Precipitation prediction method and system based on cloud parameters

Abstract

The invention discloses a precipitation prediction method and system based on cloud parameters, and the method comprises the following steps: obtaining a cloud layer data set from a remote sensing platform, carrying out the denoising and correction of the cloud layer data set, and obtaining a processed cloud layer data set; segmenting a complete cloud layer image from the processed cloud layer data set, performing feature extraction on the complete cloud layer image based on a convolutional neural network to obtain cloud layer parameters, and performing correlation analysis on the cloud layer parameters and precipitation to obtain cloud layer key parameters; and based on the cloud layer key parameters, constructing a primary function by adopting a random forest algorithm, and carrying out Boosting ensemble learning on a prediction result of the primary function to obtain a final rainfall prediction result. According to the method, the precipitation is predicted by combining the random forest algorithm and the Boosting ensemble learning technology, and the prediction accuracy and robustness are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of meteorological forecasting, and in particular relates to a precipitation forecasting method and system based on cloud parameters. Background Art

[0002] As a key part of weather forecasting, precipitation prediction has a far-reaching impact on many fields such as agricultural planning, water resources management, urban planning, disaster prevention, etc. In recent years, with the development of science and technology, especially the advancement of remote sensing technology and big data processing technology, precipitation prediction methods based on cloud parameters have shown great potential and application value.

[0003] Traditional precipitation prediction methods mostly rely on ground observation station data, satellite cloud image analysis and numerical weather forecast models, but these methods have problems such as high data acquisition cost, poor real-time performance and limited prediction accuracy. Therefore, in order to improve the real-time performance and prediction accuracy of precipitation prediction, it is urgent to propose a precipitation prediction method and system based on cloud parameters. Summary of the invention

[0004] To solve the above technical problems, the present invention proposes a precipitation prediction method and system based on cloud parameters, aiming to predict the precipitation within a certain period of time in the future by analyzing cloud characteristics with high precision, so as to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides a precipitation prediction system based on cloud parameters, comprising: a data collection module, a parameter extraction module, a precipitation prediction module and a result output module;

[0006] The data collection module is used to obtain a cloud data set from a remote sensing platform, and to denoise and correct the cloud data set to obtain a processed cloud data set;

[0007] The parameter extraction module is used to segment a complete cloud image from the processed cloud data set, and perform feature extraction on the complete cloud image based on a convolutional neural network to obtain cloud parameters, and perform correlation analysis between the cloud parameters and precipitation to obtain key cloud parameters;

[0008] The precipitation prediction module is used to construct a basis function based on key cloud parameters using a random forest algorithm, and obtain the final precipitation prediction result by performing Boosting ensemble learning on the prediction results of the basis function;

[0009] The result output module is used to output the prediction results to the user in the form of charts or reports, and supports multi-platform access.

[0010] Optionally, the data collection module includes a data collection unit and a data preprocessing unit;

[0011] The data collection unit is used to obtain multispectral images and radar echo data of clouds from remote sensing platforms including satellites, radars and drones to form a cloud data set;

[0012] The data preprocessing unit is used to perform denoising on the cloud dataset based on a median filtering algorithm, perform geometric correction on the cloud dataset based on remote sensing processing software, and perform radiation correction on the cloud dataset based on an atmospheric correction model to obtain a processed cloud dataset.

[0013] Optionally, the parameter extraction module includes an image segmentation unit, a feature extraction unit and a correlation analysis unit;

[0014] The image segmentation unit is used to segment the cloud layer from the background based on the grayscale difference between the cloud layer and the background according to a preset threshold value for the processed cloud layer data set, and then gradually expand the area based on the similarity between pixels starting from the pixel points at the edge of the segmented cloud layer to obtain a complete cloud layer image;

[0015] The feature extraction unit is used to extract cloud parameters from the complete cloud image based on a convolutional neural network;

[0016] The correlation analysis unit is used to calculate the correlation coefficient between each cloud layer parameter and precipitation, and then analyze the relationship between the interaction between several cloud layer parameters and precipitation in combination with a multivariate analysis regression method, so as to obtain key cloud layer parameters related to precipitation, wherein the key cloud layer parameters include at least cloud top height, cloud optical thickness and cloud particle effective radius.

[0017] Optionally, the precipitation prediction module includes a model building unit, an integrated learning unit and a model evaluation unit;

[0018] The model building unit is used to build a precipitation prediction model using a random forest algorithm based on the acquired key cloud parameters, and use the precipitation prediction model as a basis function; introduce historical precipitation data, divide the historical precipitation data into a training set and a validation set, and train the basis function according to a preset order based on the training set;

[0019] The ensemble learning unit is used to adjust the weight of the training parameter according to the prediction result of the previous model during the training process of the basis function based on the Boosting ensemble learning method, and then perform weighted summation of the prediction results of all basis functions according to certain weights to obtain the final precipitation prediction result;

[0020] The model evaluation unit is used to evaluate the final precipitation prediction result based on the verification set, and adjust the type of the base model or the parameters of the base model based on the evaluation result.

[0021] Optionally, the result output module includes a basis function after the result prediction unit and the result display unit, and obtains a precipitation prediction value within a preset time period through integrated learning;

[0022] The result prediction unit is used to input the key cloud layer parameters acquired in real time into the training;

[0023] The result display unit is used to output the prediction results to the user in the form of charts or reports through a visual interface or API interface, and supports multi-platform access.

[0024] The present invention also provides a precipitation prediction method based on cloud parameters and a precipitation prediction system based on cloud parameters, comprising the following steps:

[0025] Acquire a cloud dataset from a remote sensing platform, denoise and correct the cloud dataset, and obtain a processed cloud dataset;

[0026] Segment a complete cloud image from the processed cloud data set, perform feature extraction on the complete cloud image based on a convolutional neural network to obtain cloud parameters, perform correlation analysis between the cloud parameters and precipitation, and obtain key cloud parameters;

[0027] Based on the key parameters of the cloud layer, the random forest algorithm is used to construct the basis function, and the prediction results of the basis function are subjected to Boosting ensemble learning to obtain the final precipitation prediction results.

[0028] Optionally, the process of processing the cloud data set includes: denoising the cloud data set based on a median filtering algorithm, geometrically correcting the cloud data set based on remote sensing processing software, and radiation correcting the cloud data set based on an atmospheric correction model to obtain a processed cloud data set.

[0029] Optionally, the process of obtaining key cloud parameters includes: analyzing the relationship between the interactions between several cloud parameters and precipitation using a multivariate analysis regression method, thereby obtaining key cloud parameters related to precipitation, wherein the key cloud parameters include at least cloud top height, cloud optical thickness and effective radius of cloud particles.

[0030] Optionally, the process of obtaining the final precipitation prediction result includes: based on the acquired key cloud parameters, respectively using the random forest algorithm to construct a precipitation prediction model, and using the precipitation prediction model as a basis function; introducing historical precipitation data to train the basis function in a preset order; based on the Boosting ensemble learning method, during the training process of the basis function, adjusting the weights of the training parameters according to the prediction results of the previous model, and then weighted summing the prediction results of all basis functions according to certain weights to obtain the final precipitation prediction result.

[0031] Optionally, after obtaining the final precipitation forecast result, the method further includes: outputting the forecast result to the user in the form of a chart or report, and supporting multi-platform access.

[0032] Compared with the prior art, the present invention has the following advantages and technical effects:

[0033] The present invention uses a convolutional neural network to automatically extract features from cloud images, which can not only identify complete cloud images, but also deeply analyze cloud characteristics and provide key cloud parameters for precipitation prediction.

[0034] By analyzing the correlation between the extracted cloud parameters and precipitation, the system of the present invention can identify the cloud parameters that are most critical to precipitation prediction, thereby enhancing the pertinence and effectiveness of the prediction model.

[0035] The present invention combines the random forest algorithm and the Boosting ensemble learning technology to improve the accuracy and robustness of the prediction by constructing a basis function and integrating multiple prediction results.

[0036] The present invention supports displaying the prediction results in the form of charts or reports, and these results can support multi-platform access, thereby improving the usability and convenience of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0038] Figure 1 Schematic diagram of the structure of a precipitation prediction system based on cloud parameters according to an embodiment of the present invention;

[0039] Figure 2 The figure is a flow chart of a precipitation prediction method based on cloud parameters according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0042] Embodiment 1

[0043] like Figure 1As shown, this embodiment provides a precipitation prediction system based on cloud parameters, including: a data collection module, a parameter extraction module, a precipitation prediction module and a result output module;

[0044] The data collection module is used to obtain a cloud data set from a remote sensing platform, and to denoise and correct the cloud data set to obtain a processed cloud data set;

[0045] The parameter extraction module is used to segment a complete cloud image from the processed cloud data set, and perform feature extraction on the complete cloud image based on a convolutional neural network to obtain cloud parameters, and perform correlation analysis between the cloud parameters and precipitation to obtain key cloud parameters;

[0046] The precipitation prediction module is used to construct a basis function based on key cloud parameters using a random forest algorithm, and obtain the final precipitation prediction result by performing Boosting ensemble learning on the prediction results of the basis function;

[0047] The result output module is used to output the prediction results to the user in the form of charts or reports, and supports multi-platform access.

[0048] It can be implemented that the data collection module includes a data collection unit and a data preprocessing unit;

[0049] The data collection unit is used to obtain multispectral images and radar echo data of clouds from remote sensing platforms including satellites, radars and drones to form a cloud data set; the data collection unit is also used to collect multispectral images and radar echo data of clouds from a variety of remote sensing platforms, such as satellites, radars, drones, etc. These data provide the reflection, transmission and scattering characteristics of clouds at different wavelengths, as well as the three-dimensional structure information of clouds.

[0050] The data preprocessing unit is used to perform denoising on the cloud dataset based on a median filter algorithm, perform geometric correction on the cloud dataset based on remote sensing processing software, and perform radiation correction on the cloud dataset based on an atmospheric correction model to obtain a processed cloud dataset. The data preprocessing unit is also used for denoising: using a median filter algorithm to denoise the cloud dataset to reduce random noise and interference in the image and improve image quality; geometric correction: using remote sensing processing software to perform geometric correction on the image to eliminate geometric distortion of the image caused by factors such as sensor attitude and earth curvature; radiation correction: based on an atmospheric correction model, performing radiation correction on the image to eliminate the effects of atmospheric absorption, scattering, etc. on the radiation brightness of the image and restore the true reflectivity or emissivity of the ground object.

[0051] It can be implemented that the parameter extraction module includes an image segmentation unit, a feature extraction unit and a correlation analysis unit;

[0052] Among them, the image segmentation unit is used to segment the clouds from the background based on the grayscale difference between the clouds and the background and the preset threshold for the processed cloud data set, and then gradually expand the area based on the similarity between pixels starting from the pixel points at the edge of the segmented clouds to obtain a complete cloud image; the feature extraction unit is used to extract cloud parameters from the complete cloud image based on a convolutional neural network; the image segmentation unit is also used to preliminarily segment the clouds from the background based on the grayscale difference between the clouds and the background and preset one or more thresholds, and then, starting from the edge of the segmented clouds, gradually expand the area using a region growing algorithm to form a complete cloud image.

[0053] The correlation analysis unit is used to calculate the correlation coefficient between each cloud layer parameter and precipitation, and then analyze the relationship between the interaction between several cloud layer parameters and precipitation in combination with a multivariate analysis regression method, so as to obtain key cloud layer parameters related to precipitation, wherein the key cloud layer parameters include at least cloud top height, cloud optical thickness and cloud particle effective radius.

[0054] It can be implemented that the precipitation prediction module includes a model construction unit, an integrated learning unit and a model evaluation unit; the model construction unit is used to construct a precipitation prediction model based on the acquired key cloud parameters using a random forest algorithm, and use the precipitation prediction model as a basis function; historical precipitation data is introduced, and the historical precipitation data is divided into a training set and a validation set, and the basis function is trained in a preset order based on the training set; the integrated learning unit is used to adjust the weights of the training parameters according to the prediction results of the previous model during the training of the basis function based on the Boosting integrated learning method, and then weighted sum the prediction results of all basis functions according to certain weights to obtain the final precipitation prediction result; the model evaluation unit is used to evaluate the final precipitation prediction result based on the validation set, and adjust the type of the base model or the parameters of the base model based on the evaluation result.

[0055] This embodiment uses a random forest algorithm to build a precipitation prediction model based on the extracted key cloud parameters. The random forest improves the accuracy and stability of the prediction by constructing multiple decision trees and integrating their prediction results. During model training, historical precipitation data is introduced as a training set to train the random forest model to evaluate the performance of the model. During the training process, the Boosting ensemble learning method is used to dynamically adjust the training parameter weights of the subsequent model according to the prediction results of the previous model. In this way, each base model can conduct targeted learning for the shortcomings of the previous model, thereby improving the overall prediction performance. Finally, the final precipitation prediction results are evaluated based on the validation set. Commonly used evaluation indicators include mean square error (MSE), root mean square error (RMSE), correlation coefficient (R), etc. According to the evaluation results, the type of base model, such as the number and depth of decision trees or parameters, is adjusted to optimize the performance of the model.

[0056] The result output module can be implemented to include a result prediction unit and a result display unit; the result prediction unit is used to input the key cloud parameters obtained in real time into the trained basis function, and obtain the precipitation prediction value within a preset time period through integrated learning; the result display unit is used to output the prediction results to the user in the form of charts or reports through a visual interface or API interface, supporting multi-platform access.

[0057] As a specific embodiment, this embodiment provides a visual interface: the forecast results are displayed to the user in the form of charts, such as line charts, bar charts, etc. or reports through a graphical user interface (GUI), which intuitively reflects the changing trend of precipitation. An API interface is also provided to allow other systems or applications to access the precipitation forecast results programmatically to achieve data sharing and integration. Multiple access platforms are supported, such as Web browsers, mobile applications, desktop software, etc., to meet the needs of different users.

[0058] Effect: The embodiment of the present invention uses a convolutional neural network to automatically extract features from cloud images, which can not only identify complete cloud images, but also deeply analyze cloud characteristics and provide key cloud parameters for precipitation prediction.

[0059] By analyzing the correlation between the extracted cloud parameters and precipitation, the system of the embodiment of the present invention can identify the cloud parameters that are most critical to precipitation prediction, thereby enhancing the pertinence and effectiveness of the prediction model.

[0060] The embodiment of the present invention combines the random forest algorithm and the Boosting ensemble learning technology to improve the accuracy and robustness of the prediction by constructing a basis function and integrating multiple prediction results.

[0061] The embodiment of the present invention supports displaying the prediction results in the form of charts or reports, and these results can support multi-platform access, thereby improving the availability and convenience of the system.

[0062] Embodiment 2

[0063] like Figure 2 As shown, this embodiment provides a precipitation prediction method based on cloud parameters. The precipitation prediction method based on cloud parameters described below and the precipitation prediction system based on cloud parameters described above can refer to each other, and include the following steps:

[0064] Acquire a cloud dataset from a remote sensing platform, denoise and correct the cloud dataset, and obtain a processed cloud dataset;

[0065] Segment a complete cloud image from the processed cloud data set, perform feature extraction on the complete cloud image based on a convolutional neural network to obtain cloud parameters, perform correlation analysis between the cloud parameters and precipitation, and obtain key cloud parameters;

[0066] Based on the key parameters of the cloud layer, the random forest algorithm is used to construct the basis function, and the prediction results of the basis function are subjected to Boosting ensemble learning to obtain the final precipitation prediction results.

[0067] The process of processing the cloud data set can be implemented as follows: denoising the cloud data set based on a median filtering algorithm, geometrically correcting the cloud data set based on remote sensing processing software, and radiometrically correcting the cloud data set based on an atmospheric correction model to obtain a processed cloud data set.

[0068] The process of obtaining key cloud parameters can be implemented by analyzing the relationship between the interaction between several cloud parameters and precipitation using a multivariate analysis regression method, thereby obtaining key cloud parameters related to precipitation, wherein the key cloud parameters include at least cloud top height, cloud optical thickness and cloud particle effective radius.

[0069] The process of obtaining the final precipitation prediction result can be implemented, including: based on the obtained key cloud parameters, respectively using the random forest algorithm to construct precipitation prediction models, and using the precipitation prediction models as basis functions; introducing historical precipitation data to train the basis functions in a preset order; based on the Boosting ensemble learning method, in the training process of the basis functions, adjusting the weights of the training parameters according to the prediction results of the previous model, and then weighted summing the prediction results of all basis functions according to certain weights to obtain the final precipitation prediction result.

[0070] It is feasible that after obtaining the final precipitation forecast result, the forecast result is also output to the user in the form of a chart or report, and multi-platform access is supported.

[0071] It should be understood that the precipitation prediction method based on cloud parameters provided in the embodiment of the present invention has all the advantages of the precipitation prediction system based on cloud parameters provided in the above-mentioned embodiment.

[0072] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A precipitation prediction system based on cloud parameters, characterized in that: include: Data collection module, parameter extraction module and precipitation prediction module; The data collection module is used to obtain a cloud data set from a remote sensing platform, and to denoise and correct the cloud data set to obtain a processed cloud data set; The parameter extraction module is used to segment a complete cloud image from the processed cloud data set, and perform feature extraction on the complete cloud image based on a convolutional neural network to obtain cloud parameters, and perform correlation analysis between the cloud parameters and precipitation to obtain key cloud parameters; The precipitation prediction module is used to construct a basis function based on key cloud parameters using a random forest algorithm, and obtain the final precipitation prediction result by performing Boosting ensemble learning on the prediction results of the basis function.

2. The precipitation prediction system based on cloud parameters according to claim 1, characterized in that: The data collection module includes a data collection unit and a data preprocessing unit; The data collection unit is used to obtain multispectral images and radar echo data of clouds from remote sensing platforms including satellites, radars and drones to form a cloud data set; The data preprocessing unit is used to perform denoising on the cloud dataset based on a median filtering algorithm, perform geometric correction on the cloud dataset based on remote sensing processing software, and perform radiation correction on the cloud dataset based on an atmospheric correction model to obtain a processed cloud dataset.

3. The precipitation prediction system based on cloud parameters according to claim 1, characterized in that: The parameter extraction module includes an image segmentation unit, a feature extraction unit and a correlation analysis unit; The image segmentation unit is used to segment the cloud layer from the background based on the grayscale difference between the cloud layer and the background according to a preset threshold value for the processed cloud layer data set, and then gradually expand the area based on the similarity between pixels starting from the pixel points at the edge of the segmented cloud layer to obtain a complete cloud layer image; The feature extraction unit is used to extract cloud parameters from the complete cloud image based on a convolutional neural network; The correlation analysis unit is used to calculate the correlation coefficient between each cloud layer parameter and precipitation, and then analyze the relationship between the interaction between several cloud layer parameters and precipitation in combination with a multivariate analysis regression method, so as to obtain key cloud layer parameters related to precipitation, wherein the key cloud layer parameters include at least cloud top height, cloud optical thickness and cloud particle effective radius.

4. The precipitation prediction system based on cloud parameters according to claim 1, characterized in that: The precipitation prediction module includes a model building unit, an integrated learning unit and a model evaluation unit; The model building unit is used to build a precipitation prediction model using a random forest algorithm based on the acquired key cloud parameters, and use the precipitation prediction model as a basis function; introduce historical precipitation data, divide the historical precipitation data into a training set and a validation set, and train the basis function according to a preset order based on the training set; The ensemble learning unit is used to adjust the weight of the training parameter according to the prediction result of the previous model during the training process of the basis function based on the Boosting ensemble learning method, and then perform weighted summation of the prediction results of all basis functions according to certain weights to obtain the final precipitation prediction result; The model evaluation unit is used to evaluate the final precipitation prediction result based on the verification set, and adjust the type of the base model or the parameters of the base model based on the evaluation result.

5. The precipitation prediction system based on cloud parameters according to claim 1, characterized in that: It also includes a result output module, which includes a result prediction unit and a result display unit; The result prediction unit is used to input the key cloud parameters acquired in real time into the trained basis function, and obtain the precipitation prediction value within a preset time period through ensemble learning; The result display unit is used to output the prediction results to the user in the form of charts or reports through a visual interface or API interface, and supports multi-platform access.

6. A precipitation prediction method based on cloud parameters, characterized in that: The precipitation prediction system based on cloud parameters according to any one of claims 1 to 5 comprises the following steps: Acquire a cloud dataset from a remote sensing platform, denoise and correct the cloud dataset, and obtain a processed cloud dataset; Segment a complete cloud image from the processed cloud data set, perform feature extraction on the complete cloud image based on a convolutional neural network to obtain cloud parameters, perform correlation analysis between the cloud parameters and precipitation to obtain key cloud parameters; Based on the key parameters of the cloud layer, the random forest algorithm is used to construct the basis function, and the prediction results of the basis function are subjected to Boosting ensemble learning to obtain the final precipitation prediction results.

7. The precipitation prediction method based on cloud parameters according to claim 6, characterized in that: The process of processing the cloud data set includes: performing denoising on the cloud data set based on a median filtering algorithm, performing geometric correction on the cloud data set based on remote sensing processing software, and performing radiation correction on the cloud data set based on an atmospheric correction model to obtain a processed cloud data set.

8. The precipitation prediction method based on cloud parameters according to claim 6, characterized in that: The process of obtaining key cloud parameters includes: analyzing the relationship between the interaction between several cloud parameters and precipitation in combination with a multivariate analysis regression method, thereby obtaining key cloud parameters related to precipitation, wherein the key cloud parameters include at least cloud top height, cloud optical thickness and cloud particle effective radius.

9. The precipitation prediction method based on cloud parameters according to claim 6, characterized in that: The process of obtaining the final precipitation prediction result includes: based on the obtained key cloud parameters, respectively using the random forest algorithm to construct a precipitation prediction model, and using the precipitation prediction model as the basis function; introducing historical precipitation data to train the basis function in a preset order; based on the Boosting ensemble learning method, in the training process of the basis function, adjusting the weights of the training parameters according to the prediction results of the previous model, and then weighted summing the prediction results of all basis functions according to certain weights to obtain the final precipitation prediction result.

10. The precipitation prediction method based on cloud parameters according to claim 9, characterized in that: After obtaining the final precipitation forecast result, it also includes: outputting the forecast result to the user in the form of charts or reports, supporting multi-platform access.

Images