A water quality detection method, device and system for marine water quality background fluctuations

Abstract

The invention discloses a water quality detection method, equipment and system for marine water quality background fluctuations, and relates to the technical field of water quality detection, specifically comprising determining an area with abnormal water quality through water quality parameters; in the area with abnormal water quality, obtaining abnormal detection data by performing noise processing and baseline drift processing on ultraviolet spectrum data; obtaining low-dimensional abnormal detection data by performing dimensionality reduction processing on the abnormal detection data; classifying the low-dimensional abnormal detection data through a multi-classification support vector machine model, and outputting the main pollutant components of the water quality; calculating the concentration of the main pollutant components through refractive index and ultraviolet spectrum data. The invention is suitable for non-stationary and nonlinear water quality data, can accurately define the main pollutant components and their concentrations contained in the water quality, and can solve the water quality detection problem under the condition of marine water quality background fluctuations.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality detection, and more particularly to a water quality detection method, equipment and system for marine water quality background fluctuations. Background Art

[0002] The detection of marine water quality is generally carried out by ultraviolet-visible spectroscopy. The changes in the ultraviolet spectrum are used to determine whether the water contains pollutants, and the concentration of pollutants is calculated according to the Lambert-Beer Law. However, the establishment of the Lambert-Beer Law often requires a series of strict conditions, which are usually not easy to meet in actual measurements. Therefore, it is currently impossible to accurately define the main component types and concentrations of pollutants in water through ultraviolet spectroscopy.

[0003] For example, a Chinese patent with authorization announcement number CN109270016B discloses a multifunctional underwater water quality automatic monitoring system and monitoring method based on a clustering algorithm. The invention detects a variety of water quality parameters by ultraviolet-visible spectroscopy, and clusters areas with similar water quality parameters together through cluster analysis to form multiple sample clusters, and judges the water pollution situation by distinguishing between different sample clusters; however, the invention does not further determine the type and content of the material components that cause water pollution.

[0004] Therefore, in view of the deficiencies in the prior art, the present invention provides a water quality detection method, device and system for background fluctuations in marine water quality. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a water quality detection method, equipment and system for background fluctuations in marine water quality. The present invention determines the area where water quality anomalies exist through water quality parameter data, uses a multi-classification support vector machine model to determine the main pollutant components of the water quality in the area, and further uses the collected data to calculate the concentration of each main pollutant component, thereby achieving the purpose of detecting water quality data under conditions of background fluctuations in marine water quality.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for detecting water quality against background fluctuations of marine water quality, comprising:

[0008] Setting a plurality of sampling points in an area to be detected, wherein the area to be detected is located at sea;

[0009] Collecting water quality data of each sampling point within a fixed time, wherein the water quality data includes: refractive index, water quality parameter data and ultraviolet spectrum data;

[0010] In the same sampling period, taking the average of each water quality data of each sampling point to obtain the water quality data of the area to be detected within the fixed time;

[0011] Determine an area where the value of the water quality parameter data falls outside a preset parameter range as an area where water quality abnormality exists;

[0012] In the area where the water quality is abnormal, the ultraviolet spectrum data is subjected to noise processing and baseline drift processing to obtain abnormal detection data, wherein the noise processing includes single-point noise processing and smoothing filtering;

[0013] Obtaining low-dimensional anomaly detection data by performing dimensionality reduction processing on the anomaly detection data;

[0014] Classifying the low-dimensional anomaly detection data through a multi-classification support vector machine model and outputting the main pollutant components of the water quality;

[0015] The concentration of the main pollutant component is calculated by the refractive index and the ultraviolet spectrum data.

[0016] As a further improvement of the present invention, the number of the sampling points in the area to be detected is determined according to the size of the area to be detected, the sampling period of each sampling point is 30 seconds, and the fixed time is 30 days.

[0017] As a further improvement of the present invention, the ultraviolet spectrum data is the absorbance corresponding to different wavelengths, and the wavelength range is 200nm-400nm.

[0018] As a further improvement of the present invention, the step of performing dimensionality reduction processing on the anomaly detection data to obtain low-dimensional anomaly detection data includes:

[0019] Represent high-dimensional anomaly detection data as ,in is the number of sampling points, and In low dimensional The corresponding projection on the dimension is expressed as ;

[0020] right Each column vector in , using Euclidean distance as the metric, calculate and The closest vector, using express of A set of adjacent vectors;

[0021] right Each column vector in , calculate the local covariance matrix , and find the corresponding weight coefficient vector ,in for dimensional all-1 vector;

[0022] By weight coefficient vector Generate weight coefficient matrix , and calculate the matrix ,in ;

[0023] Calculating the Matrix the smallest eigenvalues ​​and calculate The eigenvalues ​​corresponding to the eigenvectors ,in is the number of eigenvalues ​​that are 0;

[0024] The first feature vectors to The feature vectors are combined into a matrix to obtain the low-dimensional anomaly detection data.

[0025] As a further improvement of the present invention, the low-dimensional anomaly detection data is classified by a multi-classification support vector machine model, and the main pollutant components of the water quality are output, including:

[0026] Randomly dividing the row vectors in the low-dimensional anomaly detection data into two groups, the group containing fewer vectors is used as training data, and the group containing more vectors is used as test data;

[0027] Using the training data to train the multi-classification support vector machine model to obtain a trained multi-classification support vector machine model;

[0028] Substitute the test data into the trained multi-classification support vector machine model, output the number of votes for each pollutant component in the water quality, and determine the pollutant component with a number of votes greater than a fixed threshold as the main pollutant component.

[0029] As a further improvement of the present invention, the multi-classification support vector machine model is trained using the training data to obtain a trained multi-classification support vector machine model, including:

[0030] Construct the multi-classification support vector machine model, the multi-classification support vector machine model is composed of The combination of two classifiers and voting mechanism is obtained. ,in is the number of the pollutant components;

[0031] Penalty factor for the multi-classification support vector machine model and the kernel parameters of the kernel function Optimize

[0032] Substitute the optimized parameters into the multi-classification support vector machine model, and use the training data to complete the training of the multi-classification support vector machine to obtain the trained multi-classification support vector machine model.

[0033] As a further improvement of the present invention, the penalty factor for the multi-classification support vector machine model is and the kernel parameters of the kernel function Optimizations include:

[0034] The training data is randomly divided into Each of them is used as a validation set, and the rest are combined into a training set;

[0035] The penalty factor The initial value range of , the kernel parameters The initial space is set to , generate parameter groups with a value interval of 0.25 ;

[0036] For the same set of parameters A total of Sub-classification accuracy test;

[0037] Will The average value of the test results is taken as the parameter of this group. The final performance evaluation index is selected, and the group of parameters with the highest average value is selected as the optimal parameter group.

[0038] As a further improvement of the present invention, the calculation of the concentration of the main pollutant component by the refractive index and the ultraviolet spectrum data includes:

[0039] In the ultraviolet spectrum data, determining the absorbance corresponding to each of the main pollutant components;

[0040] According to the Lambert-Beer law, the deviation concentration of each of the main pollutant components is obtained;

[0041] Calculate the ratio of the deviation concentrations of every two of the main pollutant components to obtain the concentration ratio of each of the main pollutant components;

[0042] Calculate the total concentration of all pollutants in the area to be detected based on the relationship between the refractive index and the water quality concentration;

[0043] The concentration of each of the main pollutant components is calculated based on the concentration ratio and the total concentration.

[0044] In a second aspect, the present invention provides a water quality detection device for marine water quality background fluctuations, which is applied to the above-mentioned water quality detection method for marine water quality background fluctuations, and is characterized by comprising:

[0045] A multi-parameter water quality sensor is used to measure water quality parameters in the area to be detected, wherein the water quality parameters include temperature, turbidity and pH value;

[0046] The ultraviolet spectrum sensor is used to irradiate the water flow in the detection area with ultraviolet light, and receive the ultraviolet light beam passing through the water flow to form an ultraviolet spectrum and obtain ultraviolet spectrum data;

[0047] A liquid refractive index sensor, used to measure the refractive index of seawater in the area to be detected;

[0048] The host computer includes a receiving module, a positioning module, a feature extraction module and a classification module;

[0049] Receiving module: used to receive data collected by the multi-parameter water quality sensor, the ultraviolet spectrum sensor and the liquid refractive index sensor;

[0050] Positioning module: used to determine the area where the value of the water quality parameter data falls outside the preset parameter range as an area with abnormal water quality;

[0051] Preprocessing module: used for performing noise processing and baseline drift processing on the ultraviolet spectrum data in the area where water quality is abnormal to obtain abnormal detection data, wherein the noise processing includes single point noise processing and smoothing filtering;

[0052] Feature extraction module: obtains low-dimensional anomaly detection data by performing dimensionality reduction processing on the anomaly detection data;

[0053] Classification module: classifies the low-dimensional anomaly detection data through a multi-classification support vector machine model, and outputs the main pollutant components of the water quality;

[0054] Calculation module: calculates the concentration of the main pollutant components through the refractive index and the ultraviolet spectrum data.

[0055] In a third aspect, the present invention provides a water quality detection system for background fluctuations in marine water quality, comprising a server and the above-mentioned detection equipment, wherein the server is communicatively connected to the detection equipment, and the server is used to output information on water quality adjustment methods based on the main pollutant components of the water quality.

[0056] Since the background fluctuation of marine water quality will cause the collected water quality data to be non-stationary and nonlinear data, and the existing technology cannot accurately define the pollutant components and their concentrations in seawater based on such data, the present invention solves the current problem of being unable to accurately identify pollutant components by reducing the dimension of nonlinear data, and calculates the accurate concentration of pollutants through the proportional relationship of pollutant deviation concentrations, thereby solving the problems existing in the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic diagram of the method steps of the present invention;

[0058] Figure 2 A schematic diagram of a process for determining the main pollutant components in the present invention;

[0059] Figure 3 It is a schematic diagram of the classification results of the present invention;

[0060] Figure 4 It is a schematic diagram of the device of the present invention;

[0061] Figure 5 It is a schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0062] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations of the technical solution of the present invention.

[0063] The term "and / or" in the following text is only a description of the association relationship between the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0064] In order to detect water quality data at sea, the present invention provides a water quality detection device, method and system for background fluctuations of water quality at sea.

[0065] like Figure 1 As shown, this embodiment provides a water quality anomaly detection method for water quality background fluctuations, including:

[0066] Setting a plurality of sampling points in an area to be detected, wherein the area to be detected is located at sea;

[0067] Collect water quality data at each sampling point within a fixed time. The water quality data includes: refractive index, water quality parameter data and ultraviolet spectrum data;

[0068] In the same sampling period, the average of each water quality data of each sampling point is taken to obtain the water quality data of the area to be tested within a fixed time;

[0069] Determine an area where the value of the water quality parameter data falls outside the preset parameter range as an area with abnormal water quality;

[0070] In the area where water quality is abnormal, the UV spectrum data is processed for noise and baseline drift to obtain abnormal detection data. The noise processing includes single-point noise processing and smoothing filtering.

[0071] By performing dimensionality reduction processing on the anomaly detection data, low-dimensional anomaly detection data is obtained;

[0072] The low-dimensional anomaly detection data is classified through a multi-classification support vector machine model, and the main pollutant components of water quality are output;

[0073] The concentrations of major pollutant components are calculated from refractive index and UV spectral data.

[0074] This embodiment determines the area with abnormal water quality through water quality parameter data, uses a multi-classification support vector machine model to determine the main pollutant components of the water quality in the area, and calculates the concentration of the main pollutant components based on the refractive index and ultraviolet spectrum data, thereby achieving the purpose of detecting water quality data under the background fluctuation conditions of marine water quality.

[0075] Among them, the preset parameter range can be determined based on the daily water quality data of the area to be tested.

[0076] In addition, the ultraviolet spectrum data is the absorbance corresponding to different wavelengths, and the wavelength range is 200nm-400nm. This is because the ultraviolet visible wavelength range is 200nm-800nm, and the ultraviolet spectrum covers a band of mainly 100nm-400nm. The range of 100nm-200nm is the far ultraviolet region, but the light waves in this region are easily absorbed by oxidation. The range of 200nm-400nm is the near ultraviolet spectrum region. The absorption spectrum in the near ultraviolet spectrum region has good sensitivity and strong selectivity. Therefore, this embodiment selects the absorption spectrum in the near ultraviolet spectrum region for analysis.

[0077] Furthermore, the number of sampling points in the area to be detected can be determined according to the size of the area to be detected, and the sampling period can be determined according to the specific equipment. For example, the UV-visible spectrum sensor can use the spectro::lyser immersion probe developed by Austrian company S::CAN. At this time, the time to scan a spectrum is 20 seconds, the interval between two spectra is 10 seconds, and the scanning period is 30 seconds. Since this embodiment is aimed at the background fluctuation of marine water quality, the fixed collection time should be relatively long, which can be determined as one month, one quarter or one year according to the detection needs.

[0078] In addition, since the spectrum collection will inevitably be affected by the scattering effect of suspended particles in the water, the scattering effect of light and random noise, which will reduce the resolution of the spectrum and even cause spectral deviation, noise processing and baseline drift processing are required.

[0079] Noise processing generally includes single-point noise processing and smoothing filtering. Single-point noise processing can adopt mean centralization and attribute centralization methods, and smoothing filtering can adopt mean method, median method, polynomial fitting method and Savitzky-Golay convolution smoothing method.

[0080] After the ultraviolet spectrum data is processed for noise and baseline drift, anomaly detection data can be obtained, but the data dimension is large, including the absorbance of hundreds of wavelengths. The absorbance of different wavelengths may have strong correlation or even redundancy, so it is necessary to reduce the dimension of the anomaly detection data and extract feature information. In the traditional field of water quality detection, principal component analysis or continuous projection algorithm is usually used for data dimension reduction and feature extraction, but the above methods are only applicable to linear data and are not applicable to the present invention.

[0081] This is because existing technology cannot accurately define the main pollutant components in seawater. An important reason is that fluctuations in the background water quality at sea can cause non-stationary and nonlinear changes in water quality data. Fluctuations in the background water quality at sea refer to changes in the background water quality caused by factors such as human activities or climate change. Such changes will cause the background water quality to show a certain volatility, and this volatility will cause non-stationary and nonlinear changes in water quality data.

[0082] Take a wavelength point as an example. In the absence of water pollution, if there is no influence of water quality background fluctuation, there should be a stable linear relationship between the absorbance corresponding to the wavelength point collected over a period of time and time. If there is the influence of water quality background fluctuation, the absorbance corresponding to the wavelength point collected over a period of time will have relatively large interference and fluctuation, that is, there is a non-stationary nonlinear relationship with time. If it is further affected by pollutants in the water, the absorbance will fluctuate more. At this time, the use of dimensionality reduction methods suitable for linear data will affect the accuracy of the data analysis results.

[0083] Therefore, this embodiment provides a dimensionality reduction method suitable for nonlinear data, such as Figure 2 As shown, by reducing the dimension of nonlinear data, low-dimensional anomaly detection data can be obtained.

[0084] This method assumes that nonlinear data is linear in a smaller range, that is, a sample data can be linearly represented by several samples around it. For example, in high-dimensional samples, You can use samples around it , , The linear representation is: ,in Represents the weight coefficient. After dimensionality reduction, we hope Projection in low-dimensional space and Projection in low-dimensional space Keeping the same linear relationship: , it can be seen that the weight coefficient is unchanged. Therefore, when the high-dimensional sample data is known, the weight coefficient can be calculated, and in the low-dimensional sample space, the low-dimensional sample data can be reconstructed according to the calculated weight coefficient.

[0085] The method specifically comprises the following steps:

[0086] Represent high-dimensional anomaly detection data as ,in is the number of sampling points, and In low dimensional The corresponding projection on the dimension is expressed as ;

[0087] right Each column vector in , using Euclidean distance as the metric, calculate and The closest vector, using express of A set of adjacent vectors;

[0088] Each column vector in , calculate the local covariance matrix , and find the corresponding weight coefficient vector ,in for dimensional all-1 vector;

[0089] By weight coefficient vector Generate weight coefficient matrix , and calculate the matrix ,in ;

[0090] Calculating the Matrix the smallest eigenvalues ​​and calculate The eigenvalues ​​corresponding to the eigenvectors ,in is the number of eigenvalues ​​that are 0;

[0091] The first feature vectors to The feature vectors form a matrix to obtain low-dimensional anomaly detection data.

[0092] In order to further understand the above steps, the loss function is introduced below for explanation. When mapping high-dimensional data to low-dimensional space, the loss function value needs to be minimized, that is, it satisfies:

[0093] ;

[0094] in represents the loss function, and Respectively Middle and column vector, Equal to in numerical terms and The weight coefficient for situation, , matrixing the loss function yields:

[0095] ;

[0096] in is the trace function, for The transpose of , and then combined with the constraints:

[0097] ;

[0098] The problem to be solved can be transformed into a standard eigendecomposition problem, that is, to find the matrix Smallest The eigenvalues ​​corresponding to However, since the eigenvalue of 0 cannot reflect the data characteristics, the first feature vectors to The final result is a matrix composed of eigenvectors, where is the number of eigenvalues ​​that are 0.

[0099] According to the above steps, this embodiment calculates the value of each absorbance in the neighborhood. The linear relationship between the nearest neighbors is obtained to obtain the linear relationship weight coefficient. The data is reconstructed in low dimension through the weight coefficient, thus realizing the dimensionality reduction of nonlinear data.

[0100] Furthermore, when selecting a classification model, this embodiment selects a multi-classification support vector machine model suitable for non-stationary nonlinear data for analysis. This model maps the data to a high-dimensional space by introducing a kernel function, so that nonlinear data that is originally difficult to separate in a low dimension becomes separable in the high-dimensional space, thereby achieving a classification effect.

[0101] Specifically, Figure 2 As shown, this embodiment provides a classification method using a multi-classification support vector machine model, and the specific steps include:

[0102] The row vectors in the low-dimensional anomaly detection data are randomly divided into two groups, the group containing fewer vectors is the training data, and the group containing more vectors is the test data;

[0103] The multi-classification support vector machine model is trained using the training data to obtain a trained multi-classification support vector machine model;

[0104] The test data is substituted into the trained multi-classification support vector machine model, and the number of votes for each pollutant component in the water quality is output. The pollutant components with votes greater than a fixed threshold are determined as the main pollutant components.

[0105] Furthermore, this embodiment also provides a specific step of model training, including:

[0106] Construct a multi-classification support vector machine model. The multi-classification support vector machine model consists of The combination of two classifiers and voting mechanism is obtained. ,in is the number of pollutant components;

[0107] Penalty factor for multi-classification support vector machine model and the kernel parameters of the kernel function Optimize

[0108] The optimized parameters are substituted into the multi-classification support vector machine model, and the training data is used to complete the training of the multi-classification support vector machine to obtain the trained multi-classification support vector machine model.

[0109] The pollutant composition and fixed threshold value in the water quality can be determined according to the daily water quality conditions in the area to be tested.

[0110] Furthermore, this embodiment also provides a specific step of parameter optimization, including:

[0111] The training data is randomly divided into Each of them is used as a validation set, and the rest are combined into a training set;

[0112] Penalty Factor The initial value range of , kernel parameters The initial space is set to , generate parameter groups with a value interval of 0.25 ;

[0113] For the same set of parameters A total of Sub-classification accuracy test;

[0114] Will The average value of the test results is taken as the parameter of this group. The final performance evaluation index is selected, and the group of parameters with the highest average value is selected as the optimal parameter group.

[0115] In order to further illustrate the accuracy of the classification of this embodiment, Figure 3 Two different classification results are shown. Figure 3 In ( ) is the result of directly using anomaly detection data for classification. Figure 3 Case (b) in the figure is the result of classification using low-dimensional anomaly detection data. , representing three different pollutant components.

[0116] All the correct classifications in the figure are concentrated on the diagonal. The darker the color, the higher the similarity between the two corresponding pollutants. If the diagonal similarity is very high, it means that the classification result of this method is better.

[0117] It can be seen that the classification effect using low-dimensional anomaly detection data is better, which shows that the dimensionality reduction method used in this embodiment can effectively reduce the impact of data redundancy and further accurately identify the components of major pollutants in seawater.

[0118] Furthermore, the main reason why the concentration of the main pollutant components cannot be accurately defined in the prior art is that when the pollutant concentration is high, or due to changes in conditions such as temperature, there is a certain nonlinearity between the absorbance of the spectral data and the pollutant concentration. Therefore, at this time, there is a large error when the concentration of each pollutant component is reversely calculated using the absorbance through the Lambert-Beer law. Although the concentration and absorbance are not linearly related, the nonlinear relationship between the concentration and absorbance corresponding to different types of pollutant components is consistent. Due to the consistency of the nonlinear relationship, the ratio relationship of the concentration of different types of pollutant components is relatively accurate. Therefore, based on the ratio relationship between the concentrations, combined with the linear relationship between the refractive index and the total concentration of all pollutant components in the prior art, the concentration of each main pollutant component can be calculated.

[0119] The linear relationship between the refractive index and the total concentration may be affected by intermolecular interactions. However, since the conditions defined in the embodiments of the present application are scenarios of marine water quality fluctuations, seawater can be equivalent to an infinitely large sea area, and therefore intermolecular interactions can be ignored in seawater.

[0120] Based on the above analysis, this embodiment further provides a method for calculating the concentration of main pollutant components through refractive index and ultraviolet spectrum data, and the specific steps include:

[0121] In the UV spectrum data, determine the absorbance corresponding to each major pollutant component;

[0122] According to the Lambert-Beer law, the deviation concentration of each major pollutant component is obtained;

[0123] Calculate the ratio of the deviation concentrations of every two main pollutant components to obtain the concentration ratio of each main pollutant component;

[0124] According to the relationship between refractive index and water quality concentration, the total concentration of all pollutant components in the area to be tested is calculated;

[0125] Calculate the concentration of each major pollutant component based on the concentration ratio and total concentration.

[0126] Since the concentration calculated directly using the Lambert-Beer law in this embodiment has a large error, in this embodiment, the concentration calculated according to the Lambert-Beer law is defined as the deviation concentration.

[0127] In this embodiment, the total concentration of all pollutant components is defined to be equal to the sum of the concentrations of each major pollutant component, that is, the influence of the concentrations of non-major pollutant components is ignored.

[0128] For example, assuming that there are three main pollutant components, according to the Lambert-Beer law, the deviation concentration of each main pollutant component is The total concentration is . Then calculate the ratio of the deviation concentration of each two main pollutant components to obtain ,but , based on the relationship between the refractive index and the water quality concentration, after calculating the total concentration, the concentration of each major pollutant component can be obtained.

[0129] From the above analysis, it can be seen that the water quality detection for marine water quality background fluctuations provided in this embodiment can accurately identify the components of major pollutants by processing nonlinear data, and accurately calculate the concentration of each major pollutant component based on the relationship between absorbance, concentration and refractive index.

[0130] The present application further provides a water quality detection device for marine water quality background fluctuations, which is applied to the above-mentioned water quality detection method for marine water quality background fluctuations, such as Figure 4 As shown, including:

[0131] Multi-parameter water quality sensor, used to measure water quality parameters in the area to be tested, including temperature, turbidity and pH value;

[0132] The ultraviolet spectrum sensor is used to irradiate the water flow in the detection area with ultraviolet light, and receive the ultraviolet light beam passing through the water flow to form an ultraviolet spectrum and obtain ultraviolet spectrum data;

[0133] Liquid refractive index sensor, used to measure the refractive index of seawater in the area to be detected;

[0134] The host computer includes a receiving module, a positioning module, a feature extraction module and a classification module;

[0135] Receiving module: used to receive data collected by multi-parameter water quality sensors, ultraviolet spectrum sensors and liquid refractive index sensors;

[0136] Positioning module: used to determine the area where the value of water quality parameter data falls outside the preset parameter range as an area with abnormal water quality;

[0137] Preprocessing module: used to perform noise processing and baseline drift processing on ultraviolet spectrum data in areas with abnormal water quality to obtain abnormal detection data. Noise processing includes single-point noise processing and smoothing filtering;

[0138] Feature extraction module: obtains low-dimensional anomaly detection data by performing dimensionality reduction processing on the anomaly detection data;

[0139] Classification module: classifies low-dimensional anomaly detection data through a multi-classification support vector machine model and outputs the main pollutant components of water quality;

[0140] Calculation module: Calculate the concentration of main pollutant components through refractive index and ultraviolet spectrum data.

[0141] Among them, the ultraviolet spectrum sensor is encapsulated in a metal sealed waterproof shell, the emitting optical element emits ultraviolet light, the receiving optical element receives the ultraviolet light and forms an ultraviolet absorption spectrum, and the detected water flows through the slit.

[0142] It can be seen from the above embodiments that the device provided in this embodiment collects data through a variety of sensor devices and performs a series of steps of data processing and analysis by a host computer, which can accurately identify the composition and concentration of major pollutants in seawater and realize water quality detection under the condition of marine water quality background fluctuations.

[0143] The present application further provides a water quality detection system for marine water quality background fluctuations, such as Figure 5 As shown, it includes a server and the above-mentioned detection equipment, the server is communicatively connected with the detection equipment, and the server is used to output information on water quality adjustment methods according to the main pollutant components of the water quality.

[0144] Furthermore, information on the main pollutant components and water quality adjustment methods can be sent to a browser or APP, allowing testers to quickly browse the test information and improve work efficiency.

[0145] The water quality detection method, device and system for marine water quality background fluctuations provided by the present invention can process non-stationary nonlinear data under water quality background fluctuation conditions, thereby improving the accuracy of defining the composition and concentration of major pollutants in seawater.

[0146] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0147] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0148] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0149] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0150] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A water quality detection method for marine water quality background fluctuations, characterized in that: include: Setting a plurality of sampling points in an area to be detected, wherein the area to be detected is located at sea; Collecting water quality data of each sampling point within a fixed time, the water quality data including: refractive index, water quality parameter data and ultraviolet spectrum data, the ultraviolet spectrum data is the absorbance corresponding to different wavelengths, and the wavelength range is 200nm-400nm; In the same sampling period, taking the average of each water quality data of each sampling point to obtain the water quality data of the area to be detected within the fixed time; Determine an area where the value of the water quality parameter data falls outside a preset parameter range as an area where water quality abnormality exists; In the area where the water quality is abnormal, the ultraviolet spectrum data is subjected to noise processing and baseline drift processing to obtain abnormal detection data, wherein the noise processing includes single-point noise processing and smoothing filtering; By performing dimensionality reduction processing on the anomaly detection data, low-dimensional anomaly detection data is obtained, specifically including: expressing the high-dimensional anomaly detection data as ,in is the number of sampling points, and In low dimensional The corresponding projection on the dimension is expressed as ;right Each column vector in , using Euclidean distance as the metric, calculate and The closest vector, using express of A set of adjacent vectors; Each column vector in , calculate the local covariance matrix , , and find the corresponding weight coefficient vector ,in for dimensional all-1 vector; through the weight coefficient vector Generate weight coefficient matrix , and calculate the matrix ,in ; Calculate the matrix the smallest eigenvalues ​​and calculate The eigenvalues ​​corresponding to the eigenvectors ,in is the number of eigenvalues ​​0; feature vectors to The feature vectors are used to form a matrix to obtain the low-dimensional anomaly detection data; The low-dimensional anomaly detection data is classified by a multi-classification support vector machine model, and the main pollutant components of the water quality are output, specifically including: randomly dividing the row vectors in the low-dimensional anomaly detection data into two groups, a group containing fewer vectors as training data, and a group containing more vectors as test data; using the training data to train the multi-classification support vector machine model to obtain a trained multi-classification support vector machine model; substituting the test data into the trained multi-classification support vector machine model, outputting the number of votes for each pollutant component in the water quality, and determining the pollutant component with a number of votes greater than a fixed threshold as the main pollutant component; The concentration of the main pollutant components is calculated by the refractive index and the ultraviolet spectrum data, specifically including: determining the absorbance corresponding to each of the main pollutant components in the ultraviolet spectrum data; obtaining the deviation concentration of each of the main pollutant components according to the Lambert-Beer law; calculating the ratio of the deviation concentrations of every two main pollutant components to obtain the concentration ratio of each main pollutant component; calculating the total concentration of all pollutant components in the area to be detected based on the relationship between the refractive index and the water quality concentration; and calculating the concentration of each main pollutant component based on the concentration ratio and the total concentration.

2. A method for detecting water quality against background fluctuations of marine water quality according to claim 1, characterized in that: The number of the sampling points in the area to be detected is determined according to the size of the area to be detected, the sampling period of each sampling point is 30 seconds, and the fixed time is 30 days.

3. A water quality detection method for marine water quality background fluctuation according to claim 1, characterized in that: The method of using the training data to train the multi-classification support vector machine model to obtain a trained multi-classification support vector machine model includes: Construct the multi-classification support vector machine model, the multi-classification support vector machine model is composed of The combination of two classifiers and voting mechanism is obtained. ,in is the number of the pollutant components; Penalty factor for the multi-classification support vector machine model and the kernel parameters of the kernel function Optimize Substitute the optimized parameters into the multi-classification support vector machine model, and use the training data to complete the training of the multi-classification support vector machine to obtain the trained multi-classification support vector machine model.

4. A water quality detection method for marine water quality background fluctuation according to claim 3, characterized in that: The penalty factor for the multi-classification support vector machine model and the kernel parameters of the kernel function Optimizations include: The training data is randomly divided into Each of them is used as a validation set, and the rest are combined into a training set; The penalty factor The initial value range of , the kernel parameters The initial space is set to , generate parameter groups with a value interval of 0.25 ; For the same set of parameters A total of Sub-classification accuracy test; Will The average value of the test results is taken as the parameter of this group. The final performance evaluation index is selected, and the group of parameters with the highest average value is selected as the optimal parameter group.

5. A water quality detection device for marine water quality background fluctuations, applied to a water quality detection method for marine water quality background fluctuations as claimed in any one of claims 1 to 4, characterized in that: include: A multi-parameter water quality sensor is used to measure water quality parameters in the area to be detected, wherein the water quality parameters include temperature, turbidity and pH value; The ultraviolet spectrum sensor is used to irradiate the water flow in the detection area with ultraviolet light, and receive the ultraviolet light beam passing through the water flow to form an ultraviolet spectrum and obtain ultraviolet spectrum data; A liquid refractive index sensor, used to measure the refractive index of seawater in the area to be detected; The host computer includes a receiving module, a positioning module, a feature extraction module and a classification module; Receiving module: used to receive data collected by the multi-parameter water quality sensor, the ultraviolet spectrum sensor and the liquid refractive index sensor; Positioning module: used to determine the area where the value of the water quality parameter data falls outside the preset parameter range as an area with abnormal water quality; Preprocessing module: used for performing noise processing and baseline drift processing on the ultraviolet spectrum data in the area where water quality is abnormal to obtain abnormal detection data, wherein the noise processing includes single point noise processing and smoothing filtering; Feature extraction module: obtains low-dimensional anomaly detection data by performing dimensionality reduction processing on the anomaly detection data; Classification module: classifies the low-dimensional anomaly detection data through a multi-classification support vector machine model, and outputs the main pollutant components of the water quality; Calculation module: calculates the concentration of the main pollutant components through the refractive index and the ultraviolet spectrum data.

6. A water quality detection system for marine water quality background fluctuations, characterized in that: It comprises a server and the detection device as claimed in claim 5, wherein the server is communicatively connected with the detection device, and the server is used to output information on water quality adjustment methods according to the main pollutant components of the water quality.

Images