Water body turbidity remote sensing inversion method, system, medium, equipment and program

Abstract

The invention relates to the technical field of water ecological environment remote sensing monitoring, and provides a water body turbidity remote sensing inversion method and system, a medium, equipment and a program. The water body turbidity remote sensing inversion method comprises the following steps: converting actually measured hyperspectral remote sensing reflectivity into satellite equivalent remote sensing reflectivity, and combining each wave band of the satellite equivalent remote sensing reflectivity to obtain a wave band combination; calculating a Pearson correlation coefficient between each single wave band and each wave band combination and the turbidity data, and screening the wave band combination meeting a set threshold value; encoding the wave band combination and the water environment data meeting a set threshold value, respectively obtaining wave band features, time features and water body type features, and constructing feature vectors; and by taking the feature vector as input and the turbidity data as output, constructing an objective function, and training a turbidity remote sensing inversion model, so as to minimize the objective function and select an optimal hyper-parameter combination in each round of iteration. The method can improve the prediction precision and reliability.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing monitoring of water ecological environment, and in particular to a method, system, medium, equipment and program for remote sensing inversion of water turbidity. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Water turbidity is one of the important indicators for measuring the quality of water ecological environment. It reflects the comprehensive impact of various factors such as particulate matter, sediments and organic pollutants in water on the water ecosystem. Turbidity monitoring is widely needed in many fields such as inland waters, estuaries and oceans. Its dynamic changes can reflect the health status of water bodies and changes in their ecological functions. Therefore, real-time monitoring and dynamic analysis of water turbidity are of great significance for water ecological environment management, water pollution control and water ecological restoration.

[0004] Traditional water turbidity monitoring methods usually rely on manual sampling and laboratory analysis at fixed monitoring sites. Although this method can provide relatively accurate turbidity data, it has obvious limitations: on the one hand, the monitoring points can usually only reflect the water quality conditions of the local area, and cannot fully reflect the turbidity changes of the water body in a wide area; on the other hand, manual sampling and laboratory analysis require a lot of time and manpower, and can only obtain a small amount of spatiotemporal data, which cannot meet the needs of long-term, large-scale, and high-frequency turbidity monitoring.

[0005] With the rapid development of remote sensing technology and machine learning algorithms, satellite remote sensing data can provide information on the optical characteristics of water bodies in a large area, thereby realizing the monitoring of water turbidity in a wide area. The high-frequency acquisition capability of remote sensing data in time and space enables the continuous observation of changes in water turbidity in different seasons and under different environmental conditions. However, existing turbidity remote sensing inversion methods generally rely on a single type of data as the input model, usually using individual bands or specific band combinations of multispectral or hyperspectral images to reflect the turbidity characteristics of water bodies. However, such methods have the following problems: on the one hand, using only spectral information may not fully capture the complex changes in water turbidity; on the other hand, most of the existing remote sensing inversion models do not take into account the hydrological information of water bodies (such as water body type, sampling time, seasonal changes, etc.), which leads to insufficient accuracy and adaptability of the model. In particular, when applied to different types of water bodies with significant temporal changes, the accuracy and portability of the model are significantly reduced. Summary of the invention

[0006] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a water turbidity remote sensing inversion method, system, medium, equipment and program. The goal of the present invention is to establish a multi-dimensional input model, comprehensively consider hydrological factors such as water body type and sampling time, while retaining the details of spectral information, so as to improve the accuracy and generalization ability of the turbidity inversion model.

[0007] In order to achieve the above object, the present invention adopts the following technical solution: A first aspect of the present invention provides a method for remote sensing inversion of water turbidity.

[0008] A water turbidity remote sensing inversion method, comprising: Acquire water environment data, measured hyperspectral remote sensing reflectance and turbidity data, convert the measured hyperspectral remote sensing reflectance into satellite equivalent remote sensing reflectance, combine each band of the satellite equivalent remote sensing reflectance to obtain a band combination; Calculate the Pearson correlation coefficient of each single band and band combination with turbidity data, and select the band combination that meets the set threshold; encode the band combination and water environment data that meet the set threshold, obtain the band characteristics, time characteristics and water type characteristics, and construct the feature vector; With the feature vector as input and turbidity data as output, the objective function is constructed and the turbidity remote sensing inversion model is trained to minimize the objective function and select the optimal hyperparameter combination in each round of iteration.

[0009] Furthermore, the method for converting the measured hyperspectral remote sensing reflectivity into the satellite equivalent remote sensing reflectivity includes: using the spectral response function of the satellite sensor to convert the measured hyperspectral remote sensing reflectivity into the satellite equivalent remote sensing reflectivity.

[0010] Furthermore, the method for combining each band of the satellite equivalent remote sensing reflectivity includes: performing band addition, subtraction, ratio, difference normalization and three-band combination on each band of the satellite equivalent remote sensing reflectivity.

[0011] Furthermore, the band combination and water environment data that meet the set threshold are encoded to obtain band characteristics, time characteristics and water body type characteristics respectively, and a feature vector is constructed; the method includes: encoding the band combination that meets the set threshold to obtain band characteristics, encoding the time information in the water environment data according to the season to obtain time characteristics; encoding the water body type information in the water environment data to obtain water body type characteristics; splicing the band characteristics, time characteristics and water body type characteristics to obtain a feature vector.

[0012] Furthermore, the objective function is expressed by the following formula:

[0013] in, represents the objective function; For the i The loss function of samples; For the t The regularization term of each tree is used to control the complexity of the decision tree in the turbidity remote sensing inversion model; n Indicates the total sample size; T Indicates that there is T A tree.

[0014] Furthermore, in each round of iteration, the objective function is minimized and the optimal hyperparameter combination is selected; the method includes: in each round of iteration, the objective function is minimized by the gradient boosting method, and the gradient and Hessian of each sample point are calculated; a greedy algorithm is used to traverse all features and find the best splitting point so that the data subset after the split has a smaller error than the data before the split; in each iteration, the turbidity prediction value of the newly generated tree is added to the turbidity prediction value of the current model; and regularization is applied for processing to select the hyperparameter combination that optimizes the model performance; the hyperparameters include: learning rate, number of trees, maximum depth of the tree, and minimum sample weight of leaf nodes.

[0015] A second aspect of the present invention provides a water turbidity remote sensing inversion system.

[0016] A water turbidity remote sensing inversion system, comprising: A data acquisition module is configured to: acquire water environment data, measured hyperspectral remote sensing reflectance and turbidity data, convert the measured hyperspectral remote sensing reflectance into satellite equivalent remote sensing reflectance, and combine each band of the satellite equivalent remote sensing reflectance to obtain a band combination; A screening and encoding module is configured to: calculate the Pearson correlation coefficient of each single band and band combination with turbidity data, and screen the band combination that meets the set threshold; encode the band combination and water environment data that meet the set threshold, respectively obtain band characteristics, time characteristics and water type characteristics, and construct a feature vector; The model training module is configured to: take the feature vector as input and the turbidity data as output, construct the objective function, train the turbidity remote sensing inversion model, so as to minimize the objective function and select the optimal hyperparameter combination in each round of iteration.

[0017] A third aspect of the present invention provides a computer-readable storage medium.

[0018] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the water body turbidity remote sensing inversion method as described in the first aspect above.

[0019] A fourth aspect of the present invention provides a computer device.

[0020] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the water body turbidity remote sensing inversion method as described in the first aspect above are implemented.

[0021] A fifth aspect of the present invention provides a computer program product or a computer program.

[0022] The present invention provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the water body turbidity remote sensing inversion method as described in the first aspect above.

[0023] Compared with the prior art, the present invention has the following beneficial effects: The present invention proposes a method, system, medium, device and program for remote sensing inversion of water turbidity, which combines spectral information and hydrological information to perform remote sensing inversion of water turbidity from a multi-dimensional and multi-factor perspective. With the powerful nonlinear fitting ability of machine learning, effective water environment information can be extracted from complex remote sensing data, and the accuracy and generalization ability of the turbidity inversion model can be improved by jointly modeling multiple factors.

[0024] The present invention provides a high-precision and robust turbidity remote sensing estimation model: the optical characteristics of water bodies in different regions are complex and have large temporal and spatial differences. Machine learning models that only input spectral information are difficult to achieve stable and high accuracy in long-term, large-scale remote sensing monitoring. The present invention can fully tap the influence of factors such as different types of water bodies and time changes, so that it can still maintain high prediction accuracy and reliability under various water body types and different environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0026] Figure 1 It is a flow chart of the water turbidity remote sensing inversion method shown in the present invention; Figure 2 is a schematic diagram of the spectral response of each band of Landsat-8 OLI shown in the present invention; Figure 3 is a schematic diagram of the measured remote sensing reflectivity shown in the present invention; Figure 4 Schematic diagram of the equivalent remote sensing reflectivity of the Landsat-8 OLI satellite shown in the present invention. DETAILED DESCRIPTION

[0027] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0028] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0029] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0030] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the methods and systems according to various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code may include one or more executable instructions for implementing the logical functions specified in each embodiment. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in connection can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of boxes in the flowchart and / or block diagram can be implemented using a dedicated hardware-based system that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0031] Embodiment 1 This embodiment provides a method for remote sensing inversion of water turbidity. The present invention integrates water type characteristics and time characteristics based on in-situ monitoring data, and constructs a water turbidity remote sensing inversion model based on the XGBoost machine learning algorithm: (1) Based on the satellite spectral response function, the measured hyperspectral data is converted into satellite equivalent multispectral data; and based on the Pearson correlation coefficient, the optimal band combination for inverting water turbidity is screened; (2) The optimal band combination, water type characteristics, and time characteristics are used as input, and turbidity is used as output to train the XGBoost model, calibrate hyperparameters such as learning rate, tree depth, and number of trees, and obtain the optimal turbidity remote sensing inversion model. Figure 1 As shown, the specific steps are as follows: Step 1: Data Collection Build a lake-reservoir water environment characteristics database: collect water environment data of various types of lakes and reservoirs, including water surface hyperspectral data and turbidity data.

[0032] Step 2: Data Processing (1) Turbidity data: Remove missing values ​​and outliers to ensure data quality, and perform interpolation and standardization on the data.

[0033] (2) Spectral data: In order to align the hyperspectral data with the spectral characteristics of satellite remote sensing, the collected hyperspectral Convolution processing is performed, and then the band average hyperspectral is calculated using the satellite remote sensing spectral response function.

[0034] The spectral response function refers to the ratio of the received radiance to the incident radiance at each wavelength, which describes the response intensity of a sensor at different wavelengths ( Figure 2 ). The radiance / reflectance at the central wavelength of the satellite image after atmospheric correction is different from the actual radiance / reflectance at this wavelength and cannot be directly compared with the measured hyperspectral data. Therefore, it is necessary to use the spectral response function to convert the measured remote sensing reflectance ( Figure 3 ) is converted into the equivalent remote sensing reflectance of the Landsat-8 OLI satellite band ( Figure 4 ):

[0035] In the formula, For satellite i Satellite equivalent remote sensing reflectivity of the band, λ 1 and λ 2 is the band range of this band, To measure the hyperspectral remote sensing reflectance, SRF (λ) is λ Spectral responsivity at wavelength.

[0036] Figure 4 B1, B2, B3, B4, and B5 are the abbreviations of the first five bands of the Landsat-8 OLI satellite, corresponding to bandi in the above formula. B1 is the nearshore band (band range 435~451 nm); B2 is the blue band (band range 452~412 nm); B3 is the green band (band range 533~590 nm); B4 is the red band (band range 636~673 nm); and B5 is the near-infrared band (band range 851~879 nm).

[0037] This approach ensures consistency between in situ measurements and satellite observations, improves the accuracy of turbidity estimates, and facilitates the integration of ground and remote sensing data for remote sensing retrieval of turbidity over large areas.

[0038] (3) Band screening and feature combination: Perform band addition, subtraction, ratio, difference normalization and three-band combination on each band of the calculated satellite equivalent remote sensing reflectance. Calculate the Pearson correlation coefficient between each single band and band combination and turbidity to evaluate the impact of the band on concentration. Select the band combination with a correlation coefficient higher than 0.6. The specific combination method is as follows:

[0039]

[0040]

[0041]

[0042] In the formula, It is a spectrum The reflectivity of the band. The best band selected by the iterative method.

[0043] (4) Temporal information processing: According to the different seasons, one-hot encoding is used to convert the season into a four-dimensional vector, namely [1, 0, 0, 0] (spring), [0, 1, 0, 0] (summer), [0, 0, 1, 0] (autumn), and [0, 0, 0, 1] (winter).

[0044] (5) Water body type information processing: According to the water body type, one-hot encoding is also used to convert different water body types into two-dimensional vectors: [1, 0] (lake), [0, 1] (reservoir).

[0045] Step 3: Model building Input: Combine the selected relevant band combinations, time information, and water body types into a feature vector X = [X b , Xb , X w ], where X b is the band characteristic, X b is the time feature, X w It is the water body type feature. It is divided into training set and test set in a ratio of 7:3 to ensure that the training data can represent the actual situation.

[0046] The core idea of ​​the XGBoost classifier is to gradually train the tree through the gradient boosting method, and use the gradient and Hessian information to optimize the loss function in the process of building each tree. At the same time, the complexity of the model is controlled through regularization and tree pruning, thereby improving the generalization ability of the model. The specific steps are as follows: 1. Initialize the model: Take the feature vector X containing hydrological and spectral information as input and the turbidity value as output to build an initialized turbidity remote sensing inversion model. The XGBoost classifier builds a series of regression trees through the gradient boosting method. The goal of each tree is to gradually improve the performance of the model by minimizing a loss function. The form of the loss function is:

[0047] in, is the turbidity value predicted by the model, y is the measured turbidity value.

[0048] 2. Train the model by optimizing the objective function: Assume that the model is composed of T The objective function is composed of trees. for:

[0049] in, For the i The loss function of samples is For the t The regularization term of the tree is used to control the complexity of the decision tree in the turbidity inversion model. It is usually defined as:

[0050] in, and is the regularization parameter, T is the number of leaves in the tree, It is the first j The weight of each leaf.

[0051] The hyperparameters that need to be calibrated include: (1) Learning rate: controls the contribution of each tree to the final turbidity remote sensing inversion model. A smaller learning rate can usually improve the generalization ability of the turbidity remote sensing inversion model, but more iterations are required to ensure the convergence of the model. A larger learning rate may cause the model to converge too quickly, but may miss the local optimal solution. It is usually adjusted between 0.01 and 0.3.

[0052] (2) Number of trees: Generally, the more trees there are, the stronger the fitting ability of the turbidity remote sensing inversion model is, but it is also more prone to overfitting. The number of trees should be selected according to the adjustment of the learning rate, generally between 50 and 200.

[0053] (3) Maximum depth of the tree: controls the complexity of the turbidity remote sensing inversion model. The greater the depth of the tree, the more complex the model is and the more prone to overfitting; the smaller the depth, the simpler the model is and the more prone to underfitting. The common adjustment range is 2 to 10.

[0054] (4) Minimum sample weight of leaf nodes: The smaller the value, the more complex the turbidity remote sensing inversion model. For smaller data sets, a smaller value can be selected, while for larger data sets, a larger value helps reduce overfitting. The range is generally between 1 and 10.

[0055] Hyperparameters are calibrated using grid search: (1) A hyperparameter grid is defined (learning rate: 0.01, 0.1, 0.2, 0.3; maximum tree depth: 2, 4, 6, 8, 10; number of trees: 50, 100, 200; minimum sample weight of leaf nodes: 1, 3, 5, 7, 9). (2) The performance of all hyperparameter combinations is tested by enumerating the hyperparameter space, and then the hyperparameter configuration that achieves the best model performance is selected.

[0056] 3. Tree splitting: Each round of training corrects the prediction error of the previous tree by building a new decision tree. During the turbidity inversion model training process: (1) In each iteration, the objective function is minimized by using the gradient boosting method to calculate the gradient and Hessian of each sample point.

[0057] (gradient) (Hessian) in, represents the gradient, represents Hessian, , They are i The predicted and measured turbidity values ​​for each data point.

[0058] (2) Use a greedy algorithm to traverse all features and find the best split point. The goal of the greedy algorithm is to select a feature and its split point at each level of the tree so that the data subset after the split has a smaller error than the data before the split. This is usually done through the following steps: 1) Calculate the gain of each candidate split point: For each node in the tree, calculate the gain of the node after selecting a certain split point of a certain feature, that is, the reduction in the objective function value after selecting the split point.

[0059] 2) Gain calculation: Assume that at a certain node t The node is split at , and after the node split, there are two child nodes L and R. The gain after the split is:

[0060] in, represents the gain after splitting, Is a node t The sum of the gradients of . , The nodes are L , R The sum of the gradients of . , The nodes are L , R The sum of the Hessians of . and is the regularization parameter.

[0061] 3) Select the split point with the largest gain: Among all candidate split points, select the split point that maximizes the gain for splitting.

[0062] (3) At each split node, the XGBoost classifier determines the weight of the leaf node based on the gradient and Hessian values ​​of all samples in the node:

[0063] in, is the sum of the node gradients, is the sum of the Hessian of the nodes, and the splitting results are optimized by adjusting these weights. is the regularization parameter.

[0064] 4. Update the model: At each iteration, the turbidity prediction value of the newly generated tree is added to the turbidity prediction value of the current model. Assume that the turbidity prediction value of the current model is , the newly generated tree output is , the updated turbidity prediction value is:

[0065] in, is the learning rate, which is used to control the influence of each tree.

[0066] 5. Apply regularization: In order to avoid overfitting of the model, the XGBoost classifier applies regularization terms at each step of training. Specifically, it penalizes the tree structure (such as the depth of the tree, the number of leaf nodes) and the weight of the leaf nodes. By increasing the L1 and L2 regularization strengths (usually in the range of 0 to 1), the risk of overfitting is reduced, thereby improving its generalization ability.

[0067] 6. Evaluation and pruning: After each round of tree training, the XGBoost classifier will evaluate the current model and calculate the error on the training set and the validation set. If the validation set error starts to increase, it means that the model may be overfitting and the training will stop. During the construction of each tree, the XGBoost classifier uses the maximum depth and minimum split gain to control the size of the tree. The tree splitting will continue until the stopping condition is met or the gain reaches the threshold.

[0068] 7. Repeat the training until stopped: XGBoost gradually improves the turbidity remote sensing inversion model by repeatedly executing the above steps. After multiple rounds of training, XGBoost will obtain a series of decision tree models. The output of each tree is a turbidity prediction value. The final model will obtain the weighted sum of all trees. To predict the turbidity output value.

[0069]

[0070] in, T is the number of trees, t The output of the tree.

[0071] Result output: prediction result It is the remote sensing predicted value of water turbidity.

[0072] Step 4: Model Evaluation The K-Fold Cross-Validation method is used to comprehensively evaluate the robustness and generalization ability of the turbidity remote sensing inversion model. The specific approach is to divide the data set into K subsets, select one of the subsets as the validation set in turn, and use the remaining subsets for training, repeat K times, and finally calculate the average performance index of each validation. This method can reduce the accidental impact caused by data division and ensure the reliability of the evaluation results. In addition, appropriate evaluation indicators (such as R² determination coefficient, RMSE root mean square error, MAE mean absolute error, MAPE mean absolute error percentage, etc.) can also be used to evaluate the performance of the model in the turbidity remote sensing inversion task. The specific indicators are as follows:

[0073]

[0074]

[0075]

[0076] In the formula, n represents the number of samples, , Respectively i The predicted and measured turbidity values ​​at each sampling point are is the average value of all turbidity measured values. 2 The closer it is to 1, the better the fitting effect. These indicators take into account the variance explanation power (R 2 ), residual magnitude (RMSE and MAE), and proportional accuracy (MAPE), thereby comprehensively evaluating the performance of the model.

[0077] Embodiment 2 This embodiment provides a water turbidity remote sensing inversion system.

[0078] A water turbidity remote sensing inversion system, comprising: A data acquisition module is configured to: acquire water environment data, measured hyperspectral remote sensing reflectance and turbidity data, convert the measured hyperspectral remote sensing reflectance into satellite equivalent remote sensing reflectance, and combine each band of the satellite equivalent remote sensing reflectance to obtain a band combination; A screening and encoding module is configured to: calculate the Pearson correlation coefficient of each single band and band combination with turbidity data, and screen the band combination that meets the set threshold; encode the band combination and water environment data that meet the set threshold, respectively obtain band characteristics, time characteristics and water type characteristics, and construct a feature vector; The model training module is configured to: take the feature vector as input and the turbidity data as output, construct the objective function, train the turbidity remote sensing inversion model, so as to minimize the objective function and select the optimal hyperparameter combination in each round of iteration.

[0079] In some embodiments, the data acquisition module is further configured to: (1) Download and preprocessing of remote sensing images: Download high-quality Landsat-8 OLI remote sensing images with little cloud and perform preprocessing such as radiation correction and FLAASH atmospheric correction.

[0080] (2) Water turbidity data collection: Collect water turbidity data synchronously or quasi-synchronously with images (3) Water surface spectral data collection: Collect water surface hyperspectral data synchronized with water turbidity measurement. Spectral measurements should be conducted under clear, cloudless weather conditions, and the measurement time is limited to between 10:30 and 15:00. The measurement range covers 400~900nm, and records hyperspectral remote sensing reflectance. R rs Spectral data of (λ). Spectral measurements should be performed according to the predetermined radiometric measurement plan to ensure the consistency and comparability of the data.

[0081] (4) Constructing a remote sensing reflectance spectral library: Based on the spectral response function of the remote sensing image, the collected water surface spectral data is converted into satellite equivalent spectral data corresponding to the satellite sensor. The converted data is used to construct a remote sensing reflectance spectral library to provide basic data for subsequent turbidity remote sensing inversion.

[0082] In some embodiments, the model training module is further configured to: 1) Construction of machine learning model Using machine learning algorithms such as XGBoost, combined with diagnostic spectral features, turbidity monitoring data, and relevant information such as water body type and sampling time, a turbidity remote sensing estimation model was developed. By integrating multiple factors and making full use of the nonlinear fitting ability of machine learning algorithms, an accurate turbidity remote sensing inversion model was established.

[0083] (2) Hyperparameter tuning and model training During the model training process, the grid search method is used to optimize and adjust key parameters. Through the predefined parameter grid, all possible hyperparameter combinations are enumerated, and the hyperparameter combination that optimizes the model performance is selected. When the maximum number of iterations is met or the residual is minimized, the model will converge to the optimal solution.

[0084] (3) Model performance evaluation and optimization The fitting accuracy of the turbidity remote sensing inversion model is comprehensively evaluated using a variety of evaluation indicators such as R², RMSE, MAE and MAPE. These indicators are used to evaluate the degree of fit of the model on the training set and the validation set to ensure that the model has high accuracy and strong robustness.

[0085] (4) Model application and turbidity inversion After the model is developed, it is applied to Landsat-8 OLI remote sensing image data for turbidity remote sensing inversion. Based on the inversion results, a high-precision turbidity remote sensing distribution dataset is generated.

[0086] Embodiment 3 This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the water turbidity remote sensing inversion method as described in the above-mentioned embodiment 1 are implemented.

[0087] Embodiment 4 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the water turbidity remote sensing inversion method described in the first embodiment are implemented.

[0088] Embodiment 5 This embodiment provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the water turbidity remote sensing inversion method described in the first embodiment. Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0089] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0090] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0092] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0093] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A water turbidity remote sensing inversion method, characterized in that: include: Acquire water environment data, measured hyperspectral remote sensing reflectance and turbidity data, convert the measured hyperspectral remote sensing reflectance into satellite equivalent remote sensing reflectance, combine each band of the satellite equivalent remote sensing reflectance to obtain a band combination; Calculate the Pearson correlation coefficient of each single band and band combination with turbidity data, and select the band combination that meets the set threshold; encode the band combination and water environment data that meet the set threshold, obtain the band characteristics, time characteristics and water type characteristics, and construct the feature vector; With the feature vector as input and turbidity data as output, the objective function is constructed and the turbidity remote sensing inversion model is trained to minimize the objective function and select the optimal hyperparameter combination in each round of iteration.

2. The water turbidity remote sensing inversion method according to claim 1, characterized in that: The method for converting the measured hyperspectral remote sensing reflectivity into satellite equivalent remote sensing reflectivity includes: using the spectral response function of the satellite sensor to convert the measured hyperspectral remote sensing reflectivity into satellite equivalent remote sensing reflectivity.

3. The water turbidity remote sensing inversion method according to claim 1, characterized in that: The method for combining each band of the satellite equivalent remote sensing reflectivity includes: performing band addition, subtraction, ratio, difference normalization and three-band combination on each band of the satellite equivalent remote sensing reflectivity.

4. The water turbidity remote sensing inversion method according to claim 1, characterized in that: The band combination and water environment data that meet the set threshold are encoded to obtain band characteristics, time characteristics and water body type characteristics respectively, and a feature vector is constructed; the method includes: encoding the band combination that meets the set threshold to obtain band characteristics, encoding the time information in the water environment data according to the season to obtain time characteristics; encoding the water body type information in the water environment data to obtain water body type characteristics; splicing the band characteristics, time characteristics and water body type characteristics to obtain a feature vector.

5. The water turbidity remote sensing inversion method according to claim 1, characterized in that: The objective function is expressed by the following formula: in, represents the objective function; For the i The loss function of samples; For the t The regularization term of each tree is used to control the complexity of the decision tree in the turbidity remote sensing inversion model; n Indicates the total sample size; T Indicates that there is T A tree.

6. The water turbidity remote sensing inversion method according to claim 5, characterized in that: In each round of iteration, the objective function is minimized and the optimal hyperparameter combination is selected; the method includes: in each round of iteration, the objective function is minimized by the gradient boosting method, and the gradient and Hessian of each sample point are calculated; a greedy algorithm is used to traverse all features and find the best splitting point so that the data subset after the split has a smaller error than the data before the split; in each iteration, the turbidity prediction value of the newly generated tree is added to the turbidity prediction value of the current model; and regularization is applied for processing to select the hyperparameter combination that optimizes the model performance; the hyperparameters include: learning rate, number of trees, maximum depth of the tree and minimum sample weight of leaf nodes.

7. A water turbidity remote sensing inversion system, characterized in that: include: A data acquisition module is configured to: acquire water environment data, measured hyperspectral remote sensing reflectance and turbidity data, convert the measured hyperspectral remote sensing reflectance into satellite equivalent remote sensing reflectance, and combine each band of the satellite equivalent remote sensing reflectance to obtain a band combination; A screening and encoding module is configured to: calculate the Pearson correlation coefficient of each single band and band combination with turbidity data, and screen the band combination that meets the set threshold; encode the band combination and water environment data that meet the set threshold, respectively obtain band characteristics, time characteristics and water type characteristics, and construct a feature vector; The model training module is configured to: take the feature vector as input and the turbidity data as output, construct the objective function, and train the turbidity remote sensing inversion model to minimize the objective function and select the optimal hyperparameter combination in each round of iteration.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the water turbidity remote sensing inversion method as described in any one of claims 1 to 6 are implemented.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the water turbidity remote sensing inversion method as described in any one of claims 1-6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps in the water turbidity remote sensing inversion method according to any one of claims 1 to 6 are implemented.

Images