Cloud removal deviation estimation method, device and equipment based on random sampling

Abstract

The invention provides a cloud removal deviation estimation method, device and equipment based on random sampling, and relates to the technical field of remote sensing data processing. The method comprises the steps of obtaining an initial remote sensing image of a research area; dividing the initial remote sensing image into a plurality of evaluation groups according to the annual average number of snow days, and randomly selecting candidate pixels from each evaluation group; performing cloud persistence analysis on the research area, and generating pseudo cloud layer data of candidate pixels based on an analysis result; superposing the pseudo cloud layer data and the initial remote sensing image to obtain a target remote sensing image; performing cloud removal processing on the target remote sensing image to obtain multiple groups of pixel filling values; and calculating a target error between each pixel filling value and a corresponding pixel true value in the initial remote sensing image. By simulating cloud coverage distribution, the deviation existing in the cloud removal process can be effectively evaluated, and then the accuracy of remote sensing data under complex climate conditions is improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of remote sensing data processing, and in particular to a cloud removal bias estimation method, device and equipment based on random sampling. Background Art

[0002] In the process of remote sensing data processing, cloud cover is a common obstacle that affects data quality and analysis accuracy. Due to cloud cover, surface information in remote sensing images is often missing or misjudged, which causes significant accuracy loss in many application fields, especially in environmental monitoring, disaster assessment, and climate change research.

[0003] At present, the commonly used cloud removal method is to supplement or reclassify the cloud coverage area through time series analysis or model-based inference. However, due to the complex spatiotemporal variation of cloud coverage, methods such as time series or model inference cannot fully and effectively estimate the deviations in the cloud removal effect.

[0004] Therefore, there is an urgent need to provide a scheme to effectively evaluate cloud removal bias in order to better correct the cloud removal method and thus improve the accuracy of remote sensing cloud removal products.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0006] The purpose of the embodiments of the present disclosure is to provide a cloud removal deviation estimation method, device and equipment based on random sampling, thereby at least to a certain extent solving the deviation problem existing in the related art that is unable to comprehensively and effectively estimate the declouding effect.

[0007] According to a first aspect of an embodiment of the present disclosure, a cloud removal bias estimation method based on random sampling is provided, comprising:

[0008] Obtain initial remote sensing images of the study area;

[0009] Dividing the initial remote sensing image into a plurality of evaluation groups according to the average number of snow days per year, and randomly selecting candidate pixels from each of the evaluation groups;

[0010] Performing cloud persistence analysis on the study area, and generating pseudo cloud layer data of the candidate pixels based on the analysis results;

[0011] Superimposing the pseudo cloud layer data with the initial remote sensing image to obtain a target remote sensing image;

[0012] De-clouding the target remote sensing image to obtain multiple groups of pixel filling values;

[0013] Calculate the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image.

[0014] In an exemplary embodiment of the present disclosure, dividing the initial remote sensing image into a plurality of evaluation groups according to the average number of snow days per year, and randomly selecting candidate pixels from each of the evaluation groups, comprises:

[0015] Extracting a first time series of snow cover in the study area from historical remote sensing data of the study area;

[0016] Calculate the annual average number of snow days corresponding to each original pixel in the initial remote sensing image according to the first time series;

[0017] The corresponding original pixels in the initial remote sensing image are divided into a plurality of evaluation groups according to the annual average number of snow days, and a preset number of candidate pixels are randomly selected from each evaluation group.

[0018] In an exemplary embodiment of the present disclosure, the analysis result includes cloud persistence frequency;

[0019] The performing cloud persistence analysis on the study area and generating pseudo cloud layer data of the candidate pixels based on the analysis result includes:

[0020] Extracting a second time series of cloud coverage of the study area from historical remote sensing data of the study area;

[0021] Obtaining the cloud persistence frequency of the study area within a preset time period according to the second time series statistics;

[0022] Calculating the cloud coverage probability of the study area within a target time period according to the cloud persistence frequency;

[0023] Based on the cloud coverage probability, pseudo cloud layer data of each candidate pixel is repeatedly generated according to a preset number of times.

[0024] In an exemplary embodiment of the present disclosure, the step of calculating the cloud coverage probability of the study area within a target time period according to the cloud persistence frequency includes:

[0025] The cloud persistence frequency is normalized and calculated to obtain the cloud coverage probability of the study area in the target time period:

[0026]

[0027] Among them, p n is the cloud coverage probability, which indicates the probability that the study area is covered by clouds for n consecutive days within the target time period, F nis the cloud persistence frequency, which indicates the frequency of cloud coverage for n consecutive days in the preset time period in the study area, and N indicates the maximum number of days with continuous cloud coverage in the target time period.

[0028] In an exemplary embodiment of the present disclosure, the step of repeatedly generating pseudo cloud layer data of each candidate pixel according to a preset number of times based on the cloud cover probability includes:

[0029] Determining whether each of the candidate pixels is in a cloud coverage state according to the cloud coverage probability of each of the candidate pixels;

[0030] The process is repeated for a preset number of times, and the cloud layer attributes corresponding to the target pixels in the candidate pixels that are in a cloud-covered state are adjusted to generate pseudo cloud layer data of the target pixels.

[0031] In an exemplary embodiment of the present disclosure, performing cloud removal processing on the target remote sensing image to obtain multiple groups of pixel padding values ​​includes:

[0032] Performing binary mask processing on the target remote sensing image to obtain a mask image;

[0033] The cloud-covered area in the mask image is removed using the trained cloud removal model to obtain a cloud-removed image, and a plurality of groups of pixel filling values ​​corresponding to the cloud-covered area in the cloud-removed image are extracted.

[0034] In an exemplary embodiment of the present disclosure, the calculating of the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image includes:

[0035] The number of candidate pixels selected from each evaluation group is fixed, and the number of pseudo cloud layer generation times of each candidate pixel is adjusted;

[0036] Calculating a first error between each of the pixel filling values ​​corresponding to each generation of a pseudo cloud layer and a corresponding pixel true value in the initial remote sensing image, and obtaining a first error combination according to each of the first errors;

[0037] Fixing the number of pseudo cloud layer generation times of each candidate pixel in each evaluation group, and adjusting the number of selected candidate pixels;

[0038] Repeatedly calculating the second errors between the pixel filling values ​​corresponding to the candidate pixels and the corresponding pixel true values ​​in the initial remote sensing image, and obtaining a second error combination according to the second errors;

[0039] If the first error combination and the second error combination meet a preset error evaluation condition, the number of target pseudo cloud layer generation times and the number of target candidate pixels are obtained;

[0040] Based on the target pseudo cloud layer generation times and the target candidate pixel quantity, a target error between each pixel filling value and a corresponding pixel true value in the initial remote sensing image is calculated.

[0041] According to a second aspect of an embodiment of the present disclosure, a cloud removal bias estimation device based on random sampling is provided, comprising:

[0042] Remote sensing image acquisition module, used to obtain the initial remote sensing image of the study area;

[0043] A candidate pixel selection module is used to divide the initial remote sensing image into multiple evaluation groups according to the average number of snow days per year, and randomly select candidate pixels from each evaluation group;

[0044] A pseudo cloud layer simulation module, used for performing cloud persistence analysis on the study area and generating pseudo cloud layer data of the candidate pixels based on the analysis results;

[0045] A target image determination module, used for superimposing the pseudo cloud layer data with the initial remote sensing image to obtain a target remote sensing image;

[0046] An image declouding processing module is used to perform declouding on the target remote sensing image to obtain multiple groups of pixel filling values;

[0047] The target error evaluation module is used to calculate the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image.

[0048] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, any step of the method described in the first aspect is implemented.

[0049] According to a fourth aspect of the present disclosure, there is provided an electronic device, including:

[0050] A processor and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, any step of the method described in the first aspect is implemented.

[0051] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0052] The cloud removal bias estimation method based on random sampling in the example embodiment of the present disclosure is achieved by acquiring an initial remote sensing image of the study area; dividing the initial remote sensing image into multiple evaluation groups according to the average number of snow days per year, and randomly selecting candidate pixels from each evaluation group; performing cloud persistence analysis on the study area, and generating pseudo cloud layer data for the candidate pixels based on the analysis results; superimposing the pseudo cloud layer data with the initial remote sensing image to obtain a target remote sensing image; performing cloud removal processing on the target remote sensing image to obtain multiple groups of pixel padding values; and calculating the target error between each pixel padding value and the corresponding pixel true value in the initial remote sensing image. On the one hand, candidate pixels are grouped according to the average annual number of snow days and randomly selected to generate pseudo cloud data, making the sample selection more representative. The impact of clouds under different environmental conditions can be accurately analyzed, which helps to improve the availability and accuracy of remote sensing data obtained under complex weather conditions. Moreover, not only the randomness of actual cloud distribution is taken into account, but also the impact of cloud cover on remote sensing images can be better understood by simulating different cloud coverage conditions. On the other hand, the target remote sensing image is declouded and multiple groups of pixel padding values ​​are generated. The pixel information can be reconstructed under pseudo cloud coverage conditions and the performance of the declouding method in different scenarios can be verified. Furthermore, by calculating the target error between each pixel padding value and the true value of the corresponding pixel in the initial remote sensing image, a multi-dimensional and multi-level deviation evaluation can be achieved, which provides a quantitative basis for evaluating the declouding effect, so as to better correct the cloud removal method and thus improve the accuracy of remote sensing declouding products.

[0053] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0055] Figure 1 A schematic diagram of the system architecture of a cloud removal bias estimation method based on random sampling to which an embodiment of the present disclosure can be applied is shown.

[0056] Figure 2 A flowchart of a cloud removal bias estimation method based on random sampling in an embodiment of the present disclosure is schematically shown.

[0057] Figure 3 A schematic diagram of a pixel random sampling process in an embodiment of the present disclosure is schematically shown.

[0058] Figure 4 A schematic diagram of a pseudo cloud layer simulation process in an embodiment of the present disclosure is schematically shown.

[0059] Figure 5 A schematic diagram of a process for determining a cloud removal error in an embodiment of the present disclosure is schematically shown.

[0060] Figure 6 A schematic diagram schematically shows the relationship between a selected pixel amount and a cloud removal deviation in an embodiment of the present disclosure.

[0061] Figure 7 A schematic diagram schematically shows the relationship between the number of cloud generation times and cloud removal deviation in an embodiment of the present disclosure.

[0062] Figure 8 A block diagram of a cloud removal bias estimation device based on random sampling according to an embodiment of the present disclosure is schematically shown.

[0063] Fig. 9 The structural diagram of an electronic device according to some embodiments of the present disclosure is schematically shown.

[0064] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0065] The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit this specification. The singular forms "a", "the" and "the" used in this specification and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0066] It should be understood that although the terms first, second, third, etc. may be used in this specification to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this specification, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0067] Figure 1 A schematic diagram of the system architecture of the random sampling-based cloud debiasing estimation method to which the present disclosure embodiment can be applied is shown.

[0068] like Figure 1As shown, the system architecture 100 may include one or more of terminal devices such as a smart phone 101, a portable computer 102, a desktop computer 103, a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the terminal device and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0069] The terminal device can be any electronic device with data processing capabilities, which has a display screen for displaying the initial remote sensing image of the study area, candidate pixels, pseudo cloud layer data, target remote sensing image and cloud removal bias assessment results to the user. The electronic device includes but is not limited to the above-mentioned desktop computers, portable computers, smart phones and tablet computers, etc.

[0070] The server 105 can communicate with the remote sensing device to obtain the initial remote sensing image of the study area and other related data. It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. According to the implementation requirements, there may be any number of terminal devices, networks and servers. For example, the server 105 may be a server cluster composed of multiple servers.

[0071] The cloud removal bias estimation method based on random sampling provided in the embodiment of the present disclosure can be executed by a terminal device, and accordingly, the cloud removal bias estimation device based on random sampling can be set in the terminal device. However, it is easy for those skilled in the art to understand that the cloud removal bias estimation method based on random sampling provided in the embodiment of the present disclosure can also be executed by the server 105, and accordingly, the cloud removal bias estimation device based on random sampling can also be set in the server 105, which is not particularly limited in this exemplary embodiment.

[0072] The embodiment of the present disclosure first provides a cloud debias estimation method based on random sampling. The method is described below using server execution as an example.

[0073] Figure 2 The flowchart of a cloud removal bias estimation method based on random sampling in an embodiment of the present disclosure is schematically shown. Figure 2 As shown, the method may include steps S210 to S260:

[0074] Step S210, obtaining an initial remote sensing image of the study area;

[0075] Step S220, dividing the initial remote sensing image into a plurality of evaluation groups according to the average number of snow days per year, and randomly selecting candidate pixels from each of the evaluation groups;

[0076] Step S230, performing cloud persistence analysis on the study area, and generating pseudo cloud layer data of the candidate pixel based on the analysis result;

[0077] Step S240, superimposing the pseudo cloud layer data with the initial remote sensing image to obtain a target remote sensing image;

[0078] Step S250, performing cloud removal processing on the target remote sensing image to obtain multiple groups of pixel filling values;

[0079] Step S260, calculating the target error between each of the pixel filling values ​​and the corresponding pixel true value in the initial remote sensing image.

[0080] The cloud removal bias estimation method based on random sampling provided by the present disclosure is executed. On the one hand, candidate pixels are grouped according to the average annual number of snow days and randomly selected to generate pseudo cloud data, so that the sample selection is more representative, and the cloud impact under different environmental conditions can be accurately analyzed, which is helpful to improve the availability and accuracy of remote sensing data obtained under complex weather conditions. Moreover, not only the randomness of the actual cloud distribution is taken into account, but also by simulating different cloud coverage conditions, the impact of cloud coverage on remote sensing images can be better understood; on the other hand, the target remote sensing image is declouded and multiple groups of pixel padding values ​​are generated, which can reconstruct pixel information under pseudo cloud coverage conditions and verify the performance of the declouding method under different scenarios. Furthermore, by calculating the target error between each pixel padding value and the true value of the corresponding pixel in the initial remote sensing image, a multi-dimensional and multi-level deviation evaluation can be achieved, which provides a quantitative basis for evaluating the declouding effect, so as to better correct the cloud removal method and thereby improve the accuracy of remote sensing declouding products.

[0081] The cloud removal bias estimation method based on random sampling in this example embodiment will be described in detail below.

[0082] In step S210, an initial remote sensing image of the study area is obtained.

[0083] In the example implementation of the present disclosure, the study area refers to the area where the remote sensing data that needs to be cloud covered is located, including but not limited to areas with various landforms such as snow cover, vegetation, and land types. The present disclosure does not specifically limit the type of the study area. The initial remote sensing image refers to image data containing surface information of the study area obtained by remote sensing satellites, drones, etc. According to the size of the study area, the observation requirements, and the required spatial and temporal resolution, a suitable remote sensing platform can be selected, and the initial remote sensing image of the study area can be obtained according to specific research objectives such as monitoring snow cover, evaluating vegetation changes, and detecting land use. It should be noted that it is also necessary to select a suitable time period according to the research objectives to ensure the timeliness of the remote sensing image. For example, if the research goal is to monitor snow cover, a time period such as winter or spring when snow may exist should be selected. If the cloud cover in the study area is obviously seasonal, it is also necessary to consider selecting a time period with less cloud cover to reduce the impact of cloud cover on image quality.

[0084] For example, if you need to monitor snow cover in the study area, you can choose a medium spatial resolution (such as 30 meters) and a medium frequency (such as monthly) temporal resolution, and obtain the initial remote sensing image of the study area from the MODIS (a remote sensing data source) platform.

[0085] After obtaining the initial remote sensing image of the study area, the image can also be preprocessed for subsequent analysis. For example, the preprocessing steps include geometric correction, radiation correction, and atmospheric correction. Geometric correction is used to align the pixel coordinates in the image with the actual geographic coordinates on the earth to ensure the geometric accuracy of the remote sensing image, radiation correction is used to eliminate sensor errors or atmospheric effects, and atmospheric correction is used to correct the atmospheric scattering and absorption effects in the remote sensing image to improve the accuracy of the surface reflectivity.

[0086] By acquiring the initial remote sensing images of the study area, we can comprehensively capture the surface observation data of the study area and provide high-resolution, multi-dimensional basic information for subsequent analysis.

[0087] In step S220, the initial remote sensing image is divided into a plurality of evaluation groups according to the average number of snow days per year, and candidate pixels are randomly selected from each of the evaluation groups.

[0088] Among them, the average annual number of snow days refers to the average number of snowfall days observed in a certain area within a year. The average annual number of snow days can reflect the frequency and seasonal changes of snowfall in the area, and can usually be obtained through meteorological station data or long-term remote sensing data (such as MODIS images) combined with ground observations.

[0089] In one example implementation, reference is made to Figure 3 As shown, step S220 may further include steps S310 to S330:

[0090] Step S310: extracting a first time series of the study area being covered by snow from historical remote sensing data of the study area.

[0091] The first time series refers to the time series data recording the snow cover status. For example, historical remote sensing images of a certain time period (such as the past five years or more) can be obtained, and specific bands (such as near infrared bands and short-wave infrared bands) in multispectral remote sensing images can be used to extract snow cover information of historical remote sensing images, such as snow cover parameters. It should be noted that the time span of historical remote sensing data should be long enough to calculate the accurate average annual number of snow days.

[0092] Specifically, the snow cover parameters in the historical remote sensing image are determined according to formula (1):

[0093]

[0094] Among them, NDSI is the snow cover parameter, Green Band is the green band data, and SWIR Band is the shortwave infrared band data. Snow-covered areas usually have higher NDSI values ​​(such as greater than 0.4), while snow-free areas have lower NDSI values.

[0095] By determining the snow cover parameters in the historical remote sensing images, the snow cover of the study area in the historical remote sensing images can be extracted, thereby obtaining a first time series representing the snow cover status of each day. For example, for a first time series of 365 days, the result is 365 data points, each of which represents whether the area is covered with snow on each day, such as 1 for snow cover and 0 for no snow cover.

[0096] Step S320, calculating the annual average number of snow days corresponding to each original pixel in the initial remote sensing image according to the first time series.

[0097] For each raw pixel in the initial remote sensing image, the number of days covered by snow in the corresponding time period is counted according to each data point in the first time series. For example, if a raw pixel is covered by snow for an average of 150 days per year in the past five years, the annual average number of snow days for the raw pixel is 150. Finally, each raw pixel corresponds to an annual average number of snow days as a representative feature of the raw pixel in the seasonality of snow cover.

[0098] Step S330, dividing the corresponding original pixels in the initial remote sensing image into a plurality of evaluation groups according to the annual average number of snow days, and randomly selecting a preset number of candidate pixels from each evaluation group.

[0099] After obtaining the average annual number of snow days corresponding to each original pixel, all the original pixels in the initial remote sensing image can be divided into multiple evaluation groups according to the average annual number of snow days, so that the cloud removal effect in areas with different snow coverage can be analyzed more specifically. For example, according to the different average annual number of snow days, all the original pixels in the study area are divided into different groups, namely 0-50 days, 50-100 days, 100-150 days and more than 150 days. Of course, the division criteria can also be customized according to the specific needs of the study area, and this disclosure does not limit this.

[0100] According to the average annual number of snow days corresponding to the location of each original pixel, the pixel data in the initial remote sensing image of the entire study area are assigned to different evaluation groups, so that the original pixels in each evaluation group have similar average annual number of snow days characteristics, thereby reducing the errors caused by differences in natural conditions in subsequent sampling and analysis.

[0101] Furthermore, after the evaluation groups are divided, in order to perform the subsequent pseudo cloud generation and cloud removal processing, a preset number of candidate pixels need to be randomly selected from each evaluation group for multiple sampling. The purpose of random selection is to ensure the representativeness of the samples while avoiding the bias of the final result due to excessive concentration in a particular group.

[0102] For example, the number of pixels to be selected from each evaluation group can be determined first, such as the size of the study area, the resolution of the initial remote sensing image, and the required sample size. Then, a preset number of candidate pixels are selected from each evaluation group using a random number generation algorithm. For example, a random number is generated for each original pixel, and the random numbers are sorted from small to large, and the original pixels ranked at the top are selected as candidate pixels. This process ensures that the selected candidate pixels have a good distribution and avoids bias towards a specific area or feature.

[0103] In the pixel selection process, it is necessary to ensure that the selected candidate pixels can cover the different terrains, land cover types and other important features in the evaluation group. If the pixel distribution in some areas is too concentrated or uneven, the selection strategy can be adaptively adjusted to increase or decrease the sample size in certain specific areas to improve the representativeness of the evaluation results.

[0104] The final selected candidate pixels will be used for subsequent pseudo cloud generation, cloud removal processing, and error assessment. It should be noted that the selected candidate pixels will have different cloud cover characteristics (affected by the average number of snow days per year), which can fully reflect the performance and bias of the cloud removal method under different climatic conditions.

[0105] In step S230, cloud persistence analysis is performed on the study area, and pseudo cloud layer data of the candidate pixels is generated based on the analysis result.

[0106] In the example implementation of the present disclosure, by performing cloud persistence analysis on the study area, the spatiotemporal distribution of clouds can be accurately simulated, and pseudo cloud layer data that conforms to the distribution characteristics can be generated for subsequent data filling and cloud removal processing. Among them, cloud persistence analysis refers to analyzing the distribution and change patterns of clouds in the study area, especially the persistence of clouds in the time dimension.

[0107] Exemplarily, the analysis result of the cloud persistence analysis includes a cloud persistence frequency. In the exemplary embodiment of the present disclosure, the cloud persistence frequency can be defined as the number of times a certain area is covered by clouds for n consecutive days within a preset time period (such as a year), denoted as F. n , which can help understand the stability, seasonality, and spatiotemporal variations of cloud cover.

[0108] refer to Figure 4 As shown, pseudo cloud layer data of candidate pixels can be generated according to steps S410 to S440:

[0109] Step S410, extracting a second time series of cloud coverage of the study area from historical remote sensing data of the study area;

[0110] The second time series refers to the time series data recording the cloud coverage status. Exemplarily, historical remote sensing images of a certain time period (such as the past year or more) can be obtained, and the cloud coverage of each day can be extracted by a cloud detection algorithm based on remote sensing images. For example, cloud coverage information of historical remote sensing images can be extracted based on the spectral characteristics of pixels, or the spectral differences between visible light and near-infrared bands can be used to distinguish between cloud and non-cloud areas.

[0111] It can be understood that the result of cloud cover detection on historical remote sensing images is a binary image, where each pixel value indicates whether the location is covered by clouds, such as 1 for cloud coverage and 0 for no cloud coverage. Then, the cloud cover results of each historical remote sensing image are arranged in chronological order to obtain a complete second time series, in which the data at each time point represents the cloud coverage situation.

[0112] Step S420, obtaining the cloud persistence frequency of the study area within a preset time period according to the second time series statistics;

[0113] For each original pixel in the initial remote sensing image, the number of times each original pixel is covered by clouds for n consecutive days within a preset time period is counted according to each data point in the second time series. For example, by performing a sliding window analysis on the second time series, the number of times the pixel is covered by clouds for n consecutive days (e.g., 3 consecutive days, 5 consecutive days, etc.) is calculated. For each pixel, a window (n days in length) is slid along the time axis, and whether all days in the window are covered by clouds is counted. Assuming that in the past year, a certain original pixel has been covered by clouds for 3 consecutive days 40 times, then the number of times the pixel is covered by clouds for 3 consecutive days is 40 times, that is, the cloud persistence frequency is 40.

[0114] Step S430, calculating the cloud coverage probability of the study area within a target time period according to the cloud persistence frequency;

[0115] According to the obtained cloud persistence frequency, the cloud coverage probability under different consecutive days (such as 1 day, 2 days, 3 days, etc.) can be generated for each original pixel, which is recorded as p n , which indicates the probability that the pixel is covered by clouds within a specific number of days. This probability reflects the persistence of cloud coverage of the pixel in different time periods and has a high spatiotemporal consistency.

[0116] For example, the cloud persistence frequency can be normalized and calculated to obtain the cloud coverage probability of the study area in the target time period:

[0117]

[0118] Among them, p n is the cloud coverage probability, which indicates the probability of cloud coverage in the study area for n consecutive days within the target time period, F n is the cloud persistence frequency, which indicates the frequency of cloud cover in the study area for n consecutive days in the preset time period, and N indicates the maximum number of days with continuous cloud cover in the target time period.

[0119] By obtaining the probability distribution of cloud coverage of different durations in the study area, data support is provided for the optimization of cloud removal methods, error assessment and other analyses.

[0120] Step S440: Based on the cloud coverage probability, repeatedly generate pseudo cloud layer data for each candidate pixel according to a preset number of times.

[0121] Among them, pseudo cloud data refers to simulated cloud coverage data, the purpose of which is to simulate real cloud cover by replacing part of the initial remote sensing image with cloud coverage. The preset number indicates the number of times pseudo cloud data is generated. The specific value of the preset number can be determined according to the experimental requirements. Each generated pseudo cloud data will simulate different cloud coverage patterns based on the same cloud coverage probability and randomness.

[0122] Exemplarily, whether each candidate pixel is in a cloud covered state can be determined according to the cloud cover probability of each candidate pixel. Repeating the process for a preset number of times, the cloud layer attribute corresponding to the target pixel in the cloud covered state among the candidate pixels is adjusted to generate pseudo cloud layer data of the target pixel.

[0123] For each candidate pixel, the corresponding cloud coverage probability is used to determine whether the pixel is in a cloud covered state. For example, assuming that the cloud coverage probability p of each candidate pixel is n It is known that for each pixel i, a random number r in the range [0,1] is randomly generated i , if the random number is less than or equal to the cloud coverage probability p of the pixel n , then the pixel is assigned a cloud coverage state, otherwise, it is assigned a cloud-free state, ensuring that the cloud coverage state of each pixel is consistent with the corresponding cloud coverage probability.

[0124] In order to simulate the randomness and uncertainty of cloud coverage, it is usually necessary to generate pseudo cloud data multiple times. By repeatedly generating multiple pseudo cloud samples, it can be ensured that different cloud coverage patterns are covered, providing more extensive and diverse test data for declouding processing and error assessment.

[0125] For each repeated generation of pseudo cloud data, it can be determined whether each pixel is in a cloud-covered state based on the cloud coverage probability and random number of each candidate pixel. If the pixel is in a cloud-covered state, the pixel is marked as cloud-covered. If it is in a cloudless state, the pixel still retains the original information of the surface. It should be noted that each time pseudo cloud data is generated, it is necessary to record the cloud properties of each pixel, including cloud density, cloud type, and spatial distribution of clouds. Among them, cloud density represents the concentration or intensity of the cloud-covered area, which can be described by indexes such as cloud optical thickness and cloud water content. Cloud types include high clouds, low clouds, and other types. The spatial distribution of clouds represents the distribution of cloud-covered areas in geographic space.

[0126] The candidate pixels marked as cloud-covered are used as target pixels, and the cloud properties of the target pixels can be adjusted. For example, if the target pixel is in a cloud-covered state, the reflectivity or brightness value of the target pixel can be adjusted according to a predetermined cloud density parameter. The denser the cloud, the greater the impact of the coverage on the surface reflection, and the cloud properties are adjusted to a higher density accordingly. The cloud type can be determined based on the spatial distribution and geographical characteristics of the cloud-covered area (such as weather conditions, seasonal changes, etc.), and the cloud properties can be adjusted according to the cloud type to simulate the real cloud occlusion effect, thereby generating pseudo cloud data for the target pixel.

[0127] The final generated pseudo cloud data will contain pixel information under different cloud coverage states. Each repeated generation of pseudo cloud data will simulate different cloud coverage scenes by adjusting the cloud properties. The pseudo cloud data will be used as the cloud occlusion area of ​​the target remote sensing image to simulate the actual cloud coverage effect.

[0128] By repeatedly generating pseudo cloud data, we can ensure that the generated data is statistically significant and reliable, provide a variety of cloud occlusion patterns for the target remote sensing images, and provide more comprehensive test data for the verification and error evaluation of the cloud removal method, which helps to evaluate the performance of the cloud removal algorithm under different pseudo cloud conditions.

[0129] In step S240, the pseudo cloud layer data is superimposed on the initial remote sensing image to obtain a target remote sensing image.

[0130] Among them, the initial remote sensing image contains the real surface information of the study area, and the pseudo cloud data simulates the affected area of ​​cloud cover. The initial remote sensing image and the pseudo cloud data are superimposed to simulate the effect of cloud occlusion in the initial remote sensing image and obtain an image that includes the influence of cloud cover as the target remote sensing image.

[0131] For example, when the pseudo cloud layer data is superimposed on each pixel value of the initial remote sensing image, assuming that a certain pixel value in the initial remote sensing image is I orig (x, y), the corresponding pseudo cloud layer data is C(x, y), then each pixel value I of the target remote sensing image target (x,y) is:

[0132] I target (x,y)=I orig (x,y)×(1-C(x,y))+C(x,y)×NoDate (3)

[0133] Among them, if there is cloud at a certain pixel position, the corresponding C(x, y) = 1, otherwise C(x, y) = 0, and NoDate represents invalid values ​​such as missing values ​​in the cloud coverage area, which usually does not contain valid surface information.

[0134] By superimposing the pseudo cloud layer data with the initial remote sensing image, it is possible to simulate the real cloud coverage scene and generate the target remote sensing image including the cloud coverage effect. It can accurately simulate the missing value situation under different cloud coverage scenes, and provide a real and controllable test environment for subsequent data filling and deviation evaluation, so as to comprehensively evaluate the performance of the cloud removal algorithm and provide support for the precise analysis of remote sensing data. In other words, the generated target remote sensing image can be used to verify the performance of different cloud removal algorithms. For example, by calculating the error between the filled value of the image after cloud removal and the true value of the initial remote sensing image, and comparing the cloud removal effects under different cloud coverage modes, it is possible to understand the stability and accuracy of the cloud removal algorithm in different regions and conditions.

[0135] In step S250, the target remote sensing image is subjected to cloud removal processing to obtain multiple groups of pixel filling values.

[0136] Among them, declouding refers to restoring the area covered by clouds in the target remote sensing image so that all pixels in the image can reflect the real information of the surface. It can be understood that declouding is usually completed by filling in missing pixel values, and the filling value represents the reflectance value, brightness value and other information of the real surface in the cloud-covered area. For example, the pixel value of the cloud-covered area can be predicted by spatial or temporal interpolation methods, or the information of surrounding pixels can be used to predict the pixel value of the cloud-covered area based on the similarity of local areas of remote sensing images. Missing pixel values ​​can also be predicted by regression analysis models, deep learning models, etc., and filled in with surrounding pixels and other features (such as time series features). The present disclosure does not specifically limit the declouding method.

[0137] In an example implementation, the target remote sensing image may be subjected to binary mask processing to obtain a mask image. The mask image is a binary image of the same size as the target remote sensing image. For example, by performing cloud cover detection on each pixel in the target remote sensing image, it is determined whether the pixel belongs to a cloud covered area, wherein the cloud covered area is marked as 1 and the non-cloud covered area is marked as 0, thereby generating a corresponding mask image. By separating the cloud covered area from the target remote sensing image, a clear partition is provided for subsequent cloud removal processing.

[0138] Then, the cloud-covered area in the mask image is removed using the trained cloud removal model to obtain a cloud-removed image, and multiple groups of pixel padding values ​​corresponding to the cloud-covered area in the cloud-removed image are extracted. The cloud removal model can be a pixel or block-based regression analysis model, a support vector machine, a random forest, a deep neural network, and other models. The cloud removal model can automatically identify and remove cloud-covered areas by learning a large amount of labeled training data. Each pixel value in the cloud-removed image is generated by processing the effective pixel information around the pixel by the trained cloud removal model, and the padding value is predicted based on contextual information (such as spectral characteristics of surrounding pixels, time series, etc.). Accordingly, the mask image is processed using the trained cloud removal model, and the padding values ​​of the cloud-covered area can be predicted, and these padding values ​​are applied to the cloud-covered area in the mask image to restore the surface information of the cloud-occluded area.

[0139] Each time the cloud-covered area in the target remote sensing image is processed, a corresponding pixel filling value will be generated. The pixel filling value represents the pixel value after the cloud-covered area is restored. The multiple sets of pixel filling values ​​obtained can represent the de-clouding recovery results under multiple different cloud coverage scenarios, which will be used for subsequent error evaluation and optimization of the cloud removal algorithm, especially in areas with more complex cloud coverage patterns.

[0140] In step S260, a target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image is calculated.

[0141] By comparing the pixel filling value with the real surface data, the error between the two is calculated and the accuracy of the cloud removal algorithm is evaluated according to the error. For example, the parameters of the cloud removal model can be adjusted according to the errors between multiple groups of pixel filling values ​​and the corresponding pixel real values ​​to improve the cloud removal effect. The error may include standard deviation and mean deviation, which is not limited in the present disclosure.

[0142] In the exemplary implementation of the present disclosure, the evaluation conditions can be optimized by adjusting the number of candidate pixels and the number of pseudo cloud layer generation times, and the optimal evaluation parameters can be determined through multiple error calculations, thereby achieving accurate restoration of the target remote sensing image. Figure 5 As shown, step S260 may further include steps S510 to S560:

[0143] Step S510, fixing the number of candidate pixels selected in each evaluation group, and adjusting the number of pseudo cloud layer generation times of each candidate pixel;

[0144] After obtaining multiple evaluation groups, the number of candidate pixels selected in each evaluation group is kept unchanged, and the number of pseudo cloud layer generation times is changed. By increasing the number of pseudo cloud layer generation times, the diversity of different cloud coverage scenes is simulated to observe the change of error. For example, a fixed number (such as 1000) of candidate pixels are randomly selected from each evaluation group. The selected candidate pixels represent the remote sensing image area under different cloud coverage conditions. By adjusting the number of pseudo cloud layer generation times, such as from 10 times to 50 times, it is tested whether the error value can be stabilized by increasing the number of pseudo cloud layer generation times, and then the influence of different pseudo cloud layer scenes on the pixel filling value is tested, and the influence of pseudo cloud layer samples on the declouding effect is evaluated.

[0145] Step S520, calculating a first error between each of the pixel filling values ​​corresponding to each generation of the pseudo cloud layer and a corresponding pixel true value in the initial remote sensing image, and obtaining a first error combination according to each of the first errors;

[0146] The corresponding pixel padding value is obtained according to the pseudo cloud layer data generated each time, and the first error is calculated by comparing the pixel padding value with the true value of the pixel in the initial remote sensing image. In other words, in the process of generating multiple pseudo cloud layers, an error value is calculated each time, and the first error combination can be composed of all the error values.

[0147] It is understandable that the first error combination may include multiple statistics, such as mean deviation, standard deviation, etc., to provide an overall performance evaluation of the cloud removal method under the current number of generation times. Among them, the standard deviation is used to measure the degree of dispersion of the error, indicating the fluctuation between the error value and the mean. The smaller the standard deviation, the smaller the fluctuation of the error, and the more consistent the performance of the cloud removal method. The mean deviation is the average of all errors. A positive deviation indicates that the cloud removal algorithm underestimates or overestimates the filling value, and a negative deviation indicates the opposite.

[0148] Step S530, fixing the pseudo cloud layer generation times of each candidate pixel in each evaluation group, and adjusting the number of selected candidate pixels;

[0149] Then, the number of pseudo cloud layer generation for each candidate pixel is kept fixed, and the number of candidate pixels selected in each evaluation group is adjusted. By changing the number of candidate pixels, it is tested whether the error can be reduced by increasing the number of sampling points, that is, the performance of the cloud removal algorithm under different sample sizes is examined to understand the stability and accuracy of the cloud removal algorithm when the amount of data increases.

[0150] Step S540, repeatedly calculating the second errors between the pixel filling values ​​corresponding to the candidate pixels and the corresponding pixel true values ​​in the initial remote sensing image, and obtaining a second error combination according to the second errors;

[0151] After adjusting the number of candidate pixels, the pseudo cloud layer is generated again, and the second error between the pixel filling value obtained in each pseudo cloud layer generation process and the corresponding pixel true value is calculated, and all the second errors form a second error combination. The second error combination can also include statistics such as mean error and standard error to help judge the performance of the cloud removal algorithm under different sample size conditions.

[0152] Step S550, if the first error combination and the second error combination meet the preset error evaluation condition, the target pseudo cloud layer generation times and the target candidate pixel quantity are obtained;

[0153] Taking the example that both the first error and the second error are standard deviations, when the number of candidate pixels or the number of pseudo cloud layer generation increases, the standard deviation gradually decreases, indicating that the error variation range is converging and the performance of the cloud removal algorithm tends to be stable. In other words, with more samples and more pseudo cloud layer generation, the error fluctuation range of the cloud removal algorithm becomes smaller, and the evaluation result of the cloud removal deviation is more reliable.

[0154] refer to Figure 6 As shown in Figure 1, a schematic diagram of the relationship between the number of selected pixels and the cloud removal deviation is given. At this time, the number of cloud generation times for each pixel is fixed to 1000 times. The cloud removal deviation includes the standard deviation and the average deviation. Figure 6 601 represents the relationship curve between the number of selected pixels in each evaluation group and the average deviation, and 602 represents the relationship curve between the number of selected pixels in each evaluation group and the standard deviation. Figure 7 As shown in Figure 1, a schematic diagram of the relationship between the number of cloud generation times and cloud removal deviation is given. At this time, the number of selected pixels in each evaluation group is fixed to 1500, and the cloud removal deviation also includes the standard deviation and the average deviation. Figure 7 701 represents the relationship curve between the number of cloud generation times and the average deviation of each pixel in each evaluation group, and 702 represents the relationship curve between the number of cloud generation times and the standard deviation of each pixel in each evaluation group.

[0155] Accordingly, in an example implementation of the present disclosure, the preset error evaluation condition may refer to the errors in the first error combination and the second error combination gradually approaching each other, and the first error in the first error combination and the second error in the second error combination gradually decreasing to a negligible value, indicating that the number of candidate pixels and the number of pseudo cloud layer generation times are large enough, and the impact on the evaluation performance is negligible. At this time, the number of candidate pixels (i.e., the target number of candidate pixels) and the number of pseudo cloud layer generation times (i.e., the target number of pseudo cloud layer generation times) are the optimal parameter settings, and at this time the cloud removal algorithm will not continuously overestimate or underestimate the filling values ​​of certain cloud-covered areas, but tends to a more accurate estimate, thereby minimizing the impact of the pseudo cloud layer on the image quality.

[0156] Step S560, based on the target pseudo cloud layer generation times and the target candidate pixel quantity, calculate the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image.

[0157] After determining the number of target pseudo cloud layer generation and the number of target candidate pixels, the pixel padding value of each candidate pixel is obtained based on a sufficiently large number of target pseudo cloud layer generation and the number of target candidate pixels, and a stable target error is calculated based on the pixel padding value and the corresponding pixel true value in the initial remote sensing image. The target error is an important indicator for the final evaluation of the declouding effect, which is used to measure the difference between the declouded image and the original image. A lower target error indicates that the declouding algorithm is more effective and can more accurately restore the surface information of the cloud-occluded area.

[0158] By continuously adjusting the number of pseudo cloud layer generation times and the number of candidate pixels, combined with error analysis, the parameter settings of cloud removal processing are optimized. By repeatedly calculating the errors, the optimal parameter configuration is found to ensure that the cloud removal method has high accuracy and stability under different conditions.

[0159] In the example implementation of the present disclosure, the authenticity of the simulated environment is improved by quantitatively analyzing the spatiotemporal distribution of cloud coverage and accurately modeling the generation of pseudo-cloud layers. Furthermore, through the reasonable division and random extraction of samples, the characteristics of the study area can be more comprehensively covered. In addition, the precise calculation of the target error and its statistical aggregation among multiple sets of data make the quantification of systematic deviations clearer, providing a quantitative basis for evaluating the declouding effect, so as to better correct the cloud removal method and thus improve the accuracy of remote sensing declouding products.

[0160] It should be noted that the cloud removal bias estimation method based on random sampling provided in the present disclosure can be applied to the analysis of snow cover remote sensing data, which can effectively estimate the impact of cloud cover removal and obtain more accurate snow cover results.

[0161] Snow-covered areas are usually accompanied by the presence of clouds, which can cause the loss of surface information in some areas, thus significantly affecting the accuracy of snow cover analysis results. In the example implementation of the present disclosure, pseudo cloud layer data matching the snow distribution characteristics is generated by analyzing the cloud persistence of historical remote sensing data of snow-covered areas. Pixels in the snow-covered area are randomly selected as candidate samples, and the impact of the cloud removal algorithm on snow cover is estimated by generating pseudo cloud layers multiple times. By calculating the deviation between the pixel filling value and the true snow cover value, a quantitative evaluation of the effect of cloud removal on snow cover results is obtained. By adjusting the number of pseudo cloud layer generation times and the number of candidate pixel samples, the performance of the cloud removal algorithm can be evaluated from multiple dimensions and scenarios to ensure the reliability of the cloud removal algorithm in different snow cover modes, and the cloud removal process is optimized based on error evaluation to improve the accuracy of snow cover remote sensing analysis results.

[0162] Applying the cloud removal bias estimation method based on random sampling to the analysis of snow cover remote sensing data can effectively deal with complex scenarios where snow cover and clouds coexist, quantitatively evaluate the bias of the cloud removal algorithm and optimize the snow cover results. This method can significantly improve the accuracy and reliability of snow cover analysis, providing strong data support for dynamic snow monitoring, climate change research and water resources assessment.

[0163] In this example implementation, a cloud removal bias estimation device based on random sampling is also provided. Figure 8 As shown, the device may include a remote sensing image acquisition module 810, a candidate pixel selection module 820, a pseudo cloud layer simulation module 830, a target image determination module 840, an image cloud removal processing module 850 and a target error evaluation module 860, wherein:

[0164] A remote sensing image acquisition module 810 is used to acquire an initial remote sensing image of a study area;

[0165] A candidate pixel selection module 820 is used to divide the initial remote sensing image into multiple evaluation groups according to the average number of snow days per year, and randomly select candidate pixels from each evaluation group;

[0166] A pseudo cloud layer simulation module 830, for performing cloud persistence analysis on the study area and generating pseudo cloud layer data of the candidate pixel based on the analysis result;

[0167] A target image determination module 840 is used to superimpose the pseudo cloud layer data with the initial remote sensing image to obtain a target remote sensing image;

[0168] An image de-clouding processing module 850 is used to perform de-clouding processing on the target remote sensing image to obtain multiple groups of pixel filling values;

[0169] The target error evaluation module 860 is used to calculate the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image.

[0170] In an exemplary embodiment of the present disclosure, when the candidate pixel selection module 820 divides the initial remote sensing image into multiple evaluation groups according to the average number of snow days per year and randomly selects candidate pixels from each of the evaluation groups, it is specifically used to:

[0171] Extracting a first time series of snow cover in the study area from historical remote sensing data of the study area;

[0172] Calculate the annual average number of snow days corresponding to each original pixel in the initial remote sensing image according to the first time series;

[0173] The corresponding original pixels in the initial remote sensing image are divided into a plurality of evaluation groups according to the annual average number of snow days, and a preset number of candidate pixels are randomly selected from each evaluation group.

[0174] In an exemplary embodiment of the present disclosure, the analysis result includes cloud persistence frequency; when the pseudo cloud layer simulation module 830 performs cloud persistence analysis on the study area and generates pseudo cloud layer data of the candidate pixel based on the analysis result, it is specifically used to:

[0175] Extracting a second time series of cloud coverage of the study area from historical remote sensing data of the study area;

[0176] Obtaining the cloud persistence frequency of the study area within a preset time period according to the second time series statistics;

[0177] Calculating the cloud coverage probability of the study area within a target time period according to the cloud persistence frequency;

[0178] Based on the cloud coverage probability, pseudo cloud layer data of each candidate pixel is repeatedly generated according to a preset number of times.

[0179] In an exemplary embodiment of the present disclosure, when the pseudo cloud layer simulation module 830 calculates the cloud coverage probability of the study area within the target time period according to the cloud persistence frequency, it is specifically used to:

[0180] The cloud persistence frequency is normalized and calculated to obtain the cloud coverage probability of the study area in the target time period:

[0181]

[0182] Among them, p n is the cloud coverage probability, which indicates the probability that the study area is covered by clouds for n consecutive days within the target time period, F n is the cloud persistence frequency, which indicates the frequency of cloud coverage for n consecutive days in the preset time period in the study area, and N indicates the maximum number of days with continuous cloud coverage in the target time period.

[0183] In an exemplary embodiment of the present disclosure, when the pseudo cloud layer simulation module 830 repeatedly generates the pseudo cloud layer data of each candidate pixel according to a preset number of times based on the cloud coverage probability, it is specifically used to:

[0184] Determining whether each of the candidate pixels is in a cloud coverage state according to the cloud coverage probability of each of the candidate pixels;

[0185] The process is repeated for a preset number of times, and the cloud layer attributes corresponding to the target pixels in the candidate pixels that are in a cloud-covered state are adjusted to generate pseudo cloud layer data of the target pixels.

[0186] In an exemplary embodiment of the present disclosure, when the image cloud removal processing module 850 performs cloud removal processing on the target remote sensing image to obtain multiple groups of pixel padding values, it is specifically used to:

[0187] Performing binary mask processing on the target remote sensing image to obtain a mask image;

[0188] The cloud-covered area in the mask image is removed using the trained cloud removal model to obtain a cloud-removed image, and a plurality of groups of pixel filling values ​​corresponding to the cloud-covered area in the cloud-removed image are extracted.

[0189] In an exemplary embodiment of the present disclosure, when calculating the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image, the target error evaluation module 860 is specifically used to:

[0190] The number of candidate pixels selected from each evaluation group is fixed, and the number of pseudo cloud layer generation times of each candidate pixel is adjusted;

[0191] Calculating a first error between each of the pixel filling values ​​corresponding to each generation of a pseudo cloud layer and a corresponding pixel true value in the initial remote sensing image, and obtaining a first error combination according to each of the first errors;

[0192] Fixing the number of pseudo cloud layer generation times of each candidate pixel in each evaluation group, and adjusting the number of selected candidate pixels;

[0193] Repeatedly calculating the second errors between the pixel filling values ​​corresponding to the candidate pixels and the corresponding pixel true values ​​in the initial remote sensing image, and obtaining a second error combination according to the second errors;

[0194] If the first error combination and the second error combination meet a preset error evaluation condition, the number of target pseudo cloud layer generation times and the number of target candidate pixels are obtained;

[0195] Based on the target pseudo cloud layer generation times and the target candidate pixel quantity, a target error between each pixel filling value and a corresponding pixel true value in the initial remote sensing image is calculated.

[0196] The specific details of each module of the above-mentioned cloud removal bias estimation device based on random sampling have been described in detail in the corresponding cloud removal bias estimation method based on random sampling, so they will not be repeated here.

[0197] The exemplary embodiments of the present disclosure also provide a computer-readable storage medium on which a program product capable of implementing the above-mentioned method of the present specification is stored. In some possible implementations, various aspects of the present disclosure may also be implemented in the form of a program product, which includes a program code, and when the program product is run on an electronic device, the program code is used to enable the electronic device to perform the steps according to various exemplary embodiments of the present disclosure described in the above-mentioned "Exemplary Method" section of the present specification.

[0198] The program product may be in the form of a portable compact disk read-only memory (CD-ROM) and include program code, and may be run on an electronic device, such as a personal computer. However, the program product of the present disclosure is not limited thereto, and in this document, a readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0199] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0200] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0201] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0202] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C#, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0203] In addition, an exemplary embodiment of the present disclosure also provides an electronic device capable of implementing the above-mentioned cloud removal bias estimation method based on random sampling.

[0204] Refer to the following Fig. 9 hereinafter describes an electronic device 900 according to such an embodiment of the present disclosure. Fig. 9 The electronic device 900 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0205] like Fig. 9 As shown, the electronic device 900 is in the form of a general computing device. The components of the electronic device 900 may include, but are not limited to: the at least one processing unit 910, the at least one storage unit 920, a bus 930 connecting different system components (including the storage unit 920 and the processing unit 910), and a display unit 940.

[0206] The storage unit 920 stores program codes, which can be executed by the processing unit 910, so that the processing unit 910 performs the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of this specification. For example, the processing unit 910 can perform the method steps in the exemplary embodiments of the present disclosure. The storage unit 920 can be used to process the initial remote sensing image of the study area in the database, the corresponding target remote sensing image, and the cloud removal deviation result obtained by evaluation, etc., which is not limited by the present disclosure.

[0207] The storage unit 920 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 921 and / or a cache storage unit (Cache) 922 , and may further include a read-only storage unit (ROM) 923 .

[0208] The storage unit 920 may also include a program / utility 924 having a set (at least one) of program modules 925, such program modules 925 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0209] Bus 930 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0210] The electronic device 900 may also communicate with one or more external devices 1000 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 900, and / or communicate with any device that enables the electronic device 900 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 950. Furthermore, the electronic device 900 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 960. As shown, the network adapter 960 communicates with other modules of the electronic device 900 via a bus 930. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 900, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0211] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

[0212] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0213] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.

[0214] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

[0215] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A cloud removal bias estimation method based on random sampling, characterized in that: include: Obtain initial remote sensing images of the study area; Dividing the initial remote sensing image into a plurality of evaluation groups according to the average number of snow days per year, and randomly selecting candidate pixels from each of the evaluation groups; Performing cloud persistence analysis on the study area, and generating pseudo cloud layer data of the candidate pixels based on the analysis results; Superimposing the pseudo cloud layer data with the initial remote sensing image to obtain a target remote sensing image; De-clouding the target remote sensing image to obtain multiple groups of pixel filling values; Calculate the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image.

2. The cloud removal bias estimation method based on random sampling according to claim 1, characterized in that: The initial remote sensing image is divided into a plurality of evaluation groups according to the average number of snow days per year, and candidate pixels are randomly selected from each of the evaluation groups, including: Extracting a first time series of snow cover in the study area from historical remote sensing data of the study area; Calculate the annual average number of snow days corresponding to each original pixel in the initial remote sensing image according to the first time series; The corresponding original pixels in the initial remote sensing image are divided into a plurality of evaluation groups according to the annual average number of snow days, and a preset number of candidate pixels are randomly selected from each evaluation group.

3. The cloud removal bias estimation method based on random sampling according to claim 2 is characterized in that: The analysis results include cloud persistence frequency; The performing cloud persistence analysis on the study area and generating pseudo cloud layer data of the candidate pixels based on the analysis result includes: Extracting a second time series of cloud coverage of the study area from historical remote sensing data of the study area; Obtaining the cloud persistence frequency of the study area within a preset time period according to the second time series statistics; Calculating the cloud coverage probability of the study area within a target time period according to the cloud persistence frequency; Based on the cloud coverage probability, pseudo cloud layer data of each candidate pixel is repeatedly generated according to a preset number of times.

4. The cloud removal bias estimation method based on random sampling according to claim 3 is characterized in that: The calculating the cloud coverage probability of the study area within the target time period according to the cloud persistence frequency includes: The cloud persistence frequency is normalized and calculated to obtain the cloud coverage probability of the study area in the target time period: Among them, p n is the cloud coverage probability, which indicates the probability that the study area is covered by clouds for n consecutive days within the target time period, F n is the cloud persistence frequency, which indicates the frequency of cloud coverage for n consecutive days in the preset time period in the study area, and N indicates the maximum number of days with continuous cloud coverage in the target time period.

5. The cloud removal bias estimation method based on random sampling according to claim 3 is characterized in that: The repeatedly generating the pseudo cloud layer data of each candidate pixel according to a preset number of times based on the cloud coverage probability includes: Determining whether each of the candidate pixels is in a cloud coverage state according to the cloud coverage probability of each of the candidate pixels; The process is repeated for a preset number of times, and the cloud layer attributes corresponding to the target pixels in the candidate pixels that are in a cloud-covered state are adjusted to generate pseudo cloud layer data of the target pixels.

6. The cloud removal bias estimation method based on random sampling according to claim 1, characterized in that: The step of performing cloud removal processing on the target remote sensing image to obtain multiple sets of pixel filling values ​​includes: Performing binary mask processing on the target remote sensing image to obtain a mask image; The cloud-covered area in the mask image is removed using the trained cloud removal model to obtain a cloud-removed image, and a plurality of groups of pixel filling values ​​corresponding to the cloud-covered area in the cloud-removed image are extracted.

7. The cloud removal bias estimation method based on random sampling according to claim 1, characterized in that: The calculating of the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image comprises: The number of candidate pixels selected from each evaluation group is fixed, and the number of pseudo cloud layer generation times of each candidate pixel is adjusted; Calculating a first error between each of the pixel filling values ​​corresponding to each generation of a pseudo cloud layer and a corresponding pixel true value in the initial remote sensing image, and obtaining a first error combination according to each of the first errors; Fixing the number of pseudo cloud layer generation times of each candidate pixel in each evaluation group, and adjusting the number of selected candidate pixels; Repeatedly calculating the second errors between the pixel filling values ​​corresponding to the candidate pixels and the corresponding pixel true values ​​in the initial remote sensing image, and obtaining a second error combination according to the second errors; If the first error combination and the second error combination meet a preset error evaluation condition, the number of target pseudo cloud layer generation times and the number of target candidate pixels are obtained; Based on the target pseudo cloud layer generation times and the target candidate pixel quantity, a target error between each pixel filling value and a corresponding pixel true value in the initial remote sensing image is calculated.

8. A cloud removal bias estimation device based on random sampling, characterized in that: include: Remote sensing image acquisition module, used to obtain the initial remote sensing image of the study area; A candidate pixel selection module is used to divide the initial remote sensing image into multiple evaluation groups according to the average number of snow days per year, and randomly select candidate pixels from each evaluation group; A pseudo cloud layer simulation module, used for performing cloud persistence analysis on the study area and generating pseudo cloud layer data of the candidate pixels based on the analysis results; A target image determination module, used for superimposing the pseudo cloud layer data with the initial remote sensing image to obtain a target remote sensing image; An image declouding processing module is used to perform declouding processing on the target remote sensing image to obtain multiple groups of pixel filling values; The target error evaluation module is used to calculate the target error between each pixel filling value and the corresponding pixel true value in the initial remote sensing image.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the cloud removal bias estimation method based on random sampling according to any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: include: processor; as well as A memory having computer-readable instructions stored thereon, wherein the computer-readable instructions, when executed by the processor, implement the cloud removal bias estimation method based on random sampling as described in any one of claims 1 to 7.

Images