An inversion method for detecting subsurface water body elements by spaceborne lidar

Abstract

The application discloses an inversion method for detecting subsurface water element by a spaceborne lidar, comprising: performing mean filtering on echo profile signals; extracting land surface data, calculating a discrete transient response function, and performing fitting; then performing transient response correction on the filtered lidar measurement signals; extracting marine neighbor data, and then sequentially calculating an integral attenuation coefficient, correcting the integral attenuation coefficient and a parallel and vertical correlation coefficient, obtaining a polarization crosstalk coefficient, and performing polarization crosstalk correction on the transient response corrected signals; sequentially calculating a total depolarization ratio of the water body and a water body backscattering coefficient; matching the spaceborne lidar and the corresponding data of the existing water body parameters according to time and latitude and longitude, then taking the total depolarization ratio, the water body backscattering coefficient and the latitude as variables, and taking the water body parameters as true values to input a deep neural network for learning, obtaining a trained deep learning network, and predicting the water body parameters. The method can quickly and accurately invert the water element.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of ocean lidar remote sensing detection, and particularly relates to an inversion method for detecting subsurface water element by a spaceborne lidar. BACKGROUND

[0002] The lidar has been widely used for the inversion of water body parameters, but due to the transient response of the detector in the receiving system of the lidar, after receiving the strong signal of the sea surface reflection, the lidar cannot immediately recover to the detection state, and a tailing effect will occur, which will extend the peak energy to the subsequent several units, so that the signal strength measured subsequently is higher than the actual backscattering signal, and then the subsequent signal processing is affected. Therefore, the water body echo signal must be deconvoluted to eliminate the influence of the transient response of the detector, so as to restore the real signal reaching the detector. In addition, for the polarization lidar, due to the non-ideal characteristics of the polarization beam splitter, part of the parallel polarized signal will be converted into the vertical polarized signal, so that the signal of the vertical channel is enhanced, and the signal of the parallel channel is weakened, and then the subsequent signal processing is affected. Therefore, the crosstalk of the polarization beam splitter must be corrected to eliminate its influence, so as to restore the real parallel channel and vertical channel signals. According to the process that the lidar signal first passes through the polarization beam splitter and then is received by the detector, before further inverting the water body parameters, the deconvolution of the transient response of the detector and the correction of the polarization crosstalk of the polarization beam splitter should be performed on the measured signal of the lidar, but the current research either corrects the transient influence of the detector or corrects the polarization crosstalk, and lacks a processing method for correcting both.

[0003] When the water body parameters are inverted, the traditional lidar water body parameter inversion method needs to use the diffuse attenuation coefficient of the passive remote sensing satellite and needs to assume the relationship between the backscattering coefficient of the water body particle and the 180° volume scattering coefficient. However, the conversion parameter between the backscattering coefficient of the water body particle and the 180° volume scattering coefficient is not a fixed value, and the uncertainty of the value introduces additional error for the inversion of the water body parameters. In addition, the conventional inversion method first converts the total depolarization ratio of the lidar signal into the water body backscattering coefficient, and then inverts the water body parameters according to the water body backscattering coefficient. However, there is a certain information loss in the conversion process. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides an inversion method for detecting subsurface water element by a spaceborne lidar, which corrects the transient response of the detector and the polarization crosstalk of the polarization beam splitter on the original measured signal, and uses a deep learning training model to accurately invert the water body parameters.

[0005] The object of the present application is achieved by the following technical solutions:

[0006] The application discloses an inversion method for detecting subsurface water body elements by a spaceborne lidar, and the method comprises the following steps.

[0007] Step one: mean filtering is performed on echo profile signals of the spaceborne lidar;

[0008] Step two: land surface data detected by the spaceborne lidar is extracted, a discrete transient response function is calculated, and a transient response analytical function is fitted; then, transient response correction is performed on the filtered lidar measurement signals according to a matrix composed of the transient response analytical functions of the detector;

[0009] Step three: marine neighbor data detected by the spaceborne lidar is extracted, an integral attenuation coefficient is calculated, and then a corrected integral attenuation coefficient and a parallel and vertical correlation coefficient are calculated in sequence to obtain a polarization crosstalk coefficient; polarization crosstalk correction is further performed on the transient response corrected signals according to the polarization crosstalk coefficient;

[0010] Step four: total depolarization ratio of the water body is calculated;

[0011] Step five: water body backscattering coefficient is calculated;

[0012] Step six: data corresponding to the spaceborne lidar and existing water body parameter data are matched according to time and latitude and longitude, then total depolarization ratio, water body backscattering coefficient and latitude are taken as variables, and water body parameters are taken as true values to input a deep neural network for learning, so that a trained deep learning network is obtained;

[0013] Step seven: total depolarization ratio, water body backscattering coefficient and latitude of the spaceborne lidar calculated in steps four and five are input into the trained deep learning network, and water body parameters are output by the network.

[0014] Further, in step one, mean filtering is performed on every 30 echo profile signals, and after filtering, echo signals with an aerosol optical depth less than 3 are selected for extraction of the land surface data, so that the influence of thick clouds on the echo profile signals is avoided.

[0015] Further, the discrete transient response function F'(z j ) is as follows:

[0016]

[0017] Wherein, β'(z j ) is a land surface neighbor signal, p is a position of a peak signal, and i and j are positions of the signal;

[0018] Further, the calculation formula of the transient response correction in step two is as follows:

[0019] β'(z)=F -1 β(z)

[0020] where β'(z) is the signal corrected by the transient response of the detector, β(z) is the signal measured by the laser radar, F is the analytical function of the transient response of the detector F(z i ) is a matrix composed of:

[0021]

[0022] where F(z i ) is calculated by the analytical function of the transient response of the detector F(z) and F(z) is:

[0023]

[0024] where z s is the position of the peak, σ1, σ2, α, σ3 are function parameters estimated by the least square method according to the discrete transient response function F'(z j ) measured by the detector.

[0025] Further, the polarization crosstalk correction in step three is:

[0026]

[0027] β ⊥,c = β ⊥,m - CT x β ||,c

[0028] where β ||,c and β ⊥,c are the signals of the parallel channel and the vertical channel after the polarization crosstalk correction respectively, β ||,m and β ⊥,m are the signals of the parallel channel and the vertical channel after the transient response correction respectively, CT is the polarization crosstalk coefficient, CT p2s is the CT that makes the correlation coefficient of the parallel and vertical signals ρ xy (CT p2s ) minimum, and the correlation coefficient of the parallel and vertical signals ρ xy (CT p2s ) is:

[0029] ρ xy (CT p2s ) = |corr(x(CT p2s ), y)|

[0030] where corr(*) represents the correlation coefficient, |*| represents the absolute value, CT p2s is the candidate crosstalk coefficient with the value from 0 to 0.02, represents the proportion of the parallel polarization converted into the vertical polarization signal, x(CT p2s ) is the correction coefficient of the integral attenuation of the vertical channel, and y is the correction coefficient of the integral attenuation of the parallel channel:

[0031] x(CTp2s ) = γ s,m - CT p2s × γ p,m

[0032] y = γ p,m

[0033] wherein γ s,m , γ p,m are the integral attenuation coefficients of the vertical and horizontal channels respectively:

[0034]

[0035] wherein β' s,m (z j ) and β' p,m (z j ) are the echo signals of the vertical and horizontal channels in the vicinity of the sea surface respectively.

[0036] Further, the total depolarization ratio δ T of the water body in step four is:

[0037]

[0038] Further, the backscattering coefficient γ of the water body in step five is:

[0039]

[0040] wherein β s is the backscattering coefficient of the sea surface:

[0041]

[0042] wherein θ is the nadir angle of the laser beam, and σ 2 is the mean square deviation of the sea wave slope:

[0043] σ 2 = 0.003 + 0.00512W

[0044] wherein W is the sea surface wind speed.

[0045] A computer device comprising a memory, a processor and a computer program stored on the memory, the processor executing the computer program to implement the above inversion method.

[0046] A computer readable storage medium having stored thereon a computer program or instructions, which, when executed by a processor, implement the above inversion method.

[0047] The beneficial effects of the present application are as follows:

[0048] (1) The application calculates the normalized continuous transient response function of the detector by function fitting, eliminates the influence of the sampling signal on the peak position of the transient response, obtains a more accurate detector transient response function, and can more accurately obtain the real signal reaching the detector;

[0049] (2) Based on the weak correlation between the parallel polarization and the vertical polarization signals, the optimal crosstalk coefficient can be estimated, so as to restore the real parallel polarization and vertical polarization signals;

[0050] (3) The detector transient response correction and polarization crosstalk correction of the measured signal can better eliminate the influence of the laser radar receiving system on the signal, so as to restore the real water body echo signal; at the same time, the total depolarization ratio of the spaceborne laser radar and the water body backscattering coefficient are used for the inversion of the water body parameters, so that more water body information can be retained, the best conversion parameter can be selected according to the spatial variation, the additional assumption can be avoided, the inversion of the water body parameters is more accurate, and the method can be used for the inversion of various water body elements, and has wide applicability. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 is a flow chart of the inversion method for detecting subsurface water body elements of a spaceborne laser radar.

[0052] Figure 2 is a detector transient response function, wherein the circle is a discrete transient response function, the solid line is a fitted analytical transient response function, the left graph is a transient response of a parallel polarization channel, and the right graph is a transient response of a vertical polarization channel.

[0053] Figure 3 is a correction result of the detector transient effect, wherein the solid line is a measured water body signal, the dashed line is a corrected water body signal, the left graph is a correction result of a parallel polarization channel, and the right graph is a correction result of a vertical polarization channel.

[0054] Figure 4 is a polarization crosstalk correction result, wherein the solid line is a measured water body signal, the dashed line is a corrected water body signal, the left graph is a correction result of a parallel polarization channel, and the right graph is a correction result of a vertical polarization channel.

[0055] Figure 5 is a comparison between the laser radar inversion result and the in-situ measured data, the left graph is a water body particle backscattering, and the right graph is chlorophyll; the discrete points represent the matching points of the laser radar inversion result and the in-situ measured data, the dashed line represents the linear fitting of the discrete points, and the solid line represents the 1:1 line of the laser radar inversion result and the in-situ measured data. DETAILED DESCRIPTION

[0056] The present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments. The objectives and effects of the present invention will become clearer as a result. The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0057] Existing research has shown that the total depolarization ratio already contains a considerable amount of water body parameter information. This invention uses deep learning to integrate the total depolarization ratio and the relationship between the water body backscattering coefficient and water body parameters. At the same time, it can automatically fit the nonlinear relationship between the lidar echo signal and water body parameters, thereby more accurately inverting water body parameters.

[0058] like Figure 1 As shown, the inversion method for detecting subsurface water elements using spaceborne lidar according to the present invention specifically includes the following steps:

[0059] Step 1: Perform mean filtering on the echo profile signal of the spaceborne lidar;

[0060] This embodiment uses the CALIPSO spaceborne lidar Level-1B V4.1 product to retrieve backscattered particulate matter from water as an example. During echo signal filtering, mean filtering is applied to every 30 echo profile signals. To avoid the influence of thick clouds on the echo profile signals, echo signals with an aerosol optical thickness of less than 3 are selected after filtering for land surface data extraction.

[0061] Step 2: Extract land surface data detected by the spaceborne lidar, calculate the discrete transient response function, and fit the transient response analytical function; then, perform transient response correction on the filtered lidar measurement signal based on the matrix composed of the detector's transient response analytical function.

[0062] Among them, the discrete transient response function F′(z) j )for:

[0063]

[0064] Wherein, β′(z j ) represents the nearest neighbor signal on the land surface, p represents the location of the peak signal, and i and j represent the cell where the signal is located. F′(z) is calculated using the lidar echo signal from January 1, 2008, at coordinates 51.61°S, 17.998°W. j )like Figure 2 As shown in the middle circle.

[0065] The formula for calculating transient response correction is:

[0066] β′(z)=F -1 β(z)

[0067] Wherein, β'(z) is the signal after correction of the instantaneous response of the detector, β(z) is the signal measured by the laser radar, F is the analytical function F(z) of the instantaneous response of the detector i ) matrix composed of:

[0068]

[0069] Wherein, F(z i ) is calculated by the analytical function F(z) of the instantaneous response of the detector, F(z) is:

[0070]

[0071] Wherein, z s is the position of the peak, σ1, σ2, α, σ3 are function parameters, which are estimated by the least square method according to the discrete instantaneous response function F'(z j ) measured by the detector. In this embodiment, the σ1, σ2, α, σ3 of the parallel polarization signal are 18.6, 149, 0.017, 12.7 respectively, and the σ1, σ2, α, σ3 of the vertical polarization signal are 19.2, 149, 0.015, 14.2 respectively. The fitting instantaneous response analytical function F(z) is shown by the solid line in Figure 2 . Figure 3 The deconvolution result of the instantaneous response correction is shown. The solid line is the measured water signal, and the dotted line is the corrected water signal. The left graph is the correction result of the parallel polarization channel, and the right graph is the correction result of the vertical polarization channel. As can be seen from the graph, after correction, the underwater signal is significantly reduced, so that the water surface signal and the water signal can be effectively separated.

[0072] Step three: extracting the marine near neighbor data detected by the spaceborne laser radar, calculating the integral attenuation coefficient, and then calculating the corrected integral attenuation coefficient and the parallel and vertical correlation coefficient in turn to obtain the polarization crosstalk coefficient; and then performing polarization crosstalk correction on the signal after the instantaneous response correction according to the polarization crosstalk coefficient;

[0073] Wherein, the calculation formula of the polarization crosstalk correction is:

[0074]

[0075] β ⊥,c =β ⊥,m -CT×β ||,c

[0076] Wherein, β ||,c and β ⊥,c are the signals of the parallel channel and the vertical channel after polarization crosstalk correction, β ||,m and β ⊥,mrespectively, CT is the polarization cross-talk coefficient, and corr is the correlation coefficient of parallel and vertical signals p2s , the correlation coefficient of parallel and vertical signals is:

[0077] ρ xy (CT p2s )=|corr(x(CT p2s ),y)|

[0078] where corr(*) represents the correlation coefficient, and |*| represents the absolute value. CT p2s is the candidate cross-talk coefficient with a value from 0 to 0.02, and represents the proportion of the parallel polarization converted into the vertical polarization signal. x(CT p2s ) is the corrected vertical channel integral attenuation coefficient, and y is the parallel channel integral attenuation coefficient:

[0079] x(CT p2s )=γ s,m -CT p2s ×γ p,m

[0080] y=γ p,m

[0081] where γ s,m and γ p,m are the integral attenuation coefficients of the vertical channel and the parallel channel, respectively:

[0082]

[0083]

[0084] where β′ s,m (z j ) and β′ p,m (z j ) are the echo signals of the vertical channel and the parallel channel near the ocean surface, respectively. In this embodiment, CT is calculated to be 0.44%. Figure 4 The polarization cross-talk correction results are shown. The solid line is the measured water signal, and the dashed line is the corrected water signal. The left graph shows the correction results of the parallel polarization channel, and the right graph shows the correction results of the vertical polarization channel. As can be seen from the graph, after correction, the vertical polarization signal is significantly reduced, and the parallel polarization signal is slightly enhanced, so that the accurate depolarization ratio can be obtained.

[0085] Step four: calculate the total depolarization ratio of the water body, and the calculation formula is as follows:

[0086]

[0087] where δ T is the total depolarization ratio of the water body.

[0088] Step five: Calculate the water backscattering coefficient γ, the calculation formula is as follows:

[0089]

[0090] Wherein, β s is the sea surface backscattering coefficient:

[0091]

[0092] Wherein, θ is the laser beam zenith angle, σ 2 is the sea wave slope variance:

[0093] σ 2 = 0.003 + 0.00512W

[0094] Wherein, W is the sea surface wind speed.

[0095] Step six: According to the time and latitude and longitude, match the corresponding data of the spaceborne lidar and the existing water parameter data, then take the total depolarization ratio, the water backscattering coefficient and the latitude as variables, and the water parameter as the true value, input the deep neural network for learning, and obtain the trained deep learning network.

[0096] In this embodiment, the water particle backscattering coefficient product of MODIS is matched with the CALIPSO data, input into the model for training to obtain the inversion model of the water particle backscattering coefficient.

[0097] Step seven: input the total depolarization ratio, the water backscattering coefficient and the latitude of all spaceborne lidars calculated in steps four and five into the trained deep learning network, and the network outputs the water parameter.

[0098] In this embodiment, the total depolarization ratio, the water backscattering coefficient and the latitude of CALIPSO are input into the inversion model of the water particle backscattering coefficient, and the water particle backscattering coefficient product of CALIPSO is obtained.

[0099] According to the comparison between the MODIS water particle backscattering monthly product and the CALIPSO inverted water particle backscattering monthly product, the two are very close, and the correlation coefficient of the two is 0.82.

[0100] The chlorophyll inversion model is established by matching the spaceborne lidar CALIPSO data and the MODIS chlorophyll product, the total depolarization ratio, the water backscattering coefficient and the latitude of CALIPSO are input into the established chlorophyll inversion model, and the CALIPSO inverted chlorophyll product is obtained. The MODIS chlorophyll monthly product and the CALIPSO inverted product are compared, and the two are very close.

[0101] Figure 5 is a comparison chart of the laser radar water body parameter inversion result and the in-situ measured result, when the discrete points are closer to the 1:1 line, it indicates that the inversion result is more accurate. Figure 5 The discrete points in the figure are concentrated near the 1:1 line, and the correlation coefficient R 2 is above 0.7. The maximum value of the backscattering coefficient in the left figure is 0.01, the root mean square error RMSE of the inversion is 0.0009, the maximum value of the in-situ measured chlorophyll in the right figure is 10, and the root mean square error RMSE of the inversion is 0.674, so the root mean square error RMSE of both is far less than the maximum value, thereby proving that the inversion method of the present application is feasible and effective.

[0102] Those skilled in the art can understand that the above description is only preferred examples of the present application and is not used to limit the present application, although the present application has been described in detail with reference to the foregoing examples, those skilled in the art can still modify the technical solutions recorded in the foregoing examples or make equivalent replacement for part of the technical features. Any modification, equivalent replacement, etc. within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An inversion method for detecting subsurface water body elements by spaceborne lidar, characterized in that, The method comprises the following steps: Step one: mean filtering of echo profile signals of the spaceborne lidar; Step two: extracting land surface data detected by the spaceborne lidar, calculating the discrete transient response function, fitting the transient response analytical function, and then performing transient response correction on the filtered lidar measurement signals according to the matrix composed of the transient response analytical function of the detector; Step three: extracting marine neighbor data detected by the spaceborne lidar, calculating the integral attenuation coefficient, and then sequentially calculating the corrected integral attenuation coefficient and the parallel and vertical correlation coefficient to obtain the polarization crosstalk coefficient; and performing polarization crosstalk correction on the transient response corrected signals according to the polarization crosstalk coefficient; Step four: calculating the total depolarization ratio of the water body; Step five: calculating the water body backscattering coefficient; Step six: matching the data corresponding to the spaceborne lidar and the existing water body parameter data according to time and latitude and longitude, and then inputting the total depolarization ratio, the water body backscattering coefficient and the latitude as variables and the water body parameter as the true value into the deep neural network for learning to obtain the trained deep learning network; Step seven: inputting all the total depolarization ratios, water body backscattering coefficients and latitudes of the spaceborne lidar calculated in steps four and five into the trained deep learning network, and outputting the water body parameter from the network.

2. The inversion method for detecting subsurface water body elements by spaceborne lidar according to claim 1, characterized in that, In step one, mean filtering is performed on every 30 echo profile signals, and after filtering, echo signals with an aerosol optical depth less than 3 are selected for extraction of land surface data, so as to avoid the influence of thick clouds on the echo profile signals.

3. The inversion method for detecting subsurface water body elements by spaceborne lidar according to claim 1, characterized in that: The discrete transient response function F'(z j ) is: wherein β'(z j ) is the signal of the land surface adjacent, p is the position of the peak signal, i, j is the unit where the signal is located; The calculation formula of the transient response correction in step two is: β'(z) = F -1 β(z) Wherein, β'(z) is the signal after correction of the instantaneous response of the detector, β(z) is the laser radar measurement signal, F is the analytical function of the instantaneous response of the detector F(z i ) matrix composed of: where F(z i ) is calculated from the analytical function F(z) of the detector transient response, F(z) being: where z s is the position of the peak, σ1, σ2, α, σ3are function parameters estimated from the detector measurement discrete transient response function F'(z j ) by least square method.

4. The inversion method for detecting subsurface water body elements by a spaceborne lidar according to claim 3, characterized in that: The polarization crosstalk correction in step three is: β ⊥,c = β ⊥,m - CT x β ||,c wherein β ||,c and β ⊥,c are the signals of the parallel and vertical channels after correction of polarization crosstalk, β ||,m and β ⊥,m are the signals of the parallel and vertical channels after correction of the transient response, CT is the polarization crosstalk coefficient, and CT p2s is the CT for which the correlation coefficient ρ xy (CT p2s ) of the parallel and vertical signals is minimized. p xy (CT p2s ) = |corr(x(CT p2s ), y)| where corr(*) denotes the correlation coefficient, and |*| denotes the absolute value; CT p2s is the candidate crosstalk coefficient with a value from 0 to 0.02, and x(CT p2s ) denotes the proportion of the parallel polarization converted into the vertical polarization signal; x(CT p2s ) is the correction coefficient of the integral attenuation of the vertical channel, and y is the integral attenuation coefficient of the parallel channel. x(CT p2s ) = γ s,m - CT p2s x γ p,m y = γ p,m where γ s,m , γ p,m are the integral attenuation coefficients of the perpendicular and parallel channels, respectively. Where, β′ s,m (z j ) and β′ p,m (z j These are the echo signals from the vertical and parallel channels near the ocean surface, respectively.

5. The inversion method for detecting subsurface water body elements by a spaceborne lidar according to claim 4, characterized in that: The total depolarization ratio δ of the water body in step four T Is:

6. The inversion method for detecting subsurface water body elements by a spaceborne lidar according to claim 5, characterized in that: The water body backscattering coefficient γ in step five is: where β s is the sea surface backscatter coefficient: where θ is the laser beam nadir angle, σ 2 is the sea wave slope variance: σ 2 = 0.003 + 0.00512W Wherein, W is the sea surface wind speed.

7. A computer device comprising a memory, a processor, and a computer program stored on the memory, wherein the computer program comprises instructions that, when executed by the processor, cause the processor to perform the method of any one of claims 1-6. The processor executes the computer program to realize the inversion method of any one of claims 1-6.

8. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions are executed by the processor to realize the inversion method of any one of claims 1-6.

Images