An ice layer parameter inversion system and method based on photoacoustic effect inversion
By employing a non-contact detection method based on photoacoustic effect, lasers are used to excite acoustic signals generated on the sea ice surface. Combined with deep learning inversion models and physical constraints, rapid and accurate ice layer parameter measurements are achieved in polar environments, solving the problems of coupling instability and measurement complexity in traditional methods.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHEASTERN UNIV AT QINHUANGDAO
- Filing Date
- 2026-05-09
- Publication Date
- 2026-07-24
Smart Images
Figure CN122150142B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of marine acoustics technology, and in particular to a system and method for inverting ice layer parameters based on photoacoustic effects. Background Technology
[0002] Since sea ice thickness is directly related to the ice surface's bearing capacity and navigation safety, and transverse wave velocity and longitudinal wave velocity reflect the ice layer's elastic modulus, density, and internal structural state, they are key indicators for assessing the mechanical properties of the sea ice surface. Therefore, obtaining sea ice thickness, longitudinal wave velocity, transverse wave velocity, and other ice layer physical parameters quickly and accurately is of great value in fields such as polar scientific research, waterway safety assessment, marine engineering in ice-covered areas, and climate change research.
[0003] However, due to the harsh polar environment and limited on-site operating conditions, accurately detecting ice parameters presents certain difficulties. Sea ice acoustic detection methods refer to a class of detection technologies that utilize the propagation characteristics of sound waves in sea ice and adjacent media (seawater, air). By emitting sound waves and receiving the reflection, refraction, and scattering signals generated inside the ice layer or at the interface, physical parameters such as sea ice thickness, sound velocity, and elastic modulus can be collected or calculated.
[0004] Currently, traditional acoustic detection methods for sea ice mainly employ contact sound sources (such as piezoelectric transducers or hammer-driven sources) in conjunction with receiving sensors. These methods require good coupling between the sound source and the ice surface in practical applications, but they suffer from several drawbacks, including operational inconvenience in low-temperature, snowy, and windy polar environments, and difficulty in maintaining stable coupling under such conditions. Furthermore, the deployment of contact sound source equipment is complex and time-consuming, hindering large-scale, rapid measurements and posing challenges to measuring ice parameters. Summary of the Invention
[0005] In view of this, this application provides an ice layer parameter inversion system and method based on photoacoustic effect inversion, which realizes non-contact, high-efficiency, end-to-end, high-precision, and physically interpretable detection of sea ice parameters, and solves the problems of reliance on manual labor, computational complexity, and poor real-time performance in the prior art.
[0006] The first aspect of this application provides an ice layer parameter inversion system based on photoacoustic effect inversion, the system comprising:
[0007] The laser excitation module is used to emit intensity-modulated laser light onto the test area on the sea ice surface;
[0008] The signal acquisition module is used to receive target acoustic waves that are related to the time-frequency characteristics of the laser; the target acoustic waves are generated by the photoacoustic effect caused by the laser irradiation of the area under test.
[0009] The data processing module is used to reconstruct the ice layer parameters of the area to be tested based on the target sound wave according to a pre-set inversion process.
[0010] Optionally, the ice layer parameters include shear wave velocity and longitudinal wave velocity, and the data processing module includes:
[0011] The physical constraint head is used to determine whether the restored ice layer parameters satisfy the condition that the P-wave velocity is greater than or equal to the S-wave velocity. When the P-wave velocity is less than the S-wave velocity, the physical laws are used to establish constraint equations to adjust the ice layer parameters and output the adjusted ice layer parameters.
[0012] The constraint equations include:
[0013] = (1); where, It is the transverse wave speed. It is the longitudinal wave velocity. It is Poisson's ratio.
[0014] Optionally, the data processing module includes a deep learning inversion model;
[0015] A deep learning inversion model, based on a deep neural network model, is used to invert the ice layer parameters of the target acoustic wave generated by the sea ice surface receiving the laser signal.
[0016] Optionally, the system further includes a virtual data generation module;
[0017] The virtual data generation module is used to simulate the acoustic wave samples generated by laser emission onto the surface of samples with different ice layer properties based on the physical characteristics of photoacoustic effect and the dispersion characteristics of asymmetric waveguides, and to synthesize multiple virtual training data. The virtual training data includes laser signal samples, ice layer parameter samples and corresponding acoustic wave samples from the sample surface.
[0018] The training process of the deep learning inversion model includes:
[0019] The deep learning inversion model integrates the features of laser signal samples and acoustic wave samples to invert the sea ice surface that receives the laser signal samples and generate the predicted ice layer parameters of the acoustic wave samples. The loss values of the predicted ice layer parameters and the ice layer parameter samples are calculated, and the model parameters are adjusted until the model converges.
[0020] Optionally, the process by which the virtual data generation module generates virtual training data includes:
[0021] Generate ice layer physical parameters to obtain multiple ice layer parameter samples;
[0022] Input multiple ice layer parameter samples into a semi-analytical finite element solver to output dispersion curve data;
[0023] A laser signal sample is set for the echo model to generate an initial waveform. The dispersion curve data is input into the echo model, and the echo model uses the dispersion curve data to modulate the initial waveform and outputs the acoustic wave sample.
[0024] Optionally, the data processing module also includes an environmental parameter acquisition module for acquiring environmental parameters;
[0025] An environment encoder is used to encode environmental parameters to obtain environmental encoded features;
[0026] Gated networks are used to output weights of environmental features based on environmental coding features;
[0027] The residual header is used to output an environmental correction amount based on the environmental coding features, so as to correct the ice layer parameters through the environmental correction amount and obtain the output prediction result.
[0028] A second aspect of this application provides a method for inverting ice layer parameters based on photoacoustic effects, applied to an ice layer parameter inversion system based on photoacoustic effects, the method comprising:
[0029] The laser excitation module emits intensity-modulated laser light into the area to be tested on the surface of the sea ice.
[0030] The signal acquisition module receives target acoustic waves that are correlated with the time-frequency characteristics of the laser; the target acoustic waves are generated by the photoacoustic effect caused by the laser irradiation of the area under test.
[0031] The data processing module reconstructs the ice layer parameters of the area to be tested based on the target sound waves, following a pre-set inversion process.
[0032] Optionally, according to a pre-set inversion process, reconstructing the ice layer parameters of the area to be tested based on the target acoustic wave includes:
[0033] Using a physical constraint head, it is determined whether the restored ice layer parameters satisfy the condition that the P-wave velocity is greater than or equal to the S-wave velocity. When the P-wave velocity is less than the S-wave velocity, a constraint equation is established using physical laws to adjust the ice layer parameters, and the adjusted ice layer parameters are output.
[0034] The constraint equations include:
[0035] = (1); where, It is the transverse wave speed. It is the longitudinal wave velocity. It is Poisson's ratio.
[0036] Optionally, the method further includes:
[0037] By using a deep learning inversion model, based on a deep neural network model, the ice layer parameters of the sea ice surface receiving the laser signal are inverted to generate the target acoustic wave.
[0038] Optionally, the method further includes:
[0039] Environmental parameters are collected through the environmental parameter acquisition module;
[0040] Environmental parameters are encoded using an environmental encoder to obtain environmental encoded features;
[0041] The weights of environmental features are output based on the environmental coding features through a gating network.
[0042] The residual head outputs an environmental correction amount based on the environmental coding features, which is then used to correct the ice layer parameters, resulting in the output prediction result.
[0043] Optionally, the process of training a deep learning inversion model includes:
[0044] By fusing the features of laser signal samples and acoustic wave samples, the predicted ice layer parameters of the acoustic wave samples are generated from the sea ice surface that received the laser signal samples. The loss values of the predicted ice layer parameters and the ice layer parameter samples are calculated, and the model parameters are adjusted until the model converges.
[0045] A third aspect of this application provides a server, comprising: a processor and a memory, the processor and the memory being connected via a communication bus; wherein the processor is used to call and execute a program stored in the memory; the memory is used to store the program, the program being used to implement the ice layer parameter inversion method based on photoacoustic effect inversion as provided in the first aspect of this application.
[0046] The fourth aspect of this application provides a computer-readable storage medium storing computer-executable instructions for performing the ice layer parameter inversion method based on photoacoustic effect inversion as provided in the first aspect of this application.
[0047] Compared to existing technologies, the ice parameter inversion system based on photoacoustic effect proposed in this application emits intensity-modulated laser light into the area to be measured on the sea ice surface, exciting the photoacoustic effect on the sea ice surface to generate acoustic signals. These acoustic signals are then collected and calculated to obtain the ice surface parameters of the area to be measured. This process avoids the coupling problem between the contact sound source and the ice surface, achieving non-contact, rapid, and real-time sea ice parameter detection, which is convenient for polar field applications.
[0048] In the training process of the deep learning inversion model in this embodiment, the model parameters are continuously optimized by comparing the physical parameters of the ice layer with the predicted ice layer parameters. The deep learning inversion model has the ability to invert and calculate the ice surface parameters of the sea ice surface irradiated by laser based on laser data and acoustic signals. Under the same excitation laser and the same environmental conditions, the acoustic signals generated by the photoacoustic effect on the same sea ice surface are the same. Therefore, the acoustic samples, laser signal samples, and ice layer parameter samples in the virtual training data generated by the virtual data generation module have a unique correspondence. When the laser emitted by the laser source is fixed during application, the deep learning inversion model trained with these samples has the function of restoring the correspondence between acoustic data and ice layer parameters, ensuring physical consistency in the process. Furthermore, by combining virtual data generation with deep learning, the dependence on measured data is reduced, and the system adaptability is improved. Attached Figure Description
[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0050] Figure 1 This is a schematic diagram of the structure of the ice layer parameter inversion system based on photoacoustic effect proposed in this application embodiment;
[0051] Figure 2 This is a schematic diagram of the execution flow of an example data processing module of this application;
[0052] Figure 3 This is a flowchart of the steps of the ice layer parameter inversion method based on photoacoustic effect proposed in this application embodiment. Detailed Implementation
[0053] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0054] In this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.
[0055] To better understand this application, the technical names involved in this application are explained below:
[0056] The laser photoacoustic effect refers to the physical phenomenon in which a material absorbs light energy and undergoes thermoelastic expansion when a laser (pulse or modulated continuous wave) of varying intensity over time is irradiated, thereby radiating sound waves or ultrasound.
[0057] Example 1
[0058] In view of the limitations of traditional sea ice acoustic detection methods, this application proposes an ice layer parameter detection technology based on laser photoacoustic effect technology, which realizes non-contact, high-efficiency, end-to-end, high-precision, and physically interpretable detection of sea ice parameters, and solves the problems of reliance on manual labor, complex calculations, and poor real-time performance in existing technologies.
[0059] Figure 1 This is a schematic diagram of the ice layer parameter inversion system based on photoacoustic effect proposed in this application embodiment, as shown below. Figure 1 As shown, the ice layer parameter inversion system based on photoacoustic effect inversion includes: a laser excitation module, a signal acquisition module, and a data processing module.
[0060] The laser excitation module can use a laser diode, and the signal acquisition module can use a fiber optic hydrophone, a membraneless optical microphone, etc. In one example, a signal conditioning device can be set between the data processing module and the signal acquisition module to perform pre-amplification, filtering, etc., on the acquired signal.
[0061] While the laser excitation module emits a laser towards the sea ice surface, it simultaneously triggers the signal acquisition module to start acquiring acoustic signals.
[0062] The laser excitation module is used to emit intensity-modulated laser light onto the test area on the sea ice surface.
[0063] The laser excitation module periodically modulates the power of the continuous laser with a sine or square wave. The periodic laser is directed onto the test area on the sea ice surface, thereby exciting the test area on the sea ice surface to generate a continuous sound wave with controllable frequency and predictable phase, thus improving the distinguishability of the sound wave.
[0064] The signal acquisition module is used to receive target acoustic waves that are associated with the time-frequency characteristics of the laser; the target acoustic waves are generated by the photoacoustic effect of the area under test being irradiated by the laser.
[0065] For example, the laser excitation module at a frequency Modulated laser light is applied to the area to be measured on the sea ice surface. The temperature rise and thermal expansion caused by the absorption of the laser light by the sea ice surface material also occur at the same frequency. The periodic changes cause the radiation of target acoustic waves that are correlated with the time-frequency characteristics of the laser; therefore, during the data acquisition phase, a lock-in amplifier is used to specifically acquire the frequency. The signal is collected to obtain the sound waves generated by the photoacoustic effect in the test area on the sea ice surface, thus avoiding interference from other sound wave signals.
[0066] Another example of this application provides another way to receive target acoustic waves associated with the time-frequency characteristics of a laser: starting from the instant the laser is emitted, the transmission time of the acoustic wave signal from the area to be measured to the signal acquisition module is calculated, and this transmission time is used as a time window. Only the signal within this time window is captured, and the signal outside the time window is regarded as interference filtering.
[0067] The data processing module is used to reconstruct the ice layer parameters of the area to be tested based on the target sound wave according to a pre-set inversion process.
[0068] The data processing module is used to deduce the ice layer parameters of the sea ice surface corresponding to the area under test based on the laser signal and the acoustic signal.
[0069] Ice parameters can include ice thickness, P-wave velocity, S-wave velocity, and ice density.
[0070] The ice parameter inversion system based on photoacoustic effect proposed in this application emits intensity-modulated laser light into the test area on the sea ice surface, exciting the photoacoustic effect on the sea ice surface to generate acoustic signals. These acoustic signals are then collected and calculated to obtain the ice surface parameters of the test area. This process avoids the coupling problem between the contact sound source and the ice surface, achieving non-contact, rapid, and real-time sea ice parameter detection, which is convenient for polar field applications.
[0071] This application also provides an implementable structure for a data processing module, which includes a virtual data generation module and a deep learning inversion model.
[0072] A deep learning inversion model, based on a deep neural network model, is used to invert the ice layer parameters of the target acoustic wave generated by the sea ice surface receiving the laser signal.
[0073] The virtual data generation module, based on the photoacoustic effect physical model and the dispersion characteristics of asymmetric waveguides, synthesizes virtual training data with physical truth labels, generates virtual training samples for training deep learning inversion models, and trains deep learning inversion models to inversely deduce ice layer parameters.
[0074] The training process for a deep learning inversion model can be as follows:
[0075] The virtual data generation module simulates the acoustic wave samples generated by laser emission onto the surface of samples with different ice layer properties, based on the physical characteristics of photoacoustic effect and the dispersion characteristics of asymmetric waveguides, and synthesizes multiple virtual training data. The virtual training data includes laser signal samples, ice layer parameter samples and corresponding acoustic wave samples from the sample surface.
[0076] The process of generating virtual training data by the virtual data generation module includes:
[0077] Step 1: Generate ice layer physical parameters and obtain multiple ice layer parameter samples.
[0078] Step 2: Input multiple ice layer parameter samples into the semi-analytical finite element solver and output dispersion curve data.
[0079] Step 3: Set the laser signal sample for the echo model to generate an initial waveform. Input the dispersion curve data into the echo model. The echo model uses the dispersion curve data to modulate the initial waveform and outputs the sound wave sample.
[0080] For example, generating ice layer parameters that exist in a real environment ( , , ,ρ), Indicates the thickness of the ice layer. Indicates the longitudinal wave velocity. Let ρ represent the transverse wave velocity and ρ represent the ice density. The ice parameters ( , , The input is a semi-analytical finite element solver (SAFE) to simulate a three-layer asymmetric waveguide consisting of air, ice, and water. SAFE solves for N frequency points. For each frequency, the eigenvalues of the wave equation are solved to obtain the corresponding wave number and phase velocity, thus determining the dispersion equation of the guided wave mode. Through the dispersion equation, dispersion curve data for the A0 and S0 modes are extracted from all guided wave modes. A laser signal sample is set for the echo model to define the laser excitation method, including pulse width, wavelength, energy, and spot size. This controls the echo model to determine the spatiotemporal distribution of the initial sound source according to the parameters corresponding to the laser signal sample. An initial waveform is generated according to this spatiotemporal distribution. The dispersion curve data is then input into the echo model, which modulates the initial waveform using the dispersion curve data to achieve a realistic dispersion broadening, outputting a sound wave sample.
[0081] Compared to the deep learning inversion model that inverts ice layer parameters from acoustic data, the above process generates acoustic signals from ice layer parameters based on photoacoustic signals. That is, the way the virtual data generation module generates data is related to and opposite to the working process of the deep learning inversion model. Therefore, the above process can quickly generate rich training data for training the deep learning inversion model.
[0082] Under the premise that the virtual data generation module generates virtual training data, the process of training a deep learning inversion model using the virtual training data includes: fusing the features of laser signal samples and acoustic wave samples; inverting the sea ice surface receiving the laser signal samples to generate the predicted ice layer parameters possessed by the acoustic wave samples; calculating the loss values of the predicted ice layer parameters and the ice layer parameter samples; and adjusting the model parameters until the model converges. General training methods, loss functions, and gradient optimization methods of deep neural networks can be used, and this application embodiment is not limited to any particular method.
[0083] Through the aforementioned training process, the model parameters are continuously optimized by comparing the physical parameters of the ice layer with the predicted ice layer parameters. The deep learning inversion model has the ability to invert and calculate the ice surface parameters of sea ice irradiated by laser based on laser data and acoustic signals. Under the same excitation laser and the same environmental conditions, the acoustic signals generated by the photoacoustic effect on the same sea ice surface are identical. Therefore, the acoustic samples, laser signal samples, and ice layer parameter samples in the virtual training data generated by the virtual data generation module have a unique correspondence. When the laser emitted by the laser source is fixed during application, the deep learning inversion model trained with these samples has the function of reconstructing the correspondence between acoustic data and ice layer parameters, ensuring physical consistency throughout the process. Furthermore, by combining virtual data generation with deep learning, the dependence on measured data is reduced, improving the system's adaptability.
[0084] Based on the deep learning inversion model obtained from the above training, this application also proposes a data processing module structure, which includes a physical constraint head, an environmental parameter acquisition module, an environmental encoder, a gating network, and a residual head.
[0085] The physical constraint head is used to determine whether the restored ice layer parameters satisfy the condition that the P-wave velocity is greater than or equal to the S-wave velocity. When the P-wave velocity is less than the S-wave velocity, the physical laws are used to establish constraint equations to adjust the ice layer parameters and output the adjusted ice layer parameters.
[0086] The constraint equations include:
[0087] = (1); where, It is the transverse wave speed. It is the longitudinal wave velocity. It is Poisson's ratio.
[0088] By using a physical constraint head, monitoring units are set up based on objective physical principles to check the rationality of the predicted ice layer parameters, ensuring that the model is both data-driven and physically interpretable, resulting in more reliable inversion results.
[0089] The environmental parameter acquisition module is used to collect environmental data and optimize the output ice layer data.
[0090] An environment encoder is used to encode environmental parameters to obtain environmental encoded features;
[0091] Gated networks are used to output weights of environmental features based on environmental coding features;
[0092] The residual header is used to output an environmental correction amount based on the environmental coding features, so as to correct the ice layer parameters through the environmental correction amount and obtain the output prediction result.
[0093] The residual header consists of a stack of fully connected layers, batch normalization layers, and activation functions, and through residual connections, it ensures that environmental information can be passed to subsequent networks without loss.
[0094] Figure 2 This is a schematic diagram of the execution flow of an example data processing module of this application, such as... Figure 2 As shown, the data processing module executes the following processes during application:
[0095] The target acoustic wave and laser parameters are input to the data processing module. The laser encoder in the data processing module encodes the laser parameters, and the waveform encoder in the data processing module encodes the waveform input of the target acoustic wave. The encoded laser parameters and waveform input are transmitted to the feature fusion layer. The feature fusion layer performs feature fusion and calculates the ice layer parameters. The ice layer parameters are output to the physical constraint head. The physical constraint head performs physical consistency constraints on the ice layer parameters to ensure that the output ice layer parameters conform to physical laws and outputs the basic ice layer parameters.
[0096] On the other hand, the environmental parameter acquisition module collects environmental parameters, the environmental encoder encodes the environmental parameters to obtain environmental coding features, the gating network outputs the weights of the environmental features based on the environmental coding features, for example, increasing the weights of the environmental features when the environment is harsh, and the residual head outputs the environmental correction amount based on the environmental coding features, and uses the environmental correction amount to correct the ice layer parameters to obtain the final predicted ice layer parameters.
[0097] The above process enables end-to-end inversion from the original signal to sea ice physical parameters (such as ice thickness and wave velocity), without the need for manual feature extraction or complex post-processing.
[0098] Example 2
[0099] Based on the ice layer parameter inversion system based on photoacoustic effect provided in Embodiment 1 of this application, correspondingly, Embodiment 2 of this application also provides an ice layer parameter inversion method based on photoacoustic effect. Figure 3 This is a flowchart illustrating the steps of the ice layer parameter inversion method based on photoacoustic effect proposed in this application. Figure 3 As shown, the method includes:
[0100] S110: Through the laser excitation module, intensity-modulated laser light is emitted towards the area to be tested on the surface of sea ice;
[0101] S111: Receive target acoustic waves related to the time-frequency characteristics of the laser through the signal acquisition module; the target acoustic waves are generated by the photoacoustic effect of the area under test being irradiated by the laser.
[0102] S112: Through the data processing module, the ice layer parameters of the area to be tested are restored based on the target sound wave according to the pre-set inversion process.
[0103] Optionally, the data processing module reconstructs the ice layer parameters of the test area based on the target sound wave according to a pre-set inversion process, including:
[0104] Using a physical constraint head, it is determined whether the restored ice layer parameters satisfy the condition that the P-wave velocity is greater than or equal to the S-wave velocity. When the P-wave velocity is less than the S-wave velocity, a constraint equation is established using physical laws to adjust the ice layer parameters, and the adjusted ice layer parameters are output.
[0105] The constraint equations include:
[0106] = (1); where, It is the transverse wave speed. It is the longitudinal wave velocity. It is Poisson's ratio.
[0107] Optionally, the method further includes:
[0108] By using a deep learning inversion model, based on a deep neural network model, the ice layer parameters of the sea ice surface receiving the laser signal are inverted to generate the target acoustic wave.
[0109] Optionally, the method further includes:
[0110] Environmental parameters are collected through the environmental parameter acquisition module;
[0111] Environmental parameters are encoded using an environmental encoder to obtain environmental encoded features;
[0112] The weights of environmental features are output based on the environmental coding features through a gating network.
[0113] The residual head outputs an environmental correction amount based on the environmental coding features, which is then used to correct the ice layer parameters, resulting in the output prediction result.
[0114] The specific principles and execution processes of each unit in the ice layer parameter inversion method based on photoacoustic effect disclosed in Embodiment 2 of this application can be found in the corresponding parts of the ice layer parameter inversion system based on photoacoustic effect disclosed in Embodiment 1 of this application, and will not be repeated here.
[0115] Example 3
[0116] Embodiment 3 of this application provides a server, including: a processor and a memory, the processor and the memory being connected via a communication bus; wherein, the processor is used to call and execute a program stored in the memory; the memory is used to store the program, the program being used to implement the ice layer parameter inversion method based on photoacoustic effect inversion as provided in Embodiment 2 of this application.
[0117] Example 4
[0118] Embodiment 4 of this application provides a computer-readable storage medium storing computer-executable instructions for executing the ice layer parameter inversion method based on photoacoustic effect provided in Embodiment 1 of this application.
[0119] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computing software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0120] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.
[0121] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A system for inverting ice layer parameters based on photoacoustic effects, characterized in that, The system includes: The laser excitation module is used to emit intensity-modulated laser light onto the test area on the sea ice surface; The signal acquisition module is used to receive target acoustic waves that are related to the time-frequency characteristics of the laser; the target acoustic waves are generated by the photoacoustic effect caused by the laser irradiation of the area under test. The data processing module is used to reconstruct the ice layer parameters of the area to be tested based on the target sound wave according to a pre-set inversion process. The data processing module includes a deep learning inversion model and a virtual data generation module; A deep learning inversion model, based on a deep neural network model, is used to invert the ice layer parameters of the sea ice surface that receives the laser signal and generates the target acoustic wave. The virtual data generation module is used to simulate the acoustic wave samples generated by laser emission onto the surface of samples with different ice layer properties based on the physical characteristics of photoacoustic effect and the dispersion characteristics of asymmetric waveguides, and to synthesize multiple virtual training data. The virtual training data includes laser signal samples, ice layer parameter samples and corresponding acoustic wave samples from the sample surface. The training process of the deep learning inversion model includes: The deep learning inversion model integrates the features of laser signal samples and acoustic wave samples to invert the sea ice surface that receives the laser signal samples and generate the predicted ice layer parameters of the acoustic wave samples. The loss values of the predicted ice layer parameters and the ice layer parameter samples are calculated, and the model parameters are adjusted until the model converges.
2. The system according to claim 1, characterized in that, The ice layer parameters include transverse wave velocity and longitudinal wave velocity, and the data processing module includes: The physical constraint head is used to determine whether the restored ice layer parameters satisfy the condition that the P-wave velocity is greater than or equal to the S-wave velocity. When the P-wave velocity is less than the S-wave velocity, the physical laws are used to establish constraint equations to adjust the ice layer parameters and output the adjusted ice layer parameters. The constraint equations include: = (1); where, It is the transverse wave speed. It is the longitudinal wave velocity. It is Poisson's ratio.
3. The system according to claim 1, characterized in that, The process by which the virtual data generation module generates virtual training data includes: Generate ice layer physical parameters to obtain multiple ice layer parameter samples; Input multiple ice layer parameter samples into a semi-analytical finite element solver to output dispersion curve data; A laser signal sample is set for the echo model to generate an initial waveform. The dispersion curve data is input into the echo model, and the echo model uses the dispersion curve data to modulate the initial waveform and outputs the acoustic wave sample.
4. The system according to claim 1, characterized in that, The data processing module also includes an environmental parameter acquisition module, used to collect environmental parameters; An environment encoder is used to encode environmental parameters to obtain environmental encoded features; Gated networks are used to output weights of environmental features based on environmental coding features; The residual header is used to output an environmental correction amount based on the environmental coding features, so as to correct the ice layer parameters through the environmental correction amount and obtain the output prediction result.
5. A method for inverting ice layer parameters based on photoacoustic effects, characterized in that, The method, applied to an ice layer parameter inversion system based on photoacoustic effects, includes: The laser excitation module emits intensity-modulated laser light into the area to be tested on the surface of the sea ice. The signal acquisition module receives target acoustic waves that are correlated with the time-frequency characteristics of the laser; the target acoustic waves are generated by the photoacoustic effect caused by the laser irradiation of the area under test. The data processing module reconstructs the ice layer parameters of the test area based on the target acoustic wave according to a pre-set inversion process. The method further includes: By using a deep learning inversion model, based on a deep neural network model, the ice layer parameters of the sea ice surface receiving the laser signal are inverted to generate the target acoustic wave. Using a virtual data generation module, based on the physical properties of photoacoustic effect and the dispersion characteristics of asymmetric waveguides, the module simulates the acoustic wave samples generated by laser emission onto the surface of samples with different ice layer properties, and synthesizes multiple virtual training data. The virtual training data includes laser signal samples, ice layer parameter samples, and corresponding acoustic wave samples from the sample surface. The training process of the deep learning inversion model includes: The deep learning inversion model integrates the features of laser signal samples and acoustic wave samples to invert the sea ice surface that receives the laser signal samples and generate the predicted ice layer parameters of the acoustic wave samples. The loss values of the predicted ice layer parameters and the ice layer parameter samples are calculated, and the model parameters are adjusted until the model converges.
6. The method according to claim 5, characterized in that, Through the data processing module, following a pre-set inversion process, the ice layer parameters of the area to be tested are reconstructed based on the target sound waves, including: Using a physical constraint head, it is determined whether the restored ice layer parameters satisfy the condition that the P-wave velocity is greater than or equal to the S-wave velocity. When the P-wave velocity is less than the S-wave velocity, a constraint equation is established using physical laws to adjust the ice layer parameters, and the adjusted ice layer parameters are output. The constraint equations include: = (1); where, It is the transverse wave speed. It is the longitudinal wave velocity. It is Poisson's ratio.
7. The method according to claim 5, characterized in that, The method further includes: Environmental parameters are collected through the environmental parameter acquisition module; Environmental parameters are encoded using an environmental encoder to obtain environmental encoded features; The weights of environmental features are output based on the environmental coding features through a gating network. The residual head outputs an environmental correction amount based on the environmental coding features, which is then used to correct the ice layer parameters, resulting in the output prediction result.