Surface water body classification map gap filling method and system based on cloud platform

Abstract

The invention provides a surface water body classification map gap filling method based on a cloud platform. The method comprises the following steps: firstly, acquiring medium-and-low-resolution optical satellite remote sensing image data, preprocessing the medium-resolution data, and then calculating water body index and vegetation index image data; constructing time or a filter, obtaining low-resolution image data, and calculating water body and vegetation indexes of the low-resolution image data. Establishing a water body and vegetation index relationship between the medium and low resolution image data by using a linear regression model; filling gap pixels generated by preprocessing in the medium-resolution image by adopting a linear regression model; and classifying the images by using the recovered image data to obtain water body classification image data. According to the method, the distribution range of the surface water body can be comprehensively obtained, the change condition of the water body can be truly reflected, the method is suitable for surface water resource monitoring and management in a large-scale area, the integrity and accuracy of the water body classification image are improved, and powerful support is provided for effective management of surface water resources.

Description

Technical Field

[0001] The present invention belongs to the technical field of water resources research, and relates to a method and system for filling gaps in a surface water classification map based on a cloud platform. Background Art

[0002] With the rapid development of remote sensing technology and the increasing number of remote sensing optical satellites, the ways to acquire remote sensing optical imagery are becoming increasingly diverse, significantly improving the openness and accessibility of data. This technology further provides a solid foundation for the detailed classification and identification of surface water using remote sensing optical imagery, making it possible to dynamically monitor surface water resources worldwide. As one of Earth's most precious natural resources, the distribution, changes, and health of surface water are directly related to the sustainable development of human society, ecological balance, and disaster early warning. Therefore, accurate and efficient monitoring of surface water resources is of great significance for scientific decision-making, resource management, and environmental protection.

[0003] Despite continuous technological advancements, numerous surface water classification methods still face challenges and limitations. One notable issue is that many traditional methods often ignore or exclude clouds, cloud shadows, voids, and other non-water pixels in optical imagery, without performing detailed processing or correction before classification. While this approach simplifies the complexity of the classification model and reduces the computational burden, it also directly leads to pixel gaps in the water images in the classification results. These ignored areas may correspond to important water bodies in real geographic space, such as small lakes, river branches, or wetlands, which are crucial for a complete understanding of surface water distribution and the assessment of water resource quantity and quality. Therefore, this ignoring strategy not only limits the accuracy and completeness of the classification results, but also makes the resulting water distribution maps inaccurately reflect the actual situation, thereby compromising the scientific and effective monitoring and management of surface water resources.

[0004] In response to the above problems, the rise of remote sensing cloud computing platforms has brought new solutions to the classification of surface water bodies. Currently, there are multiple related cloud platforms both domestically and internationally with corresponding processing capabilities. For example, Google Earth Engine (GEE), as a leader in this field, has become the preferred tool for researchers and the general public worldwide for online analysis and processing of remote sensing images due to its powerful data processing capabilities, rich remote sensing data sources, and user-friendly operation interface. The GEE platform integrates multi-source, high-resolution remote sensing image data, including Landsat series, Sentinel series, MODIS series, etc. These data cover multiple bands from visible light to infrared, providing a rich information foundation for the accurate identification of surface water bodies. Users can quickly access, process, and analyze massive amounts of remote sensing data through a simple programming interface, greatly improving research efficiency.

[0005] Using these platforms for surface water classification and mapping can not only effectively overcome the pixel gap problem in traditional methods, but also enable large-scale, large-area dynamic monitoring of water resources. By combining advanced image processing algorithms, such as deep learning and machine learning, interference factors such as clouds and cloud shadows in remote sensing images can be accurately identified and removed. At the same time, multi-temporal and multi-source data fusion technology can be used to improve the accuracy and robustness of water body classification. In addition, these cloud platforms also support spatiotemporal analysis of classification results, helping researchers to reveal the changing patterns and trends of surface water bodies and their relationship with human activities, providing a scientific basis for water resource management, flood warning, and ecological protection.

[0006] In short, with the continuous advancement of remote sensing technology and the widespread application of remote sensing cloud computing platforms, the classification and identification of surface water bodies is gradually moving towards a new stage of high precision, high efficiency, and large-scale coverage. This not only provides strong technical support for water resource management and protection, but also contributes significantly to promoting the sustainable use of global water resources and addressing the challenges of climate change. In the future, with the launch of more high-resolution satellites and the continued accumulation of remote sensing data, combined with advanced technologies such as artificial intelligence and big data, the monitoring and management of surface water resources will usher in even broader development prospects. Summary of the Invention

[0007] In view of this, the purpose of the present invention is to provide a surface water classification map gap filling method and system based on a cloud platform, so as to achieve complete and comprehensive acquisition of the distribution range of surface water bodies, thereby truly reflecting the changes and laws of surface water bodies.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] A method for filling gaps in a surface water classification map based on a cloud platform, the method comprising the following steps:

[0010] S1. Acquire medium and low resolution optical satellite remote sensing image data;

[0011] S2. performing a declouding operation on the medium-resolution optical satellite remote sensing image data and then calculating a water index band and a vegetation index band to obtain water index and vegetation index image data;

[0012] S3. Constructing a time or filter, obtaining low-resolution optical satellite remote sensing image data corresponding to the medium-resolution optical satellite remote sensing image data based on the time attribute, calculating the water index and vegetation index of the low-resolution optical satellite remote sensing image data, and constructing a relationship between the water index and vegetation index image data of the medium- and low-resolution optical satellite remote sensing image data using a linear regression model based on the information of the water index and vegetation index image data of the medium- and low-resolution optical satellite remote sensing image data;

[0013] S4. Based on the relationship between water index and vegetation index image data of different resolutions, a linear regression model is used to fill in the gap pixels generated by the cloud removal operation in the water index and vegetation index image data of the reconstructed medium-resolution optical satellite remote sensing image data;

[0014] S5. Based on the restored and filled water body index and vegetation index image data, the image is classified based on a certain rule formula to obtain water body classification image data.

[0015] Furthermore, in step S1, the optical satellite remote sensing image data of different resolutions are medium-resolution optical images and low-resolution optical images. Step S1 specifically includes the following steps:

[0016] S11. The medium-resolution optical images selected are the Landsat 7, 8, and 9 surface reflectance products from the Google Earth Engine cloud platform, with a spatial resolution of 30 meters;

[0017] S12. The low-resolution optical image selected is the MOD09A1 dataset derived from MODIS data on the Google Earth Engine cloud platform. The number of image bands in this dataset is 13, of which the spatial resolution of 6 bands (sur_refl_b01 to sur_refl_b07) is 500 meters.

[0018] Furthermore, in step S2, the following steps are specifically included:

[0019] S21. Using the QA_PIXEL band (also known as the quality control band) in the Landsat 7, 8, and 9 surface reflectance images, remove pixels contaminated by clouds and cloud shadows in the images to obtain image data without the contaminated pixels.

[0020] S22, the water index band refers to the modified normalized difference water index (mNDWI); the vegetation index band specifically includes the normalized vegetation index (NDVI) and the enhanced vegetation index (EVI);

[0021] S23. Calculate the mNDWI, NDVI, and EVI index bands of Landsat 7, 8, and 9 images respectively.

[0022] Furthermore, in step S3, the following steps are specifically included:

[0023] S31. Select a duration of 8 days, define a time window, and build a connection operation.

[0024] S32. Based on the 8-day time window, construct a maximum time difference filter;

[0025] S33. Based on the time maximum difference filter, a time right filter is constructed to obtain the "system:time_start" time attribute of the Landsat 7, 8, and 9 images. Using this attribute as a benchmark, MOD09A1 image data with a time attribute greater than this attribute and less than the maximum difference filter is obtained.

[0026] S34. Based on the time maximum difference filter, a time left filter is constructed to obtain the "system:time_start" time attribute of the Landsat 7, 8, and 9 images. Using this attribute as a benchmark, MOD09A1 image data with a time attribute less than this attribute and greater than the maximum difference filter is obtained.

[0027] S35. Based on the constructed time right and left filters, construct a time or filter that satisfies one of the two filters, and obtain the MOD09A1 image dataset after the time left and right filter operations; calculate the average mNDWI, average NDVI, and average EVI index of the MOD09A1 image dataset to obtain image data of the three index bands;

[0028] S36, construct a linear regression model. The mNDWI, NDVI, and EVI index image data of the MOD09A1 data obtained after the above steps are used as independent variables of the linear regression model. The constant independent variable is set to 1. The mNDWI, NDVI, and EVI index band data of the Landsat7, 8, and 9 images obtained in step S23 are used as dependent variables of the linear regression model. The linear regression model formula is as follows:

[0029] Landsat i =β0×Constant+β1×MOD09A1 i

[0030] Among them, i is the NDVI, mNDWI, and EVI index band; β0 is the intercept, Constant is the constant term, and β1 is the slope.

[0031] Furthermore, in step S4, the following steps are specifically included:

[0032] S41. Obtain the intercept and slope based on the linear regression model between the constructed NDVI index band image of the Landsat 7, 8, and 9 images and the NDVI of the MOD09A1 image data; fill the pixel gaps in the NDVI band data of the Landsat 7, 8, and 9 images that are affected by clouds, cloud pollution, or gaps using the intercept, slope, and NDVI index image data of the MOD09A1 data; and obtain the complete NDVI band data of the Landsat 7, 8, and 9 images.

[0033] S42. Obtain an intercept and a slope based on a linear regression model between the constructed mNDWI index band images of Landsat 7, 8, and 9 images and the mNDWI band of MOD09A1 image data; fill in the pixel gaps caused by clouds, cloud pollution, and gaps in the mNDWI index band images of Landsat 7, 8, and 9 images using the intercept, slope, and the mNDWI index band image of MOD09A1 data; and obtain a complete mNDWI index band image of Landsat 7, 8, and 9 images.

[0034] S43. Based on the constructed linear regression model between the EVI index band image of Landsat 7, 8, and 9 images and the EVI band of MOD09A1 image data, the intercept and slope are obtained; the intercept, slope, and the EVI index band image of MOD09A1 data are used to fill the pixel gaps caused by clouds, cloud pollution, and gaps in the mNDWI index band images of Landsat 7, 8, and 9 images; and the complete mNDWI index band images of Landsat 7, 8, and 9 images are obtained.

[0035] Furthermore, in step S5, it specifically includes: constructing a rule formula for determining whether a pixel in the image is a water body or a non-water body based on the reconstructed water body index and vegetation index image data. The specific formula is as follows:

[0036] Water=(mNDWI>EVI or mNDWI>NDVI)and EVI<0.1

[0037] Among them, Water means that when the mNDWI, NDVI and EVI of the image pixel meet the following rules, it is classified as a water body, otherwise it is classified as a non-water body.

[0038] The present invention also provides a surface water classification map gap filling system based on a cloud platform.

[0039] The beneficial effects of the present invention are:

[0040] The technical solution provided by the present invention utilizes the Google Earth Engine cloud platform to obtain medium-resolution Landsat7, 8, and 9 surface reflectance image data. After de-clouding, the mNDWI, NDVI, and EVI index bands are calculated. A time filter is constructed to obtain low-resolution MOD09A1 image data corresponding to the Landsat 7, 8, and 9 surface reflectance image data within a time window. A linear regression model corresponding to the mNDWI, NDVI, and EVI index bands calculated from the MOD09A1 image data and the mNDWI, NDVI, and EVI index bands calculated from the medium-resolution Landsat 7, 8, and 9 surface reflectance image data is constructed to fill in the gap pixels in the mNDWI, NDVI, and EVI index bands of Landsat 7, 8, and 9. A rule formula is used for Landsat 7, 8, and 9 surface reflectance images for water body classification, which realizes the filling of gap pixels in surface water body classification and obtains complete surface water body classification; the surface water body classification product generated by the present invention can fully and comprehensively obtain the distribution range of surface water bodies, can truly reflect the changes and laws of surface water bodies, and can be used for monitoring and management of surface water resources in large-scale areas.

[0041] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0043] Figure 1 This is a flow chart of a method for filling gaps in a surface water classification map based on the Google Earth Engine cloud platform of the present invention;

[0044] Figure 2 Comparison chart of NDWI index, NDVI index and EVI index band images of the Landsat 7 surface reflectance image before and after filling generated in the example item of the present invention;

[0045] Figure 3 Comparison chart of the mNDWI index, NDVI index, and EVI index band images of the Landsat 8 surface reflectance image before and after filling generated in the example item of the present invention;

[0046] Figure 4 Comparison chart of the mNDWI index, NDVI index, and EVI index band images of the Landsat 9 surface reflectance image before and after filling generated in the example item of the present invention;

[0047] Figure 5 This is a comparison chart of water body classification of the Landsat 7 surface reflectance image before and after filling generated in the example of the present invention;

[0048] Figure 6 This is a comparison chart of water body classification of the Landsat 8 surface reflectance image before and after filling generated in the example of the present invention;

[0049] Figure 7 This is a comparison chart of water body classification in the Landsat 9 surface reflectance image before and after filling generated in the example of the present invention. DETAILED DESCRIPTION

[0050] The technical solution of the present invention is described in detail below with reference to the accompanying drawings.

[0051] In order to fully and comprehensively reflect the distribution range of surface water bodies, solve the problem of gap pixels in the current water body classification, and fill the gap pixels, the embodiment of the present invention uses the Google Earth Engine cloud platform to obtain medium-resolution Landsat7, 8, and 9 surface reflectance image data. After de-clouding, the mNDWI index, NDVI index, and EVI index bands are calculated, a time filter is constructed, and low-resolution MOD09A1 image data corresponding to the Landsat 7, 8, and 9 surface reflectance image data within the time window is obtained. A linear regression model corresponding to the mNDWI index, NDVI index, and EVI index bands calculated from the MOD09A1 image data and the mNDWI index, NDVI index, and EVI index bands calculated from the medium-resolution Landsat 7, 8, and 9 surface reflectance image data is constructed to realize Landsat classification. The gap pixels in the mNDWI index, NDVI index and EVI index bands of Landsat 7, 8 and 9 were filled, and the water body classification of the surface reflectance images of Landsat 7, 8 and 9 was performed using a regular formula, which realized the filling of the gap pixels in the surface water body classification and obtained a complete surface water body classification.

[0052] like Figure 1 As shown, the embodiment of the present invention provides a method for filling gaps in a surface water classification map based on a cloud platform, comprising the following steps:

[0053] S1. Obtain Landsat 7, 8, and 9 remote sensing image data and perform cloud removal processing;

[0054] In this example, the present invention is based on the Google Earth Engine cloud platform and selects Landsat 7, 8, and 9 surface reflectance image data within a set time range, which is after February 18, 2000.

[0055] The present invention uses the QA_PIXEL (quality control) band in the image to remove clouds and cloud shadows in the surface reflectance image data of Landsat 7, 8, and 9, and excludes these pixels from the subsequent vegetation index and water index calculations.

[0056] S2. Construct water index bands and vegetation index bands for distinguishing water bodies from non-water bodies in Landsat 7, 8, and 9 surface emissivity images;

[0057] In this embodiment, the three index bands constructed by the present invention specifically include the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI) and the Modified Normalized Difference Water Index (mNDWI).

[0058] The calculation formula of NDVI index is:

[0059]

[0060] The P of Landsat 7 surface emissivity image is nir For the fourth band, P red The third band; the P of Landsat 8 surface emissivity image nir For the fifth band, P red The fourth band; the P of Landsat 9 surface emissivity image nir For the fifth band, P red The fourth band.

[0061] The calculation formula of the mNDWI index is:

[0062]

[0063] The P of Landsat 7 surface emissivity image is green For the second band, P swir The fifth band; the P of Landsat 8 surface emissivity image green For the third band, P swir The sixth band; the P of Landsat 9 surface emissivity image green For the third band, P swir The sixth band.

[0064] The calculation formula of the EVI index is:

[0065]

[0066] The P of Landsat 7 surface emissivity image is nir For the fourth band, P red For the third band, P blue is the first band; P of Landsat 8 surface emissivity image nir For the fifth band, P red For the fourth band, P blueis the second band; P of Landsat 9 surface emissivity image nir For the fifth band, P red For the fourth band, P blue For the second band.

[0067] S3. Construct a time or filter, and use the time filter to filter based on the "system:time_start" time attribute of Landsat 7, 8, and 9 to obtain MOD09A1 image data within the time range corresponding to the Landsat 7, 8, and 9 images. Based on the obtained MOD09A1 image data, calculate the average NDVI, average EVI, and average mNDWI index bands of MOD09A1, and construct linear regression models of the average NDVI, average EVI, and average mNDWI index bands of MOD09A1 and the NDVI, EVI, and NDVI index bands of the Landsat 7, 8, and 9 surface reflectance images obtained in step S2;

[0068] In this embodiment, the present invention first constructs a time filter to implement the screening of MOD09A1 image data.

[0069] Select an 8-day time window to define the time window, and construct a join operation and a maximum time difference filter.

[0070] Based on the time maximum difference filter, a time right filter is constructed to obtain the "system:time_start" time attribute of the Landsat 7, 8, and 9 surface emissivity images. Using this attribute as a benchmark, MOD09A1 image data with a time attribute greater than this attribute and less than the maximum difference filter is obtained.

[0071] Based on the time maximum difference filter, a time left filter is constructed to obtain the "system:time_start" time attribute of the Landsat 7, 8, and 9 surface emissivity images. Using this attribute as a benchmark, MOD09A1 image data with a time attribute less than this attribute and greater than the maximum difference filter is obtained.

[0072] Construct a time or filter that satisfies one of the two filters, obtain the MOD09A1 image dataset after the time left and right filter operations, and calculate the average mNDWI, average NDVI, and average EVI index band image data of the MOD09A1 image dataset;

[0073] A linear regression model was constructed between the average NDVI index band of MOD09A1 and the NDVI index band of the surface reflectance images of Landsat 7, 8, and 9. The NDVI index band image data of MOD09A1 data was used as the independent variable of the linear regression model. The constant independent variable was set to 1. The NDVI index band image data of the surface reflectance images of Landsat 7, 8, and 9 obtained in step S2 was used as the dependent variable of the linear regression model. The linear regression model formula is as follows:

[0074] Landsat NDVI =β0×Constant+β1×MOD09A1 NDVI

[0075] Among them, Landsat NDVI is the NDVI index band of the Landsat 7, 8, and 9 surface emissivity images; β0 is the intercept, Constant is the constant term; β1 is the slope, MOD09A1 NDVI The average NDVI calculated for MOD09A1 images after time or filter selection;

[0076] A linear regression model was constructed between the average EVI index band of MOD09A1 and the EVI index band of the surface reflectance images of Landsat 7, 8, and 9. The EVI index band image data of MOD09A1 data was used as the independent variable of the linear regression model. The constant independent variable was set to 1. The EVI index band image data of the surface reflectance images of Landsat 7, 8, and 9 obtained in step S2 was used as the dependent variable of the linear regression model. The linear regression model formula is as follows:

[0077] Landsat EVI =β0×Constant+β1×MOD09A1 EVI

[0078] Among them, Landsat EVI is the EVI index band of the Landsat 7, 8, and 9 surface emissivity images; β0 is the intercept, Constant is the constant term; β1 is the slope, MOD09A1 NDVI The average EVI calculated for MOD09A1 images filtered by time or filter;

[0079] A linear regression model was constructed between the average mNDWI index band of MOD09A1 and the mNDWI index band of the surface reflectance images of Landsat 7, 8, and 9. The mNDWI index band image data of MOD09A1 data was used as the independent variable of the linear regression model. The constant independent variable was set to 1. The mNDWI index band image data of the surface reflectance images of Landsat 7, 8, and 9 obtained in step S2 was used as the dependent variable of the linear regression model. The linear regression model formula is as follows:

[0080] Landsat mNDWI =β0×Constant+β1×MOD09A1 mNDWI

[0081] Among them, Landsat mNDWI is the mNDWI index band of the Landsat 7, 8, and 9 surface emissivity images; β0 is the intercept, Constant is the constant term; β1 is the slope, MOD09A1 mNDWI Mean mNDWI calculated for MOD09A1 images filtered by time or filter;

[0082] S4, using the linear regression model constructed in step S3 to fill in the gap pixels generated by the cloud removal operation in the Landsat 7, 8, and 9 water index mNDWI and vegetation index NDVI and EVI image data obtained in step S2;

[0083] In this example, the present invention obtains the intercept and slope based on the linear regression model between the NDVI index band image of the Landsat 7, 8, and 9 images constructed in step S3 and the average NDVI index band image of the MOD09A1 image data; the intercept, slope, and the average NDVI index band image data of the MOD09A1 data are used to fill the pixel gaps caused by clouds, cloud pollution, gaps, etc. in the NDVI index band image of the Landsat 7, 8, and 9 surface reflectance image obtained in step S2; the complete NDVI index band image of the Landsat 7, 8, and 9 surface reflectance image is obtained, and the comparison diagram before and after filling is as shown. Figure 2 As shown;

[0084] Based on the linear regression model between the mNDWI index band image of the Landsat 7, 8, and 9 images constructed in step S3 and the average mNDWI index band image of the MOD09A1 image data, the intercept and slope are obtained; the pixel gaps caused by clouds, cloud pollution, gaps, etc. in the mNDWI index band image of the Landsat 7, 8, and 9 surface reflectance images obtained in step S2 are filled using the intercept, slope, and the average mNDWI index band image data of the MOD09A1 data; the complete mNDWI index band image of the Landsat 7, 8, and 9 surface reflectance images is obtained, and the comparison before and after filling is shown in the figure below. Figure 3 As shown;

[0085] Based on the linear regression model between the EVI index band image of the Landsat 7, 8, and 9 images constructed in step S3 and the average EVI index band image of the MOD09A1 image data, the intercept and slope are obtained; the pixel gaps caused by clouds, cloud pollution, gaps, etc. in the EVI index band image of the Landsat 7, 8, and 9 surface reflectance images obtained in step S2 are filled using the intercept, slope, and the average mNDWI index band image data of the MOD09A1 data; the complete EVI index band image of the Landsat 7, 8, and 9 surface reflectance images is obtained, and the comparison before and after filling is shown in the figure below. Figure 4 As shown;

[0086] S5. Obtain complete NDVI, mNDWI, and EVI index band images of the Landsat 7, 8, and 9 surface reflectance images using step S4, and classify them based on a certain rule formula to obtain a water body classification map of the Landsat 7, 8, and 9 surface reflectance images;

[0087] In this embodiment, the rule formula is as follows:

[0088] Water=(mNDWI>EVI or mNDWI>NDVI)and EVI<0.1

[0089] Among them, when the pixel of the Landsat 7 surface reflectance image satisfies the conditions that the mNDWI value is greater than EVI and the EVI is less than 0.1 or the mNDWI value is greater than NDVI and the EVI is less than 0.1, the pixel is a water body and is assigned a value of 1; if it does not meet the conditions, the pixel is a non-water body and is assigned a value of 0, thus finally realizing the water body classification of the image. The comparison of the water body classification of the Landsat 7 surface reflectance image before and after filling is as follows: Figure 5 As shown;

[0090] When the pixel of the Landsat 8 surface reflectance image satisfies the conditions that the mNDWI value is greater than EVI and the EVI is less than 0.1, or the mNDWI value is greater than NDVI and the EVI is less than 0.1, the pixel is a water body and is assigned a value of 1; if it does not meet the conditions, the pixel is a non-water body and is assigned a value of 0, thus achieving the water body classification of the image. The comparison of the water body classification of the Landsat 8 surface reflectance image before and after filling is as follows: Figure 6 As shown;

[0091] When the pixel of the Landsat 9 surface reflectance image satisfies the conditions that the mNDWI value is greater than EVI and the EVI is less than 0.1, or the mNDWI value is greater than NDVI and the EVI is less than 0.1, the pixel is a water body and is assigned a value of 1; if it does not meet the conditions, the pixel is a non-water body and is assigned a value of 0, thus achieving the water body classification of the image. The comparison of the water body classification of the Landsat 9 surface reflectance image before and after filling is as follows: Figure 7 As shown;

[0092] In summary, the present invention combines the Landsat 7, 8, and 9 surface emissivity images on the Google Earth Engine cloud platform with a linear regression model to intelligently fill the gap pixels of surface water bodies. This can comprehensively and accurately reflect the distribution range of surface water bodies, and can be used to carry out large-scale and large-area water body classification, providing strong technical support for the management and monitoring of water resources.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for filling gaps in surface water classification maps based on a cloud platform, characterized by: The method comprises the following steps: S1. Acquire medium and low resolution optical satellite remote sensing image data; S2. Preprocessing the medium-resolution optical satellite remote sensing image data and then calculating the water index band and the vegetation index band to obtain water index and vegetation index image data; S3. Constructing a time or filter, obtaining low-resolution optical satellite remote sensing image data corresponding to the medium-resolution optical satellite remote sensing image data based on the time attribute, calculating the water index and vegetation index of the low-resolution optical satellite remote sensing image data, and constructing a relationship between the water index and vegetation index image data of the medium- and low-resolution optical satellite remote sensing image data using a linear regression model based on the information of the water index and vegetation index image data of the medium- and low-resolution optical satellite remote sensing image data; S4. Based on the relationship between water index and vegetation index image data of different resolutions, a linear regression model is used to fill in the gap pixels generated by the preprocessing operation in the reconstructed water index and vegetation index image data of the medium-resolution optical satellite remote sensing image data; S5. Based on the restored and filled water body index and vegetation index image data, the image is classified based on a certain rule formula to obtain water body classification image data.

2. The method for filling gaps in a surface water classification map based on a cloud platform according to claim 1, characterized in that: In step S1, the optical satellite remote sensing image data of different resolutions are medium-resolution optical images and low-resolution optical images. Step S1 specifically includes the following steps: S11. The medium-resolution optical images selected are the Landsat 7, 8, and 9 surface reflectance products from the Google Earth Engine cloud platform, with a spatial resolution of 30 meters; S12. The low-resolution optical image selected is the MOD09A1 dataset derived from MODIS data on the Google Earth Engine cloud platform. The number of image bands in this dataset is 13, of which the spatial resolution of 6 bands (sur_refl_b01 to sur_refl_b07) is 500 meters.

3. The method for filling gaps in a surface water classification map based on a cloud platform according to claim 2, characterized in that: In step S2, the following steps are specifically included: S21. Using the QA_PIXEL band in the Landsat 7, 8, and 9 surface reflectance images, remove pixels contaminated by clouds and cloud shadows to obtain image data without the contaminated pixels. S22, the water index band refers to the modified normalized difference water index (mNDWI); the vegetation index band specifically includes the normalized vegetation index (NDVI) and the enhanced vegetation index (EVI); S23. Calculate the mNDWI, NDVI, and EVI index bands of Landsat 7, 8, and 9 images respectively.

4. The method for filling gaps in a surface water classification map based on a cloud platform according to claim 3 is characterized by: In step S3, the following steps are specifically included: S31. Select a duration of 8 days, define a time window, and build a connection operation. S32. Based on the 8-day time window, construct a maximum time difference filter; S33. Based on the time maximum difference filter, a time right filter is constructed to obtain the "system:time_start" time attribute of the Landsat 7, 8, and 9 images. Using this attribute as a benchmark, MOD09A1 image data with a time attribute greater than this attribute and less than the maximum difference filter is obtained. S34. Based on the time maximum difference filter, a time left filter is constructed to obtain the "system:time_start" time attribute of the Landsat 7, 8, and 9 images. Using this attribute as a benchmark, MOD09A1 image data with a time attribute less than this attribute and greater than the maximum difference filter is obtained. S35. Based on the constructed time right and left filters, construct a time or filter that satisfies one of the two filters, and obtain the MOD09A1 image dataset after the time left and right filter operations; calculate the average mNDWI, average NDVI, and average EVI index of the MOD09A1 image dataset to obtain image data of the three index bands; S36, construct a linear regression model. The mNDWI, NDVI, and EVI index image data of the MOD09A1 data obtained after the above steps are used as independent variables of the linear regression model. The constant independent variable is set to 1. The mNDWI, NDVI, and EVI index band data of the Landsat7, 8, and 9 images obtained in step S23 are used as dependent variables of the linear regression model. The linear regression model formula is as follows: Landsat i =β0×Constant+β1×MOD09A1 i Among them, i is the NDVI, mNDWI, and EVI index band; β0 is the intercept, Constant is the constant term, and β1 is the slope.

5. The method for filling gaps in a surface water classification map based on a cloud platform according to claim 4, characterized in that: In step S4, the following steps are specifically included: S41. Obtain the intercept and slope based on the linear regression model between the constructed NDVI index band image of the Landsat 7, 8, and 9 images and the NDVI of the MOD09A1 image data; fill the pixel gaps in the NDVI band data of the Landsat 7, 8, and 9 images that are affected by clouds, cloud pollution, or gaps using the intercept, slope, and NDVI index image data of the MOD09A1 data; and obtain the complete NDVI band data of the Landsat 7, 8, and 9 images. S42. Obtain an intercept and a slope based on a linear regression model between the constructed mNDWI index band images of Landsat 7, 8, and 9 images and the mNDWI band of MOD09A1 image data; fill in the pixel gaps caused by clouds, cloud pollution, and gaps in the mNDWI index band images of Landsat 7, 8, and 9 images using the intercept, slope, and the mNDWI index band image of MOD09A1 data; and obtain a complete mNDWI index band image of Landsat 7, 8, and 9 images. S43. Based on the constructed linear regression model between the EVI index band image of Landsat 7, 8, and 9 images and the EVI band of MOD09A1 image data, the intercept and slope are obtained; the intercept, slope, and the EVI index band image of MOD09A1 data are used to fill the pixel gaps caused by clouds, cloud pollution, and gaps in the mNDWI index band images of Landsat 7, 8, and 9 images; and the complete mNDWI index band images of Landsat 7, 8, and 9 images are obtained.

6. The method for filling gaps in a surface water classification map based on a cloud platform according to claim 5, characterized in that: In step S5, the method specifically includes: constructing a rule formula for determining whether a pixel in the image is a water body or a non-water body based on the reconstructed water body index and vegetation index image data. The specific formula is as follows: Water=(mNDWI>EVI or mNDWI>NDVI)and EVI<0.1 Among them, Water means that when the mNDWI, NDVI and EVI of the image pixel meet the following rules, it is classified as a water body, otherwise it is classified as a non-water body.

7. A surface water classification map gap filling system based on a cloud platform, characterized by: The system adopts the method according to any one of claims 1 to 6.

Images