Method for deducing total amount of algae in water body based on remote sensing and environmental parameters

Abstract

The invention relates to a method and device for deducing the total amount of algae in a water body based on remote sensing and environmental parameters, computer equipment and a storage medium. The method comprises the following steps: acquiring a remote sensing image of a to-be-deduced area, and a water surface chlorophyll concentration and a water body profile parameter set of each preset sampling pixel position in the remote sensing image, and acquiring remote sensing image features of a water body surface according to the remote sensing image of the to-be-deduced area; determining a water surface algae density prediction model by adopting at least one machine learning algorithm, and obtaining the total amount of water surface algae through the water surface algae density prediction model; determining a water algae density prediction model by adopting at least one regression model, and obtaining the total amount of water algae through the water algae density prediction model; and obtaining the total algae amount of the to-be-deduced area based on the total water surface algae amount and the total water algae amount. By adopting the method, the accuracy and reliability of total algae amount estimation can be remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of remote sensing technology, and particularly to a method, device, computer device, and storage medium for inferring the total amount of water body algae based on remote sensing and environmental parameters. Background Art

[0002] With the development of remote sensing technology, using remote sensing means to monitor algae including cyanobacteria in water bodies has become an important tool for studying water body eutrophication, evaluating water quality, and managing water environment.

[0003] However, although traditional remote sensing technology has significant advantages in monitoring the area and concentration of algae, in a more complex water body environment, traditional methods still have limitations. For example, traditional technologies usually rely on a single data source, mainly satellite remote sensing data; traditional processing algorithms mostly focus on traditional spectral index methods or simple statistical models. However, when dealing with complex water body environments, traditional algorithms often have difficulty achieving accurate detection of algae area and volume in complex backgrounds; traditional technologies mainly focus on water surface phenomena, and the few existing technologies for estimating the total amount of algae are limited to within the euphotic layer, or use traditional bio-optical models and spectral analysis for estimation, and for accurately reflecting the true distribution of algae in water bodies, especially when the algae are thick or unevenly distributed, traditional technologies are difficult to give accurate total algae amount data. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, computer device, and storage medium for inferring the total amount of water body algae based on remote sensing and environmental parameters, which can more comprehensively reflect the spatial distribution of water body algae, especially the concentration change in the vertical direction, so as to improve the accuracy of algae monitoring.

[0005] In a first aspect, the present application provides a method for inferring the total amount of water body algae based on remote sensing and environmental parameters, including:

[0006] Step S1, obtaining a remote sensing image of the area to be inferred, as well as the water surface chlorophyll concentration and the water body profile parameter set at each preset sampling pixel position in the remote sensing image, where the water body profile parameter set includes the water body environmental parameter set of the area to be inferred and the chlorophyll concentration at the preset water depth position, and obtaining the remote sensing image features of the water surface based on the remote sensing image of the area to be inferred;

[0007] Step S2: Use at least one machine learning algorithm to train each water surface algae density prediction network structure with the remote sensing image features of the water surface in the area to be inferred as input features and the preset water surface chlorophyll concentration at each sampling pixel position as the target variable. Compare the output results of each water surface algae density prediction network structure to determine the water surface algae density prediction model, and obtain the algae density of each pixel in the remote sensing image of the area to be inferred through the water surface algae density prediction model, so as to obtain the total amount of water surface algae;

[0008] Step S3: Use at least one regression model to train each water body algae density prediction network structure with the water environment parameter set in the area to be inferred as input features and the preset chlorophyll concentration at the water depth position as the target variable. Compare the output results of each water body algae density prediction network structure to determine the water body algae density prediction model, and obtain the algae density of each pixel water column in the remote sensing image of the area to be inferred through the water body algae density prediction model, so as to obtain the total amount of water body algae;

[0009] Step S4: Obtain the total amount of algae in the area to be inferred based on the total amount of water surface algae and the total amount of water body algae.

[0010] In one embodiment, before step S3, it further includes

[0011] Divide the pixels of the remote sensing image of the area to be inferred into different sub-regions by using a spatial clustering algorithm, and obtain the water body profile parameter set of each sub-region through a water quality monitor.

[0012] In one embodiment, obtaining the water body profile parameter set of each sub-region through a water quality monitor includes:

[0013] Obtain the water body profile parameter set of each region at the preset sampling pixel positions through a water quality monitor; where the water environment parameter set in the area to be inferred in the water body profile parameter set includes algae concentration, water depth, water temperature, dissolved oxygen, nitrogen concentration, phosphorus concentration, organic matter content, transparency.

[0014] In one embodiment, in step S2, using at least one machine learning algorithm to train each water surface algae density prediction network structure with the remote sensing image features of the water surface in the area to be inferred as input features and the preset water surface chlorophyll concentration at each sampling pixel position as the target variable, and comparing the output results of each water surface algae density prediction network structure to determine the water surface algae density prediction model includes:

[0015] Prepare the water surface algae density prediction data set, including the remote sensing image features of the water surface in the area to be inferred and the preset water surface chlorophyll concentration at each sampling pixel position in the remote sensing image;

[0016] Divide the water surface algae density prediction data set into a training set and a test set;

[0017] Construct each water surface algae density prediction network structure using at least one machine learning algorithm based on the training set and the test set;

[0018] Obtain the output results of each water surface algae density prediction network structure, where the output results include the target parameters of each water surface algae density prediction network structure;

[0019] Determine the water surface algae density prediction model by comparing the target parameters of each water surface algae density prediction network structure.

[0020] In one embodiment, in step S2, the algae density of each pixel of the remote sensing image of the area to be inferred is obtained through the water surface algae density prediction model, and the total amount of water surface algae is obtained as follows:

[0021] Use the water surface algae density prediction model to predict each pixel of the remote sensing image of the area to be inferred to obtain the algae density of each pixel;

[0022] Calculate the actual area covered by each pixel according to the spatial resolution of the remote sensing image of the area to be inferred;

[0023] Obtain the total amount of water surface algae in the area to be inferred through the algae density of each pixel and the actual area covered by each pixel.

[0024] In one embodiment, in step S3, at least one regression model is used to train each water body algae density prediction network structure with the water body environment parameter set of the area to be inferred as the input feature and the chlorophyll concentration at the preset water depth position as the target variable. Comparing the output results of each water body algae density prediction network structure to determine the water body algae density prediction model includes:

[0025] Prepare the water body algae density prediction data set, including the water body environment parameter set of the area to be inferred and the chlorophyll concentration at the preset water depth position;

[0026] Divide the water body algae density prediction data set into a training set and a test set;

[0027] Based on the training set and the test set, use at least one regression model to construct each water body algae density prediction network structure;

[0028] Obtain the output results of each water body algae density prediction network structure, where the output results include the target parameters of each water body algae density prediction network structure;

[0029] Determine the water body algae density prediction model by comparing the target parameters of each water body algae density prediction network structure.

[0030] In one embodiment, step S4 includes:

[0031] Integrate each pixel water column based on the total amount of surface algae and the total amount of water column algae to obtain the total amount of algae for each pixel, and obtain the total amount of algae for each sub-region based on the total amount of algae for each pixel. The sum of the total amounts of algae for each sub-region is the total amount of algae for the region to be inferred.

[0032] In a second aspect, the present application also provides a device for inferring the total amount of water column algae based on remote sensing and environmental parameters, including:

[0033] An image data acquisition module, configured to acquire a remote sensing image of the region to be inferred, as well as the surface chlorophyll concentration and the water column parameter set at preset sampling pixel positions in the remote sensing image. The water column parameter set includes the water environment parameter set of the region to be inferred and the chlorophyll concentration at a preset water depth position, and acquire the remote sensing image features of the water surface based on the remote sensing image of the region to be inferred;

[0034] A surface algae density prediction module, configured to train each surface algae density prediction network structure by using at least one machine learning algorithm with the remote sensing image features of the water surface of the region to be inferred as input features and the surface chlorophyll concentration at preset sampling pixel positions as target variables, compare the output results of each surface algae density prediction network structure, determine the surface algae density prediction model, and obtain the algae density of each pixel of the remote sensing image of the region to be inferred through the surface algae density prediction model, so as to obtain the total amount of surface algae;

[0035] A water column algae density prediction module, configured to train each water column algae density prediction network structure by using at least one regression model with the water environment parameter set of the region to be inferred as input features and the chlorophyll concentration at a preset water depth position as target variables, compare the output results of each water column algae density prediction network structure to determine the water column algae density prediction model, and obtain the algae density of each pixel water column of the remote sensing image of the region to be inferred through the water column algae density prediction model, so as to obtain the total amount of water column algae;

[0036] An algae total amount calculation module, configured to obtain the total amount of algae in the region to be inferred based on the total amount of surface algae and the total amount of water column algae.

[0037] In a third aspect, the present application also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the method for inferring the total amount of water column algae based on remote sensing and environmental parameters are implemented.

[0038] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for inferring the total amount of water column algae based on remote sensing and environmental parameters are implemented.

[0039] The above method, device, computer equipment and storage medium for inferring the total amount of water body algae based on remote sensing and environmental parameters obtain the remote sensing image of the area to be inferred, as well as the chlorophyll concentration on the water surface and the water body profile parameter set at the preset sampling pixel positions in the remote sensing image. The water body profile parameter set includes the water body environmental parameter set of the area to be inferred and the chlorophyll concentration at the preset water depth position, and obtains the remote sensing image features on the water surface of the area to be inferred according to the remote sensing image of the area to be inferred; uses at least one machine learning algorithm to train each water surface algae density prediction network structure with the remote sensing image features on the water surface of the area to be inferred as input features and the chlorophyll concentration on the water surface at the preset sampling pixel positions as target variables, compares the output results of each water surface algae density prediction network structure, determines the water surface algae density prediction model, and obtains the algae density of each pixel in the remote sensing image of the area to be inferred through the water surface algae density prediction model to obtain the total amount of water surface algae; uses at least one regression model to train each water body algae density prediction network structure with the water body environmental parameter set of the area to be inferred as input features and the chlorophyll concentration at the preset water depth position as target variables, compares the output results of each water body algae density prediction network structure to determine the water body algae density prediction model, and obtains the algae density of each pixel water column in the remote sensing image of the area to be inferred through the water body algae density prediction model to obtain the total amount of water body algae; obtains the total amount of algae in the area to be inferred based on the total amount of water surface algae and the total amount of water body algae, which is used for inferring the total amount of algae in the water body, and takes cyanobacteria as an example. This method improves the monitoring and total amount inference accuracy of water body algae by integrating multi-source data, including high-resolution satellite images and ground observation data, and combining improved data processing algorithms. The spectral data on the water surface of the water body, that is, the remote sensing image of the area to be inferred, is obtained by using remote sensing technology, and combined with the measured environmental parameters on the spot, that is, the chlorophyll concentration on the water surface and the water body profile parameter set at the preset sampling pixel positions in the remote sensing image, to construct a water surface algae concentration model, that is, a water surface algae density prediction model; by establishing a depth distribution model of underwater algae, that is, a water body algae density prediction model, the vertical distribution of algae in the water column is inferred; combining the two models, the total amount of algae in the entire water body is calculated. The advantage of this application is that it provides three-dimensional monitoring capabilities, can more comprehensively reflect the spatial distribution of water body algae, especially the concentration change in the vertical direction, thereby improving the accuracy of algae monitoring, solving the limitation that traditional methods can only calculate the algae on the water surface, and providing information on algae data for water body management and pollution control. Description of the Drawings

[0040] Figure 1 It is a schematic flowchart of a method for inferring the total amount of water body algae based on remote sensing and environmental parameters in an embodiment;

[0041] Figure 2 It is a schematic flowchart of a method for inferring the total amount of cyanobacteria in a water body based on remote sensing and environmental parameters in an embodiment;

[0042] Figures 3(a) and 3(b) are schematic diagrams of the hyperspectral images of the drone in one embodiment;

[0043] Figures 4(a) and 4(b) are schematic diagrams of the Sentinel-2 satellite images in one embodiment;

[0044] Figure 5 It is a schematic diagram of the device for inferring the total amount of water body algae based on remote sensing and environmental parameters in one embodiment. Detailed implementation manners

[0045] With the development of remote sensing technology, using remote sensing means to monitor algae (such as cyanobacteria) in water bodies has become an important tool for studying water eutrophication, evaluating water quality, and managing water environment. Taking cyanobacteria as an example, remote sensing technology can identify and quantitatively estimate the distribution and area of cyanobacteria on the water surface by analyzing the spectral characteristics of water bodies. Especially in the season when cyanobacteria multiply in large numbers, satellite remote sensing images can provide cyanobacteria monitoring information over a large area, effectively assisting water body managers to identify high-risk areas. Traditional remote sensing technology mainly relies on spectral indices (such as normalized difference vegetation index, modified normalized difference vegetation index, floating algae index) and multispectral images to identify cyanobacteria on the water surface. These technologies have high efficiency and accuracy in estimating the area and concentration of cyanobacteria on the water surface, especially in water body monitoring at a larger scale. In addition, remote sensing technology can also monitor the dynamic changes of cyanobacteria through time series analysis to help judge the time and trend of cyanobacteria outbreaks.

[0046] However, although traditional remote sensing technology has significant advantages in monitoring the area and concentration of cyanobacteria, in a more complex water body environment, traditional methods still have limitations. Traditional technologies usually rely on a single data source, mainly satellite remote sensing data. Although this method has advantages in wide-area coverage, due to the lack of integration of multi-source information such as ground observation data and meteorological data, it often cannot provide comprehensive water body information. Traditional technologies focus on traditional spectral index methods or simple statistical models when dealing with algorithms. However, when dealing with complex water body environments, traditional algorithms are often difficult to accurately detect the area and volume of cyanobacteria in complex backgrounds. Traditional technologies mainly focus on surface phenomena of water bodies, and the few technologies for estimating the total amount of cyanobacteria are limited to within the euphotic layer or use traditional bio-optical models and spectral analysis for estimation, and it is difficult to accurately reflect the true distribution of cyanobacteria in water bodies, especially in the case of thick or uneven distribution of cyanobacteria, traditional technologies are difficult to give accurate total cyanobacteria data.

[0047] To break through the above technical bottlenecks, this application makes improvements from the following two aspects:

[0048] Fuse multi-source data to improve the estimation accuracy and adapt to water bodies with severe cyanobacteria: By fusing different types of data, such as high-resolution satellite remote sensing images and ground observation data, more comprehensive water body information can be obtained. Compared with the traditional method that only relies on remote sensing technology for the total amount of algae, after fusing multi-source data, there will be no problem of being unable to obtain deep cyanobacteria information due to the severe cyanobacteria and the reduction of the euphotic layer depth.

[0049] Improve the data processing algorithm and enhance the feature recognition ability to adapt to environments with complex or multi-factor influences: By introducing advanced data processing algorithms, such as random forest and support vector machine, the features of cyanobacteria can be extracted more effectively. In a complex background, by training a model to recognize the features of cyanobacteria, the area and volume of cyanobacteria can be detected and estimated more accurately. It solves the problem that in the case of complex water body environment or being affected by multiple factors, the traditional model may not be flexible enough to adapt to rapidly changing conditions or non-linear relationships.

[0050] In view of the above disadvantages, the present application provides a method for inferring the total amount of water body algae based on remote sensing and environmental parameters. Based on remote sensing data and environmental parameters, the present application comprehensively infers the total amount of water body algae by constructing a model. Using remote sensing technology to obtain the spectral information on the water surface and combining with the measured chlorophyll concentration, the content of algae on the water surface is first inferred. Then, through environmental parameters such as water depth, transparency, water temperature, dissolved oxygen, light intensity, nitrogen concentration, and phosphorus concentration and chlorophyll concentration data, a distribution model of underwater cyanobacteria is established. Finally, the cyanobacteria models on the water surface and underwater are combined to calculate the total amount of algae in the entire water body. In the present application, the chlorophyll concentration corresponds to the algae concentration, where the remote sensing data and environmental parameters are the remote sensing image of the area to be inferred, and the water surface chlorophyll concentration and water body profile parameter set at each preset sampling pixel position in the remote sensing image; the water surface algae model is the water surface algae density prediction model, and the underwater algae model is the water body algae density prediction model.

[0051] Specifically, the technical solution of the present application includes the following steps:

[0052] Data collection: Use remote sensing technology to obtain the spectral data on the water surface and extract the spectral information reflecting the water body characteristics, that is, obtain the remote sensing image of the area to be inferred and obtain the remote sensing image characteristics of the water surface according to the remote sensing image of the area to be inferred; measure the environmental parameters related to the water body condition, including water temperature, dissolved oxygen, light intensity, nitrogen, phosphorus, and the water surface algae concentration, that is, the water surface chlorophyll concentration and water body profile parameter set at each preset sampling pixel position in the remote sensing image, where the water body profile parameter set includes the water body environmental parameter set of the area to be inferred and the chlorophyll concentration at the preset water depth position.

[0053] Construction of the water surface algae model: Use a machine learning model and take the remote sensing data and environmental parameters as input variables to infer the concentration of algae on the water surface.

[0054] Underwater algae model construction: Measure the water depth, transparency, and algae content at different depths, and establish a relationship model between underwater algae and other parameters;

[0055] Total algae quantity calculation: Integrate the surface and underwater algae models, that is, integrate the surface algae density prediction model and the water body algae density prediction model, consider the depth distribution of the water body, and calculate the total algae quantity of the entire water body;

[0056] Optimization of measurement intervals: For different water depth conditions, select appropriate measurement depth intervals to ensure accurate capture of the vertical distribution of algae and avoid duplicate calculations.

[0057] To make the objectives, technical solutions, and advantages of this application clearer and more understandable, the following further details this application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0058] In an exemplary embodiment, as Figure 1 shown, a method for inferring the total algae quantity of a water body based on remote sensing and environmental parameters is provided, including:

[0059] Step S1, obtain the remote sensing image of the area to be inferred, as well as the surface chlorophyll concentration and the water body profile parameter set at the preset sampling pixel positions in the remote sensing image, where the water body profile parameter set includes the water body environmental parameter set of the area to be inferred and the chlorophyll concentration at the preset water depth positions, and obtain the remote sensing image features of the water body surface according to the remote sensing image of the area to be inferred;

[0060] Step S2, use at least one machine learning algorithm to train each surface algae density prediction network structure with the remote sensing image features of the water body surface of the area to be inferred as input features and the surface chlorophyll concentration at the preset sampling pixel positions as target variables, compare the output results of each surface algae density prediction network structure, determine the surface algae density prediction model, and obtain the algae density of each pixel in the remote sensing image of the area to be inferred through the surface algae density prediction model, and obtain the total surface algae quantity;

[0061] Step S3, use at least one regression model to train each water body algae density prediction network structure with the water body environmental parameter set of the area to be inferred as input features and the chlorophyll concentration at the preset water depth positions as target variables, compare the output results of each water body algae density prediction network structure to determine the water body algae density prediction model, and obtain the algae density of each pixel water column in the remote sensing image of the area to be inferred through the water body algae density prediction model, and obtain the total water body algae quantity;

[0062] Step S4, obtain the total algae quantity of the area to be inferred based on the total surface algae quantity and the total water body algae quantity.

[0063] The method provided by this application has the following beneficial effects, including: improving accuracy. By combining remote sensing data and environmental parameters, comprehensive inference of algae on the water surface and underwater is achieved, improving the estimation accuracy of the total amount of algae; reducing costs. The frequency and scope of on-site sampling are reduced, and the cost of water quality monitoring is lowered; enhancing performance. It can more comprehensively reflect the spatial distribution of algae in water bodies, especially in deep water areas where traditional methods are difficult to accurately estimate. Through the combination of multiple models, this invention provides more reliable results.

[0064] Exemplarily, the algae in this application include cyanobacteria. Taking cyanobacteria as an example, in order to infer the total amount of cyanobacteria in the water body, a method for inferring the total amount of cyanobacteria in the water body based on remote sensing and environmental parameters is obtained by combining remote sensing technology, environmental parameter monitoring, and machine learning algorithms. The schematic diagram of the method is as Figure 2 shown, including: acquisition and preprocessing of remote sensing data, acquisition of profile environmental parameters, extraction of remote sensing image features, acquisition of water surface environmental parameters, construction of an underwater cyanobacteria model for water bodies, construction of a surface cyanobacteria model for water bodies, application of the underwater cyanobacteria model for water bodies, application of the surface cyanobacteria model for water bodies, calculation of the total amount of cyanobacteria in the water body;

[0065] Among them, when acquiring and preprocessing remote sensing data, first obtain high-resolution satellite images of the water body area, that is, the remote sensing images of the area to be inferred, through official websites or drones. The drone hyperspectral images are shown in Figures 3(a) and 3(b), and the Sentinel-2 satellite images are shown in Figures 4(a) and 4(b). According to the menu display in the pictures, the drone hyperspectral images contain more bands. These images should cover the entire monitoring area and contain the band information required for cyanobacteria detection, such as infrared bands and near-infrared bands, and corresponding preprocessing is performed.

[0066] Optionally, both drone hyperspectral images and Sentinel-2 satellite images have their own advantages and disadvantages in practical applications. The advantages of drone images include: high resolution, the resolution of drone images is usually higher than that of satellite images, which can provide more detailed surface information; real-time nature, drones can be quickly deployed to obtain real-time images, which is very beneficial for the rapid response of algal bloom monitoring; flexibility, drones can adjust the flight altitude and path according to needs to conduct detailed monitoring of specific areas; not affected by weather, drones are usually not affected by clouds and can work under cloud cover, while satellite images may be affected by cloud occlusion. The disadvantages of drone images include: small observation range, the flight range of drones is limited and not suitable for large-scale monitoring; relatively high cost, compared with satellite images, the operating cost of drones is higher, especially in the case of frequent flights; complex data processing, the amount of data obtained by drones is large and requires professional data processing technology. The advantages of satellite images include: large-scale monitoring, satellite images can cover large areas and are suitable for macroscopic monitoring of lake algal blooms; periodic monitoring, satellites have a fixed revisit period and can conduct periodic monitoring; low data cost, the acquisition and update costs of satellite images are low, which is suitable for long-term and large-scale monitoring needs; global coverage, the Sentinel-2 satellite has global coverage capabilities and can be used for lake algal bloom monitoring on a global scale. The disadvantages of satellite images include: resolution limitation, the resolution of satellite images is usually lower than that of drone images and may not be able to capture small-scale algal bloom details; affected by clouds, satellite images may be affected by cloud occlusion, resulting in the inability to conduct monitoring under cloud cover; relatively weak timeliness, the acquisition of satellite images is not as fast as that of drone images and may not be able to reflect the dynamic changes of algal blooms in a timely manner. In summary, drone images and Sentinel-2 satellite images have their own advantages and limitations in lake algal bloom monitoring, and appropriate technical means can be selected according to monitoring requirements and conditions in practical applications.

[0067] Furthermore, remote sensing image features are spectral information related to water body characteristics extracted from remote sensing images, such as reflectance in different bands. These features can include the Normalized Difference Chlorophyll Index (NDCI), Modified Normalized Difference Vegetation Index (MNDVI), Floating Algae Index (FAI), or other custom spectral indices. In remote sensing data processing, the selection of bands and calculation processing for matching the actual chlorophyll concentration (especially cyanobacteria concentration) is crucial. Chlorophyll a in cyanobacteria is an important indicator of the cyanobacteria biomass in water bodies, and the reflectance characteristics of remote sensing can reveal the concentration of chlorophyll in water bodies.

[0068] NDCI is a commonly used remote sensing index for estimating chlorophyll concentration in water bodies and can also be used to monitor algal blooms including cyanobacteria. It combines the reflectance of the red light (strong absorption band) and the near-infrared band (strong reflection band). The advantage is that the NIR (Near Infrared) band is sensitive to the scattering characteristics of cyanobacteria, which can identify high concentrations of algae in water bodies. The red light band has an absorption peak of chlorophyll a near 665 nm, which can effectively detect chlorophyll concentration. Therefore, in this application, NDCI is used as the remote sensing image feature, called Chla, to match the measured chlorophyll data. To ensure the accurate detection of cyanobacteria, the present invention sets an appropriate NDCI threshold for distinguishing the reflection characteristics of cyanobacteria from other algae. Specifically, an appropriate NDCI threshold can be set according to field verification data and historical data. For example, in a specific implementation, an NDCI value greater than 0.2 can be used as a preliminary indication of the presence of cyanobacteria. In addition, the threshold can be dynamically adjusted by combining the water body environment and seasonal changes to improve the accuracy and robustness of cyanobacteria detection.

[0069] Furthermore, the acquisition of profile environmental parameters and the acquisition of water surface environmental parameters are carried out at multiple monitoring points in the study area, where the profile environmental parameters are the water body profile parameter set; water samples at different depths are collected, that is, at preset water depth positions, a multi-parameter water quality monitor is used to measure the cyanobacteria concentration, water temperature, transparency, dissolved oxygen, and the concentrations of nitrogen and phosphorus.

[0070] Exemplarily, it is necessary to measure the cyanobacteria concentration at each depth to provide data support for subsequent model construction. The selected area range should be large, and the sampling points should cover representative positions throughout the study area, including different water depths, different geographical locations, and areas with obvious changes in water body characteristics. This can ensure that the collected data can reflect the cyanobacteria concentration in the entire water body and its influencing factors. When determining the measurement depth interval, it is necessary to consider the vertical distribution characteristics of cyanobacteria in the water body to ensure that each depth layer is accurately sampled while avoiding double counting. The measurement interval should meet the following conditions, including vertical heterogeneity and measurement accuracy. Vertical heterogeneity ensures that changes in the water depth layer can be captured, and the measurement accuracy interval should not be too large to avoid missing local concentration changes.

[0071] Optionally, the recommended measurement intervals for different water depth conditions include:

[0072] For a water depth of 1 meter, the measurement interval is 0.2 meters. This smaller interval for shallow water bodies ensures that the detailed vertical distribution is captured;

[0073] For a water depth of 2 meters, the measurement interval is 0.5 meters, which is suitable for medium water depths and balances measurement accuracy and cost;

[0074] The water depth is 3 meters, and the measurement interval is from 0.5 to 1 meter. The interval can be adjusted to meet the actual needs to ensure that the entire depth of the water body is covered.

[0075] In an exemplary embodiment, in step S2, at least one machine learning algorithm is used to train each water surface algae density prediction network structure with the remote sensing image features of the water surface in the area to be inferred as input features and the preset water surface chlorophyll concentration at each sampling pixel position as target variables. By comparing the output results of each water surface algae density prediction network structure, it is determined that the water surface algae density prediction model includes:

[0076] Prepare a water surface algae density prediction data set, including the remote sensing image features of the water surface in the area to be inferred and the preset water surface chlorophyll concentration at each sampling pixel position in the remote sensing image;

[0077] Divide the water surface algae density prediction data set into a training set and a test set;

[0078] Based on the training set and the test set, use at least one machine learning algorithm to construct each water surface algae density prediction network structure;

[0079] Obtain the output results of each water surface algae density prediction network structure, where the output results include the target parameters of each water surface algae density prediction network structure;

[0080] Determine the water surface algae density prediction model by comparing the target parameters of each water surface algae density prediction network structure.

[0081] Exemplarily, construct a water surface algae density prediction model. Taking cyanobacteria as an example, first extract the chlorophyll a concentration value or the reflectance of a specific band of each pixel in the remote sensing image, which is the remote sensing image feature NDCI of the water surface, that is, Chla. By matching with the surface cyanobacteria concentration data collected in the field, where the measured data is the cyanobacteria concentration measured by a multi-parameter water quality monitor, that is, the preset water surface chlorophyll concentration and the water body profile parameter set at each sampling pixel position in the remote sensing image, a relationship model between color or Chla value and cyanobacteria density is established through regression analysis, which is the water surface algae density prediction model. Further, in order to better fit the relationship between spectral features and cyanobacteria density, this application uses a machine learning algorithm. Optionally, the machine learning algorithm includes random forest and support vector machine, and combines their respective advantages to achieve accurate prediction of cyanobacteria concentration. When the data is complex and has features, machine learning algorithms such as random forest and support vector machine can handle such problems more effectively. The advantages of machine learning algorithms (such as random forest or support vector machine) are that when the data is complex and the relationship between variables is non-linear, machine learning algorithms (such as random forest or support vector machine) can better capture complex patterns and improve the prediction ability of the model. These algorithms can handle non-linear relationships and have high robustness to noise.

[0082] Specifically, prepare the water surface algae density prediction dataset: Chla(NDVI) and NIRReflectance (near-infrared reflectance) are used as input features X, the measured water surface chlorophyll concentration at the corresponding pixels is used as the target variable y, and Chla(NDVI) and NIRReflectance (near-infrared reflectance) corresponding to other pixels of the water body are used as input features X_new, that is, the remote sensing image features of the water surface in the area to be inferred are used as input features, and the water surface chlorophyll concentration at the preset sampling pixel positions is used as the target variable; the procedure is as follows:

[0083] Let df.csv be a DataFrame containing features and target variables.

[0084] Read the CSV file

[0085] df = pd.read_csv('df.csv')

[0086] Extract features and target variables

[0087] X = df[['Chla(NDVI)', 'NIR_Reflectance']]

[0088] y = df['Chla_concentration']

[0089] Let X_new.csv be a DataFrame containing the remaining features.

[0090] Read the X_new.csv file

[0091] X_new = pd.read_csv('X_new.csv')

[0092] Furthermore, data splitting: Split the water surface algae density prediction dataset into a training set and a test set, usually in the ratio of 80% training set and 20% test set; the procedure is as follows:

[0093] X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 42)

[0094] Furthermore, create a model, that is, construct each water surface algae density prediction network structure using at least one machine learning algorithm based on the training set and the test set:

[0095] Random forest regression model

[0096] rf_model = RandomForestRegressor(n_estimators = 100, random_state = 42)

[0097] rf_model.fit(X_train, y_train)

[0098] Support vector machine regression model

[0099] svm_model = SVR(kernel = 'rbf')

[0100] svm_model.fit(X_train, y_train)

[0101] Furthermore, prediction and evaluation are carried out, that is, the output results of each water surface algae density prediction network structure are obtained, where the output results include the target parameters of each water surface algae density prediction network structure, and the water surface algae density prediction model is determined by comparing the target parameters of each water surface algae density prediction network structure:

[0102] Prediction result

[0103] rf_predictions = rf_model.predict(X_test)

[0104] svm_predictions = svm_model.predict(X_test)

[0105] Calculate MSE and R2

[0106] rf_mse = mean_squared_error(y_test, rf_predictions)

[0107] rf_r2 = r2_score(y_test, rf_predictions)

[0108] svm_mse = mean_squared_error(y_test, svm_predictions)

[0109] svm_r2 = r2_score(y_test, svm_predictions)

[0110] print(f"Random Forest MSE: {rf_mse}, R 2 : {rf_r2}")

[0111] print(f"SVM MSE: {svm_mse}, R2 :{svm_r2}")

[0112] Further, select the best model, that is, determine the water surface algal density prediction model by comparing the target parameters of each water surface algal density prediction network structure:

[0113] By comparing the output MSE, R 2 value, the accuracy of the model can be judged. The smaller the MSE, the smaller the prediction error of the model, and the higher the accuracy of the model. R 2 ranges from [0, 1]. The closer the value is to 1, the better the prediction performance of the model, and the more data variability it can explain.

[0114] Further, apply the model to predict the cyanobacteria density:

[0115] Finally, apply the best relationship model best_model, that is, the water surface algal density prediction model, and use this model to predict each pixel of the remote sensing image to obtain the corresponding cyanobacteria concentration. Specifically, convert the color or Chla value of each pixel in the remote sensing image into the corresponding cyanobacteria density. The procedure is as follows:

[0116] new_pixel_data = X_new # This data contains the Chla and NIR Reflectance of each pixel

[0117] Predict the cyanobacteria concentration of each pixel, that is, use the water surface algal density prediction model to predict each pixel of the remote sensing image in the area to be inferred, and obtain the algal density of each pixel. The procedure is as follows:

[0118] predicted_chla = best_model.predict(new_pixel_data)

[0119] Further, calculate the pixel area through the resolution of the remote sensing data:

[0120] According to the spatial resolution of the remote sensing image, calculate the actual ground area covered by each pixel, that is, calculate the actual area covered by each pixel according to the spatial resolution of the remote sensing image in the area to be inferred. The calculation formula for the pixel area is: Area = resolution × resolution. Finally, use the cyanobacteria density of the pixel and the calculated pixel area to calculate the total amount of cyanobacteria contained in each pixel, that is, obtain the total amount of water surface algae in the area to be inferred through the algal density of each pixel and the actual area covered by each pixel. Exemplarily, the formula is: Total cyanobacteria in pixel = cyanobacteria density × pixel area.

[0121] In an exemplary embodiment, before step S3, it further includes

[0122] The pixels of the remote sensing image of the area to be inferred are divided into different sub-areas by a spatial clustering algorithm, and a set of water body profile parameters for each sub-area is obtained by a water quality monitor.

[0123] Exemplarily, constructing an underwater cyanobacteria model for the water body includes: In the water body, parameters of the water body (such as water depth, water temperature, dissolved oxygen, nutrient concentration, etc.) may vary significantly spatially. Simply considering the entire water body as a unified area to calculate the chlorophyll concentration or the total amount of cyanobacteria may lead to inaccurate calculation results. At the same time, there is also a problem of excessive parameter differences, that is, the environmental parameter differences in different areas may be relatively large, which will affect the prediction of the chlorophyll concentration C (z) and further affect the calculation of the total amount of cyanobacteria M. Therefore, it is necessary to divide the area to ensure that the environmental conditions within each area are relatively uniform to reduce the prediction error. Optionally, to improve the calculation accuracy, the entire water body can be divided into multiple smaller areas. The environmental conditions (such as water temperature, water depth, dissolved oxygen, transparency, etc.) within these areas are kept as consistent as possible, so that it becomes more reliable to use a relatively simple model to calculate the chlorophyll concentration C (z) in each area. Further, the pixels are divided into several areas according to the chlorophyll concentration and water body factors, and all the pixels in the lake are divided into different areas using a spatial clustering algorithm according to the magnitude of the chlorophyll concentration and the adjacent position relationship. For each area, its average chlorophyll concentration is calculated, where the spatial clustering algorithm includes K-means or DBSCAN (Density-Based Spatial Clustering of Applications with Noise), and different sub-areas are denoted as M 1 、M 2 、M 3 … Finally, the total amount of algae M calculated in all sub-areas is integrated and summed to obtain the total amount of algae in the entire water body.

[0124] In an exemplary embodiment, obtaining a set of water body profile parameters for each sub-area by a water quality monitor includes:

[0125] Obtaining a set of water body profile parameters for each preset sampling pixel position in each area by a water quality monitor; wherein the set of water body environmental parameters of the area to be inferred in the set of water body profile parameters includes algae concentration, water depth, water temperature, dissolved oxygen, nitrogen concentration, phosphorus concentration, organic matter content, transparency.

[0126] Specifically, using the previously obtained cyanobacteria concentration profile data at different depths and the processed remote sensing data, that is, using the algae concentration profile data at different depths and the processed remote sensing data, and using the different areas divided for all pixels, that is, areas M 1 、M 2 、M 3... measure the water body profile parameter set with a multi-parameter water quality monitor, including algae concentration (cyanobacteria concentration), water depth, water temperature, dissolved oxygen, nitrogen concentration, phosphorus concentration, organic matter content, and transparency.

[0127] In an exemplary embodiment, in step S3, at least one regression model is used to train each water body algae density prediction network structure with the water body environment parameter set of the area to be inferred as the input feature and the chlorophyll concentration at the preset water depth position as the target variable. Comparing the output results of each water body algae density prediction network structure to determine the water body algae density prediction model includes:

[0128] Prepare the water body algae density prediction data set, including the water body environment parameter set of the area to be inferred and the chlorophyll concentration at the preset water depth position;

[0129] Divide the water body algae density prediction data set into a training set and a test set;

[0130] Based on the training set and the test set, use at least one regression model to construct each water body algae density prediction network structure;

[0131] Obtain the output results of each water body algae density prediction network structure, where the output results include the target parameters of each water body algae density prediction network structure;

[0132] Determine the water body algae density prediction model by comparing the target parameters of each water body algae density prediction network structure.

[0133] Exemplarily, taking cyanobacteria as an example, due to various factors, the cyanobacteria distribution in the water body profile is usually in a non-linear distribution state. Although machine learning algorithms (such as random forest, support vector machine, etc.) can capture this complex relationship more effectively, the corresponding feature formulas cannot be listed, which is not conducive to subsequent calculations. Therefore, models such as polynomial regression model and Gaussian regression model can be used.

[0134] Taking the polynomial regression model as an example below. Apply the cyanobacteria concentration of each pixel obtained from the surface cyanobacteria model, that is, the water surface algae density prediction model, to the underwater cyanobacteria model, that is, the water body algae density prediction model. The currently obtained model C (z) is derived from the obtained data set, and the water depth, water temperature, dissolved oxygen, nitrogen concentration, phosphorus concentration, organic matter content, and transparency of the water body profile in other areas are not clear. Therefore, it is necessary to convert the original model into a simplified model that is only related to the water depth, train a multi-variable regression model, extract the regression coefficient of the water depth, and regard the coefficients of other features as constants, such as the mean value in the data set or a fixed reference value, so as to obtain a single-variable regression model, that is, only the prediction of the water depth on the chlorophyll concentration.

[0135] When there is a lake dataset, that is, the water body profile parameters set includes water depth, nitrogen concentration, phosphorus concentration, organic matter content, dissolved oxygen, water temperature, transparency, and the measured chlorophyll concentration at the corresponding depth, determining the water surface algae density prediction model includes:

[0136] Data preparation and processing, that is, preparing the water body algae density prediction dataset. Optionally, taking the profile parameters obtained in one operation of the water quality detector as an example, including water depth, nitrogen concentration, phosphorus concentration, organic matter content, dissolved oxygen, water temperature, transparency, etc., the water body environment parameter set of the area to be inferred as the input feature X, and the measured chlorophyll concentration at the corresponding depth, that is, the chlorophyll concentration at the preset water depth position as the target variable y, taking the profile parameters obtained in one operation of the water quality detector as an example.

[0137] To distinguish from the df.csv file in the above embodiments, let dF.csv be a DataFrame containing features and target variables.

[0138] Read the CSV file

[0139] dF = pd.read_csv('df.csv')

[0140] Extract features and target variables

[0141] X = dF[['depth', 'nitrogen_concentration', 'phosphorus_concentration', 'organic_content', 'dissolved_oxygen', 'temperature', 'transparency']]

[0142] y = dF['Chla_concentration']

[0143] Replace other features with constants (means)

[0144] X_mean = X.mean() to obtain the mean of each feature

[0145] X_replaced = X.copy()

[0146] Replace other features except water depth with their means

[0147] for feature in X.columns:

[0148] if feature!='depth':

[0149] X_replaced[feature] = X_mean[feature]

[0150] Further, data splitting: Split the water body algae density prediction dataset into a training set and a test set, usually in the ratio of 80% training set and 20% test set.

[0151] X_train, X_test, y_train, y_test = train_test_split(X_replaced[['depth']], y, test_size = 0.2, random_state = 42)

[0152] Further, model construction, that is, based on the training set and the test set, construct the prediction network structure of the algae density of each water body using at least one regression model:

[0153] Select polynomial models of different orders

[0154] degrees = [1, 2, 3, 4, 5, 6]

[0155] best_model = None

[0156] best_mse = float('inf')

[0157] best_r2 = -float('inf')

[0158] best_degree = 0

[0159] Train and evaluate for each polynomial order

[0160] for degree in degrees:

[0161] Generate polynomial features of water depth

[0162] poly = PolynomialFeatures(degree = degree)

[0163] X_poly_train = poly.fit_transform(X_train)

[0164] X_poly_test = poly.transform(X_test)

[0165] model = LinearRegression().fit(X_poly_train, y_train)

[0166] Prediction

[0167] y_pred = model.predict(X_poly_test)

[0168] Further, select the best order, that is, determine the water body algal density prediction model by comparing the target parameters of each water body algal density prediction network structure:

[0169] Calculate MSE and R2

[0170] mse = mean_squared_error(y_test, y_pred)

[0171] r2 = r2_score(y_test, y_pred)

[0172] Output MSE and R2

[0173] print(f"Degree{degree}: MSE = {mse}, R 2 = {r2}")

[0174] # Determine whether it is the best model

[0175]

[0176]

[0177] Further, use the polynomial regression model with the simplified best order, including:

[0178] print(f"BestModel: Degree{best_degree}, MSE: {best_mse}, R 2 : {best_r2}")

[0179] Obtain the regression coefficients of the best model

[0180] coefficients = best_model.coef_

[0181] intercept = best_model.intercept_

[0182] Output the simplified regression formula

[0183]

[0184] Further, run the program to obtain the simplified best polynomial equation, including:

[0185] The model here is the above-mentioned polynomial regression model, that is, the water body algal density prediction model.

[0186] When the result is a linear term: C (z) = β + α 1 × z;

[0187] When the result is a quadratic term: C (z) = β + α 1 × z + α 2 × z 2 ;

[0188] When the result is a cubic term: C (z) = β + α 1 × z + α 2 × z 2 + α 3 × z 3 ;

[0189] When the result is an nth term: C (z) = β + α 1 × z + α 2 × z 2 + α 3 × z 3 + … + α n × z n ;

[0190] Where C (z) represents the chlorophyll concentration corresponding to different water depths; n is the degree of the above optimal polynomial; α is the regression coefficient related to the variable z; β is the constant obtained by replacing all features other than water depth, serving as the intercept term.

[0191] Exemplarily, the cyanobacteria concentration at different depths in the water column of each pixel is calculated, that is, the algae density of the water column of each pixel in the remote sensing image of the area to be inferred is obtained through the water body algae density prediction model, and the total amount of water body algae is obtained. Subsequently, by combining the results of the surface and underwater models, the total amount of cyanobacteria in the water body can be effectively estimated, that is, the total amount of algae in the area to be inferred is obtained based on the total amount of surface algae and the total amount of water body algae.

[0192] Among them, the mean square error (MSE) and the coefficient of determination (R 2 ) are used to evaluate the performance of the model. If the MSE and R 2 values of the measured water body model and the data are not ideal, by comparing with other regression models and comparing the output MSE and R 2 values, the accuracy of the model can be judged, and the best model can be selected. The smaller the MSE, the smaller the prediction error of the model and the higher the accuracy of the model. The value range of R 2 is [0, 1]. The closer the value is to 1, the better the prediction performance of the model and the more data variability it can explain.

[0193] In an exemplary embodiment, step S4 includes:

[0194] Integrate each pixel's water column based on the total amount of surface algae and the total amount of water column algae to obtain the total amount of algae for each pixel, and obtain the total amount of algae for each sub-region based on the total amount of algae for each pixel. The sum of the total amounts of algae for each sub-region is the total amount of algae for the area to be inferred.

[0195] Exemplarily, still taking cyanobacteria as an example, the specific calculation of cyanobacteria total amount estimation is as follows: By integrating the entire water column, calculate the total amount of cyanobacteria for each pixel, then calculate the total amount of cyanobacteria for each region, and finally add up the total amounts of cyanobacteria in different regions of the water body to obtain the estimation of the total amount of cyanobacteria in the entire region. Specifically, as shown in formula (1):

[0196]

[0197] M is the total amount of cyanobacteria; M j Different regions (M 1 、M 2 、M 3 …) into which the water body is divided according to pixels; k is the number of regions; n represents the number of pixels, that is, the total number of pixels in the entire area to be inferred for research; represents the sum of all pixels. By estimating the total amount of cyanobacteria in each pixel and adding up the total amounts of all pixels, the total amount of cyanobacteria in the entire research area can be obtained; h represents the maximum depth of the water column; C (z) is the underwater cyanobacteria model of the water body, that is, the water body algae density prediction model; C (z) dz represents the variation function of cyanobacteria concentration with depth z, represents the cyanobacteria concentration integration from the water surface (z = 0) to the bottom (z = h) in the water column; S is the total amount of cyanobacteria in the pixel area.

[0198] In summary, by reasonably selecting and combining different models, the present invention can not only accurately predict the distribution of surface cyanobacteria, but also estimate the total amount of cyanobacteria in the entire water column. The use of multiple linear regression and machine learning algorithms enables the model to handle complex data relationships. This comprehensive method significantly improves the accuracy and reliability of cyanobacteria total amount estimation.

[0199] Those skilled in the art can easily understand that the above are only preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included within the protection scope of the present invention.

[0200] As described above, at least some of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages can be executed at different times, or can be executed alternately or in turn with at least some of the steps or stages in other steps or other steps.

[0201] Based on the same inventive concept, an embodiment of the present application further provides a device for inferring the total amount of water body algae based on remote sensing and environmental parameters for implementing the method for inferring the total amount of water body algae based on remote sensing and environmental parameters involved above. The solution provided by this device for solving problems is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the device for inferring the total amount of water body algae based on remote sensing and environmental parameters provided below can refer to the limitations on the method for inferring the total amount of water body algae based on remote sensing and environmental parameters in the foregoing, and will not be repeated here.

[0202] In an exemplary embodiment, as Figure 5 shown, a device for inferring the total amount of water body algae based on remote sensing and environmental parameters is provided, including:

[0203] An image data acquisition module 501, configured to acquire a remote sensing image of the area to be inferred, as well as the water surface chlorophyll concentration and the water body profile parameter set at each preset sampling pixel position in the remote sensing image, where the water body profile parameter set includes the water body environmental parameter set of the area to be inferred and the chlorophyll concentration at the preset water depth position, and acquire the remote sensing image features of the water surface based on the remote sensing image of the area to be inferred;

[0204] A water surface algae density prediction module 502, configured to use at least one machine learning algorithm to train each water surface algae density prediction network structure with the remote sensing image features of the water surface of the area to be inferred as input features and the water surface chlorophyll concentration at each preset sampling pixel position as target variables, compare the output results of each water surface algae density prediction network structure to determine the water surface algae density prediction model, and obtain the algae density of each pixel of the remote sensing image of the area to be inferred through the water surface algae density prediction model, so as to obtain the total amount of water surface algae;

[0205] A water body algae density prediction module 503, configured to use at least one regression model to train each water body algae density prediction network structure with the water body environmental parameter set of the area to be inferred as input features and the chlorophyll concentration at the preset water depth position as target variables, compare the output results of each water body algae density prediction network structure to determine the water body algae density prediction model, and obtain the algae density of each pixel water column of the remote sensing image of the area to be inferred through the water body algae density prediction model, so as to obtain the total amount of water body algae;

[0206] An algae total amount calculation module 504, configured to obtain the total amount of algae in the area to be inferred based on the total amount of water surface algae and the total amount of water body algae.

[0207] Each module in the above-mentioned device for inferring the total amount of water body algae based on remote sensing and environmental parameters can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0208] In one embodiment, a computer device is further provided, which includes a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0209] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0210] For those of ordinary skill in the art, without departing from the concept of the present application, several deformations and improvements can be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims. It is easy for those skilled in the art to understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for estimating the total amount of algae in a water body based on remote sensing and environmental parameters, characterized in that: The method is used to infer the total amount of algae in a region to be inferred, and includes: Step S1, obtaining a remote sensing image of the area to be inferred, and a water surface chlorophyll concentration and a water body profile parameter set at each preset sampling pixel position in the remote sensing image, wherein the water body profile parameter set includes a water body environment parameter set of the area to be inferred, and a chlorophyll concentration at a preset water body depth position, and obtaining remote sensing image features of the water body surface based on the remote sensing image of the area to be inferred; Step S2, using at least one machine learning algorithm to train a network structure for predicting algae density on each water surface by taking the remote sensing image features of the water body surface of the area to be inferred as input features and the water surface chlorophyll concentration at each sampling pixel position as the target variable, comparing the output results of the network structure for predicting algae density on each water surface, determining a water surface algae density prediction model, and obtaining the algae density of each pixel of the remote sensing image of the area to be inferred through the water surface algae density prediction model, and obtaining the total amount of algae on the water surface; Step S3, using at least one regression model to take the water environment parameter set of the area to be inferred as input features and the chlorophyll concentration at the preset water depth position as the target variable to train and obtain the algae density prediction network structure of each water body, comparing the output results of the algae density prediction network structure of each water body to determine the water body algae density prediction model, and obtaining the algae density of each pixel water column of the remote sensing image of the area to be inferred through the water body algae density prediction model, and obtaining the total amount of water body algae; Step S4, deriving the total amount of algae in the area to be inferred based on the total amount of algae on the water surface and the total amount of algae in the water body.

2. The method according to claim 1, characterized in that The step S3 also includes The pixels of the remote sensing image of the area to be inferred are divided into different sub-areas using a spatial clustering algorithm, and the water body profile parameter set of each sub-area is obtained through a water quality monitoring instrument.

3. The method according to claim 2, characterized in that The water body profile parameter set of each sub-area obtained by the water quality monitor includes: The water quality monitoring instrument is used to obtain the water body profile parameter set of each preset sampling pixel position in each area; the water body environment parameter set of the area to be inferred in the water body profile parameter set includes algae concentration, water depth, water temperature, dissolved oxygen, nitrogen concentration, phosphorus concentration, organic matter content, and transparency.

4. The method according to claim 3, characterized in that In step S2, at least one machine learning algorithm is used to train the network structure for predicting algae density on each water surface by using the remote sensing image features of the water surface in the area to be inferred as input features and the water surface chlorophyll concentration at each sampling pixel position as the target variable, and the output results of the network structure for predicting algae density on each water surface are compared to determine the water surface algae density prediction model, including: Prepare a water surface algae density prediction data set, including the remote sensing image features of the water surface in the area to be inferred and the water surface chlorophyll concentration at each preset sampling pixel position in the remote sensing image; The water surface algae density prediction dataset is divided into a training set and a test set; Based on the training set and the test set, at least one machine learning algorithm is used to construct a network structure for predicting the algae density on each water surface; Obtaining output results of the network structure for predicting algae density on each water surface, wherein the output results include target parameters of the network structure for predicting algae density on each water surface; The water surface algae density prediction model is determined by comparing the target parameters of each water surface algae density prediction network structure.

5. The method according to claim 4, characterized in that In step S2, the algae density of each pixel of the remote sensing image of the area to be inferred is obtained by using the water surface algae density prediction model, and the total amount of water surface algae is obtained, which includes: The water surface algae density prediction model is used to predict each pixel of the remote sensing image of the inference area to obtain the algae density of each pixel; Calculate the actual area covered by each pixel based on the spatial resolution of the remote sensing image of the area to be inferred; The total amount of algae on the water surface of the area to be inferred is obtained through the algae density of each pixel and the actual area covered by each pixel.

6. The method according to claim 5, characterized in that In step S3, at least one regression model is used to train the water environment parameter set of the area to be inferred as the input feature and the chlorophyll concentration at the preset water depth position as the target variable to obtain the prediction network structure of the algae density of each water body, and the output results of the algae density prediction network structure of each water body are compared to determine the water body algae density prediction model, including: Prepare a water body algae density prediction data set, including a water body environmental parameter set for the area to be inferred and chlorophyll concentration at a preset water body depth; The water body algae density prediction dataset is divided into a training set and a test set; Based on the training set and the test set, at least one regression model is used to construct a network structure for predicting the algae density of each water body; Obtaining output results of the network structure for predicting the density of algae in each water body, wherein the output results include target parameters of the network structure for predicting the density of algae in each water body; The water body algae density prediction model is determined by comparing the target parameters of the algae density prediction network structure of each water body.

7. The method according to claim 6, characterized in that The step S4 comprises: Based on the total amount of algae on the water surface and the total amount of algae in the water body, the water column of each pixel is integrated to obtain the total amount of algae in each pixel, and the total amount of algae in each sub-area is obtained based on the total amount of algae in each pixel. The sum of the total amount of algae in each sub-area is the total amount of algae in the area to be inferred.

8. A device for estimating the total amount of algae in a water body based on remote sensing and environmental parameters, characterized in that: The device comprises: An image data acquisition module is used to acquire a remote sensing image of the area to be inferred, and a water surface chlorophyll concentration and a water body profile parameter set at each preset sampling pixel position in the remote sensing image, wherein the water body profile parameter set includes a water body environment parameter set of the area to be inferred, and a chlorophyll concentration at a preset water body depth position, and acquire remote sensing image features of the water body surface based on the remote sensing image of the area to be inferred; The water surface algae density prediction module is used to use at least one machine learning algorithm to use the remote sensing image features of the water body surface of the area to be inferred as input features and the water surface chlorophyll concentration of each sampling pixel position as the target variable to train and obtain the network structure for predicting the density of algae on each water surface, compare the output results of the network structure for predicting the density of algae on each water surface, determine the water surface algae density prediction model, and obtain the algae density of each pixel of the remote sensing image of the area to be inferred through the water surface algae density prediction model, so as to obtain the total amount of algae on the water surface; A water body algae density prediction module is used to use at least one regression model to take the water body environmental parameter set of the area to be inferred as input features and the chlorophyll concentration at the preset water body depth position as the target variable to train and obtain the prediction network structure of each water body algae density, compare the output results of each water body algae density prediction network structure to determine the water body algae density prediction model, and obtain the algae density of each pixel water column of the remote sensing image of the area to be inferred through the water body algae density prediction model, and obtain the total amount of water body algae; The algae total amount calculation module is used to obtain the total amount of algae in the area to be inferred based on the total amount of algae on the water surface and the total amount of algae in the water body.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images