Fish activity sensing method and apparatus based on underwater distributed fiber sensing

Abstract

The application discloses a fish activity sensing method and device based on underwater distributed optical fiber sensing, equipment and medium, comprising: based on the Rayleigh scattering characteristics or Brillouin scattering characteristics of the distributed optical fiber, obtaining the local disturbance signal in the target water area; obtaining the temperature data and flow field characteristic data corresponding to the local disturbance signal under the same sequence; extracting the features of the local disturbance signal, the corresponding temperature data and the corresponding flow field characteristic data to obtain the corresponding multi-dimensional features, wherein the multi-dimensional features include time domain features, frequency domain features and time-frequency features; based on the multi-dimensional features and the fish behavior corresponding to the multi-dimensional feature pair, training a deep learning model, and after the training is completed, the deep learning model is used for predicting the fish behavior. The application belongs to the field of fish behavior prediction. The application can realize accurate prediction of fish behavior in complex water bodies.

Description

Technical Field

[0001] This invention relates to the field of fish behavior prediction, and more particularly to fish activity sensing methods, devices, equipment, and media based on underwater distributed optical fiber sensing. Background Technology

[0002] Traditional methods for monitoring fish activity mainly include acoustic monitoring, underwater video surveillance, tagging and telemetry technologies, and environmental DNA (eDNA) analysis. Among these, acoustic monitoring (such as sonar, split-beam sonar, and echo sounders) effectively obtains information such as the location, quantity, size, and direction of movement of fish by emitting sound waves and receiving the echoes reflected by the fish, and has certain advantages in turbid or deep water environments. Underwater video surveillance can intuitively and realistically record the behavioral details of fish and is suitable for waters with high visibility.

[0003] However, traditional methods still face significant challenges for continuous, dynamic, and large-scale monitoring of fish behavior: sonar systems are susceptible to underwater background noise, multipath interference, and complex terrain, leading to data distortion or misjudgment; video surveillance becomes drastically less effective in low visibility, at night, or in vast bodies of water, and relies heavily on manual annotation or complex image recognition algorithms. Therefore, how to achieve continuous monitoring of fish activity in complex aquatic environments is an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a method, device, equipment, and medium for sensing fish activity based on underwater distributed optical fiber sensing, which solves the technical problem of difficulty in continuous monitoring of fish activity in complex aquatic environments in the prior art, and achieves the technical effect of continuous monitoring of fish activity in complex aquatic environments.

[0005] In a first aspect, the present invention provides a method for sensing fish activity based on underwater distributed optical fiber sensing, comprising: Based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers, local disturbance signals in the target water area are obtained. These local disturbance signals are related to fish behavior, which includes at least stillness, cruising, chasing, foraging, gathering, and escaping. Acquire temperature data and flow field characteristic data corresponding to local disturbance signals at the same time, including flow velocity, turbulence intensity, vorticity Reynolds number and shear rate. Feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain the corresponding multidimensional features, which include time-domain features, frequency-domain features, and time-frequency features. A deep learning model is trained based on multidimensional features and fish behaviors corresponding to those features. Once training is complete, the deep learning model is used to predict fish behavior.

[0006] Furthermore, based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers, local disturbance signals in the target water area are obtained, including: Using a narrow linewidth continuous laser, short pulses are generated by a pulse generator after phase or intensity modulation; Short pulses are injected into the distributed optical fiber to excite Rayleigh scattering light; Based on the Rayleigh scattered light received by the photodetector and the corresponding time series; Based on the returned Rayleigh scattered light, a perturbation trace is generated, and cross-correlation, differential or wavelet transform are performed between the perturbation trace and the first preset reference trace to obtain a local perturbation signal. Determining the location of disturbances based on time series data includes:

[0007] in, The distance between the target and the distributed optical fiber. At the speed of light, For echo duration, The core refractive index; or, Injecting pulsed light into a distributed optical fiber to excite Brillouin backscattering; The Brillouin frequency shift is compared with the second preset reference trace to obtain the comparison difference; The local perturbation signal is obtained based on the amplitude, rate of change, and contrast differences of the Brillouin frequency shift.

[0008] Furthermore, feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain time-domain features, including: Time axis synchronization and spatial registration are performed on local disturbance signals, corresponding temperature data, and corresponding flow field characteristic data; Divide time periods based on a sliding time window; For each time period, the amplitude distribution, fluctuation intensity and non-stationary characteristics of the local disturbance signal, the corresponding temperature data and the corresponding flow field characteristic data are determined. Based on amplitude distribution, fluctuation intensity, and non-stationary characteristics, coupling features are constructed. The coupling features are normalized to obtain the corresponding time-domain features.

[0009] Furthermore, feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain frequency domain features, including: For each time period, the linear potential term or low-order trend term is removed from the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data, and a window function is applied. Fourier transform is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain frequency domain parameters, which include: signal periodicity, energy distribution, and oscillation mode. Correlation indices are constructed based on frequency domain parameters, and frequency domain features are obtained by normalizing the correlation indices.

[0010] Furthermore, feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain time-frequency features, including: Based on correlation indicators and coupling characteristics, multi-source collaborative time-frequency coupling characteristics are constructed. The multi-source collaborative time-frequency coupling characteristics are normalized to obtain the corresponding time-frequency characteristics.

[0011] Furthermore, the deep learning model is trained, including: By binding the multidimensional features with the corresponding fish behaviors, several data sets were obtained. The data set is input into the deep learning model for training. When the deep learning model meets the preset training requirements, it is used to predict fish behavior.

[0012] Furthermore, the loss function of a deep learning model includes:

[0013] Where N is the number of data sets. For the i-th correct fish behavior, For the i-th predicted fish behavior, For the total loss, The regularization coefficient is . These are learnable parameters.

[0014] In a second aspect, the present invention provides a fish activity sensing device based on underwater distributed optical fiber sensing, comprising: The disturbance signal acquisition module is used to acquire local disturbance signals in the target water area based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers. The local disturbance signals are related to fish behavior, which includes at least stillness, cruising, chasing, foraging, gathering, and escaping. Simultaneous data module is used to acquire temperature data and flow field characteristic data corresponding to local disturbance signals at the same time. The flow field characteristic data includes flow velocity, turbulence intensity, vorticity Reynolds number and shear rate. The feature processing module is used to extract features from local disturbance signals, corresponding temperature data, and corresponding flow field characteristic data to obtain corresponding multidimensional features, including time-domain features, frequency-domain features, and time-frequency features. The model training module is used to train a deep learning model based on multidimensional features and fish behaviors corresponding to those multidimensional features. Once training is complete, the deep learning model is used to predict fish behavior.

[0015] Thirdly, the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to execute a fish activity sensing method based on underwater distributed optical fiber sensing, as provided in the first aspect.

[0016] Fourthly, the present invention provides a non-transitory computer-readable storage medium, wherein when the instructions in the non-transitory computer-readable storage medium are executed by a processor of an electronic device, the electronic device is able to execute a fish activity sensing method based on underwater distributed optical fiber sensing as provided in the first aspect.

[0017] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention deeply integrates local disturbance signals captured by fiber Rayleigh / Brillouin scattering with synchronously acquired temperature and multidimensional flow field characteristics. By extracting multidimensional features in the time domain, frequency domain, and time-frequency joint, it comprehensively characterizes the complex coupling relationship between fish behavior and its hydrodynamic-thermal environment. Based on this, it uses a deep learning model to establish an end-to-end mapping between multiphysics field features and behavioral labels, and introduces a regularized loss function to improve generalization ability. Compared with traditional acoustic or video monitoring methods, this invention has significant advantages such as continuous spatial and temporal coverage, resistance to electromagnetic interference, strong concealment, no need for light source illumination, and applicability to turbid or deep water environments. At the same time, it effectively improves the accuracy of behavior recognition and ecological interpretability through multi-source environmental collaborative perception. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the fish activity sensing method based on underwater distributed optical fiber sensing provided by the present invention. Detailed Implementation

[0020] This invention provides a method for sensing fish activity based on underwater distributed optical fiber sensing, which solves the technical problem of difficulty in continuously monitoring fish activity in complex aquatic environments in the prior art.

[0021] The technical solution of this invention is to solve the above-mentioned technical problems, and the overall idea is as follows: A method for sensing fish activity based on underwater distributed optical fiber sensing includes: acquiring local disturbance signals in a target water area based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers, wherein the local disturbance signals are related to fish behavior, which includes at least stillness, cruising, chasing, foraging, gathering, and escaping; acquiring temperature data and flow field characteristic data corresponding to the local disturbance signals at the same time, wherein the flow field characteristic data includes flow velocity, turbulence intensity, vorticity Reynolds number, and shear rate; extracting features from the local disturbance signals, the corresponding temperature data, and the corresponding flow field characteristic data to obtain corresponding multidimensional features, wherein the multidimensional features include time-domain features, frequency-domain features, and time-frequency features; training a deep learning model based on the multidimensional features and the fish behavior corresponding to the multidimensional feature pairs, and after training, using the deep learning model to predict fish behavior.

[0022] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0023] First, it should be clarified that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0024] The sensing optical fibers in this invention can be arranged in a variety of ways. Bottom laying is suitable for flat areas, suspended deployment can be used for monitoring the middle layer of water, while vertical multi-layer deployment is suitable for complex water structures, realizing multi-layer sensing at different depths.

[0025] The laser sensing unit generates laser pulse signals through a light source module, a photodetector module, and a control module, and propagates these signals through an optical fiber. Minor vibrations caused by fish activity lead to changes in scattering within the fiber, generating specific optical signal disturbances.

[0026] This invention provides, for example Figure 1 The method for sensing fish activity based on underwater distributed optical fiber sensing, shown in Figure 1, includes steps S11-S15: Step S11: Based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers, obtain local disturbance signals in the target water area, wherein the local disturbance signals are related to fish behavior, and fish behavior includes at least stillness, cruising, chasing, foraging, gathering, and escaping.

[0027] Based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers, local disturbance signals in the target water area are obtained, including: [Acquiring Local Disturbance Signals in Target Water Areas Based on Rayleigh Scattering Characteristics of Distributed Optical Fibers] Specifically, this includes: using a narrow-linewidth continuous-wave laser, after phase / intensity modulation, generating short pulses via a pulse generator; injecting the short pulses into a distributed optical fiber and exciting Rayleigh scattering light; receiving the returned Rayleigh scattering light and its corresponding time series based on a photodetector; generating a perturbation trace based on the returned Rayleigh scattering light, and performing cross-correlation, differential, or wavelet transform on the perturbation trace and a first preset reference trace to obtain a local perturbation signal; determining the perturbation location based on the time series, including:

[0028] in, The distance between the target and the distributed optical fiber. At the speed of light, For echo duration, The refractive index is the core refractive index.

[0029] A narrow-linewidth continuous laser can be used as the light source; then the phase or intensity of the narrow-linewidth continuous laser is modulated, and it is modulated into an optical pulse with a preset width by a pulse generator, thereby forming a resolvable spatial sampling unit in the time domain.

[0030] Short pulses are injected into distributed optical fibers deployed in the target water area. During propagation inside the optical fiber, Rayleigh backscattered light is excited due to the microscopic inhomogeneity of the medium.

[0031] When fish behavior disturbs a local area of ​​an optical fiber, it causes strain or temperature changes in that area, resulting in minute but detectable changes in the phase or intensity of the Rayleigh scattered light. The returned Rayleigh scattered light is received by a photodetector, and its corresponding time-series signal is recorded simultaneously.

[0032] Based on this time series, a disturbance response trace reflecting the distribution along the optical fiber is constructed. To further extract effective disturbance information, the disturbance trace is subjected to cross-correlation analysis, differential processing, or wavelet transform and other signal processing methods with the first preset reference trace in the undisturbed state, thereby highlighting and separating the characteristic signals caused by external disturbances.

[0033] Finally, by combining the relationship between the speed of light in the optical fiber and the echo time, the spatial location of the disturbance can be accurately determined.

[0034] [Acquiring local disturbance signals in target water areas based on the Brillouin scattering properties of distributed optical fibers] Specifically, this includes: injecting pulsed light into a distributed optical fiber to excite Brillouin backscattering; comparing the Brillouin frequency shift with a second preset reference trace to obtain the comparison difference; and obtaining a local perturbation signal based on the amplitude, rate of change, and comparison difference of the Brillouin frequency shift.

[0035] Narrow-linewidth pulsed light is injected into distributed optical fibers deployed in the target water area. As the pulses propagate in the optical fiber, they undergo inelastic interactions with acoustic phonons in the fiber material, thereby exciting Brillouin backscattered light.

[0036] When fish-like behaviors occur, they cause minute strains or temperature changes in localized areas of the optical fiber, leading to alterations in the Brillouin frequency shift at those locations. By employing high-resolution Brillouin optical time-domain analysis or Brillouin optical time-domain reflectometry, the Brillouin frequency shift values ​​at each location are measured point-by-point along the fiber's length, forming a real-time Brillouin frequency shift distribution trace.

[0037] Subsequently, the trace is compared point by point with the second preset reference trace in the undisturbed state, and the differences between the two (such as frequency shift offset, local slope change, etc.) are calculated.

[0038] Based on this, the amplitude change of Brillouin frequency shift, its rate of change with time or space, and the difference in comparison with the reference trace are analyzed. By setting thresholds, using pattern recognition or machine learning algorithms, abnormal features caused by external disturbances are extracted, and finally, local disturbance signals are generated.

[0039] Step S12: Obtain temperature data and flow field characteristic data corresponding to the local disturbance signal at the same time sequence. The flow field characteristic data includes flow velocity, turbulence intensity, vorticity Reynolds number and shear rate.

[0040] Temperature data can be obtained in real time through distributed fiber optic thermometry, with spatial resolution consistent with disturbance location; flow field characteristic parameters can be directly measured by acoustic Doppler velocity profiler or particle image velocimetry system, and parameters such as turbulence intensity, vorticity, Reynolds number and shear rate can be calculated accordingly.

[0041] Fish swimming patterns, habitat selection, feeding activities, and escape responses are highly dependent on the aquatic environment: flow field characteristics (such as flow velocity, turbulence intensity, and shear rate) directly affect fish energy consumption, movement strategies, and group structure, while water temperature regulates their metabolic rate, activity level, and vertical distribution. Therefore, combining these two types of environmental parameters can effectively distinguish the intrinsic driving mechanisms of behavioral patterns, improving the accuracy of behavioral identification and ecological interpretability.

[0042] Step S13: Extract features from the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain the corresponding multidimensional features, which include time-domain features, frequency-domain features, and time-frequency features.

[0043] Feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain time-domain features. This includes: time axis synchronization and spatial registration of the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data; time period division based on a sliding time window; determination of amplitude distribution, fluctuation intensity, and non-stationary characteristics for the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data in each time period; construction of coupling features based on amplitude distribution, fluctuation intensity, and non-stationary characteristics; and normalization of the coupling features to obtain the corresponding time-domain features.

[0044] The local disturbance signal acquired through distributed optical fiber, the synchronously measured temperature data, and the corresponding flow field characteristic data are synchronized in time and spatially to ensure that the three types of data are strictly aligned in time and space, thereby ensuring the physical consistency of subsequent analysis. A sliding time window strategy is used to divide the continuous time series into several overlapping or non-overlapping analysis periods to capture local dynamic changes. Within each period, the statistical and dynamic characteristics of the disturbance signal, temperature and flow field data are calculated, including amplitude distribution (such as mean, variance, extreme values), fluctuation intensity (such as standard deviation, energy, rate of change) and non-stationary characteristics (such as trend term, abrupt change point, time-frequency localization characteristics). By fusing features of different physical quantities that have spatiotemporal correlation, a coupling feature that can reflect the coupling effect of multiple physical fields can be constructed. The coupling feature may include the ratio of the disturbance signal RMS to the flow velocity, the correlation coefficient between the BFS change rate and the temperature gradient, or the flow field shear intensity at the time of the disturbance.

[0045] To eliminate dimensional differences and improve the model's generalization ability, all coupled features can be normalized (e.g., Z-score normalization or Min-Max scaling).

[0046] Feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain frequency domain features. This includes: removing linear potential terms or low-order trend terms from the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data for each time period, and applying a window function; performing Fourier transform on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain frequency domain parameters, which include: signal periodicity, energy distribution, and oscillation mode; constructing correlation indicators based on the frequency domain parameters, and normalizing the correlation indicators to obtain frequency domain features.

[0047] For each time period, the local disturbance signal, corresponding temperature data, and flow field characteristic data (such as flow velocity, turbulence intensity, etc.) are preprocessed: linear trend terms or low-order polynomial trends are removed (e.g., through detrending operations or high-pass filtering) to eliminate slowly changing background interference (such as diurnal temperature variation or slowly varying flow caused by tides), while retaining the dynamic fluctuation components.

[0048] To reduce spectral leakage and improve the accuracy of the Fourier transform, an appropriate window function (such as the Hanning window or Hamming window) is applied to the detrended signal. Next, a Fourier transform (FFT) is performed on each of the three processed signals to map them from the time domain to the frequency domain, thereby obtaining their respective spectral information. Frequency domain parameters are extracted from this information, including the signal's periodicity (such as the dominant frequency and its harmonic structure), energy distribution in different frequency bands (such as the low-frequency / high-frequency energy ratio and total spectral energy), and the dominant oscillation mode (such as the presence of narrowband resonants, broadband noise characteristics, or multimodal oscillations).

[0049] Construct cross-modal correlation indicators, which may include the consistency between the dominant frequency of the disturbance and the vortex shedding frequency of the flow field, the correlation between the temperature spectral entropy and the disturbance spectral entropy, or the energy ratio of the disturbance to the flow velocity within a specific frequency band.

[0050] To eliminate the dimensional differences between different indicators and improve the stability and generalization ability of subsequent models, all related indicators are normalized.

[0051] Feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain time-frequency features, including: constructing multi-source cooperative time-frequency coupling features based on correlation indicators and coupling features; and normalizing the multi-source cooperative time-frequency coupling features to obtain the corresponding time-frequency features.

[0052] Specifically, based on the obtained coupling features and correlation indicators, multi-source collaborative time-frequency coupling features are constructed. First, the two types of features are aligned according to the time window to ensure spatiotemporal consistency. Then, through feature splicing, product interaction, correlation coefficient calculation or tensor fusion, a joint representation across domains and modes (multi-source collaborative time-frequency coupling features) is established.

[0053] Multi-source cooperative time-frequency coupling characteristics can include the product of the perturbation nonstationary index and the dominant frequency energy, the energy proportion of temperature fluctuation intensity in the turbulence-dominant frequency band, and the correlation between flow field shear rate changes and perturbation wavelet coefficients at a specific scale.

[0054] To ensure consistency in the dimensions, amplitudes, and distributions of the features, and to facilitate subsequent processing by machine learning or pattern recognition algorithms, all constructed multi-source cooperative time-frequency coupled features can be normalized.

[0055] Step S14: Based on the multidimensional features and the fish behavior corresponding to the multidimensional feature pairs, train the deep learning model. After training is completed, use the deep learning model to predict fish behavior.

[0056] Training the deep learning model includes: binding multidimensional features with corresponding fish behaviors to obtain several data sets; inputting the data sets into the deep learning model for training; and when the deep learning model meets the preset training requirements, using the deep learning model to predict fish behavior.

[0057] The loss function of a deep learning model includes:

[0058] Where N is the number of data sets. For the i-th correct fish behavior, For the i-th predicted fish behavior, For the total loss, The regularization coefficient is . These are learnable parameters.

[0059] By constructing a correspondence between multidimensional features (time-domain, frequency-domain, and time-frequency features that integrate local disturbances, temperature, and flow field characteristics) and known fish behavior labels, several labeled data sets are formed for training deep learning models.

[0060] During training, the data set is input into the model, enabling it to learn the mapping relationship between multidimensional environmental features and fish behavior. The loss function uses a mean squared error term to measure the deviation between predicted and actual behavior, and a regularization term is added to constrain the model parameters to prevent overfitting. Training is complete when the total model loss converges to a preset threshold or other training termination conditions are met.

[0061] In summary, this invention deeply integrates local disturbance signals captured by fiber Rayleigh / Brillouin scattering with synchronously acquired temperature and multidimensional flow field characteristics. By systematically extracting multidimensional features in the time domain, frequency domain, and time-frequency joint, it comprehensively characterizes the complex coupling relationship between fish behavior and its hydrodynamic-thermal environment. Based on this, a deep learning model is used to establish an end-to-end mapping between multiphysics features and behavioral labels, and a regularized loss function is introduced to improve generalization ability. Compared with traditional acoustic or video monitoring methods, this invention has significant advantages such as continuous spatial coverage, resistance to electromagnetic interference, strong concealment, no need for light source illumination, and applicability to turbid or deep-water environments. Furthermore, multi-source environmental collaborative sensing effectively improves the accuracy of behavior recognition and ecological interpretability.

[0062] Based on the same inventive concept, this invention provides a fish activity sensing device based on underwater distributed optical fiber sensing, comprising: The disturbance signal acquisition module is used to acquire local disturbance signals in the target water area based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers. The local disturbance signals are related to fish behavior, which includes at least stillness, cruising, chasing, foraging, gathering, and escaping. Simultaneous data module is used to acquire temperature data and flow field characteristic data corresponding to local disturbance signals at the same time. The flow field characteristic data includes flow velocity, turbulence intensity, vorticity Reynolds number and shear rate. The feature processing module is used to extract features from local disturbance signals, corresponding temperature data, and corresponding flow field characteristic data to obtain corresponding multidimensional features, including time-domain features, frequency-domain features, and time-frequency features. The model training module is used to train a deep learning model based on multidimensional features and fish behaviors corresponding to those multidimensional features. Once training is complete, the deep learning model is used to predict fish behavior.

[0063] Based on the same inventive concept, the present invention also provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to execute a fish activity sensing method based on underwater distributed fiber optic sensing, as described above.

[0064] Based on the same inventive concept, the present invention also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to execute the fish activity sensing method based on underwater distributed optical fiber sensing as described above.

[0065] Since the electronic device described in this embodiment is an electronic device used to implement the information processing method in the embodiments of the present invention, those skilled in the art can understand the specific implementation methods and various variations of the electronic device in this embodiment based on the information processing method described in the embodiments of the present invention. Therefore, how the electronic device implements the method in the embodiments of the present invention will not be described in detail here. Any electronic device used by those skilled in the art to implement the information processing method in the embodiments of the present invention falls within the scope of protection of the present invention.

[0066] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0068] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0069] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0070] 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.

[0071] 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 method for sensing fish activity based on underwater distributed optical fiber sensing, characterized in that, include: Based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers, local disturbance signals in the target water area are obtained. These local disturbance signals are related to fish behavior, which includes at least stillness, cruising, chasing, foraging, gathering, and escaping. Acquire temperature data and flow field characteristic data corresponding to local disturbance signals at the same time, including flow velocity, turbulence intensity, vorticity Reynolds number and shear rate. Feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain corresponding multidimensional features. These multidimensional features include time-domain features, frequency-domain features, and time-frequency features. Specifically, feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain time-domain features, including: Time axis synchronization and spatial registration are performed on local disturbance signals, corresponding temperature data, and corresponding flow field characteristic data; Divide time periods based on a sliding time window; For each time period, the amplitude distribution, fluctuation intensity and non-stationary characteristics of the local disturbance signal, the corresponding temperature data and the corresponding flow field characteristic data are determined. Based on amplitude distribution, fluctuation intensity, and non-stationary characteristics, coupling features are constructed. The coupling features are normalized to obtain the corresponding time-domain features; In this process, feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain frequency domain features, including: For each time period, the linear potential term or low-order trend term is removed from the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data, and a window function is applied. Fourier transform is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain frequency domain parameters, which include: signal periodicity, energy distribution, and oscillation mode. Correlation indices are constructed based on frequency domain parameters, and frequency domain features are obtained by normalizing the correlation indices. In this process, feature extraction is performed on the local disturbance signal, the corresponding temperature data, and the corresponding flow field characteristic data to obtain time-frequency features, including: Based on the correlation index and the coupling feature, a multi-source collaborative time-frequency coupling feature is constructed. The multi-source cooperative time-frequency coupling features are normalized to obtain the corresponding time-frequency features; A deep learning model is trained based on multidimensional features and fish behaviors corresponding to those features. Once training is complete, the deep learning model is used to predict fish behavior.

2. The fish activity sensing method based on underwater distributed optical fiber sensing as described in claim 1, characterized in that, Based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers, local disturbance signals in the target water area are obtained, including: Using a narrow linewidth continuous laser, short pulses are generated by a pulse generator after phase or intensity modulation; Short pulses are injected into the distributed optical fiber to excite Rayleigh scattering light; Based on the Rayleigh scattered light received by the photodetector and the corresponding time series; Based on the returned Rayleigh scattered light, a perturbation trace is generated, and cross-correlation, differential or wavelet transform are performed between the perturbation trace and the first preset reference trace to obtain a local perturbation signal. Determining the location of disturbances based on time series data includes: in, The distance between the target and the distributed optical fiber. At the speed of light, For echo duration, The core refractive index; or, Injecting pulsed light into a distributed optical fiber to excite Brillouin backscattering; The Brillouin frequency shift is compared with the second preset reference trace to obtain the comparison difference; The local perturbation signal is obtained based on the amplitude and rate of change of the Brillouin frequency shift and the contrast difference.

3. The fish activity sensing method based on underwater distributed optical fiber sensing as described in claim 1, characterized in that, Training a deep learning model includes: By binding the multidimensional features with the corresponding fish behaviors, several data sets were obtained. The data set is input into the deep learning model for training. When the deep learning model meets the preset training requirements, it is used to predict fish behavior.

4. The fish activity sensing method based on underwater distributed optical fiber sensing as described in claim 1, characterized in that, The loss function of a deep learning model includes: Where N is the number of data sets. For the i-th correct fish behavior, For the i-th predicted fish behavior, For the total loss, The regularization coefficient is . These are learnable parameters.

5. A fish activity sensing device based on underwater distributed optical fiber sensing, characterized in that, The fish activity sensing method based on underwater distributed optical fiber sensing, applied to any one of claims 1-4, comprises: The disturbance signal acquisition module is used to acquire local disturbance signals in the target water area based on the Rayleigh scattering or Brillouin scattering characteristics of distributed optical fibers. The local disturbance signals are related to fish behavior, which includes at least stillness, cruising, chasing, foraging, gathering, and escaping. Simultaneous data module is used to acquire temperature data and flow field characteristic data corresponding to local disturbance signals at the same time. The flow field characteristic data includes flow velocity, turbulence intensity, vorticity Reynolds number and shear rate. The feature processing module is used to extract features from local disturbance signals, corresponding temperature data, and corresponding flow field characteristic data to obtain corresponding multidimensional features, including time-domain features, frequency-domain features, and time-frequency features. The model training module is used to train a deep learning model based on multidimensional features and fish behaviors corresponding to those multidimensional features. Once training is complete, the deep learning model is used to predict fish behavior.

6. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute a fish activity sensing method based on underwater distributed optical fiber sensing as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium, characterized in that, When the instructions in the non-transitory computer-readable storage medium are executed by the processor of the electronic device, the electronic device is able to perform the fish activity sensing method based on underwater distributed optical fiber sensing as described in any one of claims 1 to 4.

Images