate of tidal level change for the next 12 hours was predicted using harmonic analysis. Harmonic analysis uses eight simple harmonic components, including two semi-diurnal tides, two diurnal tides, two shallow water tides, and two long-period tides. The predicted rate of tidal level change for the next hour at the current moment was -0.18 m / h, for the next two hours it was -0.08 m / h, and for the next three hours it was 0.05 m / h. The elements of the tidal level change trend vector showed a trend of first decreasing and then increasing, which is consistent with the normal variation pattern of semi-diurnal tides. The system did not trigger an abnormal tidal warning.

[0142] After 30 days of continuous monitoring, the technical team's statistical analysis showed that the system could stably output real-time tide levels under different sea conditions. In calm sea conditions, the weighting coefficient for a 30-degree incident angle combined with a 532nm wavelength reached 0.18, resulting in optimal specular reflection echo quality for short-wavelength and small-angle channels. In rough sea conditions, the weighting coefficient for a 60-degree incident angle combined with a 1550nm wavelength increased to 0.22, leading to better stability of diffuse reflection echoes for long-wavelength and large-angle channels. The two-layer game-theoretic optimization model dynamically adjusts the wavefront fitting order and peak detection threshold based on the characteristics of the integrated echo signal, maintaining a balance between ranging accuracy and wavefront reconstruction accuracy when sea conditions change. The multi-point cross-validation mode is activated when the deviation value is greater than 0.3m, improving the reliability of identifying abnormal tidal events through spatial redundancy measurements.

[0143] This invention represents a significant technological advancement over traditional tide monitoring methods. Traditional pressure-type tide gauges calculate tide levels by measuring underwater pressure, but their accuracy and response speed are limited by variations in seawater density and wave disturbances. This invention employs a multi-frequency lidar non-contact measurement method, utilizing combinations of different wavelengths and incident angles to adapt to varying sea conditions. Through a sea state identification model, it adaptively adjusts measurement parameters and weight allocation strategies, achieving high-precision real-time monitoring in all weather conditions. The ultra-narrowband filtering function and time-gated filtering of the multi-cavity interferometric filter group effectively suppress solar background light interference, maintaining a high signal-to-noise ratio even under strong light conditions. A two-layer game-theoretic optimization model, through the coordinated optimization of wavefront reconstruction accuracy and ranging error, accurately extracts the average sea surface height under wave motion conditions, avoiding interference from instantaneous wave crests and troughs in tide level measurement. The statistical deviation correction coefficient introduced by the tide level mapping vector compensates for systematic biases caused by random wave sampling, and the tide level change trend vector provides temporal characteristic information for short-term tide level forecasts and abnormal tide warnings. These technological innovations make this invention superior to traditional methods in terms of monitoring accuracy, real-time performance, and sea condition adaptability, providing reliable tidal level information support for port scheduling, waterway management, and marine disaster early warning.

[0144] It should be noted that the variables involved in this invention are explained in detail in Tables 3 and 4.

[0145] Table 3. Variable Explanation Table (Part 1)

[0146]

[0147] Table 4. Variable Explanation Table (Part Two)

[0148]

[0149] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring of sea surface tides based on multi-frequency lidar, characterized in that, A multi-frequency lidar device is installed above the monitored sea area. The device emits laser pulses at an oblique angle to the sea surface and acquires sea surface echo signals. The detector gain coefficient and laser emission power are adjusted according to the average echo intensity. The detector gain factor and time-gated filtering are adjusted according to the sea surface echo signal intensity and background light noise level. After timestamp alignment and amplitude normalization of the sea surface echo signal, a sea surface reflection interference matrix is ​​established. This matrix is ​​input into a sea state recognition model to determine the sea surface state. The laser pulse emission frequency is adjusted according to the confidence level. The wavelength-angle response uniform matrix is ​​queried and weighted fusion is performed to obtain a comprehensive echo signal. The signal is input into a two-layer game optimization model to output the optimal wavefront fitting order and the optimal peak detection threshold. Peak detection is performed based on the optimal peak detection threshold, and the laser round-trip time is calculated. The instantaneous distance is calculated through the time-distance conversion relationship. An instantaneous distance sequence is established, and the average sea surface height is obtained by wavefront fitting based on the optimal wavefront fitting order. The real-time tide level is obtained by combining the tide level mapping vector conversion. A tide level change trend vector is established for short-term tide level forecasting and abnormal tide warning. The structure of the sea state recognition model is a time-folded multilayer perceptron. The input layer receives a 9-dimensional feature vector, and the input layer is connected to a time-folded coding layer. The time-folded coding layer folds the input features of 12 consecutive time moments in the time dimension. The temporal folding coding layer is configured with 12 parallel feature extraction sub-networks that process the input features at 12 time points respectively. The outputs of the 12 feature extraction sub-networks are weighted and fused through a temporal attention mechanism. The wavelength-angle response uniform matrix is ​​a 5-row, 9-column matrix. The row index corresponds to 5 sea surface roughness parameter levels, and the column index corresponds to 9 wavelength-angle combinations. The matrix elements represent the optimal weighting coefficients of the wavelength-angle combinations under the corresponding sea surface roughness parameter levels. The optimal weighting coefficients are determined through offline calibration experiments. The two-layer game optimization model includes an upper-layer model that aims to maximize wavefront reconstruction accuracy and a lower-layer model that aims to minimize ranging error. The upper-layer model and the lower-layer model are associated through a distance-wave coupling term.

2. The method according to claim 1, characterized in that, The multi-frequency lidar device includes three laser emitting modules with different wavelengths and a matching narrowband filter receiving module.

3. The method according to claim 2, characterized in that, The narrowband filter receiving module uses a multi-cavity interference filter group to achieve ultra-narrowband filtering function.

4. The method according to claim 3, characterized in that, The oblique angle includes three incident angles: 30 degrees, 45 degrees, and 60 degrees.

5. The method according to claim 4, characterized in that, The average echo intensity is the average value of the peak power of the sea surface echo signal received in the previous measurement period in the time domain.

6. The method according to claim 5, characterized in that, When the sea surface echo signal strength is below 10mW / And the background light noise level is higher than 200W / At the same time, the detector gain factor and startup time gating filtering are enhanced.

7. The method according to claim 6, characterized in that, The time-gated filtering process achieves background noise suppression by utilizing the temporal characteristics of the pulsed laser.

8. The method according to claim 7, characterized in that, The sea surface reflection interference matrix is ​​a 3x3 matrix, with row indices corresponding to three incident angles and column indices corresponding to three laser wavelengths. The matrix elements represent the echo signal quality assessment values ​​under different incident angles and different wavelength combinations.

9. The method according to claim 8, characterized in that, The echo signal quality assessment value is calculated by multiplying the echo peak signal-to-noise ratio by the echo waveform symmetry.

Images