Satellite imaging quality prediction method and device, electronic equipment and medium

Abstract

The invention discloses a satellite imaging quality prediction method and apparatus, an electronic device and a medium. The method comprises the steps of obtaining a to-be-predicted data set of a to-be-predicted region; the to-be-predicted data set comprises position information, time information and meteorological data; performing quality evaluation on the to-be-predicted data set through a pre-constructed classification prediction model to determine a data quality grade of the to-be-predicted data set; according to the data quality grade, inputting the to-be-predicted data set into a pre-constructed imaging prediction model of a corresponding grade to obtain an imaging prediction score; the higher the imaging prediction score is, the higher the imaging suitability of the to-be-predicted area is. Therefore, through a two-stage prediction mechanism, based on the to-be-predicted data, the imaging suitability degree of the to-be-predicted area is predicted, task requests with low imaging suitability can be eliminated in advance, the space-time matching degree and the success rate of an imaging plan are improved, invalid imaging and data transmission are reduced, and the satellite imaging efficiency and the ground station resource utilization rate are improved.

Description

Technical Field

[0001] This application relates to the field of remote sensing image technology, and in particular to methods, devices, electronic equipment and media for predicting satellite imaging quality. Background Technology

[0002] Satellite remote sensing data is widely used in many fields such as environmental monitoring, resource surveys, urban planning, disaster early warning, and agricultural yield estimation due to its advantages such as wide coverage, high timeliness, and objective and accurate observation results. However, the acquisition process of satellite remote sensing data still faces many challenges.

[0003] Specifically, satellite remote sensing imaging is affected by environmental factors. Meteorological factors such as cloud cover, rain and snow, atmospheric disturbances, changes in lighting conditions, and low visibility can lead to a decrease in the quality of acquired images, thereby affecting the success rate of imaging missions and the availability of data.

[0004] Furthermore, current remote sensing satellites generally adopt a "space-based sensing and ground-based computing" working mode, whereby after completing observations in orbit, the satellite transmits the raw data downlink to ground receiving stations for image processing such as decoding, radiometric correction, and geometric correction. In this approach, less than one-tenth of the data acquired by the satellite is transmitted downlink in a timely manner, resulting in the loss of a large amount of observational data due to the inability to transmit it back promptly. Moreover, the limited bandwidth of space-to-ground communication makes it difficult to support the real-time transmission of massive amounts of remote sensing data, further reducing the availability and informational value of remote sensing image data.

[0005] Therefore, accurately assessing image quality and predicting imaging availability before image acquisition is of great significance for improving satellite data utilization efficiency, reducing resource waste, and ensuring the success rate of imaging missions. Summary of the Invention

[0006] In view of this, one aspect of this application provides a method for predicting satellite imaging quality, the method comprising: Obtain the dataset to be predicted for the region to be predicted; the dataset to be predicted includes location information, time information, and meteorological data; The dataset to be predicted is evaluated using a pre-built classification prediction model to determine the data quality level of the dataset to be predicted. Based on the data quality level, the dataset to be predicted is input into a pre-built imaging prediction model of the corresponding level to obtain an imaging prediction score; the higher the imaging prediction score, the higher the imaging suitability of the region to be predicted.

[0007] Optionally, constructing the classification prediction model includes: Acquire historical remote sensing imagery and a pre-built initial classification prediction model; A training dataset is constructed based on the historical remote sensing images. The training dataset includes multiple sub-datasets, each with a different data quality level label. Each sub-dataset includes location information, time information, and meteorological data corresponding to the historical images. The initial classification prediction model is iteratively trained using the multiple subset datasets to obtain the classification prediction model.

[0008] Optionally, constructing the imaging prediction model includes: Construct an initial imaging prediction model; the number of the initial imaging prediction models is the same as the number of the subset datasets. The initial imaging prediction model is iteratively trained using the subset dataset to obtain the imaging prediction model.

[0009] Optionally, the iterative training includes: Determine the classification loss function of the initial classification prediction model and the regression loss function of the initial imaging prediction model; The classification loss function and the regression loss function are fused to obtain the target loss function; Based on the target loss function, the parameters of the initial classification prediction model and the initial imaging prediction model are iteratively adjusted to minimize the target loss function.

[0010] Optionally, constructing a training dataset based on the historical remote sensing images includes: Extract target image metadata from the historical remote sensing images; the target image metadata includes the location information and time information of the historical imaging. Obtain the initial meteorological data corresponding to the target image metadata; The initial meteorological data is filtered to obtain the target meteorological data; The training dataset is constructed using the target image metadata and the target meteorological data.

[0011] Optionally, the step of filtering the initial meteorological data to obtain the target meteorological data includes: After preprocessing the initial meteorological data, preprocessed meteorological data is obtained; the preprocessing includes missing value imputation and normalization. Determine the meteorological characteristics included in the preprocessed meteorological data; Determine the first correlation coefficient between the meteorological features and the target quality score; the target quality score is used to characterize the quality of satellite imaging under specified meteorological feature conditions. Among the meteorological features, those features whose first correlation coefficient is less than the first threshold are to be removed; After removing the data corresponding to the features to be removed from the preprocessed meteorological data, the target meteorological data is obtained.

[0012] Optionally, before removing the data corresponding to the feature to be removed from the preprocessed meteorological data, the method further includes: Determine a second correlation coefficient between any two meteorological features; and extract meteorological feature pairs whose second correlation coefficient is greater than a second threshold; where the second threshold is greater than the first threshold. Determine the third correlation coefficient between each meteorological feature in the meteorological feature pair and the target quality score; Among the meteorological feature pairs, the feature corresponding to the one with the smaller third correlation coefficient is selected as the feature to be removed.

[0013] Another aspect of this application provides a satellite imaging quality prediction apparatus, the apparatus comprising: The data acquisition module is used to acquire the dataset to be predicted for the region to be predicted; the dataset to be predicted includes location information, time information, and meteorological data. The quality assessment module is used to assess the quality of the dataset to be predicted using a pre-built classification prediction model, so as to determine the data quality level of the dataset to be predicted. The imaging prediction module is used to input the dataset to be predicted into a pre-built imaging prediction model of the corresponding level according to the data quality level, and obtain an imaging prediction score; the higher the imaging prediction score, the higher the imaging suitability of the region to be predicted.

[0014] Another aspect of this application provides an electronic device including a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor executes the computer program to implement the steps of the method for predicting satellite imaging quality.

[0015] Another aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for predicting satellite imaging quality.

[0016] The satellite imaging quality prediction method, apparatus, electronic equipment, and medium provided in this application have the following beneficial effects: Before the satellite imaging mission is launched, a classification prediction model is used to perform coarse-grained prediction of the data quality of the dataset to be predicted, and then an imaging prediction model is used to perform fine-grained prediction of the imaging score. Thus, through a two-stage prediction mechanism, the imaging suitability of the area to be predicted is predicted and evaluated based on the location information, time information, and meteorological data of the area to be predicted. Low-quality conditions, i.e. mission requests with low imaging suitability, can be eliminated in advance, significantly improving the spatiotemporal matching degree and success rate of the imaging plan, reducing invalid imaging and data transmission, and improving satellite imaging efficiency and ground station resource utilization.

[0017] Furthermore, phased modeling and prediction can reduce errors caused by mixed data categories, improve imaging prediction accuracy, and enable the prediction of imaging conditions at any location and time. It can be widely applied to the prediction of imaging suitability in different regions and on different satellites, providing reliable data support for remote sensing applications. Attached Figure Description

[0018] Figure 1 A schematic flowchart illustrating a method for predicting satellite imaging quality provided in an embodiment of this application; Figure 2 A schematic diagram illustrating the principle of a satellite imaging quality prediction method provided in this application embodiment; Figure 3 A schematic diagram illustrating the principle of a model training method provided in an embodiment of this application; Figure 4 This is a schematic diagram illustrating the principle of meteorological data processing provided in an embodiment of this application; Figure 5 A schematic diagram of the structure of a satellite imaging quality prediction device provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0019] The reference numerals in the attached diagram are as follows: 50 is , 51 is , 52 is , 60 is memory, 61 is processor, 62 is display screen, 63 is input / output interface, 64 is communication interface, 65 is power supply, 66 is communication bus, 601 is computer program, 602 is operating system, and 603 is data. Detailed Implementation

[0020] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0021] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0022] Figure 1 This is a flowchart illustrating a method for predicting satellite imaging quality provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes: S10: Obtain the dataset to be predicted for the area to be predicted; the dataset to be predicted includes location information, time information, and meteorological data; It is understood that, in specific embodiments, satellite imaging quality will be affected by factors such as different times, different locations and different weather conditions. Therefore, in order to determine the imaging suitability of the area to be predicted, it is necessary to first obtain the location information, time information and meteorological data of the area to be predicted in order to form a dataset to be predicted.

[0023] The location information can be the latitude and longitude coordinates of the center point of the area to be predicted, or it can be the latitude and longitude range of the entire area to be predicted; this application does not limit this. In an optional embodiment, the time information can be the imaging suitability predicted for a future period of time. Correspondingly, the meteorological data is the meteorological data for that future period of time; that is, the meteorological data is data within the time range represented by the time information.

[0024] In one optional embodiment, the acquired meteorological data includes at least weather codes, average cloud cover, sunshine duration, average sea level pressure, precipitation, precipitation duration, average temperature, average dew point temperature, average humidity, and day length. In fact, this application does not specifically limit the acquired meteorological data.

[0025] It should be noted that, in specific embodiments, there may be one or more regions to be predicted, and a different dataset to be predicted will be obtained for each region.

[0026] S11: Using a pre-built classification prediction model, assess the quality of the dataset to be predicted to determine its data quality level. Figure 2 This is a schematic diagram illustrating the principle of a satellite imaging quality prediction method provided in an embodiment of this application. In an optional embodiment, to ensure the prediction accuracy of satellite imaging quality, a two-stage prediction model is set up. Specifically, as shown... Figure 2 As shown, it includes a pre-built first-stage classification prediction model and a second-stage imaging prediction model.

[0027] In a specific embodiment, the acquired dataset to be predicted is first input into the first-stage classification prediction model for processing, thereby determining the data quality level of the dataset to be predicted. This stage can be understood as performing coarse-grained classification of the dataset to be predicted, dividing it into data of different quality levels.

[0028] It should be noted that this application does not limit the number of data quality levels. The specific number of levels is determined when the classification prediction model is pre-built. Specifically, the classification prediction model can be trained with data of multiple different data quality levels so that the classification data model has the ability to identify data quality and classify it into the corresponding quality level.

[0029] Of course, in an alternative embodiment, to conserve computing resources, the data quality level can be divided into two levels. For ease of understanding, the following will use the division of data quality levels into high-quality and low-quality levels as an example, combined with... Figure 2 Please provide a detailed explanation.

[0030] like Figure 2 As shown, after the dataset to be predicted is input into the classification prediction model, the classification prediction model processes it to identify whether the quality level of the current dataset to be predicted belongs to the low quality level or the high quality level.

[0031] S12: Based on the data quality level, input the dataset to be predicted into the pre-built imaging prediction model of the corresponding level to obtain the imaging prediction score; the higher the imaging prediction score, the higher the suitability of the imaging of the area to be predicted.

[0032] Furthermore, in an optional embodiment, to further improve the accuracy of imaging suitability prediction, different imaging prediction models of different quality levels are selected for prediction in the first stage. That is, the number of imaging prediction models is the same as the number of data quality levels. Figure 2In the illustrated embodiment, the data quality level includes two levels: low quality level and high quality level. Correspondingly, the imaging prediction model includes a first imaging prediction model and a second imaging prediction model. The first imaging prediction model is used to perform imaging prediction for low quality level data, while the second imaging prediction model is used to perform imaging prediction for high quality level data.

[0033] Similarly, it should be noted that in specific embodiments, this application does not limit the number of imaging prediction models. As long as the number of imaging prediction models is the same as the number of data quality levels, it is possible to achieve one imaging prediction model for each data quality level.

[0034] like Figure 2 As shown, based on the first-stage coarse-grained prediction, the corresponding imaging prediction model is selected according to different data quality levels to process the dataset to be predicted, thereby obtaining the imaging prediction score of the region to be predicted. The higher the obtained imaging prediction score, the higher the imaging suitability of the region to be predicted.

[0035] As shown in Figure 2, after obtaining the imaging prediction score, as an optional embodiment, the imaging suitability level of the area to be predicted can be determined based on the pre-built mapping relationship, so that users can more intuitively determine whether a satellite imaging task can be performed at present.

[0036] The mapping relationship refers to the correspondence between imaging scores and suitability levels. For example, in one optional embodiment, an imaging score of 5 is the highest score, corresponding to a suitability level of 1, which indicates excellent imaging conditions and suitability for performing imaging tasks. An imaging score of 3 to 4 corresponds to a suitability level of 2, indicating average imaging conditions and feasible imaging, but with uncertain quality. An imaging score of 0 to 2 corresponds to a suitability level of 3, indicating poor imaging conditions and unsuitability for performing imaging tasks.

[0037] In specific embodiments, the two-stage imaging prediction method provided in this application can accurately predict the imaging suitability for different meteorological data and different times. Specifically, Table 1 is a schematic table of imaging prediction results under different meteorological data conditions provided in the embodiments of this application. The imaging accuracy under different meteorological data conditions will be explained below with reference to Table 1.

[0038] Table 1 shows the imaging prediction results under different meteorological data conditions.

[0039] As shown in Table 1, the accuracy of remote sensing images under different weather phenomena was found to be the highest when the weather was clear, reaching 97.96%. Rain and snow both affected the prediction, and the accuracy decreased as the rain and snow intensified. The accuracy was the lowest when there was heavy snow, at 78.79%.

[0040] Table 2 is a schematic table of imaging prediction results for different time periods provided in the embodiments of this application. The imaging accuracy for different time periods will be explained below with reference to Table 2.

[0041] Table 2 shows the imaging prediction results for different time periods.

[0042] Table 2 shows the remote sensing image quality for different months. The image quality of different seasons, from highest to lowest, is winter > autumn > spring > summer. The accuracy rate was lowest in July at 87.34%. Analysis shows that this is because July is the peak summer season, with frequent and heavy rainfall. Especially in South and East China, the monsoon brings frequent heavy rains and thunderstorms, and thick, rapidly changing cloud cover makes it difficult to predict the imaging window. Rainy and cloudy weather further reduces the imaging success rate.

[0043] In one optional embodiment, predictions and statistics can be performed on different areas to be detected. Statistical analysis shows that for Northeast, Central, and North China, the flat terrain and stable climate make weather forecasts relatively accurate, resulting in good stability of remote sensing imaging and facilitating continuous optical remote sensing observations. In East and South China, especially coastal areas and regions with complex terrain, convective clouds develop rapidly, local weather is highly variable, forecasting is difficult, and imaging is significantly affected. Furthermore, Guangxi and Hainan, located in the humid southern region, experience high annual precipitation, high humidity, rapid cloud formation, and atmospheric instability, leading to lower accuracy. In Shaanxi's northwest to Qinling Mountains, the alternating Loess Plateau and Qinling Mountains experience spring dust storms, summer torrential rains, and autumn and winter smog, resulting in low imaging prediction accuracy. For areas like Zhoushan, Shanwei, and Zhongshan, which have maritime weather patterns, rapid convection development also affects imaging quality. Overall, the combined effects of climate zones, topography, and rapid convection lead to variable weather, posing challenges to remote sensing imaging.

[0044] In an optional embodiment, imaging prediction can be performed based on meteorological data of cloud cover percentage. Table 3 is a schematic table of imaging prediction results for different cloud cover percentages provided in the embodiments of this application. The imaging prediction results for different cloud cover percentages will be explained below with reference to Table 3.

[0045] Table 3 shows the imaging prediction results for different cloud cover percentages.

[0046] As shown in Table 3, different cloud cover percentages correspond to different imaging suitability levels, with cloudless weather conditions providing the best imaging conditions. It should be noted that in specific embodiments, the quality is visually assessed based on the cloud cover distribution in the image, allowing for some flexibility in handling complex situations. Specifically, scattered cloud distributions are calculated based on the extent of cloud clusters, thin clouds have their influence range appropriately reduced, important ground targets obscured by clouds have their range appropriately increased, and the extent of land obscured by clouds at the land-sea boundary is calculated, etc.

[0047] Based on the above description, in an optional embodiment, the prediction method provided in this application can be used to predict the imaging suitability score for different regions, times and meteorological conditions before the imaging task is executed. Combined with satellite status information and task requirements, the priority of the imaging task can be dynamically adjusted to avoid the failure of task planning due to resource changes or sudden weather changes, and to ensure the execution stability and response efficiency of the imaging task.

[0048] Specifically, it automatically identifies high-risk imaging periods and areas such as atmospheric instability, complex local circulation, and frequent severe convection, reduces the priority of imaging tasks or postpones their execution, and adjusts the weights according to different seasonal meteorological patterns, prioritizing imaging periods with stable conditions to increase the proportion of effective data throughout the year.

[0049] Furthermore, the prediction results can be applied to multi-satellite collaborative scheduling to achieve dynamic monitoring, rapid feedback and continuous coverage of specific areas, significantly improving the spatial coverage and temporal continuity of disaster emergency response, major event monitoring and long-term sequence observation.

[0050] Finally, before performing the imaging mission, low-quality data is filtered out to reduce the bandwidth usage and storage pressure of satellite and ground links, and to reduce the redundant burden of ground station reception, storage and processing. In specific embodiments, API interface calls are supported to provide stable data services for third-party imaging planning systems.

[0051] Therefore, the satellite imaging quality prediction method provided in this application, before the satellite imaging mission is issued, performs coarse-grained prediction of the data quality of the dataset to be predicted using a classification prediction model, and further performs fine-grained imaging score prediction using an imaging prediction model. This two-stage prediction mechanism, based on the location information, time information, and meteorological data of the area to be predicted, predicts and evaluates the imaging suitability of the area. This allows for the early elimination of low-quality conditions, i.e., mission requests with low imaging suitability, significantly improving the spatiotemporal matching degree and success rate of the imaging plan, reducing invalid imaging and data transmission, and improving satellite imaging efficiency and ground station resource utilization. Furthermore, the staged modeling and prediction can reduce errors caused by mixed data categories, improve imaging prediction accuracy, and enable prediction of imaging conditions at any location and time. It can be widely applied to imaging suitability prediction for different regions and different satellites, providing reliable data support for remote sensing applications.

[0052] In one alternative embodiment, constructing a classification prediction model includes: Acquire historical remote sensing imagery and a pre-built initial classification prediction model; A training dataset is constructed based on historical remote sensing images. The training dataset includes multiple sub-datasets, each with different data quality level labels. The sub-datasets include location information, time information, and meteorological data corresponding to the historical images. The initial classification prediction model is iteratively trained using multiple subsets of data to obtain the final classification prediction model.

[0053] In a specific embodiment, an initial classification and prediction model is pre-built, and a large amount of historical remote sensing image data is acquired. It is understood that remote sensing images include information such as latitude and longitude and the time of capture; therefore, the location and time information of historical imaging can be extracted from historical remote sensing images. Furthermore, corresponding meteorological data can be obtained based on the location and time information.

[0054] In one optional embodiment, meteorological information corresponding to the image capture can be obtained by calling a historical weather API interface. This historical weather API interface can support searching historical weather data by latitude, longitude, and date. The query results may include, but are not limited to, weather phenomena (e.g., sunny, rainy, and snowy), temperature, precipitation, air pressure, cloud cover, surface radiation, wind speed and direction, daylight hours, and sunshine duration.

[0055] Based on acquired historical remote sensing imagery, a training dataset can be constructed that includes the location, time, and meteorological data of the historical images. Understandably, this training dataset is used to train an initial classification and prediction model, allowing the model to learn from the data in the training dataset. This enables the trained model to be directly used subsequently to predict the data quality level of the dataset to be predicted.

[0056] Therefore, the training dataset includes multiple subsets corresponding to different data quality level labels. In other words, the data included in different subsets belong to different data quality levels. In order for the initial classification prediction model to obtain data quality levels, level labels are set for different subsets.

[0057] The number of subsets is the same as the number of data quality level labels. A single subset can include data from multiple locations, times, and weather conditions. In other words, for a large amount of historical remote sensing imagery, all data is divided according to different data quality levels. For ease of understanding, the following explanation will use low-quality and high-quality levels as examples.

[0058] For example, the training dataset can be divided into two subsets: one subset representing low-quality data and the other representing high-quality data. It should be noted that the subset division can be done manually or by using a specified large language model; this application does not impose any restrictions on this.

[0059] Figure 3 This is a schematic diagram illustrating the principle of a model training method provided in an embodiment of this application. Further, as... Figure 3 As shown, the initial classification prediction model is iteratively trained using a training dataset to obtain the final classification prediction model. In one optional instance, the classification prediction model can be a LightGBM (Light Gradient Boosting Machine) classification model. The iterative training time can be used to obtain the classification loss function of the classification prediction model. The parameters of the model are optimized with the goal of minimizing the classification loss function until the classification loss function reaches the convergence condition, thereby obtaining the final classification prediction model.

[0060] In the examples of the high and low quality levels above, the prediction results of the classification prediction model can be expressed as follows: ,in, ,when When, it indicates a low quality level, when When the value is specified, it indicates a high quality level. In an optional embodiment, after training the classification prediction model, the classification prediction model pkl file is stored for subsequent model deployment and inference.

[0061] Based on the above embodiments, as an optional embodiment, an imaging prediction model is constructed, including: Construct initial imaging prediction models; the number of initial imaging prediction models is the same as the number of subsets. By using a subset of data, the initial imaging prediction model is iteratively trained in a one-to-one correspondence to obtain the imaging prediction model.

[0062] In a specific embodiment, firstly, multiple initial imaging prediction models are constructed. The specific number of models constructed is the same as the number of subsets obtained in the above embodiments, that is, the same as the number of data quality level labels in the above embodiments.

[0063] To enable more targeted imaging predictions for data of varying quality, this application establishes multiple imaging prediction models corresponding to different quality levels. In fact, for high-quality data, the accuracy of model predictions is very high, and imaging tasks can be directly executed after prediction. However, in reality, cloudless weather is rare in specific embodiments; most of the time, environmental weather conditions are not ideal, resulting in lower quality data for prediction. Dividing the data into different imaging prediction models allows for the extraction of usable data from low-quality data, as much as possible. In other words, selecting different imaging prediction models for different quality levels aims not only to obtain high-quality data but, more importantly, to extract usable data from low-quality data, as much as possible, for the imaging task.

[0064] Therefore, in this embodiment, subsets of data with different quality level labels are input into different initial imaging prediction models for iterative training, resulting in a final imaging prediction model that can be directly used for imaging suitability prediction. For ease of understanding, the following explanation will use two quality levels, high and low, as examples.

[0065] In one optional embodiment, the data quality level labels include low-quality level labels and high-quality level labels, corresponding to two subsets of datasets. The initial imaging prediction model is a LightGBM regression model, and this model includes a first initial imaging prediction model for predicting low-quality data. And a second initial imaging prediction model for predicting high-quality data. .

[0066] The prediction results of the initial imaging predictive model can be expressed as follows: , , The two prediction results can be combined as follows: Similarly, the trained imaging prediction model is stored as a pkl file for subsequent model deployment and inference.

[0067] In one alternative embodiment, iterative training includes: Determine the classification loss function of the initial classification prediction model and the regression loss function of the initial imaging prediction model; The classification loss function and the regression loss function are fused to obtain the target loss function; Based on the target loss function, the parameters of the initial classification prediction model and the initial imaging prediction model are iteratively adjusted to minimize the target loss function.

[0068] In a specific embodiment, the loss functions of the classification prediction model and the imaging prediction model can be determined separately, and the parameters of the two models can be optimized separately with the goal of minimizing the loss function to obtain the final model. However, it is understood that there is a close relationship between the two models. In a specific embodiment, the output of the classification prediction model is directly input into the imaging prediction model. Therefore, in an optional embodiment, to further improve the imaging prediction accuracy, the two models are jointly optimized. That is, the loss function adopts a standardized weighted combination of classification and regression losses to improve prediction accuracy.

[0069] Specifically, such as Figure 3 As shown, during model training, the classification loss function of the initial classification model and the regression loss function of the initial imaging prediction model are calculated. In an optional embodiment, the classification loss function can be the cross-entropy loss function, calculated using the following formula: ,in, The total number of samples, Indicates the first The data quality level label of the input training samples, for example, in the two types of data quality level labels in the above embodiment, when the data quality level label is high quality, the representation is a positive class result. When the data quality level is marked as low, the representation is a negative result. It should be noted that the location information, time information, and meteorological data included in a historical image dataset constitute a single input sample. Represents all input samples The prediction is positive, that is The probability of.

[0070] In another alternative embodiment, the regression loss function can be the mean squared error loss function, calculated as follows: ,in, Indicates the first The prediction quality level of each input training sample. This indicates the actual level of data quality.

[0071] After obtaining the classification loss function and the regression loss function, as follows: Figure 3 As shown, the two loss functions are fused to obtain the target loss function. Specifically, in an optional embodiment, the loss functions of the two stages are aligned in terms of dimensions, and the specific calculation formula is as follows: , ,in, The standardized classification loss, The regression loss is the standardized value. The average value of the classification loss. The average value of the regression loss. The standard deviation of the classification loss. The standard deviation of the regression loss. To prevent division by zero, the coefficient is a very small positive number in the specific implementation; for example, it could be... .

[0072] After performing two alignments, in one optional embodiment, the classification loss and regression loss metrics fall within the range of... Within the range of values, it resembles a normal distribution, which facilitates joint optimization training of the subsequent model.

[0073] Therefore, the expression for the target loss function can be obtained: ,in, For weighting coefficients, in an optional embodiment, the weighting coefficients are... It can be set to 0.5 to maintain a balance between classification loss and regression loss.

[0074] like Figure 3 As shown, after obtaining the target loss function, the model parameters of the classification prediction model and the imaging prediction model are continuously iterated and optimized with the goal of minimizing the target loss function until the target loss function reaches the convergence condition, thereby obtaining a two-stage prediction model. This model can be returned through a standardized API interface for remote sensing imaging planning systems or other third-party task planning systems to call.

[0075] In a specific embodiment, a two-stage model is loaded, for example, as follows: Figure 2 As shown, a coarse-grained classification is first performed using a one-stage classification prediction model, and then a corresponding imaging prediction model is used for accurate prediction based on the classification results.

[0076] After predicting the image quality, the predicted values ​​are compared with the actual values, and the image quality prediction accuracy rates for different cities, under different meteorological conditions, and at different times are calculated.

[0077] In one optional embodiment, experiments are conducted to compare multiple different prediction methods to analyze their accuracy. Table 4 is a schematic table showing the comparison results of imaging accuracy of different prediction methods provided in the embodiments of this application. The following description will be based on Table 4.

[0078] Table 4 shows a comparison of the imaging accuracy of different prediction methods.

[0079] Comparing the root mean square error (RMSE) of the single-stage regression model and the two-stage model provided in this application, and... As shown in Table 4, the two-stage model provided in this application reduces the average prediction error by 0.144 compared to the single-stage Light_GBM model. It also improves the explained variance by 11.2%, resulting in a better overall model fit. Therefore, the two-stage method provided in this application reduces local errors through piecewise modeling and effectively enhances robustness.

[0080] Therefore, the satellite imaging quality prediction method provided in this application constructs a joint classification and regression prediction model based on historical image quality data, which improves the stability and accuracy of meteorological element determination such as cloud cover and humidity, reduces the misjudgment and misscreening problems of traditional cloud cover threshold methods, and enhances the ability to distinguish between good and bad imaging conditions.

[0081] In one optional embodiment, the initial meteorological data is filtered to obtain target meteorological data, including: After preprocessing the initial meteorological data, preprocessed meteorological data is obtained; the preprocessing includes missing value imputation and normalization. Identify the meteorological characteristics included in the preprocessed meteorological data; Determine the first correlation coefficient between meteorological features and target quality score; the target quality score is used to characterize the quality of satellite imaging under specified meteorological features. In acquiring meteorological features, features whose first correlation coefficient is less than the first threshold are to be removed. After removing the data corresponding to the features to be removed from the preprocessed meteorological data, the target meteorological data is obtained.

[0082] It is understandable that the initial meteorological data obtained initially includes a lot of incomplete and poor-quality data, which will directly affect the training effect of the model. Therefore, in order to improve the training accuracy of the model, in one optional embodiment, the initial meteorological data is preprocessed after it is obtained.

[0083] Specifically, for meteorological features with missing values, the data is filled with the average value of the corresponding meteorological feature, and continuous meteorological features are normalized. The processed data can be saved to the PostGIS spatial database, supporting direct data retrieval from the database for model training and validation.

[0084] Figure 4 This is a schematic diagram illustrating the principle of meteorological data processing provided in an embodiment of this application, as shown below. Figure 4 As shown, after preprocessing the initial meteorological data to obtain preprocessed meteorological data, the meteorological data screening stage begins.

[0085] Specifically, in one optional embodiment, the meteorological characteristics included in the preprocessed meteorological data are first determined, wherein the meteorological characteristics may include, but are not limited to, maximum, minimum, and average wind speeds, reference total evapotranspiration, cloud cover, dew point temperature, ambient temperature, sunshine duration, day length, and air pressure.

[0086] It is understandable that, in specific embodiments, the data corresponding to meteorological characteristics such as maximum, minimum, and average wind speeds and reference evapotranspiration have little impact on image quality and are therefore invalid data. In order to improve the efficiency and accuracy of model training, these data need to be screened and removed.

[0087] Therefore, in an alternative embodiment, such as Figure 4 As shown, the first correlation coefficient between each meteorological feature and the target quality score is calculated using the following formula: ,in, The first correlation coefficient, For the first A meteorological feature, for example, temperature. The target quality score is used to characterize the quality of satellite imaging under specified meteorological conditions. It is a specific numerical score. For example, under specified meteorological conditions including cloud cover, precipitation, temperature, humidity and air pressure, the target quality score is 3 points, which reflects the quality of imaging under those meteorological conditions and also the suitability of imaging under those meteorological conditions.

[0088] Meteorological characteristics With target quality score Covariance between Meteorological characteristics standard deviation For the target quality score The standard deviation of . In a specific embodiment, the first correlation coefficient Used to measure meteorological characteristics With target quality score The degree of linear correlation, when At that time, meteorological characteristics With target quality score Perfectly positive correlation, when At that time, meteorological characteristics With target quality score Perfectly negative correlation, when At that time, meteorological characteristics With target quality score Irrelevant.

[0089] Furthermore, after calculating the first correlation coefficient, in an optional embodiment, features among the meteorological features whose first correlation coefficient is less than a first threshold are extracted as features to be removed. For example... Figure 4 As shown, after removing the data to be removed from the preprocessed meteorological data, the target meteorological data is obtained.

[0090] Based on the above embodiments, as an optional embodiment, before removing the data corresponding to the features to be removed from the preprocessed meteorological data, the following steps are also included: Determine the second correlation coefficient between any two meteorological features; and extract meteorological feature pairs whose second correlation coefficient is greater than a second threshold; the second threshold is greater than a first threshold; Determine the third correlation coefficient between each meteorological feature and the target quality score in the meteorological feature pair; Among the meteorological features, the feature corresponding to the one with the smaller third correlation coefficient is selected as the feature to be removed.

[0091] In a specific embodiment, the preprocessed meteorological data may include multiple similar meteorological features, such as all being temperature data or all being air pressure data. For meteorological features with the same meaning, in order to ensure computational efficiency and save computational resources, one of the similar meteorological features can be deleted.

[0092] In fact, when there are too many meteorological features, the trained model has poor scalability, making it unsuitable for widespread application in various scenarios. Furthermore, a large number of meteorological features leads to poor model performance and low efficiency. At the same time, filtering meteorological features ensures that the amount of data obtained meets the requirements.

[0093] Based on the above description, in one optional embodiment, such as Figure 4 As shown, the screening of the first correlation coefficient in the above embodiment is regarded as the first stage screening of preprocessed meteorological data, and the multicollinearity detection provided in the embodiment of this application is regarded as the second stage screening, thereby further improving the training efficiency and accuracy of the model.

[0094] Specifically, such as Figure 4 As shown, when performing multicollinearity detection, a correlation matrix can be calculated for all meteorological features. This involves first calculating the second correlation coefficient between any two meteorological features. The calculation of the second correlation coefficient is the same as the calculation of the first correlation coefficient in the above embodiment, and will not be repeated here. Simultaneously, meteorological feature pairs with a second correlation coefficient greater than a second threshold are extracted; that is, two highly similar meteorological features are selected. It is important to note that highly similar features are selected here, therefore the second threshold is greater than the first threshold.

[0095] After extracting highly similar meteorological feature pairs, the third correlation coefficient between each meteorological feature in the pair and the target quality score is further analyzed, and the feature corresponding to the smaller third correlation coefficient is selected as the feature to be removed. In other words, for two highly similar meteorological features, only the feature more similar to the target quality score is retained.

[0096] It should be noted that, in specific embodiments, Figure 4 The two screening stages shown can be performed simultaneously, or either stage can be performed first; this application does not impose any restrictions on this. Figure 4 This is only one feasible embodiment. In a specific embodiment, after two stages of screening, the target meteorological features obtained may include, but are not limited to, weather codes, average cloud cover, sunshine duration, average sea level pressure, precipitation, precipitation duration, average temperature, average dew point temperature, average humidity, and daylight hours.

[0097] It is worth noting that, in one optional embodiment, the initial meteorological data is screened in order to build a training dataset later. In a specific embodiment, different quality data models focus on different meteorological features. Therefore, after obtaining the training dataset, the meteorological features of data of different quality levels can be sorted, and the final target training data can be selected based on the sorting results.

[0098] For example, the ranking results for low-quality meteorological characteristics are air pressure, day length, temperature, dew point temperature, sunshine duration, and cloud cover. The ranking results for high-quality meteorological characteristics are cloud cover, dew point temperature, ambient temperature, sunshine duration, day length, and air pressure. When selecting target training data, one can choose the top preset percentage of the ranked data, or the top 70% of the ranked data. This application does not limit the final selection method.

[0099] In one optional embodiment, the latest historical remote sensing images can be acquired at regular intervals. After data filtering as described in the above embodiments, the classification prediction model and imaging prediction model are retrained and optimized to continuously adapt to environmental changes and ensure the accuracy of imaging prediction.

[0100] Therefore, the satellite imaging quality prediction method provided in this application supports weather imaging quality forecasts for any location and any time. It is encapsulated as an API interface service for third-party systems, such as remote sensing imaging planning systems. Based on the prediction score, it is divided into three categories: high (e.g., 5 points, suitable for imaging), medium (e.g., 3 to 4 points, average imaging conditions), and low (e.g., 0 to 2 points, unsuitable for imaging). This significantly reduces invalid imaging tasks and supports intelligent task planning and autonomous decision-making.

[0101] In the above embodiments, the method for predicting satellite imaging quality has been described in detail. This application also provides an embodiment of a satellite imaging quality prediction device.

[0102] Figure 5 This is a schematic diagram of the structure of a satellite imaging quality prediction device provided in an embodiment of this application, as shown below. Figure 5 As shown, the device includes: The data acquisition module 50 is used to acquire the dataset to be predicted for the area to be predicted; the dataset to be predicted includes location information, time information and meteorological data; The quality assessment module 51 is used to assess the quality of the dataset to be predicted using a pre-built classification prediction model, so as to determine the data quality level of the dataset to be predicted. The imaging prediction module 52 is used to input the dataset to be predicted into a pre-built imaging prediction model of the corresponding level according to the data quality level, and obtain the imaging prediction score; the higher the imaging prediction score, the higher the imaging suitability of the area to be predicted.

[0103] Furthermore, the satellite imaging quality prediction device provided in this application embodiment also includes: The first acquisition module is used to acquire historical remote sensing images and a pre-built initial classification prediction model; The training dataset construction module is used to build a training dataset based on historical remote sensing images. The training dataset includes multiple sub-datasets, each with different data quality level labels. The sub-datasets include location information, time information, and meteorological data corresponding to the historical images. The iterative training module is used to iteratively train the initial classification prediction model using multiple subsets of data to obtain the classification prediction model.

[0104] The initial imaging prediction model building module is used to build the initial imaging prediction model; the number of initial imaging prediction models is the same as the number of subsets. The iterative training module is also used to perform one-to-one iterative training on the initial imaging prediction model using a subset of datasets to obtain the imaging prediction model.

[0105] The loss function determination module is used to determine the classification loss function of the initial classification prediction model and the regression loss function of the initial imaging prediction model. The loss function fusion module is used to fuse the classification loss function and the regression loss function to obtain the target loss function; The parameter optimization module is used to iteratively adjust the parameters of the initial classification prediction model and the initial imaging prediction model according to the target loss function in order to minimize the target loss function.

[0106] The metadata extraction module is used to extract target image metadata from historical remote sensing images; the target image metadata includes the location and time information of historical imaging. The initial meteorological data acquisition module is used to acquire the initial meteorological data corresponding to the target image metadata. The meteorological data filtering module is used to filter the initial meteorological data to obtain the target meteorological data; The dataset building module is used to build a training dataset using target image metadata and target meteorological data.

[0107] The preprocessing module is used to preprocess the initial meteorological data to obtain preprocessed meteorological data; the preprocessing includes missing value imputation and normalization. The meteorological feature determination module is used to determine the meteorological features included in the preprocessed meteorological data; The first correlation coefficient determination module is used to determine the first correlation coefficient between meteorological features and target quality score; the target quality score is used to characterize the quality of satellite imaging under specified meteorological feature conditions. The feature to be removed module is used to obtain meteorological features whose first correlation coefficient is less than the first threshold. The data removal module is used to remove data corresponding to the features to be removed from the preprocessed meteorological data to obtain the target meteorological data.

[0108] The second correlation coefficient determination module is used to determine the second correlation coefficient between any two meteorological features; and to extract meteorological feature pairs whose second correlation coefficient is greater than a second threshold; the second threshold is greater than a first threshold. The third correlation data determination module is used to determine the third correlation coefficient between each meteorological feature and the target quality score in the meteorological feature pair; The feature to be removed module is also used to identify the feature with the smaller third correlation coefficient among the meteorological feature pairs as the feature to be removed.

[0109] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application, such as... Figure 6 As shown, the electronic device includes: a memory 60 for storing computer programs; The processor 61 is configured to execute a computer program to implement the steps of the satellite imaging quality prediction method as described in the above embodiments.

[0110] The electronic devices provided in this embodiment may include, but are not limited to, laptops or desktop computers.

[0111] The processor 61 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 61 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 61 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 61 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 61 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.

[0112] The memory 60 may include one or more computer-readable storage media, which may be non-transitory. The memory 60 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 60 is used to store at least the following computer program 601, which, after being loaded and executed by the processor 61, is capable of implementing the relevant steps of the satellite imaging quality prediction method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 60 may also include an operating system 602 and data 603, and the storage method may be temporary or permanent storage. The operating system 602 may include Windows, Unix, Linux, etc. The data 603 may include, but is not limited to, relevant data involved in the satellite imaging quality prediction method.

[0113] In some embodiments, the electronic device may further include a display screen 62, an input / output interface 63, a communication interface 64, a power supply 65, and a communication bus 66.

[0114] Those skilled in the art will understand that Figure 6 The structures shown do not constitute a limitation on electronic devices and may include more or fewer components than those shown.

[0115] The electronic device provided in this application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the satellite imaging quality prediction method in the above embodiments.

[0116] It should be noted that although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

Claims

1. A method for predicting satellite imaging quality, characterized in that, The method includes: Obtain the dataset to be predicted for the region to be predicted; the dataset to be predicted includes location information, time information, and meteorological data; The dataset to be predicted is evaluated using a pre-built classification prediction model to determine the data quality level of the dataset to be predicted. Based on the data quality level, the dataset to be predicted is input into a pre-built imaging prediction model of the corresponding level to obtain an imaging prediction score; the higher the imaging prediction score, the higher the imaging suitability of the region to be predicted.

2. The method for predicting satellite imaging quality as described in claim 1, characterized in that, Constructing the classification prediction model includes: Acquire historical remote sensing imagery and a pre-built initial classification prediction model; A training dataset is constructed based on the historical remote sensing images. The training dataset includes multiple sub-datasets, each with a different data quality level label. Each sub-dataset includes location information, time information, and meteorological data corresponding to the historical images. The initial classification prediction model is iteratively trained using the multiple subset datasets to obtain the classification prediction model.

3. The method for predicting satellite imaging quality as described in claim 2, characterized in that, Constructing the imaging prediction model includes: Construct an initial imaging prediction model; the number of the initial imaging prediction models is the same as the number of the subset datasets. The initial imaging prediction model is iteratively trained using the subset dataset to obtain the imaging prediction model.

4. The method for predicting satellite imaging quality as described in claim 3, characterized in that, The iterative training includes: Determine the classification loss function of the initial classification prediction model and the regression loss function of the initial imaging prediction model; The classification loss function and the regression loss function are fused to obtain the target loss function; Based on the target loss function, the parameters of the initial classification prediction model and the initial imaging prediction model are iteratively adjusted to minimize the target loss function.

5. The method for predicting satellite imaging quality as described in claim 2, characterized in that, The step of constructing a training dataset based on the historical remote sensing images includes: Extract target image metadata from the historical remote sensing images; the target image metadata includes the location information and time information of the historical imaging. Obtain the initial meteorological data corresponding to the target image metadata; The initial meteorological data is filtered to obtain the target meteorological data; The training dataset is constructed using the target image metadata and the target meteorological data.

6. The method for predicting satellite imaging quality as described in claim 5, characterized in that, The process of filtering the initial meteorological data to obtain the target meteorological data includes: After preprocessing the initial meteorological data, preprocessed meteorological data is obtained; the preprocessing includes missing value imputation and normalization. Determine the meteorological characteristics included in the preprocessed meteorological data; Determine the first correlation coefficient between the meteorological features and the target quality score; the target quality score is used to characterize the quality of satellite imaging under specified meteorological feature conditions. Among the meteorological features, those features whose first correlation coefficient is less than the first threshold are to be removed; After removing the data corresponding to the features to be removed from the preprocessed meteorological data, the target meteorological data is obtained.

7. The method for predicting satellite imaging quality as described in claim 6, characterized in that, Before removing the data corresponding to the feature to be removed from the preprocessed meteorological data, the process further includes: Determine a second correlation coefficient between any two meteorological features; and extract meteorological feature pairs whose second correlation coefficient is greater than a second threshold; where the second threshold is greater than the first threshold. Determine the third correlation coefficient between each meteorological feature in the meteorological feature pair and the target quality score; Among the meteorological feature pairs, the feature corresponding to the one with the smaller third correlation coefficient is selected as the feature to be removed.

8. A device for predicting satellite imaging quality, characterized in that, The device includes: The data acquisition module is used to acquire the dataset to be predicted for the region to be predicted; the dataset to be predicted includes location information, time information, and meteorological data. The quality assessment module is used to assess the quality of the dataset to be predicted using a pre-built classification prediction model, so as to determine the data quality level of the dataset to be predicted. The imaging prediction module is used to input the dataset to be predicted into a pre-built imaging prediction model of the corresponding level according to the data quality level, and obtain an imaging prediction score; the higher the imaging prediction score, the higher the imaging suitability of the region to be predicted.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the satellite imaging quality prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the satellite imaging quality prediction method according to any one of claims 1 to 7.

Images