Fresh water ecosystem health monitoring method and system

Abstract

The invention discloses a fresh water ecosystem health monitoring method and system. The method comprises the following steps: acquiring a historical remote sensing image, and acquiring a physical index; acquiring historical ground measured data, and extracting chemical indexes and biological indexes; establishing a training set and a test set based on each index; a classification machine learning model is selected to construct a sub-model for each index, the training set is used as input for training, and a plurality of water quality evaluation sub-models are obtained; performing accuracy analysis on each water quality evaluation sub-model, calculating relevancy and stability between different models, and obtaining fusion weights of the corresponding water quality evaluation sub-models; combining all the water quality evaluation sub-models based on the fusion weight to construct a comprehensive water quality evaluation model; and performing fresh water ecological water quality evaluation through the comprehensive water quality evaluation model to generate an evaluation report. According to the invention, accurate evaluation of fresh water ecological water quality is realized by combining remote sensing and ground measured data.

Description

Technical Field

[0001] The present invention relates to the field of ecological environment remote sensing technology, and more specifically to a freshwater ecosystem health monitoring method and system. Background Art

[0002] At present, population growth and the improvement of industrialization level have intensified the contradiction between population, resources and environment. Ecological and environmental problems such as water pollution and desertification continue to emerge. Accurate evaluation of the ecological and environmental conditions will help to protect the ecological environment.

[0003] Remote sensing refers to the technology of obtaining information about the earth's surface through non-contact sensors. With the continuous development of satellite technology and sensor technology, the spatial resolution, spectral resolution and temporal resolution of remote sensing data have been greatly improved. Although remote sensing technology has many advantages, it also has certain limitations. Remote sensing data is mainly based on information such as the reflectance spectrum of the water surface, and the measurement of some deep water quality parameters is not accurate enough. Ground-measured data can make up for the shortcomings of remote sensing technology. By sampling at different depths in the water body and using professional water quality analysis instruments, various water quality parameters can be accurately measured. At the same time, ground-measured data can also be used to calibrate and verify remote sensing models.

[0004] Therefore, how to combine remote sensing and ground-based measured data to achieve freshwater ecological water quality assessment is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides a freshwater ecosystem health monitoring method and system, which realizes accurate assessment of freshwater ecological water quality by combining remote sensing and ground-based measured data.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] A method for monitoring the health of a freshwater ecosystem, comprising:

[0008] Obtain historical remote sensing images, construct turbidity detection parameters, and obtain water turbidity based on the historical remote sensing images; at the same time, obtain suspended matter concentration through a pre-trained suspended matter concentration prediction model and the historical remote sensing images; the water turbidity and suspended matter concentration are used as physical indicators; obtain historical ground measured data, extract chemical indicators and biological indicators of the water body; establish training sets and test sets based on the physical indicators, chemical indicators and biological indicators;

[0009] Select a classification machine learning model to construct a sub-model for each of the physical indicators, chemical indicators and biological indicators, use the training set as input for training, and obtain several water quality assessment sub-models;

[0010] Perform accuracy analysis on each of the water quality assessment sub-models, calculate the correlation and stability between different models, and obtain the fusion weight of the corresponding water quality assessment sub-model; based on the fusion weight of the water quality assessment sub-model, combine all water quality assessment sub-models to construct a comprehensive water quality assessment model;

[0011] Acquire the remote sensing images and ground measured data to be tested, conduct freshwater ecological water quality assessment based on the comprehensive water quality assessment model, and generate an evaluation report.

[0012] Preferably, the method of obtaining the turbidity of the water body specifically includes: extracting the quasi-plate measurement values ​​of the historical remote sensing images, the diffuse scattered light from the sky and the radiance off the water to establish a reflectance curve of the sensitive wavelength, and constructing turbidity detection parameters based on the reflectance curve to perform water turbidity detection.

[0013] Preferably, the step of establishing a reflectivity curve of a sensitive wavelength specifically includes:

[0014] ρ p =30%;

[0015] ρ = 0.028;

[0016] Among them, ρ p is the reflectivity of the standard plate, E d (0 + ) represents the total incident irradiance on the water surface, ρ represents the reflectivity of the air-water interface to the sky light, and L p is the measured value of the standard plate, L w is the water-leaving radiance, L sky Sky diffuse light, R rs is the reflectivity curve.

[0017] Preferably, the construction of turbidity detection parameters specifically includes extracting the reflectance of infrared band wavelengths and near-infrared band wavelengths to construct a water surface turbidity index SWTI:

[0018]

[0019] Among them, λ a is the central wavelength of the near-infrared band, λ b is the central wavelength of the infrared band, a1 and b1 are the fitting coefficients.

[0020] Preferably, the construction process of the suspended matter concentration prediction model is:

[0021] Collect particulate matter or sediment samples from water bodies in historical remote sensing images, and conduct matching experiments through reflectance spectrum measurement containers. At the same time, conduct multispectral drone aerial photography and suspended matter sample collection experiments, calculate the reflectivity of the multispectral image of the suspended matter sample in the matching experiment, and use the weighing method to obtain the corresponding concentration of the suspended matter sample;

[0022] According to the reflectivity of the multispectral image of the suspended matter sample in the proportioning experiment and the corresponding concentration of the suspended matter sample, a fitting analysis of the sensitive band is performed, the best sensitive factor to the change of suspended matter concentration is screened, a scatter plot of the sensitive factor and the suspended matter concentration is drawn, and the fitting function with the highest fitting correlation coefficient is determined as the suspended matter concentration prediction model.

[0023] Preferably, the chemical indicators and biological indicators of the water body specifically include: the biological indicators include the water organism content and chlorophyll concentration of the water body; the chemical indicators include the nitrogen content and phosphorus content of the water body.

[0024] Preferably, the obtaining of several water quality assessment sub-models specifically includes: selecting any one of the random forest, gradient boosting tree, and support vector machine classification machine learning models to construct a sub-model, and using each of the physical indicators, chemical indicators, and biological indicators as the input of the sub-model, to obtain several water quality assessment sub-models.

[0025] Preferably, the acquisition of the fusion weight of the corresponding water quality assessment sub-model specifically includes: analyzing the assessment accuracy of each sub-model through a test set as the first indicator of the fusion weight; calculating the mutual information between the training sets generated by different indicators as the second indicator of the fusion weight; calculating the stability of different indicators as the third indicator of the fusion weight; and normalizing the three weights and adding them up to obtain the final fusion weight of each sub-model.

[0026] Preferably, the construction of a comprehensive water quality assessment model specifically includes: determining the importance of the sub-model generated by each indicator in the fusion model by the fusion weight of the water quality assessment sub-model; multiplying the classification result of each sub-model by the weight, and combining them to obtain a comprehensive water quality assessment model.

[0027] A freshwater ecosystem health monitoring system, comprising:

[0028] The data set establishment module obtains historical remote sensing images, constructs turbidity detection parameters, and obtains water turbidity based on the historical remote sensing images; at the same time, the suspended matter concentration is obtained through the pre-trained suspended matter concentration prediction model and the historical remote sensing images; the water turbidity and suspended matter concentration are used as physical indicators; historical ground measured data are obtained to extract chemical indicators and biological indicators of the water body; training sets and test sets are established based on the physical indicators, chemical indicators and biological indicators;

[0029] A sub-model building module selects a classification machine learning model to construct a sub-model for each of the physical indicators, chemical indicators and biological indicators, uses the training set as input for training, and obtains a number of water quality assessment sub-models;

[0030] The water quality assessment sub-model fusion module performs accuracy analysis on each of the water quality assessment sub-models, calculates the correlation and stability between different models, and obtains the fusion weights of the corresponding water quality assessment sub-models; based on the fusion weights of the water quality assessment sub-models, all water quality assessment sub-models are combined to construct a comprehensive water quality assessment model;

[0031] The evaluation report generation module obtains the remote sensing images to be tested and the ground measured data, and generates an evaluation report based on the freshwater ecological water quality based on the comprehensive water quality assessment model.

[0032] Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses a method and system for monitoring the health of freshwater ecosystems, which comprehensively considers physical, chemical and biological indicators to fully reflect the health status of freshwater ecosystems. Physical indicators include water turbidity and suspended solids concentration, which are obtained through historical remote sensing images and specific models, and can reflect the physical state of water bodies from a macroscopic perspective; chemical indicators such as nitrogen content and phosphorus content, biological indicators such as aquatic organism content and chlorophyll concentration, these indicators can deeply reflect the ecological characteristics and water quality of water bodies, use historical remote sensing images and ground-measured data to establish training sets and test sets, train multiple water quality assessment sub-models, and then fuse them to construct a comprehensive water quality assessment model. The entire process is efficient and systematic. It only needs to obtain the remote sensing images to be tested and the ground-measured data to quickly generate an evaluation report based on the comprehensive water quality assessment model, which provides a timely and effective means for monitoring the health of freshwater ecosystems. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0034] Figure 1 A diagram of the method steps provided by the present invention;

[0035] Figure 2 A schematic diagram of the structure provided by the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] The embodiment of the present invention discloses a method for monitoring the health of a freshwater ecosystem. Figure 1 As shown, including:

[0038] Obtain historical remote sensing images, construct turbidity detection parameters, and obtain water turbidity based on historical remote sensing images; at the same time, obtain suspended solids concentration through pre-trained suspended solids concentration prediction model and historical remote sensing images; water turbidity and suspended solids concentration are used as physical indicators; obtain historical ground measured data, extract chemical indicators and biological indicators of water bodies; establish training sets and test sets based on physical indicators, chemical indicators and biological indicators;

[0039] Select a classification machine learning model to construct a sub-model for each physical indicator, chemical indicator, and biological indicator, use the training set as input for training, and obtain several water quality assessment sub-models;

[0040] Conduct accuracy analysis on each water quality assessment sub-model, calculate the correlation and stability between different models, and obtain the fusion weight of the corresponding water quality assessment sub-model; Based on the fusion weight of the water quality assessment sub-model, combine all water quality assessment sub-models to build a comprehensive water quality assessment model;

[0041] Acquire the remote sensing images and ground measured data to be tested, conduct freshwater ecological water quality assessment based on the comprehensive water quality assessment model, and generate an evaluation report.

[0042] In a specific embodiment, based on the surface water environmental quality standard (GB 3838-2002), water turbidity, suspended matter concentration, aquatic organism content, chlorophyll concentration, nitrogen content and phosphorus content are collected to obtain the corresponding water quality measurement standard as training data.

[0043] In a specific embodiment, obtaining water turbidity specifically includes: extracting the quasi-plate measurement values ​​of historical remote sensing images, establishing a reflectance curve of sensitive wavelengths based on the diffuse scattered light from the sky and the radiance off the water, and constructing turbidity detection parameters based on the reflectance curve to perform water turbidity detection.

[0044] Among them, the parameter L pUse remote sensing image processing software to open the image data source file collected by the ground-based hyperspectral imager, select any end band to form the monitoring area image, select the entire white board in the image as the region of interest, use the ROI tool to read the maximum value of the region of interest, form the maximum value set of standard plate measurements in each band, and export it as a text document;

[0045] Parameter L sky Use remote sensing image processing software to open the image data source file collected by the ground-based hyperspectral imager, select any band to form an image of the monitoring area, select the farthest sky in the image as the region of interest, and use the ROI tool to read the average value of the region of interest to form a spectral curve containing each band, and export it as a text document;

[0046] Parameter L w Use remote sensing image processing software to open the image data source file collected by the ground-based hyperspectral imager, select any band to form the monitoring area image, select the water body without white caps, ripples, reflections, and ships from the target waters as the area of ​​interest, and use the ROI tool to read the average value of the area of ​​interest to form a spectral curve containing each band, and export it as a text document;

[0047] In a specific embodiment, establishing a reflectivity curve of a sensitive wavelength specifically includes:

[0048] ρ p =30%;

[0049] ρ = 0.028;

[0050] Among them, ρ p is the reflectivity of the standard plate, E d (0 + ) represents the total incident irradiance on the water surface, ρ represents the reflectivity of the air-water interface to the sky light, and L p is the measured value of the standard plate, L w is the water-leaving radiance, L sky Sky diffuse light, R rs is the reflectivity curve.

[0051] In a specific embodiment, constructing turbidity detection parameters specifically includes extracting infrared band wavelength reflectance and near infrared band wavelength reflectance, and constructing a water surface turbidity index SWTI:

[0052]

[0053] Among them, λ ais the central wavelength of the near-infrared band, λ b is the central wavelength of the infrared band, a1 and b1 are the fitting coefficients.

[0054] In a specific embodiment, the reflectance of a specific wavelength for distinguishing the turbidity of a water body is extracted by using a sensitive wavelength reflectance curve, wherein the central wavelength of the red and near infrared bands is λ b =660nm and λ a =830nm, the corresponding wavelength reflectivity is R rs (λ b ) and R rs (λ a ), the detection parameter for determining water turbidity constructed by the red and near-infrared central wavelength reflectance is the water surface turbidity index SWTI. In order to obtain the fitting coefficient in the formula, in the sample area, λ b =660nm and λ a =The logarithmic ratio of the reflectance value corresponding to the wavelength of 830nm is imported into the Excel table. At the same time, the logarithmic value of the turbidity value of the target water sample set is also imported into the Excel table. Using the tool of the Excel table, linear fitting is performed to obtain a1=3.19 and b1=0.98.

[0055] In a specific embodiment, the construction process of the suspended matter concentration prediction model is as follows:

[0056] Collect particle or sediment samples from water bodies in historical remote sensing images, and conduct matching experiments through reflectance spectrum measurement containers. At the same time, conduct multispectral drone aerial photography and suspended matter sample collection experiments, calculate the reflectivity of the multispectral image of the suspended matter sample in the matching experiment, and use the weighing method to obtain the corresponding concentration of the suspended matter sample;

[0057] According to the reflectivity of the multispectral image of the suspended matter sample in the proportioning experiment and the corresponding concentration of the suspended matter sample, the fitting analysis of the sensitive band is carried out, the best sensitive factor to the change of suspended matter concentration is screened, a scatter plot of the sensitive factor and the suspended matter concentration is drawn, and the fitting function with the highest fitting correlation coefficient is determined as the suspended matter concentration prediction model.

[0058] Among them, the basic requirements for drone aerial photography are:

[0059] (1) Shooting mode: single-scene continuous shooting mode under visible conditions, which can obtain visible light photos and multispectral images;

[0060] (2) UAV flight altitude: determined according to the scope of the water quality assessment area and the flight restrictions in the flight area, generally 400 to 500 meters;

[0061] (3) Shooting angle: normal (-90°);

[0062] (4) Shooting time: the time period when the solar altitude angle is higher than 30°, generally 9:30 to 15:30;

[0063] (5) Shooting range: the possible diffusion area of ​​suspended matter.

[0064] In a specific embodiment, the chemical indicators and biological indicators of the water body specifically include: the biological indicators include the water organism content and chlorophyll concentration of the water body; the chemical indicators include the nitrogen content and phosphorus content of the water body.

[0065] In a specific embodiment, obtaining several water quality assessment sub-models specifically includes: selecting any one of the random forest, gradient boosting tree, and support vector machine classification machine learning models to construct a sub-model, and using each physical indicator, chemical indicator, and biological indicator as the input of the sub-model, to obtain several water quality assessment sub-models.

[0066] In a specific embodiment, obtaining the fusion weight of the corresponding water quality assessment sub-model specifically includes: analyzing the assessment accuracy of each sub-model through a test set as the first indicator of the fusion weight; calculating the mutual information between the training sets generated by different indicators as the second indicator of the fusion weight; calculating the stability of different indicators as the third indicator of the fusion weight; normalizing the three weights and adding them together to obtain the final fusion weight of each sub-model.

[0067] In a specific embodiment, constructing a comprehensive water quality assessment model specifically includes: determining the importance of the sub-model generated by each indicator in the fusion model by the fusion weight of the water quality assessment sub-model; multiplying the classification result of each sub-model by the weight, and combining them to obtain a comprehensive water quality assessment model.

[0068] A freshwater ecosystem health monitoring system, such as Figure 2 As shown, including:

[0069] The data set establishment module obtains historical remote sensing images, constructs turbidity detection parameters, and obtains water turbidity based on historical remote sensing images; at the same time, the suspended solids concentration is obtained through the pre-trained suspended solids concentration prediction model and historical remote sensing images; water turbidity and suspended solids concentration are used as physical indicators; historical ground measured data are obtained to extract chemical indicators and biological indicators of water bodies; training sets and test sets are established based on physical indicators, chemical indicators and biological indicators;

[0070] The sub-model building module selects the classification machine learning model to construct a sub-model for each physical indicator, chemical indicator and biological indicator, uses the training set as input for training, and obtains several water quality assessment sub-models;

[0071] The water quality assessment sub-model fusion module analyzes the accuracy of each water quality assessment sub-model, calculates the correlation and stability between different models, and obtains the fusion weight of the corresponding water quality assessment sub-model; based on the fusion weight of the water quality assessment sub-model, all water quality assessment sub-models are combined to build a comprehensive water quality assessment model;

[0072] The evaluation report generation module obtains the remote sensing images to be tested and the ground measured data, and generates an evaluation report based on the freshwater ecological water quality based on the comprehensive water quality assessment model.

[0073] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0074] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring the health of a freshwater ecosystem, characterized in that: include: Acquire historical remote sensing images, construct turbidity detection parameters, and acquire water turbidity based on the historical remote sensing images; At the same time, the suspended solids concentration is obtained through the pre-trained suspended solids concentration prediction model and the historical remote sensing images; the water turbidity and suspended solids concentration are used as physical indicators; historical ground measured data are obtained to extract chemical indicators and biological indicators of the water body; training sets and test sets are established based on the physical indicators, chemical indicators and biological indicators; Select a classification machine learning model to construct a sub-model for each of the physical indicators, chemical indicators and biological indicators, use the training set as input for training, and obtain several water quality assessment sub-models; Perform accuracy analysis on each of the water quality assessment sub-models, calculate the correlation and stability between different models, and obtain the fusion weight of the corresponding water quality assessment sub-model; Based on the fusion weights of the water quality assessment sub-models, all water quality assessment sub-models are combined to construct a comprehensive water quality assessment model; Acquire the remote sensing images and ground measured data to be tested, conduct freshwater ecological water quality assessment based on the comprehensive water quality assessment model, and generate an evaluation report.

2. A freshwater ecosystem health monitoring method according to claim 1, characterized in that: The method of obtaining the turbidity of the water body specifically includes: extracting the quasi-plate measurement value of the historical remote sensing image, the diffuse scattered light from the sky and the off-water radiance to establish a reflectance curve of the sensitive wavelength, and constructing turbidity detection parameters based on the reflectance curve to perform water turbidity detection.

3. A freshwater ecosystem health monitoring method according to claim 2, characterized in that: The establishment of the reflectivity curve of the sensitive wavelength specifically includes: Among them, ρ p is the reflectivity of the standard plate, E d (0 + ) represents the total incident irradiance on the water surface, ρ represents the reflectivity of the air-water interface to the sky light, and L p is the measured value of the standard plate, L w is the water-leaving radiance, L sky Sky diffuse light, R rs is the reflectivity curve.

4. A freshwater ecosystem health monitoring method according to claim 3, characterized in that: The construction of turbidity detection parameters specifically includes extracting the reflectance of infrared band wavelengths and near-infrared band wavelengths to construct a water surface turbidity index SWTI: Among them, λ a is the central wavelength of the near-infrared band, λ b is the central wavelength of the infrared band, a1 and b1 are the fitting coefficients.

5. A freshwater ecosystem health monitoring method according to claim 1, characterized in that: The construction process of the suspended matter concentration prediction model is as follows: Collect particulate matter or sediment samples from water bodies in historical remote sensing images, and conduct matching experiments through reflectance spectrum measurement containers. At the same time, conduct multispectral drone aerial photography and suspended matter sample collection experiments, calculate the reflectivity of the multispectral image of the suspended matter sample in the matching experiment, and use the weighing method to obtain the corresponding concentration of the suspended matter sample; According to the reflectivity of the multispectral image of the suspended matter sample in the proportioning experiment and the corresponding concentration of the suspended matter sample, a fitting analysis of the sensitive band is performed, the best sensitive factor to the change of suspended matter concentration is screened, a scatter plot of the sensitive factor and the suspended matter concentration is drawn, and the fitting function with the highest fitting correlation coefficient is determined as the suspended matter concentration prediction model.

6. A freshwater ecosystem health monitoring method according to claim 1, characterized in that: The chemical indicators and biological indicators of the water body specifically include: the biological indicators include the water organism content and chlorophyll concentration of the water body; the chemical indicators include the nitrogen content and phosphorus content of the water body.

7. A freshwater ecosystem health monitoring method according to claim 1, characterized in that: The obtaining of several water quality assessment sub-models specifically includes: selecting any one of the random forest, gradient boosting tree, and support vector machine classification machine learning models to construct a sub-model, and using each of the physical indicators, chemical indicators, and biological indicators as inputs of the sub-model to obtain several water quality assessment sub-models.

8. A freshwater ecosystem health monitoring method according to claim 1, characterized in that: The method of obtaining the fusion weight of the corresponding water quality assessment sub-model specifically includes: analyzing the assessment accuracy of each sub-model through a test set as the first indicator of the fusion weight; calculating the mutual information between the training sets generated by different indicators as the second indicator of the fusion weight; calculating the stability of different indicators as the third indicator of the fusion weight; and normalizing the three weights and adding them together to obtain the final fusion weight of each sub-model.

9. A freshwater ecosystem health monitoring method according to claim 1, characterized in that: The construction of the comprehensive water quality assessment model specifically includes: determining the importance of the sub-model generated by each indicator in the fusion model by the fusion weight of the water quality assessment sub-model; multiplying the classification result of each sub-model by the weight, and combining them to obtain a comprehensive water quality assessment model.

10. A freshwater ecosystem health monitoring system, using a freshwater ecosystem health monitoring method according to any one of claims 1 to 9, characterized in that: include: The data set establishment module obtains historical remote sensing images, constructs turbidity detection parameters, and obtains water turbidity based on the historical remote sensing images; at the same time, the suspended matter concentration is obtained through the pre-trained suspended matter concentration prediction model and the historical remote sensing images; the water turbidity and suspended matter concentration are used as physical indicators; historical ground measured data are obtained to extract chemical indicators and biological indicators of the water body; training sets and test sets are established based on the physical indicators, chemical indicators and biological indicators; A sub-model building module selects a classification machine learning model to construct a sub-model for each of the physical indicators, chemical indicators and biological indicators, uses the training set as input for training, and obtains a number of water quality assessment sub-models; The water quality assessment sub-model fusion module performs accuracy analysis on each of the water quality assessment sub-models, calculates the correlation and stability between different models, and obtains the fusion weight of the corresponding water quality assessment sub-model; Based on the fusion weights of the water quality assessment sub-models, all water quality assessment sub-models are combined to construct a comprehensive water quality assessment model; The evaluation report generation module obtains the remote sensing images to be tested and the ground measured data, and generates an evaluation report based on the freshwater ecological water quality based on the comprehensive water quality assessment model.

Images