Water body monitoring method and equipment based on satellite remote sensing data and storage medium

Abstract

The invention provides a water body monitoring method and equipment based on satellite remote sensing data and a storage medium, and relates to the technical field of water body monitoring, the method comprises the following steps: firstly, acquiring remote sensing image data and water quality measured data (including water quality parameters and geographic positions) of a target area at least covering two spectral bands; performing radiation correction on the remote sensing image data, and extracting target image data corresponding to the water body by using the image segmentation model; matching the water quality parameters with the target remote sensing image data according to the geographic position information, and constructing a training set of time-space alignment; using a multilayer model fusion strategy to train a preset water quality inversion model based on the training set to obtain a target model; and finally, inputting an image to be detected to the target model to obtain water quality parameters, and mapping the water quality parameters to a geographic space to generate a water quality two-dimensional distribution diagram. According to the invention, the water body monitoring precision and monitoring details based on satellite remote sensing data can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of water body monitoring, and in particular, to a method, device, and storage medium for water body monitoring based on satellite remote sensing data. Background Art

[0002] Currently, the method of using satellite remote sensing technology for water body monitoring is relatively common. It mainly uses the remote sensing image data of conventional satellites, combines basic band combinations such as red, green, and blue, and establishes an inversion model for water quality parameters. This method is technically mature and has a wide coverage, and is suitable for roughly evaluating the conventional water quality of large lakes or rivers.

[0003] However, the above technologies have obvious shortcomings in practical applications. On the one hand, the band design of ordinary satellites is relatively broad, making the spectral signals of algae and turbid substances easy to be mixed, resulting in distorted monitoring results; on the other hand, most of the inversion models in related technologies adopt a single algorithm (such as linear regression, simple neural network), which can only be trained for a single problem. In the face of complex situations where multiple pollutions coexist, it is difficult to accurately monitor the water body. Summary of the Invention

[0004] The problem solved by the present invention is how to improve the accuracy of water body monitoring based on satellite remote sensing data.

[0005] To solve the above problems, the present invention provides a method, device, and storage medium for water body monitoring based on satellite remote sensing data.

[0006] In a first aspect, the present invention provides a method for water body monitoring based on satellite remote sensing data, including:

[0007] Obtaining remote sensing image data of a target area and measured water quality data of the water body area in the target area, where the remote sensing image data includes at least two spectral bands adapted to the reflectance characteristics of the water body, and the measured water quality data includes water quality parameter data and geographical location data;

[0008] Performing radiometric correction processing on the remote sensing image data, and extracting target remote sensing image data corresponding to the water body area in the radiometrically corrected remote sensing image data through an image segmentation model;

[0009] Matching the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data to construct a spatiotemporally aligned training data set;

[0010] Training a preset water quality inversion model based on the training data set to obtain a target water quality inversion model, where the preset water quality inversion model is constructed based on a multi-layer model fusion strategy;

[0011] Input the image data to be measured into the target water quality inversion model to obtain the water quality parameter data of the area to be measured, and map the water quality parameter data of the area to be measured to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

[0012] Optionally, the remote sensing image data includes at least 4 spectral bands with a central wavelength in the range of 410 nanometers to 590 nanometers, the bandwidth of each spectral band is 20 nanometers to 30 nanometers, the water quality parameter data includes chlorophyll concentration, suspended solid concentration and solvent organic matter content, and the geographical location data includes the position coordinates in the geodetic coordinate system.

[0013] Optionally, the radiation correction process for the remote sensing image data and the extraction of the target remote sensing image data corresponding to the water body area from the radiometrically corrected remote sensing image data through an image segmentation model include:

[0014] Perform atmospheric correction and sensor radiometric calibration on the remote sensing image data;

[0015] Input the remote sensing image data after the atmospheric correction and the sensor radiometric calibration into the input layer of the image segmentation model, and perform feature extraction on the remote sensing image data through a convolutional layer, a batch normalization layer, a multi-scale pooling layer and an activation function in sequence to obtain a first multi-scale feature map;

[0016] Process the multi-scale feature map based on the dual-channel attention module of the image segmentation model to obtain a second multi-scale feature map;

[0017] Perform upsampling on the second multi-scale feature map through transposed convolution and combine padding through the convolutional layer to output a binary segmentation probability map of the water body area;

[0018] Perform threshold segmentation on the binary segmentation probability map, and map the segmentation result to the image coordinate system where the geographical location data is located based on the geographic coordinate transformation matrix to generate the target remote sensing image data corresponding to the water body area.

[0019] Optionally, the matching of the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data to construct a spatio-temporally aligned training dataset includes:

[0020] Match the sampling point coordinates of the water quality parameter data with the pixel geographical coordinates in the target remote sensing image data according to the geographical location data;

[0021] Extract the characteristic band reflectance values of the target remote sensing image data, establish a fitting relationship between the concentration and the characteristic band reflectance values for each water quality parameter data, and generate a band water quality parameter concentration mapping table.

[0022] Perform spatio-temporal consistency screening based on the timestamp of the target remote sensing image data and the detection time of the water quality parameter data;

[0023] Associate and match the water quality parameter data, the pixel geographic coordinates, the timestamp, and the target remote sensing image data after spatio-temporal consistency screening in the band water quality parameter concentration mapping table to construct a spatio-temporally aligned training dataset.

[0024] Optionally, the performing spatio-temporal consistency screening based on the timestamp of the target remote sensing image data and the detection time of the water quality parameter data includes:

[0025] Compare the timestamp of the target remote sensing image data with the detection time of the water quality parameter data. If the difference between the timestamp and the detection time is greater than a preset threshold, then eliminate the corresponding target remote sensing image data and water quality parameter data.

[0026] Optionally, the preset water quality inversion model is an ensemble learning regression model. The base learners of the preset water quality inversion model include at least two of random forest, gradient boosting decision tree, support vector machine, and extreme gradient boosting algorithm. The meta-learner of the preset water quality inversion model includes a multiple linear regression algorithm.

[0027] Optionally, the inputting the to-be-detected image data into the target water quality inversion model to obtain the water quality parameter data of the to-be-detected area, and mapping the water quality parameter data of the to-be-detected area to the to-be-detected area to generate the two-dimensional water quality spatial distribution map of the to-be-detected area includes:

[0028] Calculate the sensitive band value of each pixel in the water body area according to the to-be-detected image data;

[0029] Perform feature engineering processing on the sensitive band value of each pixel;

[0030] Input the sensitive band value of each pixel after feature processing into the target water quality inversion model to generate the water quality parameter prediction value of each pixel;

[0031] Map the water quality parameter prediction value to the to-be-detected area to generate the two-dimensional water quality spatial distribution map of the to-be-detected area.

[0032] In a second aspect, the present invention provides a water body monitoring device based on satellite remote sensing data, including:

[0033] A data acquisition module for acquiring remote sensing image data of a target area and measured water quality data of a water area in the target area, where the remote sensing image data includes at least two spectral bands adapted to the water body reflectance characteristics, and the measured water quality data includes water quality parameter data and geographical location data;

[0034] A calibration and segmentation module for performing radiometric calibration processing on the remote sensing image data and extracting target remote sensing image data corresponding to the water area in the remote sensing image data through an image segmentation model;

[0035] A training dataset construction module for matching the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data to construct a spatiotemporally aligned training dataset;

[0036] A model training module for training a preset water quality inversion model based on the training dataset to obtain a target water quality inversion model, where the preset water quality inversion model is constructed based on a multi-layer model fusion strategy;

[0037] An output module for inputting the image data to be measured into the target water quality inversion model to obtain water quality parameter data of the area to be measured, and mapping the water quality parameter data of the area to be measured to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

[0038] In a third aspect, the present invention provides an electronic device, including a memory and a processor;

[0039] The memory is used for storing a computer program;

[0040] The processor is used for implementing the above-mentioned water body monitoring method based on satellite remote sensing data when executing the computer program.

[0041] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned water body monitoring method based on satellite remote sensing data is implemented.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows: By adopting at least two spectral bands adapted to the water body reflectivity characteristics, the optical characteristics of the water body can be specifically matched, thereby enhancing the extraction ability of water quality characteristics in remote sensing images and providing high-discrimination input data for subsequent water body monitoring; Subsequently, through radiometric correction processing of the remote sensing image data, atmospheric scattering and sensor noise interference can be eliminated, and the water body area can be accurately extracted through an image segmentation model, thereby effectively removing the interference of land and Yuncheng on the remote sensing image data and avoiding the misleading of non-water body data on water quality inversion; Based on geographical location data, matching water quality parameters with the coordinates and time attributes of remote sensing images, and constructing a spatio-temporal consistency training data set can solve the data deviation problem caused by spatial resolution differences and time asynchrony, ensuring the credibility of the spatio-temporal dimensions of the model input; Then, the water quality inversion model constructed based on the multi-layer fusion strategy comprehensively improves the inversion robustness of parameters such as chlorophyll and suspended solids in complex water body environments through the complementarity of different algorithms; Finally, mapping the inversion results to a two-dimensional water quality spatial distribution map can visually display the spatial gradient changes of the concentrations of various monitored substances in the water body. The present invention can effectively improve the water body monitoring accuracy and monitoring detail based on satellite remote sensing data. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 FIG. is a schematic flowchart of a water body monitoring method based on satellite remote sensing data provided by an embodiment of the present invention;

[0044] Figure 2 FIG. is a schematic flowchart of step S12 in a water body monitoring method based on satellite remote sensing data provided by an embodiment of the present invention;

[0045] Figure 3 FIG. is a schematic structural diagram of a water body monitoring device based on satellite remote sensing data provided by an embodiment of the present invention;

[0046] Figure 4 FIG. is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of specific embodiments of the present invention is provided in conjunction with the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments described herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not used to limit the protection scope of the present invention.

[0048] It should be understood that the various steps described in the method embodiments of the present invention can be executed in different orders and / or executed in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this regard.

[0049] As used herein, the term "comprising" and its variants are open-ended, that is, "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts such as "first" and "second" mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0050] It should be noted that the modifications of "one" and "multiple" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly specified in the context, it should be understood as "one or more".

[0051] Currently, the method of using satellite remote sensing technology for water body monitoring is relatively common. The conventional method relies on the combination of basic bands (such as red, green, and blue) to obtain image data, and based on this, an inversion model of water quality parameters is established. This method is suitable for rough evaluation of large-scale water areas. However, its deficiencies are as follows: First, the spectral bands designed by ordinary satellite sensors are relatively wide, resulting in the spectral signals specific to algae and turbid substances being mixed with each other, affecting the monitoring accuracy; second, most of the existing inversion models are constructed using a single algorithm, such as linear regression or simple neural network models. These models are usually designed for specific conditions or single types of water quality problems, and it is difficult to provide accurate monitoring results when facing complex water quality conditions with multiple pollutants coexisting. In short, the applicability and accuracy of traditional methods are limited in complex pollution situations and need to be improved to meet the needs of more complex water quality monitoring.

[0052] Based on this, the present invention provides a water body monitoring method, device and storage medium based on satellite remote sensing data.

[0053] Referring to Figure 1 , the present invention provides a water body monitoring method based on satellite remote sensing data, including:

[0054] S11. Obtain the remote sensing image data of the target area and the measured water quality data of the water body area in the target area. The remote sensing image data includes at least two spectral bands adapted to the water body reflectance characteristics, and the measured water quality data includes water quality parameter data and geographical location data.

[0055] Specifically, obtain the remote sensing image data of the target area through the satellite sensor of the remote sensing satellite. The remote sensing image data contains at least two spectral bands to adapt to the water body reflectance characteristics, so as to better identify and analyze the water body. At the same time, select several representative locations in the target area for on-site water quality measurement, and collect water quality parameter data including but not limited to turbidity, chlorophyll concentration, dissolved organic matter content, etc. and the corresponding geographical location data.

[0056] It should be noted that in order to improve the matching of the remote sensing image data and the measured water quality data and improve the accuracy of water body monitoring, the on-site water quality measurement and the remote sensing image data need to be spatio-temporally corresponding, that is, the time difference between the two needs to be controlled within a certain range; further, when the dates and coverage of the on-site water quality measurement and the remote sensing image data meet the requirements, the light conditions during the water quality measurement should also be ensured, that is, it is best to conduct the water quality measurement when there are no clouds over the target area and the water surface is not frozen, so as to avoid the influence of weather environment factors on the water body monitoring accuracy.

[0057] S12. Perform radiometric correction processing on the remote sensing image data, and extract the target remote sensing image data corresponding to the water body area in the radiometrically corrected remote sensing image data through an image segmentation model.

[0058] Perform radiometric correction processing on the obtained remote sensing image data to eliminate the influence of factors such as atmospheric absorption and scattering on the quality of the remote sensing image data, and ensure the accuracy of subsequent water body monitoring. After completing the radiometric correction, use an advanced image segmentation model (such as a semantic segmentation algorithm based on deep learning) to accurately extract the target remote sensing image data corresponding to the water body area from the corrected image.

[0059] S13. Match the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data, and construct a spatio-temporally aligned training dataset.

[0060] Use the geographical location information in the measured water quality data obtained in step S11 to spatially match the actually measured water quality parameter data with the target remote sensing image data in the same coordinate system. This process realizes the spatial position correspondence between the water quality parameter data and the remote sensing image, constructs a spatio-temporally aligned training dataset, and provides a basis for the training of subsequent water quality inversion models.

[0061] S14. Train a preset water quality inversion model based on the training data set to obtain a target water quality inversion model, where the preset water quality inversion model is constructed based on a multi-layer model fusion strategy.

[0062] Based on the training data set constructed in step S13, train a preset water quality inversion model. This preset model adopts a multi-layer model fusion strategy, which can comprehensively consider feature information at different levels, improving the generalization ability and prediction accuracy of the model. After multiple iterations of optimization, a target water quality inversion model that can accurately reflect the water quality status is finally obtained.

[0063] S15. Input the image data to be measured into the target water quality inversion model to obtain water quality parameter data for the area to be measured, and map the water quality parameter data of the area to be measured to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

[0064] When it is necessary to evaluate the water quality of the area to be measured, input the remote sensing image data corresponding to this area into the target water quality inversion model, and the model outputs the estimated values of the water quality parameters of this area. Subsequently, map these water quality parameters back to the corresponding positions in the original remote sensing image to generate a detailed two-dimensional spatial distribution map of the water quality. This not only facilitates an intuitive understanding of the water quality distribution in the entire monitoring area but also provides a scientific basis for further environmental management and pollution control.

[0065] Compared with the prior art, the beneficial effects of the present invention are as follows: By using at least two spectral bands adapted to the characteristics of water body reflectance, it is possible to specifically match the optical characteristics of the water body, thereby enhancing the extraction ability of water quality characteristics in remote sensing images and providing high-discrimination input data for subsequent water body monitoring; Subsequently, through radiometric correction processing of the remote sensing image data, it is possible to eliminate atmospheric scattering and sensor noise interference, and accurately extract the water body area through an image segmentation model, thereby effectively removing the interference of land and Yuncheng on the remote sensing image data and avoiding the misleading of non-water body data on water quality inversion; By matching water quality parameters with the coordinate and time attributes of remote sensing images based on geographical location data, a spatio-temporal consistent training data set is constructed, which can solve the data deviation problem caused by spatial resolution differences and time asynchrony, ensuring the credibility of the spatio-temporal dimensions of the model input; Then, the water quality inversion model constructed based on a multi-layer fusion strategy comprehensively improves the inversion robustness of parameters such as chlorophyll and suspended matter in a complex water body environment through the complementarity of different algorithms; Finally, mapping the inversion results to a two-dimensional spatial distribution map of water quality can visually display the spatial gradient changes of the concentrations of various monitored substances in the water body. The present invention can effectively improve the monitoring accuracy and detail of water bodies based on satellite remote sensing data.

[0066] In one embodiment, the remote sensing image data includes at least 4 spectral bands with central wavelengths in the range of 410 nanometers to 590 nanometers, the bandwidth of each spectral band is 20 nanometers to 30 nanometers, the water quality parameter data includes chlorophyll concentration, suspended sediment concentration, and dissolved organic matter content, and the geographical location data includes position coordinates in the geodetic coordinate system.

[0067] This embodiment further refines the spectral band characteristics of the remote sensing image data and the water quality parameter data, that is, the remote sensing image data selects at least four spectral bands with central wavelengths in the range of 410 nanometers to 590 nanometers. These bands are specifically selected to capture the reflectance characteristics of water bodies, which helps to more accurately analyze the water quality. The width of each selected spectral band can be set to 20 nanometers to 30 nanometers to ensure that the influence of different components on the reflection signal can be effectively distinguished. For example, four bands with central wavelengths of 410nm, 443nm, 490nm, and 555nm can be selected, which correspond to the blue to green regions respectively, and are conducive to detecting water quality indicators such as chlorophyll-a concentration and suspended sediment concentration.

[0068] Referring to Figure 2 , performing radiometric correction processing on the remote sensing image data, and extracting target remote sensing image data corresponding to the water body region in the radiometrically corrected remote sensing image data through an image segmentation model, including:

[0069] S121. Perform atmospheric correction and sensor radiometric calibration processing on the remote sensing image data.

[0070] First, perform atmospheric correction on the acquired remote sensing image data to eliminate the influence of effects such as atmospheric absorption and scattering, and ensure that the reflectance value accurately reflects the true situation of the ground surface. Then, perform sensor radiometric calibration processing to correct the response characteristics of the sensor itself and ensure data consistency between different times or different satellites. The processing in this embodiment enables the remote sensing image data to truly reflect the reflection characteristics of water bodies.

[0071] S122. Input the remote sensing image data after the atmospheric correction and the sensor radiometric calibration processing into the input layer of the image segmentation model, and sequentially perform feature extraction on the remote sensing image data through a convolutional layer, a batch normalization layer, a multi-scale pooling layer, and an activation function to obtain a first multi-scale feature map.

[0072] Specifically, the remotely sensed image data after atmospheric correction and sensor radiometric calibration processing is input into a pre-trained image segmentation model. The model first extracts the basic features of the image through a series of convolutional layers; then, uses batch normalization layers to stabilize the input distribution of each layer, accelerate the training process, and improve the model performance; next, adopts multi-scale pooling layers to capture the feature information at different scales; finally, uses activation functions (such as ReLU) to enhance the non-linear expression ability and generate the first multi-scale feature map with rich details.

[0073] S123. Process the multi-scale feature map through a dual-channel attention module based on the image segmentation model to obtain a second multi-scale feature map;

[0074] S124. Upsample the second multi-scale feature map through transposed convolution and combine padding through the convolutional layer to output a binary segmentation probability map of the water body area.

[0075] In this embodiment, transposed convolution technology is used to upsample the second multi-scale feature map, restoring the original resolution while retaining key features. At the same time, necessary padding operations are performed by combining convolutional layers to ensure the integrity of boundary information. Finally, the image segmentation model outputs a binary segmentation probability map representing the probabilities of water body and non-water body areas, where the probability value of each pixel reflects the likelihood of it belonging to the water body.

[0076] S125. Perform threshold segmentation processing on the binary segmentation probability map and map the segmentation result to the image coordinate system where the geographic location data is located based on the geographic coordinate transformation matrix to generate the target remotely sensed image data corresponding to the water body area.

[0077] Specifically, apply a threshold segmentation method to the generated binary segmentation probability map to convert the probability value into a clear classification result (i.e., water body or non-water body). Subsequently, according to the geographic coordinate transformation matrix, map the segmentation result back to the geographic location coordinate system of the original remotely sensed image to ensure the consistency of spatial positions.

[0078] This embodiment realizes the effective conversion from the original remotely sensed image data to the target remotely sensed image data with accurately located water body areas, and can effectively improve the accuracy and efficiency of water body monitoring.

[0079] In one embodiment, the matching of the water quality parameter data with the target remotely sensed image data in the same coordinate system according to the geographic location data to construct a spatio-temporally aligned training dataset includes:

[0080] Match the sampling point coordinates of the water quality parameter data with the pixel geographical coordinates in the target remote sensing image data according to the geographical location data.

[0081] First, use the known geographical location information (such as latitude and longitude coordinates) to find the sampling point coordinates corresponding to the water quality parameter data collected in the field. Then, find the specific pixel geographical coordinates corresponding to these sampling points in the processed target remote sensing image data. This step can ensure that the ground-measured water quality data can accurately correspond to the corresponding positions in the remote sensing image, thus establishing a direct spatial connection.

[0082] Extract the characteristic band reflectance values of the target remote sensing image data, establish a fitting relationship between the concentration and the characteristic band reflectance values for each water quality parameter data, and generate a band water quality parameter concentration mapping table.

[0083] Specifically, extract the reflectance values of specific bands from the target remote sensing image data (such as at least four spectral bands in the range of 410nm to 590nm mentioned above). For each water quality parameter (such as chlorophyll concentration, suspended sediment concentration, etc.), establish a mathematical fitting relationship between its concentration and the reflectance values of the selected bands through statistical analysis or machine learning methods. The purpose of this step is to find the spectral characteristics that can best reflect the changes of specific water quality parameters, and then generate a band water quality parameter concentration mapping table, providing basic data for subsequent model training.

[0084] Perform spatio-temporal consistency screening based on the timestamp of the target remote sensing image data and the detection time of the water quality parameter data.

[0085] To ensure the consistency and accuracy of the data, it is necessary to perform spatio-temporal consistency screening based on the timestamp of the target remote sensing image data and the actual detection time of the water quality parameter data. When the remote sensing image shooting time and the water quality sample collection time are close enough, it can be determined as valid data, thus avoiding data inaccuracy problems caused by time differences, ensuring that all data used for training are obtained within the same time period, and improving the reliability of model prediction.

[0086] Associate and match the water quality parameter data, the pixel geographical coordinates, the timestamp, and the target remote sensing image data after spatio-temporal consistency screening in the band water quality parameter concentration mapping table to construct a spatio-temporally aligned training dataset.

[0087] Finally, comprehensively integrate the water quality parameter data, pixel geographic coordinates, timestamps, and target remote sensing image data after spatio-temporal consistency screening. Specifically, correlate and match the above data according to a unified format and standard to form a comprehensive dataset containing multi-dimensional information (including spatial location, time information, and water quality parameters). The comprehensive dataset not only includes the actual measured values of water quality parameters but also the corresponding remote sensing image features, providing comprehensive and accurate data support for subsequent training of the water quality inversion model.

[0088] In one embodiment, the spatio-temporal consistency screening based on the timestamps of the target remote sensing image data and the detection times of the water quality parameter data includes:

[0089] Compare the timestamps of the target remote sensing image data with the detection times of the water quality parameter data. If the difference between the timestamp and the detection time is greater than a preset threshold, then eliminate the corresponding target remote sensing image data and water quality parameter data.

[0090] First, collect the timestamps of all target remote sensing image data (i.e., the specific dates and times when the satellites captured these images) and the detection times corresponding to the water quality parameter data collected in the field (i.e., the specific dates and times when the water quality parameters were actually sampled and measured); then, define a reasonable preset time difference threshold, which depends on the speed of water quality change and the specific situation of the study area. For example, in some rapidly changing water body environments, a shorter time threshold (such as a few hours to one day) may need to be set, while in relatively stable environments, a longer time threshold (such as a few days to one week) can be used; furthermore, for each pair of remote sensing image data and water quality parameter data, calculate the difference between their timestamps and detection times. If this difference exceeds the preset time difference threshold, it is considered that this pair of data does not match in time, so it should be eliminated. This means that these data will not be included in the subsequent training dataset. Conversely, if the difference is within the threshold range, then retain this pair of data as a valid sample. Finally, if the time difference of the data pair exceeds the set threshold, remove the data pair whose time difference exceeds the preset time difference threshold from the dataset. This embodiment ensures that the final remaining dataset contains remote sensing image data and water quality parameter data that are closely related in time, ensuring that the basic data for model training has high temporal and spatial consistency, and thus improving the accuracy and reliability of the water quality inversion model.

[0091] In one embodiment, the preset water quality inversion model is an ensemble learning regression model, the base learners of the preset water quality inversion model include at least two of random forest, gradient boosting decision tree, support vector machine, and extreme gradient boosting algorithm, and the meta-learner of the preset water quality inversion model includes a multiple linear regression algorithm.

[0092] It should be noted that an ensemble learning regression model can be selected as the preset water quality inversion model. This preset water quality inversion model combines the prediction results of multiple base models (base learners) and uses another model (meta-learner) to integrate the above prediction results. Among them, the base learner is the basic component of the ensemble learning model, and the meta-learner is the model used to integrate the prediction results of the base learners.

[0093] Furthermore, the training process of the water quality inversion model includes:

[0094] Input the spatiotemporally aligned training dataset, including the reflectance values of the characteristic bands of remote sensing images, the concentrations of water quality parameters, etc.;

[0095] Train the selected base learners (such as random forest, XGBoost, etc.) respectively, and each base learner will generate a set of prediction results for water quality parameters;

[0096] Take the prediction results of each base learner as new features and input them into the meta-learner (multiple linear regression) for training. The meta-learner learns how to best combine these prediction results to generate the final predicted values of water quality parameters;

[0097] Evaluate the model performance through methods such as cross-validation, and adjust the selection of base learners, parameter settings, or the design of the meta-learner as needed until satisfactory prediction accuracy is achieved.

[0098] This embodiment improves the accuracy and robustness of the water quality inversion model by comprehensively utilizing the advantages of multiple machine learning algorithms.

[0099] In one embodiment, the step of inputting the image data to be measured into the target water quality inversion model to obtain the water quality parameter data of the area to be measured, and mapping the water quality parameter data of the area to be measured to the area to be measured to generate the two-dimensional spatial distribution map of the water quality of the area to be measured includes:

[0100] Calculate the sensitive band values of each pixel in the water body area according to the image data to be measured.

[0101] First, obtain the remote sensing image data of the area to be measured. Based on the spectral bands previously determined to be particularly sensitive to water quality changes (such as specific bands in the range of 410nm to 590nm), calculate the reflectance values of each pixel in the water body area under these sensitive bands. The sensitive band values can reflect the different components and their concentration changes in the water body.

[0102] Perform feature engineering processing on the sensitive band values of each pixel.

[0103] Specifically, perform feature engineering on the sensitive band values of each pixel, aiming to improve the quality of the data and the performance of the model; the feature engineering includes but is not limited to: normalization processing, feature selection, and dimensionality reduction processing.

[0104] Input the sensitive band values of each pixel after feature processing into the target water quality inversion model to generate the predicted values of water quality parameters for each pixel.

[0105] It should be explained that the sensitive band values of each pixel after feature engineering are input into the trained target water quality inversion model. The target water quality inversion model will output the corresponding predicted values of water quality parameters according to the input data, such as chlorophyll concentration, suspended solid concentration, etc. For each pixel, the target water quality inversion model will output specific estimated values of one or more water quality parameters.

[0106] Map the predicted values of the water quality parameters to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

[0107] Finally, map the predicted values of the water quality parameters of each pixel obtained above back to the original remote sensing image coordinate system, that is, reassign them to the corresponding geographical coordinate positions according to their geographical locations. This step is used to visually display the spatial distribution of water quality parameters in the entire area to be measured. Among them, different water quality parameter values can be represented in the form of grayscale images or color-coded maps, where different colors or gray levels correspond to different water quality concentration levels. Finally, a clear two-dimensional spatial distribution map of water quality is generated to help users quickly identify the water quality problem areas and their severity.

[0108] This embodiment can efficiently and accurately complete the water quality monitoring task of large water areas, providing a scientific basis for water resource management and environmental protection.

[0109] Refer to Figure 3 , the present invention provides a water body monitoring device 30 based on satellite remote sensing data, including:

[0110] A data acquisition module 31 for acquiring remote sensing image data of a target area and measured water quality data of the water body area in the target area, the remote sensing image data including at least two spectral bands adapted to the water body reflectance characteristics, and the measured water quality data including water quality parameter data and geographical location data;

[0111] A calibration and segmentation module 32 for performing radiometric calibration processing on the remote sensing image data and extracting the target remote sensing image data corresponding to the water body area in the remote sensing image data through an image segmentation model;

[0112] A training dataset construction module 33, configured to match the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data, and construct a spatiotemporally aligned training dataset;

[0113] A model training module 34, configured to train a preset water quality inversion model based on the training dataset to obtain a target water quality inversion model, where the preset water quality inversion model is constructed based on a multi-layer model fusion strategy;

[0114] An output module 35, configured to input the image data to be measured into the target water quality inversion model to obtain water quality parameter data of the area to be measured, and map the water quality parameter data of the area to be measured to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

[0115] Refer to Figure 4 , the present invention provides an electronic device 40, including a memory 41 and a processor 42;

[0116] The memory 41 is configured to store a computer program;

[0117] The processor 42 is configured to implement the above-mentioned water body monitoring method based on satellite remote sensing data when executing the computer program.

[0118] Or, an electronic device 40, including a memory 41 and a processor 42 coupled to the memory 41; the memory 41 is configured to store a computer program; the processor 42 is configured to perform the following operations when executing the computer program:

[0119] Obtain remote sensing image data of a target area and measured water quality data of a water body area in the target area, where the remote sensing image data includes at least two spectral bands adapted to the water body reflectance characteristics, and the measured water quality data includes water quality parameter data and geographical location data;

[0120] Perform radiometric correction processing on the remote sensing image data, and extract target remote sensing image data corresponding to the water body area in the radiometrically corrected remote sensing image data through an image segmentation model;

[0121] Match the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data, and construct a spatiotemporally aligned training dataset;

[0122] Train a preset water quality inversion model based on the training dataset to obtain a target water quality inversion model, where the preset water quality inversion model is constructed based on a multi-layer model fusion strategy;

[0123] Input the image data to be measured into the target water quality inversion model to obtain the water quality parameter data of the area to be measured, and map the water quality parameter data of the area to be measured to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

[0124] The present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned water body monitoring method based on satellite remote sensing data is implemented.

[0125] Or, a non-volatile computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the processor performs the following operations:

[0126] Obtain the remote sensing image data of the target area and the measured water quality data of the water body area in the target area. The remote sensing image data includes at least two spectral bands adapted to the water body reflectance characteristics, and the measured water quality data includes water quality parameter data and geographical location data;

[0127] Perform radiometric correction processing on the remote sensing image data, and extract the target remote sensing image data corresponding to the water body area in the radiometrically corrected remote sensing image data through an image segmentation model;

[0128] Match the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data to construct a spatiotemporally aligned training data set;

[0129] Train a preset water quality inversion model based on the training data set to obtain a target water quality inversion model. The preset water quality inversion model is constructed based on a multi-layer model fusion strategy;

[0130] Input the image data to be measured into the target water quality inversion model to obtain the water quality parameter data of the area to be measured, and map the water quality parameter data of the area to be measured to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

[0131] Now, the electronic device 40 that can be a server or a client of the present invention will be described. It is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device 40 is intended to represent various forms of digital electronic computer devices, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device 40 can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0132] The electronic device 40 includes a computing unit that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) or a computer program loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The computing unit, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0133] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc. In the present application, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiments of the present invention. In addition, the functional units in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0134] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will all fall within the protection scope of the present invention.

Claims

1. A water body monitoring method based on satellite remote sensing data, characterized in that, Including: Obtaining remote sensing image data of a target area and measured water quality data of a water body area in the target area, where the remote sensing image data includes at least two spectral bands adapted to the water body reflectance characteristics, and the measured water quality data includes water quality parameter data and geographical location data; Performing radiometric correction processing on the remote sensing image data, and extracting target remote sensing image data corresponding to the water body area in the radiometrically corrected remote sensing image data through an image segmentation model; Matching the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data to construct a spatiotemporally aligned training dataset; Training a preset water quality inversion model based on the training dataset to obtain a target water quality inversion model, where the preset water quality inversion model is constructed based on a multi-layer model fusion strategy; Inputting the image data to be measured into the target water quality inversion model to obtain water quality parameter data of the area to be measured, and mapping the water quality parameter data of the area to be measured to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

2. The water body monitoring method based on satellite remote sensing data according to claim 1, characterized in that The remote sensing image data includes at least 4 spectral bands with central wavelengths in the range of 410 nanometers to 590 nanometers, the bandwidth of each spectral band is 20 nanometers to 30 nanometers, the water quality parameter data includes chlorophyll concentration, suspended solid concentration and solvent organic matter content, and the geographical location data includes position coordinates in the geodetic coordinate system.

3. The water body monitoring method based on satellite remote sensing data according to claim 1, wherein, The performing radiometric correction processing on the remote sensing image data, and extracting target remote sensing image data corresponding to the water body area in the radiometrically corrected remote sensing image data through an image segmentation model includes: Performing atmospheric correction and sensor radiometric calibration processing on the remote sensing image data; Inputting the remote sensing image data after the atmospheric correction and the sensor radiometric calibration processing into the input layer of the image segmentation model, and sequentially performing feature extraction on the remote sensing image data through a convolutional layer, a batch normalization layer, a multi-scale pooling layer and an activation function to obtain a first multi-scale feature map; Processing the multi-scale feature map based on the dual-channel attention module of the image segmentation model to obtain a second multi-scale feature map; Performing upsampling processing on the second multi-scale feature map through a transposed convolution, and performing channel connection filling with the shallow feature map of the convolutional layer through a skip connection to output a binary segmentation probability map of the water body area; Performing threshold segmentation processing on the binary segmentation probability map, and mapping the segmentation result to the image coordinate system where the geographical location data is located based on a geographical coordinate transformation matrix to generate target remote sensing image data corresponding to the water body area.

4. The water body monitoring method based on satellite remote sensing data according to claim 1, characterized in that, The matching the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographical location data to construct a spatiotemporally aligned training dataset includes: Matching the sampling point coordinates of the water quality parameter data with the pixel geographical coordinates in the target remote sensing image data according to the geographical location data; Extract the characteristic band reflectance values of the target remote sensing image data, establish the fitting relationship between the concentration and the characteristic band reflectance values for each of the water quality parameter data, and generate a band water quality parameter concentration mapping table; Perform spatio-temporal consistency screening based on the timestamp of the target remote sensing image data and the detection time of the water quality parameter data; Associate and match the water quality parameter data, the pixel geographic coordinates, the timestamp, and the target remote sensing image data after spatio-temporal consistency screening in the band water quality parameter concentration mapping table to construct a spatio-temporally aligned training dataset.

5. The water body monitoring method based on satellite remote sensing data according to claim 4, wherein The spatio-temporal consistency screening based on the timestamp of the target remote sensing image data and the detection time of the water quality parameter data includes: Compare the timestamp of the target remote sensing image data with the detection time of the water quality parameter data. If the difference between the timestamp and the detection time is greater than a preset threshold, the corresponding target remote sensing image data and the water quality parameter data are excluded.

6. The water body monitoring method based on satellite remote sensing data according to claim 1, wherein The preset water quality inversion model is an ensemble learning regression model. The base learners of the preset water quality inversion model include at least two of random forest, gradient boosting decision tree, support vector machine, and extreme gradient boosting algorithm. The meta-learner of the preset water quality inversion model includes multiple linear regression algorithm.

7. The water body monitoring method based on satellite remote sensing data according to claim 1, characterized in that The step of inputting the to-be-tested image data into the target water quality inversion model to obtain the water quality parameter data of the to-be-tested area and mapping the water quality parameter data of the to-be-tested area to the to-be-tested area to generate the two-dimensional water quality spatial distribution map of the to-be-tested area includes: Calculate the sensitive band values of each pixel in the water body area of the to-be-tested area according to the to-be-tested image data; Perform feature engineering processing on the sensitive band values of each pixel; Input the sensitive band values of each pixel after feature processing into the target water quality inversion model to generate the water quality parameter prediction values of each pixel in the water body area; Map the water quality parameter prediction values to the water body area in the to-be-tested area to generate the two-dimensional water quality spatial distribution map of the to-be-tested area.

8. A water body monitoring device based on satellite remote sensing data, characterized in that, including: A data acquisition module for acquiring remote sensing image data of a target area and the measured water quality data of the water body area in the target area. The remote sensing image data includes at least two spectral bands adapted to the water body reflectance characteristics. The measured water quality data includes water quality parameter data and geographic location data; A calibration and segmentation module for performing radiometric calibration processing on the remote sensing image data and extracting the target remote sensing image data corresponding to the water body area in the remote sensing image data through an image segmentation model; A training dataset construction module for matching the water quality parameter data with the target remote sensing image data in the same coordinate system according to the geographic location data to construct a spatio-temporally aligned training dataset; A model training module for training a preset water quality inversion model based on the training dataset to obtain a target water quality inversion model. The preset water quality inversion model is constructed based on a multi-layer model fusion strategy; An output module, configured to input image data to be measured into a target water quality inversion model, obtain water quality parameter data of the area to be measured, and map the water quality parameter data of the area to be measured to the area to be measured to generate a two-dimensional spatial distribution map of the water quality of the area to be measured.

9. An electronic device, characterized in that, It includes a memory and a processor; The memory is used for storing a computer program; The processor is configured to, when executing the computer program, implement the water body monitoring method based on satellite remote sensing data according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the storage medium, and when the computer program is executed by the processor, the water body monitoring method based on satellite remote sensing data according to any one of claims 1 to 7 is implemented.

Images