In-situ accurate sampling investigation method and system for protosalanx hyalocranius spawning site

Abstract

The invention belongs to the technical field of protosalanx hyalocranius spawning, and provides a protosalanx hyalocranius spawning site in-situ accurate sampling investigation method and system, and the method comprises the steps: deploying a sensor in a collection water area to collect water body parameters, and dividing interference grades based on an analytic hierarchy process; calculating a signal propagation effective distance according to the interference level, determining a hydrophone array node spacing, forming a hydrophone monitoring matrix, and carrying out stability verification; after unifying the timestamps, collecting the audio frequency of the protosalanx hyalocranius in the breeding period, extracting a Mel-frequency cepstral coefficient, and constructing frequency domain, time domain and space domain feature vectors; constructing a sound behavior comprehensive recognition model based on the multi-dimensional feature vector to recognize a protosalanx hyalocranius positive event; all the positive events are imported into a geographic information system to generate an acoustic activity thermodynamic diagram, matching verification is conducted through water body environment DNA sampling, and finally the protosalanx hyalocranius spawning site range is determined. According to the invention, accurate, lossless and in-situ monitoring and evaluation of the protosalanx hyalocranius spawning site are realized.

Description

Technical Field

[0001] This invention belongs to the field of large silverfish spawning technology, specifically a method and system for in-situ precise sampling and investigation of large silverfish spawning grounds. Background Technology

[0002] The silverfish is an important economic fish resource in my country, and the accurate location and assessment of its spawning grounds are crucial for resource protection and propagation.

[0003] Traditional methods for surveying spawning grounds mainly rely on trawling and underwater observation, which have problems such as high workload, strong interference, and low accuracy.

[0004] Especially in turbid water environments, optical observation methods are limited, making effective monitoring difficult. Although acoustic detection technology has been applied to fish behavior research, it still faces many challenges in monitoring large silverfish: large silverfish are small and transparent, resulting in weak acoustic reflection signals; the tapping sound during the breeding season is prone to spectral overlap with the acoustic signals of other fish; complex underwater environments interfere with acoustic signals; and existing technologies cannot simultaneously achieve accurate identification and precise positioning.

[0005] Therefore, there is an urgent need to develop a new method that can overcome the above-mentioned technical difficulties and realize in-situ, accurate, and non-destructive investigation of the spawning grounds of giant silverfish.

[0006] Therefore, the present invention provides a method and system for in-situ precise sampling and investigation of spawning grounds of large silverfish. Summary of the Invention

[0007] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0008] The technical solution adopted by this invention to solve its technical problem is: On the one hand, this invention provides a method for in-situ precise sampling and investigation of spawning grounds of large silverfish, including: S1: Deploy sensors in the water area to collect water parameters, and classify the interference level based on the water parameters using the analytic hierarchy process. S2: Based on different interference levels, analyze the effective distance of signal propagation in the water area, determine the spacing between hydrophone array nodes according to the effective distance of signal propagation, form a hydrophone monitoring matrix, and verify the stability of the hydrophone monitoring matrix. S3: If the hydrophone monitoring matrix is ​​stable and reliable, then unify the timestamp of the hydrophone monitoring matrix, collect the sound frequency of the silverfish breeding season through the hydrophone monitoring matrix, extract the Mel frequency cepstral coefficients of the sound frequency of the silverfish breeding season, construct the frequency domain feature vector, time domain feature vector, and use the hydrophone monitoring matrix to construct the spatial domain feature vector. S4: Construct a comprehensive behavioral sound behavior recognition model based on frequency domain feature vectors, time domain feature vectors, and spatial domain feature vectors to identify positive events of large silverfish; S5: Import all positive events of giant silverfish into the geographic information system to generate a heat map of the acoustic activity of giant silverfish reproduction. The matching of the heat map of the acoustic activity of giant silverfish reproduction is verified by DNA sampling of the aquatic environment to determine the range of giant silverfish spawning grounds.

[0009] In this invention, the specific process of collecting water parameters is as follows: A laser turbidity sensor, a miniature ADCP flow velocity sensor, and an integrated sound velocity meter are deployed in the water area to collect water parameters, including water turbidity, water flow velocity, and sound velocity profile. The laser turbidity sensor is used to monitor and acquire water turbidity, the miniature ADCP flow velocity sensor is used to monitor and acquire water flow velocity, and the integrated sound velocity meter is used to monitor the sound velocity profile and acquire the sound velocity distribution of different water layers.

[0010] In this invention, the specific process of classifying interference levels using the analytic hierarchy process is as follows: Based on the water parameters collected in the sampling area, a hierarchical analysis structure is constructed. The target layer is the interference level L, and the criterion layer includes three core indicators: water turbidity, water flow velocity, and sound velocity profile. The collected water parameters are normalized, and a judgment matrix is ​​constructed using an expert scoring method. The importance of the three core indicators to the interference level is compared, and the scientific validity is ensured through consistency verification. The three core indicators after normalization are superimposed according to their weights through weighted linear combination to obtain a comprehensive interference index. The interference level is then classified based on the comprehensive interference index.

[0011] In this invention, the specific process of determining the spacing between hydrophone array nodes based on the effective signal propagation distance is as follows: Based on different interference levels, the turbidity (NTU) and flow velocity (c) of the water body parameters corresponding to different interference levels are obtained, and the effective signal propagation distance (D) of the collected water area is calculated using the following formula: ,in, For calibration coefficients, The reference sound velocity is given in standard clear water environment, in m / s. 500 is the turbidity value of the reference water area, used to normalize the effect of water turbidity on sound propagation. If the turbidity (NTU) of the water body being sampled is less than 500, the effective propagation distance can be calculated. Then the node spacing of the hydrophone array is set to 200m; If the water turbidity (NTU) of the sampled water area is greater than or equal to 500 and less than 1500, the effective propagation distance can be calculated. Then the node spacing of the hydrophone array is set to 100m; If the water turbidity (NTU) of the sampled water body is greater than or equal to 1500, the effective propagation distance can be calculated. The node spacing of the hydrophone array is set to 50m, and a double-layer array is used, deployed on the surface and middle layers respectively. The horizontal propagation attenuation is compensated by superimposing the vertical acoustic signals.

[0012] In this invention, a two-dimensional rectangular coordinate system is established with the geometric center of the hydrophone monitoring matrix as the origin; the node spacing D of the hydrophone array is matched with the current water interference level to determine the placement spacing d of the audio simulator. The spacing d between the sampling points is no greater than 1 / 2 of the node spacing D, ensuring that the spatial sampling rate satisfies the Nyquist sampling theorem and can effectively capture the spatial response characteristics of the monitoring matrix. Based on the spacing d, a two-dimensional grid is formed within the coverage area of ​​the monitoring matrix, and an audio simulator is deployed at each grid point. Control all audio simulators to play preset audio signals for the breeding season of large silverfish in sequence; record whether each node in the hydrophone array successfully receives the signal when each audio simulator plays the signal; For any audio simulator node Si, if there exists a circular region centered on Si with a radius of R, where R is the effective communication radius of the hydrophone (its value can be determined through prior experiments or provided by the equipment manufacturer), and at least K hydrophone nodes within this region can receive the signal emitted by Si, then the signal coverage of node Si is deemed to meet the requirements. If the signal coverage of all audio simulator nodes meets the requirements, then the signal coverage of the entire hydrophone monitoring matrix is ​​determined to be without blind spots; otherwise, blind spots exist. The waveform of the audio signal received by each hydrophone node is recorded synchronously; The received signal is compared with the original standard signal output by the audio simulator to extract key acoustic feature parameters, including but not limited to: center frequency deviation Δf, signal-to-noise ratio SNR, and waveform similarity NCC. Set allowable thresholds for each characteristic parameter. If all characteristic parameters of the signals received by all hydrophone nodes are within their corresponding allowable threshold ranges, the signal transmission is considered to be faithful; otherwise, it is considered to be unfaithful. If the signal coverage is without blind spots and the signal transmission is of high fidelity, then the stability of the hydrophone monitoring matrix is ​​reliable.

[0013] In this invention, the specific process of constructing the frequency domain feature vector, time domain feature vector, and spatial domain feature vector is as follows: The specific process of constructing frequency domain feature vectors is as follows: The acoustic frequencies of large silverfish during their breeding season were collected using a hydrophone monitoring matrix. The 13th-order Mel-frequency cepstral coefficients, spectral centroid SC, spectral bandwidth SB, and spectral entropy SE of these acoustic frequencies were extracted to form a frequency domain feature vector. ; The specific process of constructing time-domain feature vectors is as follows: An adaptive thresholding method was used to segment pulses in the audio frequency of large silverfish during their breeding season: a signal amplitude threshold was set. , The standard deviation of background noise will be continuously exceeded. A signal segment is determined as a pulse, and the interval between adjacent pulses is > Then it is considered a new sequence; Constructing temporal feature vectors Where IPI is the pulse interval, the time difference between adjacent pulses. AMP is the pulse duration, the duration of a single pulse; AMP is the amplitude variation, the coefficient of variation of the pulse amplitude. The specific process of constructing spatial feature vectors is as follows: Using the time-difference localization method with hydrophone monitoring matrices, the source coordinates of each acoustic signal event are calculated. Let the coordinates of any three hydrophone monitoring matrices be 3D coordinate 1, 3D coordinate 2, and 3D coordinate 3, respectively. The measured signal arrival times are respectively... , , The time difference between any two three-dimensional coordinates is calculated. By solving the Euclidean distance system with respect to the distance between the hydrophone monitoring matrices, the sound source coordinates are obtained. Based on the sound source coordinates, a spatial feature vector is constructed. ,in, The number of acoustic signal events per unit area. The average water depth of the acoustic signal event. This represents the coordinate changes between adjacent events.

[0014] In this invention, the specific process of constructing the behavior-sound behavior integrated recognition model is as follows: The spatial feature vector, temporal feature vector, and spatial feature vector are first manually labeled to identify positive events of the silverfish. Positive events of the silverfish include: silverfish acoustic time, silverfish courtship event, and silverfish spawning event. The acoustic analysis branch uses a one-dimensional convolutional neural network (1D-CNN) to extract local features in the spatial feature vector, and combines a long short-term memory network (LSTM) to learn the time dependence of pulse intervals in the temporal feature vector. The context analysis branch inputs spatial feature vectors and data on water temperature, water depth, dissolved oxygen, and pH into a fully connected neural network; The output features of the two branches are fused together, and finally a classifier is used to determine the probability that the acoustic event belongs to the silverfish sound event, silverfish courtship event, and silverfish spawning event in the silverfish positive event, thus constructing a comprehensive behavioral sound behavior recognition model.

[0015] In this invention, the specific process of generating the acoustic activity heat map of the breeding of large silverfish is as follows: All spatiotemporal coordinates of the positive events of large silverfish identified by the behavior-sound behavior integrated recognition model were imported into the Geographic Information System (GIS) and the Kernel Density Estimation (KDE) analysis method was used to generate a heat map of the acoustic activity of large silverfish reproduction.

[0016] In this invention, the specific process of determining the range of the large silverfish spawning grounds is as follows: Based on the acoustic activity heatmap of the breeding of large silverfish, water environment DNA was sampled from the high, medium and low density activity areas shown in the acoustic activity heatmap of the breeding of large silverfish. The concentration of large silverfish DNA in the water samples was quantitatively detected by laboratory qPCR technology. If the high, medium, and low density activity regions shown in the acoustic activity heatmap of silverfish reproduction are positively correlated with the concentration of silverfish DNA in the water sample, then the aquatic environmental DNA matches the acoustic activity heatmap of silverfish reproduction. Based on the matching of aquatic environmental DNA with the acoustic activity heat map of giant silverfish reproduction, a density threshold is pre-set according to the density in the acoustic activity heat map of giant silverfish reproduction. In the geographic information system, a continuous boundary that meets the density threshold condition is drawn. The area within this boundary is the range of giant silverfish spawning grounds.

[0017] On the other hand, the present invention provides an in-situ precision sampling and survey system for spawning grounds of large silverfish, comprising: Parameter sensing module: Deploy sensors in the water area to collect water parameters, and classify the interference level based on the water parameters using the analytic hierarchy process (AHP). Hydrophone array stability verification module: Based on different interference levels, analyze the effective distance of signal propagation in the water area, determine the spacing between hydrophone array nodes according to the effective distance of signal propagation, form a hydrophone monitoring matrix, and verify the stability of the hydrophone monitoring matrix. Feature extraction module: If the hydrophone monitoring matrix is ​​stable and reliable, then unify the timestamp of the hydrophone monitoring matrix, collect the sound frequency of the silverfish breeding season through the hydrophone monitoring matrix, extract the Mel frequency cepstral coefficients of the sound frequency of the silverfish breeding season, construct frequency domain feature vectors and time domain feature vectors, and use the hydrophone monitoring matrix to construct spatial domain feature vectors. Sound Behavior Comprehensive Recognition Model Construction Module: Based on frequency domain feature vectors, time domain feature vectors, and spatial domain feature vectors, a sound behavior comprehensive recognition model is constructed to identify positive events of large silverfish; Spawning ground range determination module: Import all positive events of giant silverfish into the geographic information system to generate a heat map of the acoustic activity of giant silverfish reproduction. The matching verification of the heat map of the acoustic activity of giant silverfish reproduction is carried out by DNA sampling of the aquatic environment to determine the range of giant silverfish spawning grounds.

[0018] The beneficial effects of this invention are as follows: This invention scientifically classifies water area interference levels through multi-sensor integration and hierarchical analysis, providing a theoretical basis for the optimized deployment of hydrophone arrays and overcoming the interference problem of complex underwater environments on acoustic monitoring. It adopts an adaptive hydrophone array configuration method based on interference levels, which significantly improves the efficiency and reliability of acoustic signal acquisition through dynamic adjustment of node spacing and dual-layer array design. An innovative multi-dimensional feature extraction method was proposed, which analyzes acoustic signals from the perspectives of frequency domain, time domain, and spatial domain. This effectively solves the problem of identifying the overlapping spectrum of the sound signals of large silverfish tapping with those of other fish. The constructed sound behavior comprehensive recognition model combines the advantages of one-dimensional convolutional neural networks and long short-term memory networks, and can accurately identify the sound frequency events, courtship events, and spawning events of large silverfish, greatly improving the monitoring accuracy. By employing a dual verification mechanism of acoustic thermal mapping and aquatic environmental DNA sampling, an evidence chain of acoustics and molecular biology was formed, which greatly improved the scientificity and reliability of the delineation of spawning grounds. It also enabled in-situ, precise, and non-destructive monitoring of the spawning grounds of giant whitebait, providing an effective technical means for the protection, management, and sustainable utilization of giant whitebait resources. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart of the steps of a method for in-situ precise sampling and investigation of spawning grounds of large silverfish according to the present invention; Figure 2 This is a system module diagram of an in-situ precision sampling and investigation system for spawning grounds of large silverfish, as described in this invention. Detailed Implementation

[0021] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0022] Example 1 like Figure 1 As shown in the embodiment of the present invention, a method for in-situ precise sampling and investigation of spawning grounds of large silverfish includes: S1: Deploy sensors in the water area to collect water parameters, and classify the interference level based on the water parameters using the analytic hierarchy process. In S1, specifically, sensors are deployed in the water area for data collection, including: a laser turbidity sensor, a miniature ADCP flow velocity sensor, and an integrated sound velocity meter. In S1, the second specific process for collecting water body parameters is as follows: A laser turbidity sensor, a miniature ADCP flow velocity sensor, and an integrated sound velocity meter are deployed in the water area to collect water parameters, including water turbidity, water flow velocity, and sound velocity profile. The laser turbidity sensor is used to monitor and acquire water turbidity, the miniature ADCP flow velocity sensor is used to monitor and acquire water flow velocity, and the integrated sound velocity meter is used to monitor the sound velocity profile and acquire the sound velocity distribution of different water layers, avoiding the propagation time calculation error caused by using a fixed sound velocity value.

[0023] In S1, the third specific process, based on water body parameters and using the analytic hierarchy process (AHP) to classify disturbance levels, is as follows: Based on the water parameters collected in the sampling area, a hierarchical analysis structure is constructed. The target layer is the interference level L, and the criterion layer includes three core indicators: water turbidity, water flow velocity, and sound velocity profile. The collected water parameters are normalized, a judgment matrix is ​​constructed using expert scoring, the importance of the three core indicators to the interference level is compared, and the scientific validity is ensured through consistency verification. The three core indicators after normalization are superimposed according to their weights through weighted linear combination to obtain the comprehensive interference index. The interference level is then classified according to the comprehensive interference index. For example, the collected water parameters are normalized, and the three core indicator values ​​are mapped to the [0,1] interval. A judgment matrix is ​​constructed using an expert scoring method, such as the water turbidity weight obtained according to the expert scoring method. Water flow velocity weight Weight of sound velocity profile Then the judgment matrix is: 1 represents equal importance, 2 represents the former being more important than the latter. The largest eigenvalue can be calculated through matrix operations. Checking the random consistency index, when n=3, the random consistency index is equal to 0.58. Calculating the consistency ratio CR, since the consistency ratio CR<0.1, the judgment matrix meets the consistency requirements, and the weight allocation is scientific and reliable. By using a weighted linear combination, the dimensionless indicators are superimposed according to their weights to obtain the comprehensive interference index: , The turbidity value of the water body after normalization. The normalized water flow velocity value. The normalized sound velocity profile value is used to classify the interference level according to the comprehensive interference index. For example, water parameters collected from a certain sampling area are normalized to obtain... , , Substituting the values ​​into the comprehensive interference index, we obtain a comprehensive interference index of 0.4068. According to the zoning rules, this area is classified as a medium interference zone. It should be noted that, since the water area being collected is subject to significant natural disturbances, this invention does not specifically limit the level of disturbance or the zoning rules. S2: Based on different interference levels, analyze the effective distance of signal propagation in the water area, determine the spacing between hydrophone array nodes according to the effective distance of signal propagation, form a hydrophone monitoring matrix, and verify the stability of the hydrophone monitoring matrix. In S2, the first specific step is to analyze the effective signal propagation distance of the water area based on different interference levels, and then determine the spacing between hydrophone array nodes based on the effective signal propagation distance. Based on different interference levels, the turbidity (NTU) and flow velocity (c) of the water body parameters corresponding to different interference levels are obtained, and the effective signal propagation distance (D) of the collected water area is calculated using the following formula: ,in, For calibration coefficients, The reference sound velocity is given in standard clear water environment, in m / s. 500 is the turbidity value of the reference water area, used to normalize the effect of water turbidity on sound propagation. If the turbidity (NTU) of the water body being sampled is less than 500, the effective propagation distance can be calculated using the formula. The hydrophone array node spacing is set to 200m, and it is deployed in the core water layer during the breeding season of the giant silverfish, such as the water depth. Place; If the turbidity (NTU) of the water sampled is greater than or equal to 500 but less than 1500, the effective propagation distance can be calculated using the formula. The hydrophone array node spacing is set to 100m, and it is deployed in the core water layer during the breeding season of the giant silverfish, such as the water depth. Place; If the water turbidity (NTU) of the sampled water body is greater than or equal to 1500, the effective propagation distance can be calculated using the formula. The node spacing of the hydrophone array is set to 50m, and a double-layer array is used, which is deployed in the surface layer and the middle layer respectively. The horizontal propagation attenuation is compensated by superimposing the vertical acoustic signals. In S2, the second specific hydrophone monitoring matrix includes: On the hydrophone array nodes, integrated water temperature, water depth, dissolved oxygen, and pH sensors are deployed to synchronously record water temperature, water depth, dissolved oxygen, and pH data when each sound signal occurs, for subsequent data filtering. In S2, the third specific step, the process of verifying the stability of the hydrophone monitoring matrix, is as follows: A two-dimensional rectangular coordinate system is established with the geometric center of the hydrophone monitoring matrix as the origin; the node spacing D of the hydrophone array is matched with the current water interference level to determine the placement spacing d of the audio simulator. The spacing d between the sampling points is no greater than 1 / 2 of the node spacing D, ensuring that the spatial sampling rate satisfies the Nyquist sampling theorem and can effectively capture the spatial response characteristics of the monitoring matrix. Based on the spacing d, a two-dimensional grid is formed within the coverage area of ​​the monitoring matrix, and an audio simulator is deployed at each grid point. Control all audio simulators to play preset audio signals for the breeding season of large silverfish in sequence; record whether each node in the hydrophone array successfully receives the signal when each audio simulator plays the signal; For any audio simulator node Si, if there exists a circular region centered on Si with a radius of R, where R is the effective communication radius of the hydrophone, the value of which can be determined through previous experiments or provided by the equipment manufacturer, such that at least K hydrophone nodes (K is a preset value, such as K=3, but this invention does not limit K) can receive the signal emitted by Si, then it is determined that the signal coverage of node Si meets the requirements. If the signal coverage of all audio simulator nodes meets the requirements, then the signal coverage of the entire hydrophone monitoring matrix is ​​determined to be without blind spots; otherwise, blind spots exist. The waveform of the audio signal received by each hydrophone node is recorded synchronously; The received signal is compared with the original standard signal output by the audio simulator to extract key acoustic feature parameters, including but not limited to: center frequency deviation Δf, signal-to-noise ratio SNR, and waveform similarity NCC. Set allowable thresholds for each characteristic parameter. If all characteristic parameters of the signals received by all hydrophone nodes are within their corresponding allowable threshold ranges, the signal transmission is considered to be faithful; otherwise, it is considered to be unfaithful. In this invention, the permissible threshold values ​​are, for example: center frequency deviation Δf ≤ 5Hz, signal-to-noise ratio SNR ≥ 20dB, and waveform similarity NCC ≥ 0.9. This invention does not limit these values. Comprehensive verification: If the above signal coverage has no blind spots and the signal transmission is of high fidelity, then the hydrophone monitoring matrix is ​​stable and reliable. S3: If the hydrophone monitoring matrix is ​​stable and reliable, then unify the timestamp of the hydrophone monitoring matrix, collect the sound frequency of the silverfish breeding season through the hydrophone monitoring matrix, extract the Mel frequency cepstral coefficients of the sound frequency of the silverfish breeding season, construct the frequency domain feature vector, time domain feature vector, and use the hydrophone monitoring matrix to construct the spatial domain feature vector. A unified hydrophone monitoring matrix timestamp is used to collect audio frequencies of large silverfish during their breeding season, providing accurate data for the subsequent construction of frequency domain feature vectors, time domain feature vectors, and spatial domain feature vectors. In S3, the specific process of constructing the frequency domain feature vector is as follows: The acoustic frequencies of large silverfish during their breeding season were collected using a hydrophone monitoring matrix. The 13th order Mel-frequency cepstral coefficients, spectral centroid (SC), spectral bandwidth (SB), and spectral entropy (SE) of the acoustic frequencies during the breeding season were extracted to form a frequency domain feature vector. ; In S3, the second specific step, the process of constructing the temporal feature vector, is as follows: An adaptive thresholding method was used to segment pulses in the audio frequency of large silverfish during their breeding season: a signal amplitude threshold was set. , The standard deviation of background noise will be continuously exceeded. A signal segment is determined as a pulse, and the interval between adjacent pulses is > ( If the value is a preset value (e.g., T=100ms, which is not limited in this invention), it is considered a new sequence; Constructing temporal feature vectors Where IPI is the pulse interval, the time difference between adjacent pulses. AMP is the pulse duration, the duration of a single pulse; AMP is the amplitude variation, the coefficient of variation of the pulse amplitude. In S3, the third specific step, the process of constructing the spatial feature vector, is as follows: Using the time-difference localization method with hydrophone monitoring matrices, the source coordinates of each acoustic signal event are calculated. Let the coordinates of any three hydrophone monitoring matrices be 3D coordinate 1, 3D coordinate 2, and 3D coordinate 3, respectively. The measured signal arrival times are respectively... , , The time difference between any two three-dimensional coordinates is calculated. By solving the Euclidean distance system with respect to the distance between the hydrophone monitoring matrices, the sound source coordinates are obtained. Based on the sound source coordinates, a spatial feature vector is constructed. ,in, The number of acoustic signal events per unit area. The average water depth of the acoustic signal event. This represents the coordinate changes between adjacent events; For example, the time difference localization method using a hydrophone monitoring matrix is ​​used to calculate the sound source coordinates for each acoustic signal event. Let the coordinates of the three hydrophone nodes be... , , The measured signal arrival times were as follows: , , Calculate the time difference between any two three-dimensional coordinates. By using Euclidean distance to solve the simultaneous equations regarding the distances between the hydrophone monitoring matrices: Where c is the propagation speed of the sound signal in the current water, the coordinates of the sound source are obtained by solving. Constructing spatial feature vectors based on sound source coordinates ; S4: Construct a comprehensive behavioral sound behavior recognition model based on frequency domain feature vectors, time domain feature vectors, and spatial domain feature vectors to identify positive events of large silverfish; In S4, the specific process of constructing a comprehensive behavior-sound behavior recognition model based on frequency domain feature vectors, time domain feature vectors, and spatial domain feature vectors is as follows: The spatial feature vector, temporal feature vector, and spatial feature vector are first manually labeled to identify positive events of the silverfish. Positive events of the silverfish include: silverfish acoustic time, silverfish courtship event, and silverfish spawning event. The acoustic analysis branch uses a one-dimensional convolutional neural network (1D-CNN) to extract local features in the spatial feature vector and combines it with a long short-term memory network (LSTM) to learn the time dependence of the pulse interval in the temporal feature vector. The context analysis branch inputs spatial feature vectors and data on water temperature, water depth, dissolved oxygen, and pH into a fully connected neural network; The output features of the two branches are fused together, and finally a classifier is used to determine the probability that the acoustic event belongs to the silverfish sound event, silverfish courtship event, and silverfish spawning event in the silverfish positive event, thus constructing a comprehensive behavioral sound behavior recognition model. In S4, the second specific process for identifying positive events in large silverfish is as follows: Based on the constructed behavior and sound behavior integrated recognition model, the probability of large silverfish sound events, large silverfish courtship events, and large silverfish spawning events in the large silverfish positive event is identified, and the three-dimensional coordinates of large silverfish sound events, large silverfish courtship events, and large silverfish spawning events are standardized. S5: Import all positive events of giant silverfish into the geographic information system to generate a heat map of the acoustic activity of giant silverfish reproduction. The matching verification of the heat map of the acoustic activity of giant silverfish reproduction is carried out by DNA sampling of the aquatic environment to determine the range of giant silverfish spawning grounds. In S5, the first specific step, importing all positive events of large silverfish into the geographic information system to generate a heat map of the acoustic activity of large silverfish reproduction, is as follows: All spatiotemporal coordinates of the large silverfish positive events identified by the behavior-sound behavior integrated recognition model were imported into a geographic information system (GIS), and a heat map of the acoustic activity of large silverfish reproduction was generated using the kernel density estimation (KDE) analysis method. The darker the area in the diagram, the denser the acoustic events, indicating the core area of ​​reproductive activity.

[0024] In S5, the second specific step, which involves verifying the matching of the acoustic activity heatmap of the large silverfish's reproduction through DNA sampling of the aquatic environment, is as follows: Based on the acoustic activity heatmap of the breeding of large silverfish, water environment DNA was sampled from the high, medium and low density activity areas shown in the acoustic activity heatmap of the breeding of large silverfish. The concentration of large silverfish DNA in the water samples was quantitatively detected by laboratory qPCR technology. If the high, medium, and low density activity regions shown in the acoustic activity heatmap of silverfish reproduction are positively correlated with the concentration of silverfish DNA in the water sample, then the aquatic environmental DNA matches the acoustic activity heatmap of silverfish reproduction. In S5, the third specific process for determining the extent of the large silverfish spawning grounds is as follows: Based on the matching of aquatic environmental DNA with the acoustic activity heat map of giant silverfish reproduction, a density threshold is pre-set according to the density in the acoustic activity heat map of giant silverfish reproduction. In the geographic information system, a continuous boundary that meets the density threshold condition is drawn. The area within this boundary is the range of giant silverfish spawning grounds. It should be noted that the density threshold is a reference value set by technical personnel in this industry based on historical experience; Example 2 like Figure 2 As shown in Example 1, this invention provides an in-situ precision sampling and survey system for large silverfish spawning grounds, comprising: Parameter sensing module: Deploy sensors in the water area to collect water parameters, and classify the interference level based on the water parameters using the analytic hierarchy process (AHP). Hydrophone array stability verification module: Based on different interference levels, analyze the effective distance of signal propagation in the water area, determine the spacing between hydrophone array nodes according to the effective distance of signal propagation, form a hydrophone monitoring matrix, and verify the stability of the hydrophone monitoring matrix. Feature extraction module: If the hydrophone monitoring matrix is ​​stable and reliable, then unify the timestamp of the hydrophone monitoring matrix, collect the sound frequency of the silverfish breeding season through the hydrophone monitoring matrix, extract the Mel frequency cepstral coefficients of the sound frequency of the silverfish breeding season, construct frequency domain feature vectors and time domain feature vectors, and use the hydrophone monitoring matrix to construct spatial domain feature vectors. Sound Behavior Comprehensive Recognition Model Construction Module: Based on frequency domain feature vectors, time domain feature vectors, and spatial domain feature vectors, a sound behavior comprehensive recognition model is constructed to identify positive events of large silverfish; Spawning ground range determination module: Import all positive events of giant silverfish into the geographic information system to generate a heat map of the acoustic activity of giant silverfish reproduction. The matching verification of the heat map of the acoustic activity of giant silverfish reproduction is carried out by DNA sampling of the aquatic environment to determine the range of giant silverfish spawning grounds.

[0025] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for in-situ accurate sampling investigation of spawning ground of Pseudobagrus, characterized in that: The method comprises the following steps: S1: deploying sensors in the collection water area to collect water body parameters, and dividing interference levels by using the analytic hierarchy process based on the water body parameters; S2: based on different interference levels, analyzing the signal propagation effective distance of the collection water area, determining the hydrophone array node spacing according to the signal propagation effective distance, forming a hydrophone monitoring matrix, and verifying the stability of the hydrophone monitoring matrix; S3: if the stability of the hydrophone monitoring matrix is reliable, the time stamp of the hydrophone monitoring matrix is unified, the sound frequency of the icefish breeding period is collected by the hydrophone monitoring matrix, the mel frequency cepstrum coefficient of the sound frequency of the icefish breeding period is extracted, the frequency domain feature vector and the time domain feature vector are constructed, and the spatial domain feature vector is utilized; S4: according to the frequency domain feature vector, the time domain feature vector and the spatial domain feature vector, an action sound behavior comprehensive recognition model is constructed to identify the icefish positive events; S5: all icefish positive events are imported into a geographic information system to generate an icefish breeding acoustic activity heat map, and the icefish breeding acoustic activity heat map is verified for matching by water body environment DNA sampling to determine the range of icefish egg field.

2. The method according to claim 1, wherein the method is characterized by: The specific process of collecting water body parameters is as follows: Laser turbidity sensors, micro ADCP flow rate sensors and integrated sound velocity meters are arranged in the collection water area to collect water body parameters, and the water body parameters specifically include water turbidity, water flow rate and sound velocity profile, wherein the laser turbidity sensor is used to monitor and obtain the water turbidity, the micro ADCP flow rate sensor is used to monitor and obtain the water flow rate, and the integrated sound velocity meter is used to monitor the sound velocity profile and obtain the sound velocity distribution of different water layers.

3. The method according to claim 1, wherein the method is characterized by: The specific process of dividing interference levels by using the analytic hierarchy process is as follows: According to the water body parameters collected in the collection water area, an analytic hierarchy structure is constructed, the target layer is the interference level L, and the criterion layer includes three core indexes of water turbidity, water flow rate and sound velocity profile; The collected water body parameters are normalized, a judgment matrix is constructed by expert scoring method, the importance of the three core indexes to the interference level is compared, and the scientificity is ensured by consistency test, the normalized three core indexes are superimposed according to the weight by weighted linear combination to obtain a comprehensive interference index, and the interference level is divided according to the comprehensive interference index.

4. The method according to claim 1, wherein the method is characterized by: The specific process of determining the hydrophone array node spacing according to the signal propagation effective distance is as follows: Based on different interference levels, different interference levels are obtained Corresponding to the water body parameter of water area turbidity NTU, water flow velocity c, the signal propagation effective distance D of the collected water area is calculated, and the specific formula is: Wherein, is a calibration coefficient, is the reference sound speed in the standard clear water environment, unit m / s, 500 is the reference water area turbidity value, used for normalizing the influence of water area turbidity on sound propagation; If the water turbidity NTU of the collection water area is less than 500, the effective propagation distance is calculated and the hydrophone array node spacing is set to 200 m. If the water turbidity NTU of the collected water area is greater than or equal to 500 and less than 1500, the effective propagation distance is calculated The hydrophone array node spacing is set to 100 m. If the water turbidity NTU of the collected water area is greater than or equal to 1500, the effective propagation distance is calculated The hydrophone array node spacing is set to 50 m, a double-layer array is adopted, and is respectively arranged in the surface layer and the middle layer, and the horizontal propagation attenuation is compensated by vertical direction sound signal superposition.

5. The method according to claim 1, wherein the method is characterized by: A two-dimensional rectangular coordinate system is established with the geometric center of the hydrophone monitoring matrix as the origin, the node spacing D of the hydrophone array under the current water area interference level is matched to determine the point spacing d of the sound frequency simulator; The point spacing d is not greater than 1 / 2 of the node spacing D to ensure that the spatial sampling rate meets the Nyquist sampling theorem and can effectively capture the spatial response characteristics of the monitoring matrix; According to the point spacing d, a two-dimensional grid array is formed in the monitoring matrix coverage area, and a sound frequency simulator is arranged at each grid point; All sound frequency simulators play the preset icefish breeding period sound frequency signal in turn; Whether each node of the hydrophone array successfully receives the signal when each sound frequency simulator plays the signal is recorded. For any one of the audio simulator node Si, if there is a circular area with it as the center and a radius of R, R is the effective communication radius of the hydrophone, which can be determined by the pre-experiment or provided by the equipment manufacturer, so that at least K hydrophone nodes in the area can receive the signal sent by Si, it is determined that the signal coverage of node Si meets the requirements; If the signal coverage of all audio simulator nodes meets the requirements, it is determined that the signal coverage of the entire hydrophone monitoring matrix has no blind area; Otherwise, there is a blind area; Synchronously record the acoustic signal waveform received by each hydrophone node; Compare the received signal with the original standard signal output by the audio simulator, and extract key acoustic feature parameters, including but not limited to: center frequency deviation Δf, signal-to-noise ratio SNR, waveform similarity NCC; Set the allowed threshold of each feature parameter, if all the feature parameters of the signals received by all the hydrophone nodes are within the corresponding allowed threshold range, it is determined that the signal transmission is faithful; otherwise, it is determined to be unfaithful; If the signal coverage has no blind area and the signal transmission is faithful, the stability of the hydrophone monitoring matrix is reliable.

6. The method according to claim 1, wherein the method is characterized by: The specific process of constructing the frequency domain feature vector, the time domain feature vector, and the spatial domain feature vector is: The specific process of constructing the frequency domain feature vector is: The sound frequency of the breeding period of the icefish is collected by a hydrophone monitoring matrix, and the 13-order coefficients of the Mel frequency cepstrum, the spectral centroid SC, the spectral bandwidth SB and the spectral entropy SE of the sound frequency of the breeding period of the icefish are extracted to form a frequency domain feature vector ; The specific process of constructing the time domain feature vector is: An adaptive thresholding method was used to segment pulses in the audio frequency of silverfish during their breeding season: a signal amplitude threshold was set. , The standard deviation of background noise will be continuously exceeded. A signal segment is determined as a pulse, and the interval between adjacent pulses is > Then it is considered a new sequence; Constructing time-domain feature vectors where IPI is the inter-pulse interval, the time difference between adjacent pulses, is the pulse duration, the length of time of a single pulse, and AMP is the amplitude variation, the coefficient of variation of pulse amplitude. The specific process of constructing the spatial domain feature vector is: The time difference positioning method of the hydrophone monitoring matrix is used to calculate the sound source coordinates of each sound signal event. Assuming that the coordinates of any three hydrophone monitoring matrices are three-dimensional coordinate 1, three-dimensional coordinate 2 and three-dimensional coordinate 3, the measured signal arrival times are , , The time difference between any two three-dimensional coordinates is calculated, and the distance between the hydrophone monitoring matrices is solved by the Euclidean distance to obtain the sound source coordinates. The spatial feature vector is constructed based on the sound source coordinates , wherein is the number of sound signal events per unit area, is the average water depth of the sound signal event, is the coordinate change of adjacent events.

7. The method according to claim 1, wherein the method is characterized by: The specific process of constructing the behavior sound behavior comprehensive recognition model is: First, manually label the spatial domain feature vector, the time domain feature vector, and the spatial domain feature vector to determine the silverfish positive event, which includes: silverfish audio time, silverfish courtship event, and silverfish spawning event; The acoustic analysis branch uses a one-dimensional convolutional neural network 1D-CNN to extract local features in the spatial domain feature vector, and combines a long short-term memory network LSTM to learn the time-dependent relationship of pulse intervals in the time domain feature vector; The context analysis branch inputs the spatial domain feature vector and water temperature, water depth, dissolved oxygen, and pH value data into a fully connected neural network; Fuse the output features of the two branches, and finally determine the probability of the acoustic event belonging to the silverfish audio event, the silverfish courtship event, and the silverfish spawning event in the silverfish positive event by a classifier, to construct the behavior sound behavior comprehensive recognition model.

8. The method according to claim 1, wherein the method is characterized by: The specific process of generating the silverfish reproductive acoustic activity heat map is: Import all the spatio-temporal coordinate points recognized as silverfish positive events by the behavior sound behavior comprehensive recognition model into a geographic information system GIS, and use the kernel density estimation KDE analysis method to generate the silverfish reproductive acoustic activity heat map.

9. The method according to claim 1, wherein the method is characterized by: The specific process of determining the silverfish egg field range is: Based on the silverfish reproductive acoustic activity heat map, perform water body environmental DNA sampling on the high, medium, and low activity areas displayed in the silverfish reproductive acoustic activity heat map, and quantitatively detect the concentration of silverfish DNA in the water sample through laboratory qPCR technology; If the high, medium, and low activity areas displayed in the silverfish reproductive acoustic activity heat map are positively correlated with the concentration of silverfish DNA in the water sample, the water body environmental DNA matches the silverfish reproductive acoustic activity heat map. Based on the matching of water body environmental DNA and the acoustic activity thermogram of icefish reproduction, the density threshold is set in advance according to the density in the acoustic activity thermogram of icefish reproduction, and the continuous boundary meeting the density threshold condition is drawn in the geographic information system, so that the egg field range of icefish is determined.

10. A precise sampling investigation system in situ of an egg laying ground of Pseudobagrus, characterized in that: Comprise: Parameter perception module: deploy sensors in the water area, collect water parameters, and divide the interference level based on the water parameters through the analytic hierarchy process; The stability verification module of the hydrophone array: based on different interference levels, analyze the signal propagation effective distance of the water area, determine the node spacing of the hydrophone array according to the signal propagation effective distance, form a hydrophone monitoring matrix, and verify the stability of the hydrophone monitoring matrix; Feature extraction module: if the stability of the hydrophone monitoring matrix is reliable, unify the time stamp of the hydrophone monitoring matrix, collect the sound frequency of the icefish breeding period through the hydrophone monitoring matrix, extract the mel frequency cepstrum coefficient of the sound frequency of the icefish breeding period, construct the frequency domain feature vector, time domain feature vector, and use the hydrophone monitoring matrix, spatial feature vector; Sound behavior comprehensive recognition model construction module: according to the frequency domain feature vector, time domain feature vector, and spatial feature vector, construct a sound behavior comprehensive recognition model for identifying icefish positive events; Egg field range determination module: import all icefish positive events into the geographic information system to generate an acoustic activity thermogram of icefish reproduction, and perform matching verification on the acoustic activity thermogram of icefish reproduction through water body environmental DNA sampling to determine the egg field range of icefish.

Images