Water body phosphate concentration prediction method based on multispectral image

Abstract

The invention relates to the technical field of water quality data analysis, in particular to a multispectral image-based water phosphate concentration prediction method, which is characterized by comprising the steps of data acquisition, image preprocessing, absorption characteristic determination, key wave band selection and characteristic extraction, model establishment and training and real-time prediction and optimization. The continuity and the precision in the prediction area are improved; according to the invention, a high-precision, real-time and online phosphate concentration prediction system is constructed by utilizing a method of combining laboratory simulation and field data and directly utilizing the specific absorption characteristic of a complex generated by reacting phosphate with metal ions such as iron and aluminum in a red edge or near-infrared band; through multiple links of image preprocessing, key wave band selection, spectral feature extraction, nonlinear regression model establishment, spatio-temporal data optimization and the like, the defects that a traditional remote sensing method depends on indirect indexes and is greatly interfered by the environment and the like are successfully overcome, and the accuracy and reliability of water quality monitoring are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality data analysis, and in particular to a method for predicting phosphate concentration in water based on multispectral imaging. Background Art

[0002] At present, the problem of eutrophication of water bodies is becoming increasingly serious. Among them, phosphate is one of the main nutrients. Its excessive presence often causes abnormal algae reproduction and algal bloom, which in turn threatens the ecological balance and water quality safety. Traditional water phosphate monitoring mainly relies on laboratory chemical analysis methods, such as spectrophotometry, ion chromatography and electrochemical sensing technology. Although these methods are accurate, they have defects such as long sample collection, transportation, and detection cycle and limited spatial coverage. It is difficult to meet the needs of large-scale, real-time online monitoring. In recent years, remote sensing technology has been widely used in water quality monitoring. Monitoring methods based on multispectral imaging have attracted much attention due to their wide coverage and strong real-time performance.

[0003] However, due to the lack of obvious spectral response of phosphate itself, traditional multispectral remote sensing methods usually rely on indirect indicators (such as chlorophyll concentration, suspended matter concentration, etc.) for inference, resulting in low prediction accuracy and greater environmental interference. Existing technologies have not fully utilized the specific absorption characteristics of phosphate in water bodies when it reacts with metal ions such as iron and aluminum to form complexes in the red edge or near-infrared bands as a basis for directly reflecting changes in phosphate concentration, thereby limiting the application effect of remote sensing technology in water quality monitoring. To solve the above problems, we propose a method for predicting phosphate concentration in water bodies based on multispectral images. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method for predicting phosphate concentration in water based on multispectral images to solve the problems raised in the background technology.

[0005] The above technical objectives of the present invention are achieved through the following technical solutions:

[0006] A method for predicting phosphate concentration in water based on multispectral images, comprising:

[0007] S1. Data acquisition, used to collect multi-phase multi-spectral image data of the target water body and on-site environmental parameters, where the environmental parameters include pH value, temperature, dissolved oxygen content, and concentration of metal ions such as iron and aluminum;

[0008] S2, image preprocessing, used to perform atmospheric, geometric and radiation correction on the collected images, and use water mask technology to segment the water area of ​​the image, so as to extract the spectral reflectance data of the effective area of ​​the water body;

[0009] S3, absorption characteristics determination, used to determine the absorption peak position, absorption depth and bandwidth parameters of the complex formed by the reaction of metal ions such as iron and aluminum with phosphate in the red edge or near infrared band under laboratory conditions by simulating the water environment and adjusting the metal ion concentration and pH value;

[0010] S4, key band selection and feature extraction, according to the results of S3, the key band corresponding to the absorption peak of the complex is adaptively selected, the reflectance data of each key band is extracted from the image data extracted by S2, and smoothing filtering is used to improve the data signal-to-noise ratio, and the spectral feature vector of the water area is constructed;

[0011] S5, model building and training, is used to use the phosphate concentration data obtained from field sampling, combined with the metal ion concentration and pH value collected in S1, to establish a mathematical model between the phosphate concentration and the spectral features extracted by S4 through a nonlinear least squares regression method, and to optimize the model parameters using cross-validation;

[0012] S6, real-time prediction and optimization, is used to input the spectral features of the latest multispectral image data extracted by S4 into the S5 model, calculate the phosphate concentration in each area of ​​the water body in real time, and use Kriging interpolation and time series analysis to perform spatiotemporal coupling optimization on the preliminary prediction results to improve the continuity and accuracy within the prediction area.

[0013] By adopting the above technical solution, the multispectral image data collected in step S1 includes visible light, red edge and near infrared band data, and the on-site environmental parameters are collected in real time through portable water quality monitoring instruments to ensure data synchronization and accuracy.

[0014] By adopting the above technical solution, the image preprocessing used in step S2 includes:

[0015] Use atmospheric correction algorithm to eliminate atmospheric scattering interference;

[0016] Use geometric correction technology to eliminate image distortion;

[0017] The water mask algorithm based on threshold segmentation or edge detection is used to accurately distinguish water areas from land and clouds.

[0018] By adopting the above technical solution, in the S3 step, by performing spectral scanning under the conditions of controlling temperature, pH value and metal ion concentration, the absorption peak position of the complex (such as wavelength λ≈680–720nm), absorption depth and bandwidth parameters are determined, and their quantitative relationship with changes in metal ion concentration and pH value is recorded.

[0019] By adopting the above technical solution, the spectral feature vector extracted in the step S4 is normalized after being smoothed and filtered to eliminate the reflectivity deviation caused by changes in ambient light, thereby ensuring that the proposed feature reflects the complex absorption characteristics.

[0020] By adopting the above technical solution, the calibration model established in step S5 is a nonlinear least squares regression model, in which the model parameters are determined by cross-validation, and the metal ion concentration and pH value are used as covariates in the model to improve the prediction accuracy.

[0021] By adopting the above technical solution, the spatiotemporal optimization adopted in step S6 includes:

[0022] The spatial distribution data are interpolated using the Kriging interpolation method to ensure the continuity of the predicted values ​​in each area of ​​the water body;

[0023] Combined with time series analysis, multi-temporal data are smoothed and trend corrected to eliminate the impact of short-term fluctuations on prediction accuracy.

[0024] By adopting the above technical solution, the water phosphate concentration prediction result after S6 optimization is output in the form of spatiotemporal distribution diagram and statistical diagram, and data visualization is achieved through the GIS system.

[0025] In summary, the present invention mainly has the following beneficial effects:

[0026] 1. The present invention simulates the water environment under laboratory conditions, adjusts the concentration of metal ions such as iron and aluminum and the pH value, accurately determines the absorption peak position, absorption depth and bandwidth parameters in the red edge or near-infrared band when phosphate reacts with these metal ions to form a complex, establishes a quantitative relationship between it and the phosphate concentration, directly utilizes the absorption characteristics of the complex in a specific band, and makes up for the defect that traditional remote sensing technology cannot directly reflect the change of phosphate concentration.

[0027] 2. The present invention extracts spectral data that directly reflects the absorption characteristics of the complex from multispectral images by adaptively selecting key bands corresponding to laboratory measurement results, combines atmospheric, geometric and radiation corrections and water masking technology to ensure the accuracy and regional continuity of the extracted data, and then establishes an accurate phosphate concentration prediction model to achieve real-time and direct monitoring of phosphate concentration in water bodies. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of the process of the method for predicting phosphate concentration in water body of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0030] The following examples are used to illustrate the present invention, but they cannot be used to limit the scope of protection of the present invention. The conditions in the examples can be further adjusted according to specific conditions. Simple improvements to the method of the present invention under the premise of the concept of the present invention belong to the scope of protection claimed in the present invention.

[0031] Example 1

[0032] refer to Figure 1 A method for predicting phosphate concentration in water based on multispectral images. This embodiment uses a reservoir as the target water body and verifies the feasibility of the technology for predicting phosphate concentration in water based on multispectral images by combining laboratory and field data. The method specifically includes the following steps:

[0033] S1. Data acquisition: After selecting the target reservoir area on site, high-resolution multispectral remote sensing instruments are used to collect image data of the area. The collected images cover three bands: visible light, red edge, and near-infrared, which can capture the reflection characteristics of water bodies under different spectra. At the same time, portable water quality monitoring instruments are used to record the environmental parameters of the water body in real time, including pH value, temperature, dissolved oxygen, and concentrations of metal ions such as iron and aluminum, to ensure the synchronization of image data with on-site water quality parameters in time and space, providing a reliable basis for subsequent analysis.

[0034] S2. Image preprocessing: The collected raw image data needs to be preprocessed to eliminate the interference caused by the environment and instruments. First, the atmospheric correction technology is used to remove the atmospheric scattering and absorption effects and restore the true reflectivity of the water body. Then, the geometric correction technology is used to eliminate the image distortion caused by sensor movement and terrain undulations. Finally, the water body mask is generated through the automated segmentation algorithm to accurately distinguish the water area from the land and clouds, so that only the reflectivity data in the water area is extracted for subsequent processing.

[0035] S3. Determination of absorption characteristics: Under laboratory conditions, a simulated water environment is built. By adjusting the temperature, pH value, and the concentration of metal ions such as iron and aluminum, the absorption characteristics of the complex formed by the reaction of phosphate and metal ions are studied. The simulated water is scanned with a spectrometer, focusing on the absorption characteristics of the red edge or near-infrared band (such as the 680-720 nanometer region). The absorption peak position, absorption depth, and bandwidth parameters of the complex are obtained through experiments, and the relationship between these parameters and changes in metal ion concentration and pH value is recorded, providing a scientific basis for the selection of key bands.

[0036] S4. Key band selection and feature extraction: According to the absorption characteristics measured in the laboratory, the key bands related to the absorption of the complex in the red edge or near-infrared region are determined. The reflectance information of these key bands is extracted using the preprocessed image data, and a vector representing the spectral characteristics of the water body is constructed. To improve the reliability of the data, the reflectance data of each band is first smoothed to reduce noise interference; then normalized to eliminate the deviation caused by different ambient lighting conditions, so that the extracted features can truly reflect the absorption characteristics of the complex.

[0037] S5. Model establishment and training. A nonlinear regression model is constructed using the actual phosphate concentration data collected on site, as well as the spectral features and environmental parameters (such as iron ion concentration and pH value) extracted from the images. The model continuously adjusts the model parameters by minimizing the error between the predicted value and the actual value. To ensure that the model has strong generalization ability and robustness, the cross-validation method is used to tune the model parameters. Through this process, a mathematical model that can reflect the relationship between phosphate concentration in water bodies and key spectral features and environmental parameters is constructed.

[0038] S6. Real-time prediction and optimization. After the model is established, the newly collected multispectral images and on-site environmental data are used to predict the phosphate concentration in each area of ​​the water body in real time through the trained model. In order to ensure the continuity and accuracy of the prediction results in space, the Kriging interpolation method is used to spatially optimize the prediction values ​​of each region. At the same time, combined with time series analysis, multi-phase data are smoothed and trend corrected to effectively eliminate the interference caused by short-term environmental fluctuations, thereby obtaining more stable and reliable spatiotemporal prediction results.

[0039] The prediction results after time and space optimization are then integrated and displayed through the GIS system to generate intuitive time and space distribution maps and statistical graphs. Users can use these visual graphics to intuitively understand the dynamic changes in phosphate concentration in various areas of the water body, facilitating the timely implementation of water quality management and early warning measures.

[0040] In this embodiment, a method combining laboratory simulation with field data is used to directly utilize the specific absorption characteristics of complexes generated by the reaction of phosphate with metal ions such as iron and aluminum in the red edge or near-infrared band to construct a high-precision, real-time online phosphate concentration prediction system. Through multiple links such as image preprocessing, key band selection, spectral feature extraction, nonlinear regression model establishment, and spatiotemporal data optimization, the defects of traditional remote sensing methods such as reliance on indirect indicators and being greatly affected by environmental interference are successfully solved, which significantly improves the accuracy and reliability of water quality monitoring.

[0041] Example 2

[0042] In order to verify the superiority of the method of the present invention, a comparative experiment was carried out in the same water body area, using the improved method of the present invention and the conventional remote sensing method based on multispectral imaging to predict the phosphate concentration of the water body, and comparing and analyzing the water sample detection data collected on site.

[0043] Experimental design

[0044] Data collection: In the same reservoir area, the same multispectral remote sensing instrument was used to collect image data. The collected data included visible light, red edge and near-infrared bands. Portable water quality monitoring instruments were used on site to record the environmental parameters of the water body (pH value, temperature, dissolved oxygen and concentrations of metal ions such as iron and aluminum), and water samples were collected for laboratory phosphate concentration testing as the actual value;

[0045] Conventional method: mainly uses the indicators that indirectly reflect the water status in the image (such as chlorophyll concentration and suspended solids concentration) to establish a regression model to predict phosphate concentration. Since phosphate itself lacks obvious spectral response, this method has large prediction errors under different lighting and environmental interference;

[0046] Example 1 Method: By simulating the water environment in the laboratory, the concentration of metal ions such as iron and aluminum and the pH value are adjusted to obtain the specific absorption characteristics of the complex in the red edge or near-infrared band. Subsequently, the spectral features reflecting the absorption characteristics of the complex are directly extracted from the image data by adaptive key band selection and spectral smoothing and normalization processing, and a nonlinear regression model is established in combination with the on-site environmental parameters to achieve direct prediction of phosphate concentration.

[0047] The experimental results and comparisons are shown in Table 1:

[0048] Table 1

[0049] Prediction accuracy: Experimental results show that the prediction results of conventional methods under different environmental conditions have large deviations from the actual water sample detection values, and the error fluctuations are obvious;

[0050] The method of the present invention directly utilizes the absorption characteristics of the complex, so that the prediction results are more consistent with the actual test data and the overall error is significantly reduced;

[0051] Environmental adaptability: Since conventional methods rely on indirect indicators, their prediction results are easily affected by environmental factors such as light and weather, and there is a large degree of uncertainty;

[0052] The method of the present invention determines the key absorption characteristics in the laboratory and applies the corresponding correction strategy to the field data, which effectively eliminates the influence of ambient light changes and other interferences on the prediction results and improves the stability of the system.

[0053] Spatiotemporal continuity: When making spatial predictions for water areas, conventional methods do not have smooth spatial distribution due to large data fluctuations;

[0054] The method of the present invention combines Kriging interpolation and time series smoothing, which not only achieves a spatially continuous and uniform prediction effect, but also effectively corrects short-term fluctuations, making the prediction result more stable in the time domain.

[0055] Through comparative experiments, it can be seen that the method of the present invention has higher prediction accuracy, better environmental adaptability and better spatiotemporal continuity in the real-time monitoring of phosphate concentration in water bodies compared with the traditional remote sensing method based on multispectral imaging, which provides more reliable technical support for actual water quality monitoring and early warning.

[0056] Although the embodiments of the present invention have been shown and described, it is understandable to those skilled in the art that, unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the usual meanings understood by persons with ordinary skills in the field to which the present invention belongs. The words "including" or "comprising" and the like used in the present invention mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. The words "connect" or "connected" and the like are not limited to physical or mechanical connections, but may also include electrical connections, whether direct or indirect. "Up", "down", "left", "right", etc. are only used to indicate relative position relationships. When the absolute position of the described object changes, the relative position relationship may also change accordingly.

[0057] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting phosphate concentration in water based on multispectral images, characterized in that: include: S1. Data acquisition, used to collect multi-phase multi-spectral image data of the target water body and on-site environmental parameters, where the environmental parameters include pH value, temperature, dissolved oxygen content, and iron and aluminum metal ion concentrations of the water body; S2, image preprocessing, used to perform atmospheric, geometric and radiation correction on the collected images, and use water mask technology to segment the water area of ​​the image, so as to extract the spectral reflectance data of the effective area of ​​the water body; S3, absorption characteristics determination, used to determine the absorption peak position, absorption depth and bandwidth parameters of the complex formed by the reaction of iron and aluminum metal ions with phosphate in the red edge or near infrared band under laboratory conditions by simulating the water environment and adjusting the metal ion concentration and pH value; S4, key band selection and feature extraction, according to the results of S3, the key band corresponding to the absorption peak of the complex is adaptively selected, the reflectance data of each key band is extracted from the image data extracted by S2, and smoothing filtering is used to improve the data signal-to-noise ratio, and the spectral feature vector of the water area is constructed; S5, model building and training, is used to use the phosphate concentration data obtained from field sampling, combined with the metal ion concentration and pH value collected in S1, to establish a mathematical model between the phosphate concentration and the spectral features extracted by S4 through a nonlinear least squares regression method, and to optimize the model parameters using cross-validation; S6, real-time prediction and optimization, is used to input the spectral features of the latest multispectral image data extracted by S4 into the S5 model, calculate the phosphate concentration in each area of ​​the water body in real time, and use Kriging interpolation and time series analysis to perform spatiotemporal coupling optimization on the preliminary prediction results to improve the continuity and accuracy within the prediction area.

2. The method for predicting phosphate concentration in water based on multispectral imaging according to claim 1, characterized in that: The multispectral image data collected in step S1 includes visible light, red edge and near infrared band data, and the on-site environmental parameters are collected in real time through a portable water quality monitoring instrument to ensure data synchronization and accuracy.

3. The method for predicting phosphate concentration in water based on multispectral imaging according to claim 1, characterized in that: The image preprocessing used in step S2 includes: Use atmospheric correction algorithm to eliminate atmospheric scattering interference; Use geometric correction technology to eliminate image distortion; The water mask algorithm based on threshold segmentation or edge detection is used to accurately distinguish water areas from land and clouds.

4. The method for predicting phosphate concentration in water based on multispectral imaging according to claim 1, characterized in that: In the step S3, the absorption peak position, absorption depth and bandwidth parameters of the complex are determined by performing spectral scanning under the conditions of controlling temperature, pH value and metal ion concentration, and the quantitative relationship between the absorption peak position, absorption depth and bandwidth parameters as the metal ion concentration and pH value change is recorded.

5. The method for predicting phosphate concentration in water based on multispectral imaging according to claim 1, characterized in that: The spectral feature vector extracted in step S4 is normalized after being smoothed and filtered to eliminate the reflectivity deviation caused by changes in ambient light, ensuring that the proposed feature reflects the complex absorption characteristics.

6. The method for predicting phosphate concentration in water based on multispectral imaging according to claim 1, characterized in that: The calibration model established in step S5 is a nonlinear least squares regression model, in which the model parameters are determined by cross-validation, and the metal ion concentration and pH value are used as covariates in the model to improve the prediction accuracy.

7. The method for predicting phosphate concentration in water based on multispectral imaging according to claim 1, characterized in that: The spatiotemporal optimization used in step S6 includes: The spatial distribution data are interpolated using the Kriging interpolation method to ensure the continuity of the predicted values ​​in each area of ​​the water body; Combined with time series analysis, multi-temporal data are smoothed and trend corrected to eliminate the impact of short-term fluctuations on prediction accuracy.

8. The method for predicting phosphate concentration in water based on multispectral imaging according to claim 1, characterized in that: The water body phosphate concentration prediction results after S6 optimization are output in the form of spatiotemporal distribution diagrams and statistical diagrams, and data visualization is achieved through a GIS system.

Images