Processing method, device and processing equipment of deep and large reservoir water bloom remote sensing prediction model

Abstract

The invention provides a deep and large reservoir water bloom remote sensing prediction model processing method, device and processing equipment, which are used for innovatively applying a long and short term memory network based on a self-attention mechanism to water bloom remote sensing prediction work, and the model integrates the self-attention mechanism and the long and short term memory network. According to the method, key features in long-time-series data can be more effectively identified and captured, so that the deep and large reservoir water bloom remote sensing prediction model is obtained through training of a model training scheme built by a long and short-term memory network based on a self-attention mechanism; the scheme effect of remarkably improving the accuracy and reliability in water bloom remote sensing prediction work is achieved, important decision basis and technical support can be provided for early-stage water bloom early warning and later-stage water bloom prevention and control intervention of deep and large reservoirs, and then the progress of fine management of regional water resources and water ecology protection work is effectively promoted.

Description

Technical Field

[0001] The present application relates to the field of water quality monitoring, and specifically to a processing method, device, and processing equipment for a remote sensing prediction model of algal blooms in the Shenzhen Reservoir. Background Art

[0002] Compared with ordinary lakes and small and medium-sized reservoirs, the Shenzhen Reservoir is famous for its vast water area and large storage capacity. It is a key role in regional water resource management and undertakes important functions such as water supply, flood control, and irrigation.

[0003] The water level of the Shenzhen Reservoir is more significantly affected by artificial regulation, resulting in large fluctuations. During the process of water level rise and fall, the formation of the drawdown zone promotes the material exchange between the reservoir drawdown zone, water body, and surrounding land. For example, excessive nutrients such as nitrogen and phosphorus accumulated in the drawdown zone are washed into the reservoir. This exchange not only changes the water quality of the reservoir but also has a profound impact on the ecological balance and the occurrence of algal blooms.

[0004] This complex material exchange and ecological impact, especially in the backwater area of the inflowing rivers of the Shenzhen Reservoir, due to its unique geographical location and the comprehensive impact of reservoir operation, exhibit special hydrodynamic characteristics, making this area a high-incidence zone for water eutrophication and algal bloom phenomena. Different from traditional rivers with strong fluidity, rapid water body renewal, and difficult accumulation of nutrients, after the Shenzhen Reservoir stores water, the water body in the backwater area is affected by the backwater effect of the reservoir water body, resulting in slowed flow velocity, sediment deposition, and nutrient accumulation.

[0005] In addition, the changes in surrounding human activities and meteorological conditions also exacerbate the risk of algal blooms in the backwater area, such as agricultural non-point source pollution and domestic sewage discharge, continuously injecting excessive nutrients such as nitrogen and phosphorus into the water body.

[0006] Therefore, the outbreak of algal blooms in the tributary inflows not only threatens the overall water quality and function of the reservoir, affects water supply safety, reduces the availability of water resources, but may also cause long-term destructive effects on the ecological environment. It is particularly important to monitor and prevent algal blooms in the backwater area of the inflowing rivers of the Shenzhen Reservoir.

[0007] Remote sensing technology, with its wide coverage, fast monitoring speed, and powerful dynamic monitoring ability, has become a powerful tool for algal bloom monitoring and early warning.

[0008] However, the inventors of the present application found that when predicting algal blooms through relevant neural network models based on remote sensing images, in terms of actual performance, there are still certain defects in prediction accuracy, which affects the implementation of the management work of the Shenzhen Reservoir. Summary of the Invention

[0009] The present application provides a processing method, device and processing equipment for a remote sensing prediction model of algal blooms in the Shenzhen University Reservoir, which innovatively applies the Key-Attention Network - Long Short-Term Memory Network (KAN-LSTM) based on the self-attention mechanism to the work of remote sensing prediction of algal blooms. This model combines the self-attention mechanism and the long short-term memory network, and can more effectively identify and capture key features in long-time series data. Therefore, the remote sensing prediction model of algal blooms in the Shenzhen University Reservoir trained by the model training scheme based on the KAN-LSTM has achieved the effect of significantly improving the accuracy and reliability in the work of remote sensing prediction of algal blooms, and can provide important decision-making basis and technical support for the early warning of algal blooms and the subsequent prevention and control intervention of algal blooms in the Shenzhen University Reservoir, thus effectively promoting the progress of refined management of regional water resources and the work of water ecological protection.

[0010] In the first aspect, the present application provides a processing method for a remote sensing prediction model of algal blooms in the Shenzhen University Reservoir. The method includes: For the target Shenzhen University Reservoir, obtain sample multispectral satellite images; Based on the preset algal bloom characterization indicators, calculate the indicator values at each position of the sample multispectral satellite images; Combined with the indicator values and indicator thresholds at each position of the sample multispectral satellite images, determine whether there is an algal bloom situation at each position of the multispectral satellite images, and form a binary distribution layer of algal blooms corresponding to the sample multispectral satellite images; Based on the sample multispectral satellite images and the binary distribution layer of algal blooms, quantify the influence of the cumulative effect of the preset indicators on the algal bloom situation and the influence of the multi-day average value of the preset indicators on the algal bloom situation, where the preset indicators include water quality parameters, reservoir hydrological rhythms and meteorological parameters; Based on the quantization results, use the redundancy analysis method to screen out the target indicators that meet the conditions for the relative importance of remote sensing prediction of algal blooms from the preset indicators; Configure the training samples with the indicator data corresponding to the target indicators and the binary distribution layer of algal blooms, and train the remote sensing prediction model of algal blooms in the Shenzhen University Reservoir. The remote sensing prediction model of algal blooms in the Shenzhen University Reservoir is specifically a long short-term memory network based on the self-attention mechanism. The remote sensing prediction model of algal blooms in the Shenzhen University Reservoir is configured with 7 output layers, and each output layer corresponds to the prediction output for different days within the next 7 days. The remote sensing prediction model of algal blooms in the Shenzhen University Reservoir is used to predict whether there is an algal bloom situation within the next 7 days based on the indicator data input by the model corresponding to the target indicators.

[0011] In the second aspect, the present application provides a processing device for a remote sensing prediction model of algal blooms in the Shenzhen University Reservoir. The device includes: An acquisition unit, configured to acquire a sample multispectral satellite image for a target Shenzhen University Reservoir; A calculation unit, configured to calculate the index values of each position of the sample multispectral satellite image based on a preset water bloom characterization index; A determination unit, configured to determine whether there is a water bloom situation at each position of the multispectral satellite image by combining the index values and index thresholds of each position of the sample multispectral satellite image, and form a water bloom binary distribution layer corresponding to the sample multispectral satellite image; A quantification unit, configured to quantify the influence of the cumulative effect of a preset index on the water bloom situation and the influence of the multi-day average value of the preset index on the water bloom situation based on the sample multispectral satellite image and the water bloom binary distribution layer, where the preset index includes water quality parameters, reservoir hydrological rhythm, and meteorological parameters; A screening unit, configured to screen out target indexes that meet the conditions for the relative importance of water bloom remote sensing prediction from the preset indexes by a redundancy analysis method based on the quantification result; A training unit, configured to configure training samples with the index data corresponding to the target indexes and the water bloom binary distribution layer, and train a water bloom remote sensing prediction model for the Shenzhen University Reservoir. Specifically, the water bloom remote sensing prediction model for the Shenzhen University Reservoir is a long short-term memory network based on a self-attention mechanism. The water bloom remote sensing prediction model for the Shenzhen University Reservoir is configured with 7 output layers, and each output layer corresponds to the prediction output for different days within the next 7 days. The water bloom remote sensing prediction model for the Shenzhen University Reservoir is used to predict whether there is a water bloom situation within the next 7 days based on the index data input by the model corresponding to the target indexes.

[0012] In a third aspect, the present application provides a processing device, including a processor and a memory. A computer program is stored in the memory. When the processor calls the computer program in the memory, it executes the method provided in the first aspect of the present application or any possible implementation manner of the first aspect of the present application.

[0013] In a fourth aspect, the present application provides a computer-readable storage medium. The computer-readable storage medium stores multiple instructions, and the instructions are suitable for being loaded by a processor to execute the method provided in the first aspect of the present application or any possible implementation manner of the first aspect of the present application.

[0014] From the above content, the following beneficial effects of the present application can be obtained: For the objective of remote sensing prediction of algal blooms in the Shenda Reservoir, this application innovatively applies the long short-term memory network based on the self-attention mechanism to the remote sensing prediction of algal blooms. This model combines the self-attention mechanism and the long short-term memory network, which can more effectively identify and capture key features in long-term time series data. Therefore, the remote sensing prediction model of algal blooms in the Shenda Reservoir trained by the model training scheme based on the long short-term memory network with self-attention mechanism has achieved the effect of significantly improving the accuracy and reliability in the remote sensing prediction of algal blooms, which can provide important decision-making basis and technical support for the early warning of algal blooms and the subsequent prevention and control intervention of algal blooms in the Shenda Reservoir, and thus effectively promote the progress of the refined management of regional water resources and the protection of water ecology. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0016] Figure 1 It is a schematic flowchart of a processing method of the remote sensing prediction model of algal blooms in the Shenda Reservoir of this application; Figure 2 It is a schematic structural diagram of a processing device of the remote sensing prediction model of algal blooms in the Shenda Reservoir of this application; Figure 3 It is a schematic structural diagram of a processing device of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the drawings in the embodiments of this application. Obviously, the described embodiments are only some, rather than all, embodiments of this application. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative efforts fall within the scope of protection of this application.

[0018] In the description and claims of this application and the above-mentioned drawings, terms such as "first" and "second" are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order different from that shown or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or modules does not necessarily have to be limited to those steps or modules clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of steps that appear in this application does not mean that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering. The named or numbered process steps can be changed in the order of execution according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0019] The division of modules that appears in this application is a logical division. In actual implementation, there may be other division methods. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the couplings, direct couplings or communication connections shown or discussed between each other can be through some interfaces. The indirect couplings or communication connections between modules can be electrical or other similar forms, which are not limited in this application. And the modules or sub-modules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed to multiple circuit modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this application.

[0020] Before introducing the processing method of the bloom remote sensing prediction model provided by this application, the background content involved in this application is first introduced.

[0021] The processing method, device and computer-readable storage medium of the bloom remote sensing prediction model provided by this application can be applied to a processing device, and are used to innovatively apply the long short-term memory network based on the self-attention mechanism to bloom remote sensing prediction work. This model combines the self-attention mechanism and the long short-term memory network, and can more effectively identify and capture key features in long-time series data. Therefore, the bloom remote sensing prediction model trained by the model training scheme built with the long short-term memory network based on the self-attention mechanism in this application has obtained the solution effect of significantly improving the accuracy and reliability in bloom remote sensing prediction work, and can provide important decision-making basis and technical support for the early bloom warning and the later bloom prevention and control intervention of the Shenzhen Reservoir, thereby effectively promoting the progress of the refined management of regional water resources and the work of water ecological protection.

[0022] The processing method of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir mentioned in this application can have its execution entity as the processing device of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir, or different types of processing devices such as a server, a physical host, or a user equipment (UE) that integrates the processing device of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir. Among them, the processing device of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir can be implemented in a hardware or software manner. The UE can specifically be a terminal device such as a smart phone, a tablet computer, a laptop computer, a desktop computer, or a personal digital assistant (PDA). The processing device can be set up in the form of a device cluster.

[0023] In specific applications, considering that the main purpose of the solution in this application is to train the remote sensing prediction model for water blooms in the Shenzhen University Reservoir based on existing data, therefore, the processing device that executes the processing method of the remote sensing prediction model for water blooms in this application or that carries the application service corresponding to the processing method of the remote sensing prediction model for water blooms in this application usually only needs to meet the required data processing capabilities, and its specific device type and device deployment form are relatively flexible.

[0024] If it also involves the acquisition of source data, that is, multi-spectral satellite images, or if it also involves the actual application of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir, then further adaptive adjustments need to be made to the processing device.

[0025] As an example, the processing device can specifically include a first device that executes the training task of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir in a laboratory environment and a second device that executes the actual application of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir on-site.

[0026] It can be easily seen that the processing device can adjust the specific device type and device deployment form of the processing device according to the application scenarios involved and the data processing links involved in the actual situation.

[0027] Next, the processing method of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir provided in this application will be introduced.

[0028] First, refer to Figure 1 , Figure 1 FIG. shows a schematic flow chart of a processing method of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir provided in this application. The processing method of the remote sensing prediction model for water blooms in the Shenzhen University Reservoir provided in this application can specifically include the following steps S101 to step S106: Step S101, for the target Shenzhen University Reservoir, obtain sample multi-spectral satellite images; It can be understood that the model training scheme involved in the solution of this application is to target the Shenzhen Reservoir, for which it is necessary to predict whether a bloom has occurred, and to specifically configure training samples to train a remote sensing prediction model for the Shenzhen Reservoir with high adaptability to the target Shenzhen Reservoir.

[0029] Among them, the target Shenzhen Reservoir can be any reservoir that meets the Shenzhen Reservoir determination criteria. For the sample multispectral satellite images of the target Shenzhen Reservoir, they can be actual images, such as past relevant satellite telemetry data products, or images obtained through corresponding data processing based on the actual images of the target Shenzhen Reservoir, or images directly prepared for the target Shenzhen Reservoir. These are all possible in actual situations.

[0030] As an example, multispectral satellite images of satellite telemetry data products such as HJ-2A / B, Sentinel-2, and GF series (GF1, GF1-B, GF1-C, GF1-D, GF6) can be specifically obtained.

[0031] It can be seen that in actual operation, the sample multispectral satellite images can also have the characteristics of multiple sources or different data sources. Therefore, further adjustments can be made according to the differences between different data sources to form fused multi-source remote sensing images to adapt to the model input after the actual application planned by the subsequent remote sensing prediction model for the Shenzhen Reservoir bloom.

[0032] In addition, it can be understood that the remote sensing prediction model for the Shenzhen Reservoir bloom to be trained in this application involves temporal features in the prediction process. Therefore, for the sample multispectral satellite images, it usually involves a relatively large time range, which is continuous multispectral satellite images.

[0033] As an example, continuous multispectral satellite images with a time span of 3 years can be used.

[0034] Step S102: Calculate the index values of each position of the sample multispectral satellite image based on the preset bloom characterization index; It can be understood that corresponding to the current image, during the model training process, corresponding annotations or true values are involved, that is, whether a bloom has occurred and whether there is a bloom situation, which can be determined in combination with the preset bloom characterization index.

[0035] In this way, the specific index values of each position of the sample multispectral satellite image under the preset bloom characterization index can be calculated. Among them, the position is usually in units of pixels. Correspondingly, what this application aims to achieve is the remote sensing prediction effect of water blooms at the pixel level or in terms of pixel granularity.

[0036] In this way, the delicate index values at various positions of the target Shenzhen Reservoir are used to lay a foundation for the subsequent production of the water bloom binary distribution layer.

[0037] Step S103: Combine the index values and index thresholds at various positions of the sample multispectral satellite image to determine whether there is a water bloom at each position of the multispectral satellite image, and form a water bloom binary distribution layer corresponding to the sample multispectral satellite image. It can be understood that the index value quantifies the degree of algae in the water body. Whether a water bloom has occurred in the actual situation needs to be divided by combining the corresponding threshold (which can be denoted as K).

[0038] Therefore, based on the index values at various positions of the sample multispectral satellite image obtained previously, combined with the index threshold, it can be determined whether there is a water bloom at each position of the multispectral satellite image. The determination result of whether there is a water bloom at each position of the multispectral satellite image itself corresponds to a binary processing, that is, there is a water bloom situation and there is no water bloom situation. In this case, the present application can continue to produce a corresponding water bloom binary distribution layer reflecting the overall situation.

[0039] Among them, the water bloom binary distribution layer is not only convenient for display but also convenient for the model to perform recognition processing, and it can be used as the corresponding annotation or true value involved in the model training process.

[0040] Step S104: On the basis of the sample multispectral satellite image and the water bloom binary distribution layer, quantify the influence of the cumulative effect of the preset index on the water bloom situation and the influence of the multi-day average value of the preset index on the water bloom situation, where the preset index includes water quality parameters, reservoir hydrological rhythm, and meteorological parameters. It can be understood that the present application believes that in the actual situation, different water bloom characterization indexes may have different contributions to the water bloom remote sensing prediction work of different Shenzhen Reservoirs. Therefore, screening can be carried out through the importance specifically designed in the present application to obtain the water bloom remote sensing prediction target adapted to the target Shenzhen Reservoir this time.

[0041] Under this design, the present application can, for the preset index in the preliminary stage, on the basis of the sample multispectral satellite image, on the one hand, quantify the influence of the cumulative effect of the preset index on the water bloom situation, and on the other hand, quantify the influence of the multi-day average value of the preset index on the water bloom situation. Among them, the multi-day average value corresponds to the configuration of the subsequent model and can involve the average values of 2 days, 3 days,..., 7 days.

[0042] In this step, it can be seen that this application not only discloses the three major aspects of parameter indicators (i.e., water quality parameters, reservoir hydrological rhythms, and meteorological parameters) designed in this application that can contribute to the remote sensing prediction of algal blooms, but also gives the two specific aspects that these parameter indicators can contribute to the remote sensing prediction of algal blooms (i.e., cumulative effect and the impact brought by multi-day means).

[0043] For the former, different water quality parameters and different meteorological parameters, it can be understood that they generally fall within the scope of the prior art. In specific operations, existing indicators can be directly adopted, or optimized (increased, decreased, replaced, etc.) based on existing indicators, or novel indicators developed independently can be used. For the reservoir hydrological rhythm, it is a new indicator additionally introduced in this application, which reflects the specific relevant laws of the Shenzhen University Reservoir itself in terms of hydrological changes and helps to provide a more suitable and accurate contribution to the remote sensing prediction of algal blooms.

[0044] As an example, water quality parameters can specifically involve total phosphorus, total nitrogen, chemical oxygen demand, chlorophyll a, etc., and meteorological parameters can specifically involve air temperature, air pressure, etc.

[0045] For the latter, this application does not directly capture the mapping relationship between the changes in the basic indicators and the changes in the algal bloom situation like conventional schemes / thinking. Instead, a cumulative effect and multi-day means are configured between the two, so as to more delicately and accurately infer the algal bloom situation from a deeper perspective.

[0046] Step S105, based on the quantification result, screen out target indicators that meet the conditions for the relative importance of remote sensing prediction of algal blooms from the preset indicators through a redundancy analysis method; After deeply quantifying the impacts of different aspects on hydrological prediction through the above steps, based on the corresponding quantification results, the response relationship between different indicator variables in the preset indicators at the initial stage and the binary distribution layer of algal blooms can be determined, and the relative importance of different indicator variables in the preset indicators at the initial stage for the hydrological prediction of the target Shenzhen University Reservoir can be determined. Then, under the specific redundancy analysis method adopted, multiple indicators that meet the preset relative importance conditions such as the relative importance being greater than the threshold or the relative importance ranking within the preset sequence are screened out and output as target indicators. These target indicators correspond to the model input content to be configured for the subsequent remote sensing prediction model of algal blooms in the Shenzhen University Reservoir.

[0047] Step S106: Configure training samples with the index data corresponding to the target index and the bloom binary distribution layer, and train the bloom remote sensing prediction model for the Shenzhen University Reservoir. Specifically, the bloom remote sensing prediction model for the Shenzhen University Reservoir is a long short-term memory network (LSTM) based on the self-attention mechanism. The bloom remote sensing prediction model for the Shenzhen University Reservoir is configured with 7 output layers, each output layer corresponding to the prediction output for different days within the next 7 days. The bloom remote sensing prediction model for the Shenzhen University Reservoir is used to predict whether there is a bloom situation in the next 7 days based on the index data input by the model corresponding to the target index.

[0048] It can be understood that, based on artificial intelligence (AI) technology, when configuring the bloom remote sensing prediction model for the Shenzhen University Reservoir through a corresponding machine learning model, the novel long short-term memory network based on the self-attention mechanism, namely KAN-LSTM, is specifically introduced to configure the bloom remote sensing prediction model for the Shenzhen University Reservoir.

[0049] Specifically, in the face of complex time series data, traditional machine learning models often struggle to uncover the deep-seated relationships between time series factors. Although recurrent neural networks, especially LSTM (Long Short-Term Memory Network), have performed well in the field of time series prediction and are widely used for the monitoring and early warning of blooms, the inventors of this application have found that it may overlook key early information when processing long time series data, which will limit the model's comprehensive grasp of global information and thus affect the accuracy of bloom remote sensing prediction.

[0050] To break through the above limitations, this application introduces a hybrid prediction model for the first time in the field of bloom remote sensing prediction, namely the long short-term memory network based on the self-attention mechanism (KAN-LSTM). This new model combines the self-attention mechanism, which can more effectively identify and capture key features in long time series data. It can not only finely process local features in the time series, effectively capture the time dependence of bloom occurrence, but also comprehensively focus on global information, and use the self-attention mechanism of key variables to improve the prediction performance, so as to more deeply understand the complex mechanism of bloom occurrence and significantly improve the accuracy and reliability of prediction.

[0051] For the long short-term memory network based on the self-attention mechanism introduced into the bloom remote sensing prediction work of this application, the time step of the model input layer is 7, and the model output layer is set to 7 neurons, that is, 7 output layers, corresponding to the bloom remote sensing predictions for the next 1 - 7 days respectively. Specifically, the sigmoid activation function is used to output 0 or 1 to indicate whether a bloom has occurred (usually 0 indicates no bloom, and 1 indicates a bloom, which corresponds to the marking method of the previous bloom binary distribution layer).

[0052] During the model training process, it generally includes the following content: In each round of model training, a training sample is input into the model, enabling the model to perform corresponding remote sensing prediction processing of algal blooms, achieving forward propagation. Then, based on the model output results, the loss function is calculated in combination with the binary distribution layer of algal blooms (true value), and the model parameters are optimized according to the calculation results of the loss function, achieving backpropagation. In this way, through a large number of rounds of training, when the model training requirements such as training duration, number of training times, and prediction accuracy are met, the training of the model can be completed, and a remote sensing prediction model of algal blooms in the Shenzhen University Reservoir that can be put into actual use is obtained.

[0053] It can be understood that for the specific model training scheme and the loss function adopted during the training process, either the existing scheme can be used, or it can be further optimized based on the existing scheme, or a novel self-developed scheme can be adopted.

[0054] As an example, during the training process, the cross-validation method can be used to divide the training set and the validation set, and key indicators such as accuracy, precision, recall rate, F1 score, and confusion matrix can be selected to evaluate the model performance to better display the model performance.

[0055] In addition, for the remote sensing prediction model of algal blooms in the Shenzhen University Reservoir, in addition to directly inputting the specific data set corresponding to the target indicators, it can also be configured with the processing ability to extract the specific data set corresponding to the target indicators from multi-spectral satellite images, which corresponds to two model input methods that can be adopted in actual situations.

[0056] From Figure 1 As can be seen from the illustrated embodiments, for the remote sensing prediction target of algal blooms in the Shenzhen University Reservoir, the present application innovatively applies the long short-term memory network based on the self-attention mechanism to the remote sensing prediction of algal blooms. This model integrates the self-attention mechanism and the long short-term memory network, and can more effectively identify and capture key features in long-time series data. Therefore, the remote sensing prediction model of algal blooms in the Shenzhen University Reservoir trained by the model training scheme based on the long short-term memory network with self-attention mechanism obtained a scheme effect of significantly improving the accuracy and reliability in the remote sensing prediction of algal blooms, and can provide important decision-making basis and technical support for the early warning of algal blooms and the subsequent prevention and control intervention of algal blooms in the Shenzhen University Reservoir, thereby effectively promoting the progress of the refined management of regional water resources and the work of water ecological protection.

[0057] Continue to elaborate in detail on each step of the above Figure 1 illustrated embodiments and their possible embodiments in actual applications.

[0058] As an exemplary embodiment, step S101, for the target Shenzhen University Reservoir, obtains the sample multi-spectral satellite image, which may specifically include: 1) For the target Shenzhen Reservoir, obtain the first sample multispectral satellite images that meet the requirements regarding image quality, cloud coverage, and basin coverage; It can be understood that the purpose here is to select multispectral satellite images with excellent image quality, less cloud coverage, and as comprehensive coverage of the research basin as possible during the acquisition of source data.

[0059] 2) Perform preprocessing including radiometric calibration and atmospheric correction on the first sample multispectral satellite images to obtain the second sample multispectral satellite images; It can be understood that the purpose here is to further effectively improve the image quality through preprocessing such as radiometric calibration and atmospheric correction. The preprocessing that can be adopted usually can directly follow the existing schemes. Of course, it does not exclude the possibility of adopting optimized schemes of the existing schemes or novel self-developed schemes.

[0060] 3) Using the pre-configured standard multispectral satellite images as a reference, perform georegistration on the second sample multispectral satellite images to obtain the sample multispectral satellite images.

[0061] It can be understood that the purpose here is to use the standard multispectral satellite images as a reference for georegistration, accurately adjust the image position, so as to obtain a more accurate high-quality image dataset, which ensures the spatial accuracy of the image data and provides a solid foundation for subsequent remote sensing prediction and analysis of algal blooms.

[0062] Among them, the pre-set standard multispectral satellite images can be either selected existing satellite observation data products that meet the high-quality requirements, such as Sentinel-2 satellite images, or multispectral satellite images processed based on existing satellite observation data products, or multispectral satellite images directly obtained at the theoretical level. These can all be flexibly configured according to specific requirements in actual situations.

[0063] In this way, through the scheme setting of this embodiment in this application, a practical implementation scheme is designed for how to obtain high-quality sample multispectral satellite images for the target Shenzhen Reservoir.

[0064] Furthermore, this application can also involve a super-resolution link to increase the resolution of the sample multispectral satellite images obtained in the initial stage to a high level. This corresponds to the situation where in actual cases, there is no need to directly obtain high-level images as sample multispectral satellite images and as model inputs after the model is put into actual use, which can effectively reduce the cost of scheme deployment.

[0065] Correspondingly, as an exemplary embodiment, between step S101 and step S102, the method of this application can further include: Perform super-resolution processing on the sample multispectral satellite image to increase the resolution of the sample multispectral satellite image to that of the standard multispectral satellite image. The temporal resolution of the standard multispectral satellite image is 24 hours (h) or 1 day, and the spatial resolution is 10 meters (m).

[0066] As can be seen from the embodiments herein, the water bloom remote sensing prediction effect to be achieved by this application specifically involves a high spatio-temporal resolution of 24 hours and 10 meters.

[0067] Furthermore, the super-resolution processing additionally introduced in this application can also be completed by a corresponding machine learning model, so as to efficiently and accurately complete the model processing task through the powerful processing performance of the AI model.

[0068] Specifically, as an exemplary embodiment, the super-resolution processing of the sample multispectral satellite image may specifically include: Perform super-resolution processing on the sample multispectral satellite image through a super-resolution processing model, where the super-resolution processing model specifically uses the RealESR-GAN model; The training process of the super-resolution processing model includes the following processing contents: Taking the standard multispectral satellite image as a reference, pair the low-resolution image with the corresponding standard multispectral satellite image to form a high-low spatial resolution training set; Use the high-low spatial resolution training set to train the super-resolution processing model, and during the training process, continuously optimize the generator and discriminator in the model through an adversarial process, and evaluate the quality of the generated image through corresponding visual evaluation indicators and quantitative indicators until the training requirements are met.

[0069] It can be seen that in the embodiments herein, this application introduces the Real Enhanced Super Resolution Generative Adversarial Network (RealESR-GAN) to perform the super-resolution processing designed by this application. After the Real Enhanced Super Resolution Generative Adversarial Network is trained, the generator in it is used to perform high-performance super-resolution processing on the image input to the network, so as to realize the construction of a sample multispectral satellite remote sensing image with a high temporal resolution (1 day) and a high spatial resolution (10 meters).

[0070] In addition, it can be understood that the multi-spectral satellite remote sensing images of the samples obtained in this application mentioned above can themselves involve different data sources, that is, multi-source remote sensing images. In the super-resolution processing process here, obviously, while effectively improving the spatio-temporal resolution, it also helps to promote the fusion between images from different data sources and obtain a multi-source remote sensing image dataset with high temporal / spatial resolution.

[0071] Meanwhile, corresponding to the algal bloom characterization index involved in step S102, as an exemplary embodiment, its index value can be specifically calculated by the following formula: Q = 0.8×F1×F1 / F2 - F1 + 0.2×F1×F3 / F2, where Q is the output index value, F1 is the remotely sensed reflectance after atmospheric correction in the red-edge band, F2 is the remotely sensed reflectance after atmospheric correction in the red band, and F3 is the remotely sensed reflectance after atmospheric correction in the green band.

[0072] It can be understood that the embodiment here starts from the formula level, focuses on the remotely sensed reflectance of different bands, and gives a set of specific implementation schemes for the algal bloom characterization index, which has good practical significance.

[0073] In addition, for the index threshold used in step S103, it can be understood that in actual applications, for the target Shenzhen University Reservoir this time, it can be either a pre-adapted fixed threshold, or a threshold adapted according to the actual situation within the selected time range, or a threshold adapted in real time when using the index threshold, in order to achieve the best timeliness and promote a more accurate binary processing of the algal bloom occurrence situation, and obtain an algal bloom binary distribution layer reflecting a deeper situation.

[0074] Furthermore, for the latter scheme setting, as an exemplary embodiment, the method of this application can further include: Obtaining the algal density data measured synchronously by the ground monitoring points and the sample multi-spectral satellite images; Based on the algal density data and the index values at each position of the sample multi-spectral satellite images, constructing a scatter plot fitting model, and determining the index threshold adapted to the current situation in combination with the algal bloom degree classification standard for algal density evaluation in the specification.

[0075] It can be noted that new data collection is involved here, that is, the algal density data (ground measured data) measured synchronously by the ground monitoring points deployed at the site of Shenzhen University Reservoir. The detection instruments involved usually directly adopt existing / general equipment.

[0076] It can be seen that in the embodiment here, this application provides a set of index threshold processing schemes that are easy to operate and can obtain high adaptability.

[0077] In addition, for the multispectral satellite images and algal density data involved in the above solution content, in terms of details, at least one of further parameter inversion, spatial interpolation, and resampling can be performed to ensure that data in different aspects can be aligned and matched, so as to better carry out the corresponding data processing above.

[0078] After completing the training of the bloom remote sensing prediction model for the target Shenzhen University Reservoir adapted to this time, obviously, it can be put into specific practical applications.

[0079] Correspondingly, as an exemplary embodiment, the method of the present application may further include: Obtain the multispectral satellite image to be predicted of the target Shenzhen University Reservoir according to the bloom remote sensing prediction requirement; Input the multispectral satellite image to be preset into the bloom remote sensing prediction model of the Shenzhen University Reservoir; Extract the bloom remote sensing prediction result output by the bloom remote sensing prediction model of the Shenzhen University Reservoir.

[0080] It can be understood that the multispectral satellite image to be predicted can either be manually input by the staff according to the requirement, or be obtained by the staff manually triggering the system to initiate a bloom remote sensing prediction task, or be obtained by the system according to the preset bloom remote sensing prediction task or a bloom remote sensing prediction task generated in real time independently. The specific acquisition of the multispectral satellite image corresponds to the sample multispectral satellite image before.

[0081] In this way, after obtaining the multispectral satellite image to be predicted that needs to be processed currently, it can be input into the bloom remote sensing prediction model of the Shenzhen University Reservoir to enable it to carry out the corresponding bloom remote sensing prediction processing. After the processing is completed, the bloom remote sensing prediction model of the Shenzhen University Reservoir will output the corresponding bloom remote sensing prediction result. At this time, the bloom remote sensing prediction result can be extracted, meeting the on-site or remote bloom monitoring requirements for the target Shenzhen University Reservoir, thereby providing important decision-making basis and technical support for the early bloom warning and later bloom prevention and control intervention of the target Shenzhen University Reservoir, and further effectively promoting the progress of the refined management of regional water resources and the work of water ecological protection.

[0082] In this case, for the bloom remote sensing prediction result, corresponding data application processing may also be involved. For example, it can be stored locally, stored remotely, the result can be displayed, a prompt for the completion of the prediction can be output, or further data analysis can be performed, etc. It can be understood that the specific data application processing can be flexibly adjusted according to the pre-configured and real-time configured data application strategies / rules.

[0083] The above is an introduction to the processing method of the Shenzhen University Reservoir algal bloom remote sensing prediction model. To facilitate the better implementation of the processing method of the Shenzhen University Reservoir algal bloom remote sensing prediction model provided in this application, this application also provides a processing device for the Shenzhen University Reservoir algal bloom remote sensing prediction model from the perspective of functional modules.

[0084] Refer to Figure 2 , Figure 2 , which is a schematic structural diagram of a processing device for the Shenzhen University Reservoir algal bloom remote sensing prediction model in this application. In this application, the processing device 200 for the Shenzhen University Reservoir algal bloom remote sensing prediction model may specifically include the following structures: An acquisition unit 201, configured to acquire sample multispectral satellite images for a target Shenzhen University Reservoir; A calculation unit 202, configured to calculate the index values of each position of the sample multispectral satellite images based on a preset algal bloom characterization index; A determination unit 203, configured to determine whether there is an algal bloom situation at each position of the multispectral satellite images by combining the index values and index thresholds of each position of the sample multispectral satellite images, and form a binary algal bloom distribution layer corresponding to the sample multispectral satellite images; A quantification unit 204, configured to quantify the influence of the cumulative effect of preset indexes on the algal bloom situation and the influence of the multi-day average value of the preset indexes on the algal bloom situation based on the sample multispectral satellite images and the binary algal bloom distribution layer, where the preset indexes include water quality parameters, reservoir hydrological rhythms, and meteorological parameters; A screening unit 205, configured to screen out target indexes that meet the conditions for the relative importance of algal bloom remote sensing prediction from the preset indexes by using a redundancy analysis method based on the quantification results; A training unit 206, configured to configure training samples with the index data corresponding to the target indexes and the binary algal bloom distribution layer to train the Shenzhen University Reservoir algal bloom remote sensing prediction model. The Shenzhen University Reservoir algal bloom remote sensing prediction model is specifically a long short-term memory network based on a self-attention mechanism. The Shenzhen University Reservoir algal bloom remote sensing prediction model is configured with 7 output layers, and each output layer corresponds to the prediction output for different days within the next 7 days. The Shenzhen University Reservoir algal bloom remote sensing prediction model is used to predict whether there is an algal bloom situation within the next 7 days based on the index data input into the model corresponding to the target indexes.

[0085] In an exemplary embodiment, the acquisition unit 201 is specifically configured to: For a target Shenzhen University Reservoir, acquire first sample multispectral satellite images that meet the requirements regarding image quality, cloud coverage, and basin coverage; Perform preprocessing on the first sample multispectral satellite images, including radiometric calibration and atmospheric correction, to obtain second sample multispectral satellite images; Georegister the second sample multispectral satellite image with a pre-configured standard multispectral satellite image as a reference to obtain a sample multispectral satellite image.

[0086] In yet another exemplary embodiment, the apparatus further includes a processing unit 207 for: Perform super-resolution processing on the sample multispectral satellite image to increase the resolution of the sample multispectral satellite image to the resolution of the standard multispectral satellite image, where the temporal resolution of the standard multispectral satellite image is 24 hours and the spatial resolution is 10 meters.

[0087] In yet another exemplary embodiment, the processing unit 207 is specifically configured to: Perform super-resolution processing on the sample multispectral satellite image through a super-resolution processing model, where the super-resolution processing model specifically uses the RealESR-GAN model; The training process of the super-resolution processing model includes the following processing contents: Taking the standard multispectral satellite image as a reference, pair the low-resolution image with the corresponding standard multispectral satellite image to form a high-low spatial resolution training set; Use the high-low spatial resolution training set to train the super-resolution processing model, and during the training process, continuously optimize the generator and discriminator in the model through an adversarial process, and evaluate the quality of the generated image through corresponding visual evaluation indicators and quantitative indicators until the training requirements are met.

[0088] In yet another exemplary embodiment, the index value is specifically calculated by the following formula: Q = 0.8×F1×F1 / F2 - F1 + 0.2×F1×F3 / F2, where Q is the index value as an output, F1 is the remotely sensed reflectance after atmospheric correction in the red-edge band, F2 is the remotely sensed reflectance after atmospheric correction in the red band, and F3 is the remotely sensed reflectance after atmospheric correction in the green band.

[0089] In yet another exemplary embodiment, the determination unit 203 is further configured to: Obtain the algae density data measured synchronously with the sample multispectral satellite image at the ground monitoring point; Based on the algae density data and the index values at each position of the sample multispectral satellite image, construct a scatter plot fitting model, and determine the index threshold suitable for the current situation in combination with the water bloom degree classification standard for algae density evaluation in the specification.

[0090] In yet another exemplary embodiment, the apparatus further includes an application unit 208 for: Obtain the multispectral satellite image to be predicted of the target deep reservoir according to the water bloom remote sensing prediction requirement; Input the multi-spectral satellite images to be preset into the remote sensing prediction model for algal blooms in Shenzhen University Reservoir; Extract the remote sensing prediction results of algal blooms output by the remote sensing prediction model for algal blooms in Shenzhen University Reservoir.

[0091] This application also provides a processing device from the perspective of hardware structure. Refer to Figure 3 , Figure 3 which shows a schematic structural diagram of the processing device of this application. Specifically, the processing device of this application may include a processor 301, a memory 302, and an input / output device 303. When the processor 301 executes the computer program stored in the memory 302, it realizes the steps of the processing method of the remote sensing prediction model for algal blooms in Shenzhen University Reservoir in the corresponding embodiment; or, when the processor 301 executes the computer program stored in the memory 302, it realizes the functions of each unit in the corresponding embodiment as described in Figure 1 The memory 302 is used to store the computer program required for the processor 301 to execute the processing method of the remote sensing prediction model for algal blooms in Shenzhen University Reservoir in the above corresponding embodiment. Figure 2 Figure 1

[0092] Exemplarily, the computer program can be divided into one or more modules / units. One or more modules / units are stored in the memory 302 and executed by the processor 301 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of completing specific functions, and these instruction segments are used to describe the execution process of the computer program in the computer device.

[0093] The processing device may include, but is not limited to, the processor 301, the memory 302, and the input / output device 303. Those skilled in the art can understand that the schematic diagram is only an example of the processing device, and does not constitute a limitation on the processing device. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, the processing device may also include a network access device, a bus, etc. The processor 301, the memory 302, the input / output device 303, etc. are connected through the bus.

[0094] ​​The processor 301 can be a Central Processing Unit (CPU), or other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the processing device, connecting all parts of the entire device through various interfaces and lines.

[0095] The memory 302 can be used to store computer programs and / or modules. The processor 301 realizes various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 302, and by calling the data stored in the memory 302. The memory 302 mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc.; the data storage area can store data created according to the use of the processing device, etc. In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disks, memory, plug-in hard disks, Smart Media Cards (SMCs), Secure Digital (SD) cards, Flash Cards, at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices.

[0096] When the processor 301 is used to execute the computer program stored in the memory 302, the following functions can be specifically realized: For the target deep reservoir, obtain sample multispectral satellite images; Based on the preset water bloom characterization index, calculate the index values of each position of the sample multispectral satellite image; Combining the index values of each position of the sample multispectral satellite image and the index threshold, determine whether there is a water bloom at each position of the multispectral satellite image, and form a binary distribution layer of water bloom corresponding to the sample multispectral satellite image; Based on the sample multispectral satellite image and the binary distribution layer of water bloom, quantify the influence of the cumulative effect of the preset index on the water bloom situation and the influence of the multi-day average value of the preset index on the water bloom situation, where the preset index includes water quality parameters, reservoir hydrological rhythms, and meteorological parameters; Based on the quantization results, target indicators that meet the conditions for the relative importance of bloom remote sensing prediction are screened from the preset indicators through a redundancy analysis method; Configure training samples with the indicator data corresponding to the target indicators and the bloom binary distribution layer, and train the bloom remote sensing prediction model for the Shenzhen University Reservoir. Among them, the bloom remote sensing prediction model for the Shenzhen University Reservoir is specifically a long short-term memory network based on the self-attention mechanism. The bloom remote sensing prediction model for the Shenzhen University Reservoir is configured with 7 output layers, and each output layer corresponds to the prediction output for different days within the next 7 days. The bloom remote sensing prediction model for the Shenzhen University Reservoir is used to predict whether there is a bloom situation in the next 7 days based on the indicator data input by the model corresponding to the target indicators.

[0097] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the processing device, processing equipment and their corresponding units of the bloom remote sensing prediction model for the Shenzhen University Reservoir described above can refer to Figure 1 the description of the processing method of the bloom remote sensing prediction model for the Shenzhen University Reservoir in the corresponding embodiment, and will not be elaborated here specifically.

[0098] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or by controlling relevant hardware through instructions. The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0099] Therefore, the present application provides a computer-readable storage medium, which stores multiple instructions that can be loaded by a processor to execute the steps of the processing method of the bloom remote sensing prediction model for the Shenzhen University Reservoir in the present application as Figure 1 described in the corresponding embodiment. The specific operations can refer to Figure 1 the description of the processing method of the bloom remote sensing prediction model for the Shenzhen University Reservoir in the corresponding embodiment, and will not be elaborated here.

[0100] Among them, the computer-readable storage medium may include: Read Only Memory (ROM), Random Access Memory (RAM), magnetic disk or optical disk, etc.

[0101] Since the instructions stored in the computer-readable storage medium can execute the steps of the processing method of the bloom remote sensing prediction model for the Shenzhen University Reservoir in the present application as Figure 1 described in the corresponding embodiment, the beneficial effects that can be achieved by the processing method of the bloom remote sensing prediction model for the Shenzhen University Reservoir in the present application can be realized. For details, please refer to the previous description and will not be elaborated here. Figure 1

[0102] ​The above has introduced in detail the processing method, device, processing equipment and computer-readable storage medium of the bloom remote sensing prediction model for the Shenzhen University Reservoir provided by this application. Specific examples are used in this article to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the core idea of this application; at the same time, for those skilled in the art, according to the idea of this application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A processing method for a remote sensing prediction model of algal blooms in deep and large reservoirs, characterized in that, The method includes: For a target deep and large reservoir, obtaining sample multispectral satellite images; Based on a preset water bloom characterization index, calculating the index values at each position of the sample multispectral satellite images; Combining the index values and index thresholds at each position of the sample multispectral satellite images to determine whether there is a water bloom situation at each position of the multispectral satellite images, and forming a binary water bloom distribution layer corresponding to the sample multispectral satellite images; Based on the sample multispectral satellite images and the binary water bloom distribution layer, quantifying the influence of the cumulative effect of preset indicators on the water bloom situation and the influence of the multi-day average value of the preset indicators on the water bloom situation, where the preset indicators include water quality parameters, reservoir hydrological rhythms, and meteorological parameters; Based on the quantification results, screening out target indicators that meet the conditions for the relative importance of water bloom remote sensing prediction from the preset indicators through a redundancy analysis method; Configuring training samples with the index data corresponding to the target indicators and the binary water bloom distribution layer to train a water bloom remote sensing prediction model for the deep and large reservoir, where the water bloom remote sensing prediction model for the deep and large reservoir is specifically a long short-term memory network based on a self-attention mechanism, the water bloom remote sensing prediction model for the deep and large reservoir is configured with 7 output layers, each output layer corresponding to the prediction output for different days within the next 7 days, and the water bloom remote sensing prediction model for the deep and large reservoir is used to predict whether there is a water bloom situation within the next 7 days based on the index data input into the model corresponding to the target indicators.

2. The method according to claim 1, characterized in that, For the target deep and large reservoir, obtaining the sample multispectral satellite images includes: For the target deep and large reservoir, obtaining a first sample multispectral satellite image that meets the requirements regarding image quality, cloud coverage, and basin coverage; Performing preprocessing on the first sample multispectral satellite image, including radiometric calibration and atmospheric correction, to obtain a second sample multispectral satellite image; Using a pre-configured standard multispectral satellite image as a reference to perform georegistration on the second sample multispectral satellite image to obtain the sample multispectral satellite image.

3. The method according to claim 2, wherein The method further includes: Performing super-resolution processing on the sample multispectral satellite image to increase the resolution of the sample multispectral satellite image to the resolution of the standard multispectral satellite image, where the time resolution of the standard multispectral satellite image is 24 hours and the spatial resolution is 10 meters.

4. The method according to claim 3, characterized in that Performing the super-resolution processing on the sample multispectral satellite image includes: Performing the super-resolution processing on the sample multispectral satellite image through a super-resolution processing model, where the super-resolution processing model specifically uses the RealESR-GAN model; The training process of the super-resolution processing model includes the following processing contents: Using the standard multispectral satellite image as a benchmark, pairing the low-resolution image with the corresponding standard multispectral satellite image to form a high-low spatial resolution training set; Train the super-resolution processing model using the high-low spatial resolution training set, and during the training process, continuously optimize the generator and discriminator in the model through an adversarial process, and evaluate the quality of the generated images through corresponding visual evaluation metrics and quantitative metrics until the training requirements are met.

5. The method according to claim 1, wherein The specific value of the index is calculated by the following formula: Q = 0.8×F1×F1 / F2 - F1 + 0.2×F1×F3 / F2, where Q is the value of the index as the output, F1 is the remote sensing reflectance after atmospheric correction in the red-edge band, F2 is the remote sensing reflectance after atmospheric correction in the red band, and F3 is the remote sensing reflectance after atmospheric correction in the green band.

6. The method according to claim 5, characterized in that, The method further includes: Obtain the algae density data measured synchronously with the sample multispectral satellite image at the ground monitoring point; Based on the algae density data and the index values at each position of the sample multispectral satellite image, construct a scatter plot fitting model, and determine the index threshold suitable for the current situation in combination with the water bloom degree classification standard for algae density evaluation in the specification.

7. The method according to claim 1, characterized in that The method further includes: According to the water bloom remote sensing prediction requirement, obtain the multispectral satellite image to be predicted of the target Shenzhen Reservoir; Input the multispectral satellite image to be preset into the water bloom remote sensing prediction model of the Shenzhen Reservoir; Extract the water bloom remote sensing prediction result output by the water bloom remote sensing prediction model of the Shenzhen Reservoir.

8. A processing device for a remote sensing prediction model of algal blooms in deep and large reservoirs, characterized in that, The device includes: An acquisition unit for acquiring a sample multispectral satellite image for the target Shenzhen Reservoir; A calculation unit for calculating the index value at each position of the sample multispectral satellite image based on a preset water bloom characterization index; A determination unit for determining whether there is a water bloom situation at each position of the multispectral satellite image in combination with the index value and the index threshold at each position of the sample multispectral satellite image, and forming a water bloom binary distribution layer corresponding to the sample multispectral satellite image; A quantification unit for quantifying the influence of the cumulative effect of the preset index on the water bloom situation and the influence of the multi-day average value of the preset index on the water bloom situation based on the sample multispectral satellite image and the water bloom binary distribution layer, where the preset index includes water quality parameters, reservoir hydrological rhythm, and meteorological parameters; A screening unit for screening out target indexes that meet the conditions for the relative importance of water bloom remote sensing prediction from the preset indexes through a redundancy analysis method based on the quantification result; A training unit for configuring the index data corresponding to the target index and the water bloom binary distribution layer as training samples to train the water bloom remote sensing prediction model of the Shenzhen Reservoir, where the water bloom remote sensing prediction model of the Shenzhen Reservoir is specifically a long short-term memory network based on the self-attention mechanism, the water bloom remote sensing prediction model of the Shenzhen Reservoir is configured with 7 output layers, each output layer corresponding to the prediction output for different days within the next 7 days, and the water bloom remote sensing prediction model of the Shenzhen Reservoir is used to predict whether there is a water bloom situation within the next 7 days based on the index data input by the model corresponding to the target index.

9. A processing device, characterized in that, It includes a processor and a memory, and a computer program is stored in the memory. When the processor calls the computer program in the memory, it executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores multiple instructions, and the instructions are suitable for being loaded by a processor to execute the method according to any one of claims 1 to 7.

Images