Method, device, computer equipment and medium for monitoring non-point source pollution during flood season

Abstract

The embodiment of the present invention provides a method, device, computer equipment and medium for monitoring non-point source pollution during flood season, which relates to the field of environmental monitoring technology. The method includes obtaining multiple water quality spectral data of continuous time points including flood season at multiple water body basin monitoring points, the water quality spectral data being the absorbance of water samples within a preset wavelength range; marking the pollution time during flood season and the non-point source pollution type for multiple water quality spectral data; extracting water quality spectral features from the marked multiple water quality spectral data respectively; generating a three-dimensional water quality spectral feature map corresponding to each water quality spectral data based on the water quality spectral features; using multiple three-dimensional water quality spectral feature maps and the marked contents as training data to train a monitoring model; and monitoring and identifying the pollution time during flood season and the non-point source pollution type of the water body basin to be monitored based on the trained monitoring model. The scheme improves the recognition precision and accuracy of non-point source pollution and reduces the recognition error.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and in particular to a method, device, computer equipment and medium for monitoring non-point source pollution during flood season. Background Art

[0002] Non-point source pollution during the flood season is the main source of river pollution. How to determine which pollution comes from non-point source pollution is an urgent problem that needs to be solved. The commonly used methods for determining non-point source pollution include: the first is to use flow thresholds to determine non-point source pollution data: this method uses flood season flow to determine which moments are at high water levels during the flood season, and then determines that pollution under high flow conditions belongs to flood season pollution. This method has a large error, that is, the increase in river flow is not necessarily caused by river confluence, and may be related to upstream sluices. The second is to determine non-point source pollution data in combination with precipitation data: this method uses precipitation data recorded by local rain gauges to determine the flood season time, and then marks some pollution events as non-point source pollution. This method also has errors. Since it takes time to produce runoff, the extreme precipitation time and the high flow time of the river channel may not coincide. In addition, the rain gauge can only represent the precipitation conditions in some areas, and the differences in regional precipitation will lead to large errors. Summary of the invention

[0003] In view of this, an embodiment of the present invention provides a method for monitoring non-point source pollution during flood season to solve the technical problem of large errors in the judgment method of non-point source pollution during flood season in the prior art. The method comprises:

[0004] Obtaining a plurality of water quality spectral data of a plurality of water body basin monitoring points at consecutive time points including the flood season, wherein the water quality spectral data is the absorbance of the water sample within a preset wavelength range;

[0005] Marking the pollution time during the flood season and the type of non-point source pollution for the plurality of water quality spectral data;

[0006] Extracting water quality spectrum features from the marked water quality spectrum data respectively;

[0007] Generate a three-dimensional water quality spectral characteristic map corresponding to each of the water quality spectral data based on the water quality spectral characteristics;

[0008] The monitoring model is trained using the plurality of three-dimensional water quality spectral characteristic maps and corresponding labels as training data, wherein the labels include marked flood season pollution moments and non-point source pollution types;

[0009] Based on the trained monitoring model, the pollution time during the flood season and the type of non-point source pollution in the monitored water basin are monitored and identified.

[0010] The embodiment of the present invention also provides a flood season non-point source pollution monitoring device to solve the technical problem of large errors in the judgment method of flood season non-point source pollution in the prior art. The device includes:

[0011] An acquisition module, used to obtain a plurality of water quality spectral data of a plurality of water body basin monitoring points at consecutive time points including the flood season, wherein the water quality spectral data is the absorbance of the water sample within a preset wavelength range;

[0012] A marking module, used for marking the pollution time during the flood season and the type of non-point source pollution for the plurality of water quality spectral data;

[0013] A feature extraction module, used for respectively extracting water quality spectrum features from the marked plurality of water quality spectrum data;

[0014] A spectrum generating module, used for generating a three-dimensional water quality spectrum characteristic spectrum corresponding to each of the water quality spectrum data based on the water quality spectrum characteristics;

[0015] A training module, used to train a monitoring model using a plurality of the three-dimensional water quality spectral characteristic maps and corresponding labels as training data, wherein the labels include marked flood season pollution moments and non-point source pollution types;

[0016] The identification module is used to monitor and identify the flood season pollution time and the non-point source pollution type of the monitored water basin based on the trained monitoring model.

[0017] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any of the above-mentioned methods for monitoring non-point source pollution during flood season is implemented to solve the technical problem of large errors in the methods for judging non-point source pollution during flood season in the prior art.

[0018] An embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program for executing any of the above-mentioned flood season non-point source pollution monitoring methods, so as to solve the technical problem of large errors in the judgment method of flood season non-point source pollution in the prior art.

[0019] Compared with the prior art, the beneficial effects that can be achieved by at least one of the above-mentioned technical solutions adopted in the embodiments of this specification include at least: obtaining multiple water quality spectral data of multiple water body basin monitoring points including continuous time points during the flood season, the water quality spectral data being the absorbance of the water sample within a preset wavelength range; marking the pollution moments during the flood season and the types of non-point source pollution for the multiple water quality spectral data; extracting water quality spectral features from the marked multiple water quality spectral data respectively; generating a three-dimensional water quality spectral feature map corresponding to each water quality spectral data based on the water quality spectral features; using the multiple three-dimensional water quality spectral feature maps and the corresponding labels as training data to train the monitoring model, wherein the labels include the marked pollution moments during the flood season and the types of non-point source pollution; and monitoring and identifying the pollution moments during the flood season and the types of non-point source pollution of the monitored water body basin based on the trained monitoring model. The non-point source pollution monitoring method of the present application captures comprehensive spectral characteristics within multiple wavelength ranges. Compared with traditional single-wavelength monitoring, this method provides richer data dimensions, which helps to improve the accuracy of pollutant identification and reduce judgment errors. The monitoring model is optimized and trained through a large amount of data, thereby improving the recognition ability and accuracy. Real-time monitoring and rapid response can be achieved through the monitoring model. Compared with traditional periodic monitoring, this method can detect pollution problems earlier and take timely measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 is a flow chart of a method for monitoring non-point source pollution during flood season provided by an embodiment of the present invention;

[0022] Figure 2 is a schematic diagram of wavelet decomposition provided by an embodiment of the present invention;

[0023] Figure 3 Schematic diagram of convolution operation in the wavelet decomposition process provided by an embodiment of the present invention;

[0024] Figure 4 It is a schematic diagram of a three-dimensional water quality spectrum characteristic spectrum provided by an embodiment of the present invention;

[0025] Figure 5 is a structural block diagram of a computer device provided by an embodiment of the present invention;

[0026] Figure 6 It is a structural block diagram of a flood season non-point source pollution monitoring device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0028] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0029] In an embodiment of the present invention, a method for monitoring non-point source pollution during flood season is provided, such as Figure 1 As shown, the method includes:

[0030] Step S101, obtaining a plurality of water quality spectral data of a plurality of water basin monitoring points including continuous time points during the flood season, wherein the water quality spectral data is the absorbance of the water sample within a preset wavelength range;

[0031] Step S102, marking the flood season pollution time and non-point source pollution type for the plurality of water quality spectral data;

[0032] Step S103, extracting water quality spectrum features from the marked plurality of water quality spectrum data respectively;

[0033] Step S104: generating a three-dimensional water quality spectral characteristic graph corresponding to each of the water quality spectral data based on the water quality spectral characteristics;

[0034] Step S105, using a plurality of three-dimensional water quality spectral characteristic graphs and corresponding labels as training data to train the monitoring model, wherein the labels include marked flood season pollution time and non-point source pollution type;

[0035] Step S106: monitor and identify the pollution time during the flood season and the type of non-point source pollution in the monitored water basin based on the trained monitoring model.

[0036] Non-point source pollution during the flood season has specific pollution characteristics, which are inconsistent with the characteristics of domestic sewage and industrial sewage discharge. Non-point sources during the flood season are mainly formed by the erosion of farmland and roads. Therefore, non-point source pollution can be judged by the specific typical pollutant characteristics. Non-point source pollution types generally include agricultural non-point source pollution, urban non-point source pollution, forest and wasteland non-point source pollution, etc. Agricultural non-point source pollution is the fertilizers, pesticides and animal feces used in agricultural production entering the water body with rainwater runoff, causing eutrophication of water bodies and other problems; urban non-point source pollution is the rainwater in urban areas that washes the sediments, oil, chemicals and other pollutants on the street surface, and is directly discharged into rivers, lakes and other water bodies without treatment; although forest and wasteland non-point source pollution is relatively small, activities such as deforestation and wasteland development may also lead to soil erosion and increased sediments, affecting water quality.

[0037] This embodiment captures comprehensive spectral features within a preset wavelength range. Compared with traditional single-wavelength monitoring, this method provides richer data dimensions, which helps to improve the accuracy of pollutant identification and reduce judgment errors. The monitoring model is optimized and trained through a large amount of data, thereby improving recognition ability and accuracy. Real-time monitoring and rapid response can be achieved through the monitoring model. Compared with traditional periodic monitoring, this method can detect pollution problems earlier and take timely measures.

[0038] In one embodiment, a spectral device is used to determine whether it is non-point source pollution during the flood season by the spectral characteristics of non-point source pollution during the flood season. Preferably, a high-precision ultraviolet-visible absorption spectroscopy (UV-vis) device is installed at a selected location near a key water basin monitoring station. The device has a remote data transmission function and has a high spectral resolution and sensitivity to ensure that the spectral characteristics of pollutants in the water body can be accurately captured.

[0039] The spectral equipment adopts in-situ monitoring or water pumping monitoring. The in-situ monitoring probe or water intake should be located as close to the center of the section as possible, and the depth should be maintained at 0.5 to 1.0 meters below the water surface. It should be suspended and anchored by a buoy to ensure that the water position and depth of the monitoring point remain unchanged, so that the collected water samples are representative.

[0040] In specific implementation, during the flood season, surface pollution samples are collected manually and the samples are scanned in detail using UV-vis equipment. The collected data is used to generate a three-dimensional water quality spectral characteristic map database covering different periods of the flood season to ensure the timeliness and continuity of the data. Water quality sampling can be carried out at equal time intervals.

[0041] In one embodiment, the preset wavelength range can be set to 200-720nm. By performing a detailed scan within the wavelength range through a UV-vis device, comprehensive spectral features can be captured. Compared with traditional single wavelength monitoring, the method of this embodiment provides richer data dimensions, which helps to improve the accuracy of pollutant identification.

[0042] In one embodiment, the method further comprises:

[0043] Performing data preprocessing on the marked water quality spectrum data, wherein the data preprocessing includes wavelet decomposition to remove noise and data normalization processing;

[0044] Water quality spectral features are extracted respectively based on the plurality of water quality spectral data after data preprocessing.

[0045] In the specific implementation, a non-point source polluted water sample from a certain area during the flood season was selected, and the absorbance information of the water sample was extracted using a WG-3 spectrum stick, with a test wavelength range of 200-700nm. The obtained absorbance data was converted into a linear superposition of sinusoidal signals of different frequencies through data transformation, and the data frequency and other characteristics of the spectrum were extracted to construct the water quality spectrum characteristics, which specifically included the wavelet decomposition process and the wavelet reconstruction process.

[0046] For the wavelet decomposition and noise removal process, the water quality absorbance spectrum data collected by the spectral equipment at multiple consecutive time points is used to form a spectral sequence S. The spectral sequence S is decomposed into a series of simple signals with different frequencies using multi-time scale analysis. After i transformations, the spectral sequence S is transformed to obtain a set of low-frequency approximate components S that can reflect the long-term trend. i and i groups of high-frequency detail components W reflecting short-term fluctuations j , j is the number of decomposition levels, j=1,2,3...i, the decomposition results are as follows Figure 2 As shown, where S j and W j They are the low-frequency approximate component and high-frequency detail component after decomposition at the jth layer. Wavelet mainly includes two steps, namely wavelet decomposition and wavelet reconstruction.

[0047] The wavelet decomposition mainly includes the following steps:

[0048] ① Edge extension. In order to overcome the edge effect in the decomposition process of the finite-length sequence of water absorbance spectra, the spectral sequence X to be transformed is firstly subjected to periodic edge extension. The decomposition process is extended at the head of the spectral sequence X to be transformed. The number of extensions is equal to the number of decomposition layers, which is set as j. The specific extension method is as follows:

[0049] Spectral sequence to be transformed: ;

[0050] Edge extension sequence:

[0051] Among them, j is the number of decomposition levels.

[0052] ② Filter interpolation. Use the Harr wavelet function as the base wavelet to construct the filter. Because the decomposition is completely reconstructed and the factors are evenly distributed to two filters (low-frequency decomposition filter and high-frequency decomposition filter), the filter coefficients are multiplied by , then the parameters of the low-frequency analysis filter and the high-frequency analysis filter are , ,dec l Represents the low-frequency decomposition filter parameters, dec h Represents the high-frequency decomposition filter parameters. Insert j-1 zeros in the middle of the decomposition filter parameters to form the j-th layer decomposition filter coefficients. After interpolation, the j-th layer low and high frequency decomposition filter coefficients are and , where the number of 0s in the middle is ji.

[0053] ③ Convolution operation. The j-th layer decomposition filter coefficients are convolved with the edge extension sequence. The specific process is as follows: Figure 3 As shown, C mn =C m xC n , m is the length of the decomposition filter coefficient, n is the length of the edge extension sequence; the subscript b is the individual items of the filter coefficient, and the subscript d is the sum of the corresponding columns. After the spectrum sequence X to be transformed passes through the low-frequency decomposition filter and the high-frequency decomposition filter, a convolution operation is performed, and finally a low-frequency sequence to be reconstructed RS is formed. j and high frequency sequence to be reconstructed RW j , the sequence to be reconstructed is the sequence formed by convolution of the edge extension sequence and different decomposition filters.

[0054] For wavelet reconstruction, wavelet reconstruction is the inverse process of wavelet decomposition. j and high frequency sequence to be reconstructed RW j Reconstruct and obtain the low-frequency approximate component S after wavelet transform j and high frequency detail component W j The reconstruction process is roughly the same as the decomposition process, but the details are different. The specific process of wavelet reconstruction is as follows:

[0055] ① Edge extension. The wavelet decomposition process is to extend the head of the sequence. The difference is that the wavelet reconstruction process is to extend the tail of the sequence. The number of extensions is also equal to the number of decomposition layers, set as j. The sequence after the reconstruction edge extension is It should be noted that the sequence X here should be replaced by the corresponding low-frequency sequence to be reconstructed or the high-frequency sequence to be reconstructed, and the corresponding edge extension should be performed.

[0056] ② Filter interpolation. The low and high frequency parameters of the reconstruction filter are , , rec l is the low-frequency reconstruction filter parameter, rec h is the high frequency reconstruction filter parameter. Its interpolation method refers to the interpolation method of the decomposition filter.

[0057] ③ Convolution operation. The reconstructed convolution operation process refers to the decomposed convolution operation process. After sequence reconstruction, the low-frequency approximate component S of the j-th layer transformation is obtained. j and high frequency detail component W j .

[0058] According to the transformation process, S after i transformations j and W j satisfy:

[0059] , j=1,2,3...,i.

[0060] Through the above-mentioned process of wavelet decomposition and denoising of water quality spectrum data, the following effects can be achieved:

[0061] Remove random noise: Spectral data is often affected by random noise, which will interfere with the true characteristics of the signal. Wavelet transform can effectively separate the noise components in the signal, thereby improving the clarity of the signal; Enhance useful information: Wavelet decomposition can highlight important features in the signal, such as the peak position and intensity in the spectrum, which is very important for subsequent data analysis; Simplify data processing: The denoised data is cleaner, which helps to simplify the subsequent data processing process; Improve analysis accuracy: Reducing the impact of noise can improve the accuracy of subsequent analysis; Improve visualization effect: The spectral curve after denoising is smoother, which is conducive to better display and understanding of data trends and details.

[0062] In one embodiment, the normalization processing of water quality spectrum data is described in detail. In order to avoid the influence of water quality concentration and unify the statistical distribution of samples, it is necessary to process the low-frequency baselines of different concentrations of the same water sample, and obtain the water quality spectrum data after data preprocessing by normalizing them.

[0063] The specific normalization process includes: the initial matrix is ​​formed by the low-frequency baselines of different concentrations of the same water sample as [Y 1 ,Y 2 , …,Y k ], k is the number of samples with different concentrations of the same sample. kThe low-frequency baseline in the spectral data obtained at the kth concentration of a sample is selected by selecting a column where a specific wavelength is located as the reference column Y for the normalization processing of the initial matrix. x (x<k), the entire initial matrix [Y 1 , Y 2 , …,Y k ] and divide by Y x , and obtain the transition matrix [Y 1x , Y 2x , …,Y kx ], and the absorbance data of the same wavelength in the transition matrix are averaged to generate the corresponding basic spectrum matrix [Y 200 , Y 202 , …,Y 710 ], the subscript of Y in the basic spectrum matrix is ​​the corresponding wavelength.

[0064] For the above normalization processing, the method of calculating the difference between dimensions is mainly used to construct a closed curve based on the same difference, generate a spectrum that characterizes the pollution type characteristics with systematicity, characteristic and stability, and finally form the water quality spectrum data after data preprocessing. Through normalization processing, the impact of different concentration differences can be eliminated, making the data between different concentrations comparable; it can accelerate the monitoring model training process and help avoid the gradient descent problem caused by differences in data magnitude; it can reduce random fluctuations in the data, thereby reducing the impact of noise on the final results.

[0065] In one embodiment, extracting water quality spectral features from the labeled plurality of water quality spectral data respectively comprises:

[0066] Extracting the peak position of absorbance, the trough position of absorbance and the trend characteristics of absorbance changing with wavelength for each of the marked water quality spectrum data;

[0067] The peak position of the absorbance, the trough position of the absorbance, and the trend characteristics of the absorbance changing with the wavelength of each water quality spectrum data constitute the water quality spectrum characteristics of each water quality spectrum data.

[0068] In specific implementation, the water quality spectrum feature extraction is mainly the trend characteristics of the peak and trough positions and absorbance changing with wavelength. By using correlation analysis to determine the contribution rate of different bands, the bands with large contribution rates are extracted as characteristic bands, the bands with strong correlation are eliminated, and the bands with weak correlation are extracted. Among them, trend feature extraction: t1=y[p+1]-y[p], t1 represents the absorbance trend characteristic function with wavelength, y[p] is the absorbance at wavelength p, y[p+1] is the absorbance at wavelength p+1, t1>0 indicates that the absorbance intensity is strengthened, and t1<0 indicates that the absorbance is weakened. Identification of peaks and troughs: t2=y[p]+y[p+2]-2×y[p+1], t2 represents the peak and trough identification function, y[p+2] is the absorbance at wavelength p+2, t2>0 indicates that p+1 is a trough, and t2<0 indicates that p+1 is a peak. The method is used to extract water quality spectral characteristics and prepare data for generating a three-dimensional water quality spectral characteristic map.

[0069] In one embodiment, generating a three-dimensional water quality spectral characteristic map corresponding to each of the water quality spectral data based on the water quality spectral characteristics includes:

[0070] Performing Fourier transformation on the water quality spectrum characteristics to obtain a basic wave corresponding to each of the water quality spectrum data, an absorbance vector change relative to the basic wave, and a peak position of the absorbance;

[0071] The basic wave is represented by a first coordinate, the vector change relative to the basic waveband is represented by a second coordinate, and the peak position of the absorbance is represented by a third coordinate, wherein the first coordinate, the second coordinate and the third coordinate are located in a planar three-dimensional map, the first coordinate is perpendicular to the second coordinate, and the third coordinate is perpendicular to the second coordinate;

[0072] A three-dimensional water quality spectrum characteristic spectrum corresponding to each of the water quality spectrum data is formed through the planar three-dimensional spectrum.

[0073] When implementing it, refer to Figure 4, the water quality spectrum characteristic data obtained by the above feature extraction is converted into three-dimensional water quality spectrum characteristics through Fourier transformation, etc. The three-dimensional water quality spectrum characteristic map mainly consists of three parts: the basic wave, the absorbance vector change (frequency) relative to the basic wave, and the peak position of the absorbance, which is displayed in the plane as a three-dimensional spectrum. Among them, the basic wave and the vector change are respectively displayed by the horizontal coordinate (first coordinate, wavelength) and vertical coordinate (second coordinate, frequency) information in the figure, and the relative position of the absorption peak is represented by the upper horizontal coordinate (third coordinate, wavelength). The color depth in the figure represents the speed of absorbance change. At the same time, the generated three-dimensional water quality spectrum characteristic map corresponds to the marked flood season pollution time and non-point source pollution type. A standard database can be generated by obtaining the three-dimensional water quality spectrum characteristic map including the flood season, and the monitoring model is trained using the standard database. The trained monitoring model is used to monitor non-point source pollution in the monitored waters.

[0074] In specific implementation, the monitoring model is a convolutional neural network (CNN) model, which performs supervised learning on the three-dimensional water quality spectral feature map. By continuously optimizing the CNN model, the model's recognition accuracy for different pollution types and pollution moments during the flood season is improved, ensuring the robustness and reliability of the model in practical applications.

[0075] In one embodiment, the monitoring and identifying of the flood season pollution time and the non-point source pollution type of the water body basin to be monitored based on the trained monitoring model includes:

[0076] The instantaneous water sample spectrum data of the water body to be monitored is obtained, and the instantaneous water sample spectrum data is identified using the trained monitoring model. If the instantaneous water sample spectrum data is identified as non-point source pollution data during the flood season, the pollution time during the flood season and the type of non-point source pollution are output.

[0077] In one embodiment, the step of using the trained monitoring model to identify the instantaneous water sample spectrum data includes:

[0078] Acquire a three-dimensional water quality spectral characteristic spectrum to be monitored corresponding to the instantaneous water sample spectrum data;

[0079] Using the trained monitoring model to perform image comparison and high-dimensional data feature comparison on the three-dimensional water quality spectral feature map to be monitored;

[0080] Perform comprehensive analysis on the image comparison results and high-dimensional data feature comparison results to obtain comprehensive analysis results;

[0081] Based on the comprehensive analysis result, it is identified whether the instantaneous water sample map data belongs to flood season non-point source pollution data.

[0082] In specific implementation, the trained monitoring model is used to perform image comparison and high-dimensional data feature comparison on the three-dimensional water quality spectral feature map to be monitored, which mainly includes the following steps:

[0083] Setting preset thresholds for comprehensive analysis results in the monitoring model;

[0084] When the comprehensive analysis result of the instantaneous water sample spectrum data of the monitored area exceeds the preset threshold, the pollution type of the area is judged to be non-point source pollution.

[0085] In the specific implementation, the three-dimensional water quality spectral characteristic spectrum to be monitored corresponding to the instantaneous water sample spectrum data is input into the trained monitoring model, and the three-dimensional water quality spectral characteristic spectrum to be monitored is compared with the three-dimensional water quality spectral characteristic spectrum (standard spectrum) in the monitoring model to obtain a comprehensive analysis result. Among them, for image comparison, a low-dimensional comparison algorithm is used to perform simple similarity judgment from the perspectives of the absorption peak position of absorbance, the absorbance change rate or the shape of the spectrum, and the judgment result of the image comparison is expressed by a similarity coefficient. For example, a set of absorbance data consists of 2000 numbers. When two spectra are compared, if 1000 of them are similar, their similarity coefficient is 50%, and the result of the image comparison is 50%. For high-dimensional data feature comparison, low-dimensional data is converted into high-dimensional data through mapping, that is, one-dimensional data is converted into two-dimensional data or above, and data features at high latitudes are extracted. The similarity between the actual water sample (water sample in the monitored area) and the standard spectrum is measured by comparing the centroid distance to achieve the purpose of traceability comparison. The comparison between high-latitude data can make the comparison results more accurate. The final judgment of the flood season non-point source pollution data needs to be made after the comprehensive comparison of image comparison results and high-dimensional data feature comparison results.

[0086] In one embodiment, the method further comprises:

[0087] The trained monitoring model is calibrated according to environmental changes and the drift of water quality spectrum data acquisition equipment.

[0088] In this embodiment, a self-correction function is set for the monitoring model, which can be adjusted according to environmental changes and equipment drift to ensure long-term operational stability and data reliability, thereby ensuring the system's adaptability in various complex environments.

[0089] The method of the present application captures comprehensive spectral features within a preset wavelength range. Compared with traditional single-wavelength monitoring, this method provides richer data dimensions, which helps to improve the accuracy of pollutant identification and reduce judgment errors. The monitoring model is optimized and trained through a large amount of data, thereby improving recognition ability and accuracy. Real-time monitoring and rapid response can be achieved through the monitoring model. Compared with traditional periodic monitoring, this method can detect pollution problems earlier and take timely measures.

[0090] In this embodiment, a computer device is provided, such as Figure 5 As shown, it includes a memory 501, a processor 502, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any of the above-mentioned flood season non-point source pollution monitoring methods is implemented.

[0091] Specifically, the computer device may be a computer terminal, a server or a similar computing device.

[0092] In this embodiment, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program for executing any of the above-mentioned methods for monitoring non-point source pollution during flood season.

[0093] Specifically, computer-readable storage media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer-readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable storage media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0094] Based on the same inventive concept, an embodiment of the present invention also provides a flood season non-point source pollution monitoring device, as described in the following embodiments. Since the principle of solving the problem by the flood season non-point source pollution monitoring device is similar to that of the flood season non-point source pollution monitoring method, the implementation of the flood season non-point source pollution monitoring device can refer to the implementation of the flood season non-point source pollution monitoring method, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements predetermined functions. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0095] Figure 6 is a structural block diagram of a flood season non-point source pollution monitoring device according to an embodiment of the present invention. Figure 6 As shown, it includes: an acquisition module 601, a marking module 602, a feature extraction module 603, a graph generation module 604, a training module 605 and a recognition module 606. The structure is described below.

[0096] An acquisition module 601 is used to obtain a plurality of water quality spectral data of a plurality of water body basin monitoring points at consecutive time points including the flood season, wherein the water quality spectral data is the absorbance of the water sample within a preset wavelength range;

[0097] A marking module 602, for marking the pollution time during the flood season and the type of non-point source pollution for the plurality of water quality spectrum data;

[0098] A feature extraction module 603 is used to extract water quality spectrum features from the marked water quality spectrum data respectively;

[0099] A spectrum generating module 604, configured to generate a three-dimensional water quality spectrum characteristic spectrum corresponding to each of the water quality spectrum data based on the water quality spectrum characteristics;

[0100] A training module 605 is used to train a monitoring model using the plurality of three-dimensional water quality spectral characteristic maps as training data;

[0101] The identification module 606 is used to monitor and identify the pollution time during the flood season and the type of non-point source pollution in the monitored water basin based on the trained monitoring model.

[0102] In one embodiment, the feature extraction module 603 is further used to:

[0103] Performing data preprocessing on the marked water quality spectrum data, wherein the data preprocessing includes wavelet decomposition to remove noise and data normalization processing;

[0104] Water quality spectral features are extracted respectively based on the plurality of water quality spectral data after data preprocessing.

[0105] In one embodiment, the feature extraction module 603 is further used to:

[0106] Extracting the peak position of absorbance, the trough position of absorbance and the trend characteristics of absorbance changing with wavelength for each of the marked water quality spectrum data;

[0107] The peak position of the absorbance, the trough position of the absorbance, and the trend characteristics of the absorbance changing with the wavelength of each water quality spectrum data constitute the water quality spectrum characteristics of each water quality spectrum data.

[0108] In one embodiment, the graph generation module 604 is further used to:

[0109] Performing Fourier transformation on the water quality spectrum characteristics to obtain a basic wave corresponding to each of the water quality spectrum data, an absorbance vector change relative to the basic wave, and a peak position of the absorbance;

[0110] The basic wave is represented by a first coordinate, the vector change relative to the basic waveband is represented by a second coordinate, and the peak position of the absorbance is represented by a third coordinate, wherein the first coordinate, the second coordinate and the third coordinate are located in a planar three-dimensional map, the first coordinate is perpendicular to the second coordinate, and the third coordinate is perpendicular to the second coordinate;

[0111] A three-dimensional water quality spectrum characteristic spectrum corresponding to each of the water quality spectrum data is formed through the planar three-dimensional spectrum.

[0112] In one embodiment, the identification module 606 is further configured to:

[0113] The instantaneous water sample spectrum data of the water body to be monitored is obtained, and the instantaneous water sample spectrum data is identified using the trained monitoring model. If the instantaneous water sample spectrum data is identified as non-point source pollution data during the flood season, the pollution time during the flood season and the type of non-point source pollution are output.

[0114] In one embodiment, the identification module 606 is further configured to:

[0115] Acquire a three-dimensional water quality spectral characteristic spectrum to be monitored corresponding to the instantaneous water sample spectrum data;

[0116] Using the trained monitoring model to perform image comparison and high-dimensional data feature comparison on the three-dimensional water quality spectral feature map to be monitored;

[0117] Perform comprehensive analysis on the image comparison results and high-dimensional data feature comparison results to obtain comprehensive analysis results;

[0118] Based on the comprehensive analysis result, it is identified whether the instantaneous water sample map data belongs to flood season non-point source pollution data.

[0119] In one embodiment, the apparatus further comprises:

[0120] The correction module is used to correct the trained monitoring model according to environmental changes and the drift of the water quality spectrum data acquisition equipment.

[0121] The embodiment of the present invention achieves the following technical effects: obtaining multiple water quality spectral data of multiple water body basin monitoring points including continuous time points during the flood season, the water quality spectral data being the absorbance of the water sample within a preset wavelength range; marking the pollution time during the flood season and the type of non-point source pollution for the multiple water quality spectral data; extracting water quality spectral features from the marked multiple water quality spectral data respectively; generating a three-dimensional water quality spectral feature map corresponding to each water quality spectral data based on the water quality spectral features; using the multiple three-dimensional water quality spectral feature maps and corresponding labels as training data to train a monitoring model, wherein the labels include the marked pollution time during the flood season and the type of non-point source pollution; and monitoring and identifying the pollution time during the flood season and the type of non-point source pollution of the water body basin to be monitored based on the trained monitoring model. The non-point source pollution monitoring method of the present application captures comprehensive spectral characteristics within multiple wavelength ranges. Compared with traditional single-wavelength monitoring, this method provides richer data dimensions, which helps to improve the accuracy of pollutant identification and reduce judgment errors. The monitoring model is optimized and trained through a large amount of data, thereby improving the recognition ability and accuracy. Real-time monitoring and rapid response can be achieved through the monitoring model. Compared with traditional periodic monitoring, this method can detect pollution problems earlier and take timely measures.

[0122] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present invention are not limited to any specific combination of hardware and software.

[0123] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for monitoring non-point source pollution during flood season, characterized in that: include: Obtaining a plurality of water quality spectral data of a plurality of water body basin monitoring points at consecutive time points including the flood season, wherein the water quality spectral data is the absorbance of the water sample within a preset wavelength range; Marking the pollution time during the flood season and the type of non-point source pollution for the plurality of water quality spectral data; Extracting water quality spectrum features from the marked water quality spectrum data respectively, including: extracting the peak position of absorbance, the trough position of absorbance and the trend feature of absorbance changing with wavelength from each of the marked water quality spectrum data; the peak position of absorbance, the trough position of absorbance and the trend feature of absorbance changing with wavelength of each of the water quality spectrum data constitute the water quality spectrum feature of each of the water quality spectrum data; Generating a three-dimensional water quality spectral characteristic spectrum corresponding to each of the water quality spectral data based on the water quality spectral characteristics, including: performing Fourier transform on the water quality spectral characteristics to obtain a basic wave corresponding to each of the water quality spectral data, a vector change of absorbance relative to the basic wave, and a peak position of absorbance; representing the basic wave by a first coordinate, representing the vector change relative to the basic waveband by a second coordinate, and representing the peak position of absorbance by a third coordinate, wherein the first coordinate, the second coordinate, and the third coordinate are located in a planar three-dimensional spectrum, the first coordinate is perpendicular to the second coordinate, and the third coordinate is perpendicular to the second coordinate; forming a three-dimensional water quality spectral characteristic spectrum corresponding to each of the water quality spectral data through the planar three-dimensional spectrum; The monitoring model is trained using the plurality of three-dimensional water quality spectral characteristic maps and corresponding labels as training data, wherein the labels include marked flood season pollution moments and non-point source pollution types; Based on the trained monitoring model, the pollution time during the flood season and the type of non-point source pollution in the monitored water basin are monitored and identified.

2. The method for monitoring non-point source pollution during flood season according to claim 1, characterized in that: The step of extracting water quality spectrum features from the labeled water quality spectrum data respectively includes: Performing data preprocessing on the marked water quality spectrum data, wherein the data preprocessing includes wavelet decomposition to remove noise and data normalization processing; Water quality spectral features are respectively extracted based on the plurality of water quality spectral data after data preprocessing.

3. The method for monitoring non-point source pollution during flood season according to claim 1, characterized in that: The monitoring and identification of the pollution time during the flood season and the type of non-point source pollution in the water basin to be monitored based on the trained monitoring model includes: The instantaneous water sample spectrum data of the water body to be monitored is obtained, and the instantaneous water sample spectrum data is identified using the trained monitoring model. If the instantaneous water sample spectrum data is identified as non-point source pollution data during the flood season, the pollution time during the flood season and the type of non-point source pollution are output.

4. The method for monitoring non-point source pollution during flood season according to claim 3, characterized in that: The step of using the trained monitoring model to identify the instantaneous water sample spectrum data includes: Acquire a three-dimensional water quality spectral characteristic spectrum to be monitored corresponding to the instantaneous water sample spectrum data; Using the trained monitoring model to perform image comparison and high-dimensional data feature comparison on the three-dimensional water quality spectral feature map to be monitored; Perform comprehensive analysis on the image comparison results and high-dimensional data feature comparison results to obtain comprehensive analysis results; Based on the comprehensive analysis result, it is identified whether the instantaneous water sample map data belongs to flood season non-point source pollution data.

5. The method for monitoring non-point source pollution during flood season according to claim 1, characterized in that: The method further comprises: The trained monitoring model is calibrated according to environmental changes and the drift of water quality spectrum data acquisition equipment.

6. A flood season non-point source pollution monitoring device, characterized in that: include: An acquisition module, used to obtain a plurality of water quality spectral data of a plurality of water body basin monitoring points at consecutive time points including the flood season, wherein the water quality spectral data is the absorbance of the water sample within a preset wavelength range; A marking module, used for marking the pollution time during the flood season and the type of non-point source pollution for the plurality of water quality spectral data; A feature extraction module is used to extract water quality spectrum features from the marked water quality spectrum data, including: extracting the peak position of absorbance, the trough position of absorbance and the trend feature of absorbance changing with wavelength for each of the marked water quality spectrum data; the peak position of absorbance, the trough position of absorbance and the trend feature of absorbance changing with wavelength of each of the water quality spectrum data constitute the water quality spectrum feature of each of the water quality spectrum data; A spectrum generation module, for generating a three-dimensional water quality spectrum characteristic spectrum corresponding to each of the water quality spectrum data based on the water quality spectrum characteristics, comprising: performing Fourier transform on the water quality spectrum characteristics to obtain a basic wave corresponding to each of the water quality spectrum data, a vector change of absorbance relative to the basic wave, and a peak position of absorbance; representing the basic wave by a first coordinate, representing the vector change relative to the basic wave band by a second coordinate, and representing the peak position of absorbance by a third coordinate, wherein the first coordinate, the second coordinate, and the third coordinate are located in a planar three-dimensional spectrum, the first coordinate is perpendicular to the second coordinate, and the third coordinate is perpendicular to the second coordinate; forming a three-dimensional water quality spectrum characteristic spectrum corresponding to each of the water quality spectrum data through the planar three-dimensional spectrum; A training module, used to train a monitoring model using a plurality of the three-dimensional water quality spectral characteristic maps and corresponding labels as training data, wherein the labels include marked flood season pollution moments and non-point source pollution types; The identification module is used to monitor and identify the flood season pollution time and the non-point source pollution type of the monitored water basin based on the trained monitoring model.

7. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for monitoring non-point source pollution during flood season according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the flood season non-point source pollution monitoring method according to any one of claims 1 to 5.

Images