A river and lake shore zone remote sensing monitoring method and product

Abstract

The embodiment of the application provides a kind of river and lake shore zone remote sensing monitoring method and product, the river and lake shore zone remote sensing monitoring method includes: obtaining historical medium resolution remote sensing image data and historical high resolution remote sensing image data;From the historical medium resolution remote sensing image data, the end-member spectrum of typical ground object is extracted;The end-member spectrum is corrected according to the high resolution remote sensing image data and the interference of shadow or cloud layer is eliminated, and the target end-member spectrum matrix is obtained;According to the target end-member spectrum matrix, pixel unmixing is carried out on the medium resolution remote sensing image data to be decomposed, and the pixel unmixing result is obtained;According to the pixel unmixing result and water flooding frequency, water level amplitude area is determined;The shore zone monitoring range is determined according to the water level amplitude area;The image data corresponding to the shore zone monitoring range is input into wavelet transform U-Net deep convolutional neural network model, and the shore zone land use classification result is obtained;According to the shore zone land use classification result, the ecological shore zone ratio result of each river section is quantified.

Description

Technical Field

[0001] This application relates to the field of land type identification, and specifically, the embodiments of this application relate to a remote sensing monitoring method and product for river and lake shorelines. Background Technology

[0002] River and lake shorelines, as important transitional areas where water and land meet, have multiple ecological functions such as maintaining biodiversity, water conservation, and intercepting non-point source pollution, playing a key role in regional ecological security.

[0003] With the deepening of aquatic ecological environment governance, ecological spatial identification and dynamic monitoring of river and lake riparian zones have become a key area for the application of remote sensing technology. In particular, in practices such as watershed ecological restoration, refined land management, and non-point source pollution control, there is an urgent need to obtain high-precision and timely information on the distribution of riparian land use types to support scientific decision-making.

[0004] Existing land use classification methods still face challenges such as insufficient accuracy, high misjudgment rate, and weak generalization ability when dealing with the problem of identifying non-ecological land along river and lake shorelines. Summary of the Invention

[0005] The purpose of this application is to provide a remote sensing monitoring method and product for river and lake shorelines. The embodiments of this application provide a remote sensing quantification method for the ecological shoreline ratio of shorelines based on hybrid pixel decomposition and deep learning. The improved hybrid pixel decomposition algorithm accurately extracts the boundary of the water level fluctuation zone, and the deep convolutional neural network (such as the wavelet transform U-Net deep convolutional neural network model) automatically identifies ecological land, solving the problems of blurred boundaries and low classification efficiency in traditional methods.

[0006] In a first aspect, embodiments of this application provide a remote sensing monitoring method for river and lake riparian zones. The method includes: acquiring historical medium-resolution remote sensing image data and historical high-resolution remote sensing image data; extracting endmember spectra of typical land features from the historical medium-resolution remote sensing image data, wherein the typical land features include: water bodies, vegetation, and bare land; correcting the endmember spectra based on the high-resolution remote sensing image data and removing shadows or cloud interference to obtain a target endmember spectral matrix, wherein the target endmember spectral matrix is ​​used to provide a spectral decomposition reference; performing pixel unmixing on the medium-resolution remote sensing image data to be decomposed based on the target endmember spectral matrix to obtain pixel unmixing results, wherein the medium-resolution remote sensing image data to be decomposed is used to provide the spectra of the pixels to be decomposed; and determining the water level fluctuation based on the pixel unmixing results and the flooding frequency. The process involves: determining the monitoring range of the riparian zone based on the water level fluctuation area; acquiring image data corresponding to the monitoring range of the riparian zone and obtaining data results for four high-resolution bands, including red, green, blue, and near-infrared bands; inputting the data from the four high-resolution bands into a wavelet transform U-Net deep convolutional neural network model to automatically identify the land use type of the riparian zone monitoring range and obtain the riparian zone land use classification results; wherein the wavelet transform U-Net deep convolutional neural network model uses discrete wavelet transform for downsampling processing, performs cross-layer cascading of the multi-scale features output by the discrete wavelet transform through 3×3 convolution, and combines channel attention and spatial attention for attention fusion; and quantifying the ecological riparian zone ratio of each river section based on the riparian zone land use classification results.

[0007] The inventors of this application discovered in their research that medium-resolution images (such as Sentinel-2) are affected by the mixed pixel effect, resulting in blurred water body boundaries. Meanwhile, high-resolution images have low coverage frequency, making it difficult to achieve long-term continuous monitoring. Therefore, the embodiments of this application use high-resolution images to assist endmember correction and employ an improved demixing algorithm to improve boundary continuity, thereby obtaining a more accurate monitoring range and improving the accuracy of land use type identification in the target area.

[0008] In some embodiments, the step of unmixing pixels of the medium-resolution remote sensing image data to be decomposed based on the target endmember spectral matrix to obtain pixel unmixing results includes: obtaining a spatially continuous water abundance distribution map based on the target endmember spectral matrix and the medium-resolution remote sensing image data to be decomposed; the step of determining the water level fluctuation region based on the pixel unmixing results and the flooding frequency includes: acquiring high-resolution remote sensing image data of the spatial range corresponding to the medium-resolution remote sensing image data to be decomposed, and obtaining a water body binary map based on the high-resolution remote sensing image data corresponding to the medium-resolution remote sensing image data to be decomposed; and obtaining the water level fluctuation region based on the water body binary map and the water abundance distribution map.

[0009] Some embodiments of this application define the water level variation range by using spectral unmixing and flooding frequency analysis, thereby improving the accuracy of the obtained water level variation range results.

[0010] In some embodiments, obtaining a spatially continuous water abundance distribution map based on the target endmember spectral matrix and the medium-resolution remote sensing image data to be decomposed includes: performing a linear spectral mixture model solution on the medium-resolution remote sensing image data to be decomposed pixel by pixel based on the target endmember spectral matrix to obtain the abundance vector of each pixel, wherein the abundance vector is used to quantitatively characterize the abundance distribution of endmembers of each typical land cover in each pixel; extracting the initial water body extent using the Normalized Differential Water Index (NDWI) thresholding method, and obtaining the water abundance components in each pixel based on the initial water body extent and the abundance vector to obtain the water endmember abundance value α. 水体 ； The water body endmember abundance value α 水体 Pixels with an NDWI greater than 0 and less than 0.1 are forcibly assigned a value of 0, while the remaining pixels retain the water endmember abundance value α. 水体 The spatially continuous water abundance distribution map is obtained.

[0011] The embodiments of this application are based on the pixel-level water abundance value α obtained by solving a linear mixture model. 水体 Furthermore, the calculated water abundance values ​​are corrected based on NDWI, with the correction principle being that values ​​satisfying α... 水体 Pixels with an NDWI value greater than 0 (indicating the presence of water components in the unmixed model) and an NDWI value less than 0.1 (indicating that the optical characteristics do not match the typical reflection characteristics of water) have their water abundance values ​​forcibly reset to 0. By correcting the NDWI, interference from "false water bodies" such as shadows / dark buildings can be eliminated, significantly improving the accuracy of water areas on the water abundance distribution map.

[0012] In some embodiments, the step of performing a linear spectral mixing model to solve the medium-resolution remote sensing image data to be decomposed pixel by pixel based on the target endmember spectral matrix to obtain the abundance vector of each pixel includes: based on the linear spectral mixing model, under the dual constraints of "abundance is non-negative" and "abundance sum is 1", applying the least squares method to perform spectral unmixing on each pixel to obtain the abundance vector of each pixel.

[0013] Some embodiments of this application obtain the abundance vector of each pixel through two principles, thereby improving the accuracy of the obtained abundance vector.

[0014] In some embodiments, obtaining the water level fluctuation region based on the water body binary map and the water body abundance distribution map includes: spatially resampling the abundance distribution map to obtain a spatially sampled water body abundance distribution map, ensuring that the resolution of the sampling result is consistent with the high-resolution remote sensing image data; interpolating the spatially sampled water body abundance distribution map using bilinear interpolation to obtain a target water body abundance distribution map; constructing a long-term water body raster dataset based on the water body binary map and multiple target water body abundance distribution maps, wherein the long-term water body raster dataset is a spatiotemporal cube obtained from multi-temporal target water body abundance distribution maps, and the pixel values ​​in the long-term water body raster dataset are used to store the sub-pixel level water body coverage ratio and to quantify the dynamic characteristics of the water level fluctuation region; calculating the flooding frequency of each pixel based on the long-term water body raster dataset to obtain a flooding frequency distribution map; and obtaining the water level fluctuation region based on the flooding frequency distribution map.

[0015] The embodiments of this application spatially resample the water abundance distribution map (which is the result of water body identification from medium-resolution remote sensing image data) to make its resolution consistent with that of high-resolution remote sensing images (such as GF-1 / GF-6). Furthermore, the embodiments of this application employ bilinear interpolation to maintain the continuity of abundance values ​​after sampling (avoiding pixel value jumps), thus ensuring accurate spatial matching during subsequent overlay analysis. The embodiments of this application determine sub-pixel flooding (i.e., through α... 水体 By combining statistical frequency thresholds (i.e., flooding frequency), it accurately captures the gradual transition zone between land and water that is ignored by traditional methods, providing irreplaceable dynamic baseline data for coastal zone management.

[0016] In some embodiments, obtaining the water level fluctuation region based on the flooding frequency distribution map includes: extracting all pixels satisfying 0.25≤WIF≤0.75 from the flooding frequency distribution map to obtain the water level fluctuation region.

[0017] The embodiments of this application improve the accuracy of the obtained water level fluctuation range by setting a reasonable flooding frequency range.

[0018] Secondly, some embodiments of this application provide a remote sensing monitoring device for river and lake riparian zones. The device includes: a historical data acquisition module configured to acquire historical medium-resolution remote sensing image data and historical high-resolution remote sensing image data; an endmember spectrum acquisition module configured to extract endmember spectra of typical land features from the historical medium-resolution remote sensing image data, wherein the typical land features include: water bodies, vegetation, or bare land; a preprocessing module configured to correct the endmember spectra based on the high-resolution remote sensing image data and remove shadows or cloud interference to obtain a target endmember spectral matrix, wherein the target endmember spectral matrix is ​​used to provide a spectral decomposition reference; a pixel unmixing module configured to perform pixel unmixing on the medium-resolution remote sensing image data to be decomposed based on the target endmember spectral matrix to obtain a pixel unmixing result, wherein the medium-resolution remote sensing image data to be decomposed is used to provide the spectra of the pixels to be decomposed; and a water level fluctuation zone determination module configured to determine the water level fluctuation zone based on the pixel unmixing result and the flooding frequency. The system comprises the following modules: a water level fluctuation zone determination module; a riparian zone monitoring range determination module, configured to determine the riparian zone monitoring range based on the water level fluctuation zone; a land use type identification module, configured to acquire image data corresponding to the riparian zone monitoring range and obtain data for four high-resolution bands, including red, green, blue, and near-infrared bands; inputting the data of the four high-resolution bands into a wavelet transform U-Net deep convolutional neural network model to automatically identify the riparian zone land use classification results from the high spatial resolution satellite remote sensing images of the riparian zone monitoring range; wherein the wavelet transform U-Net deep convolutional neural network model uses discrete wavelet transform for downsampling processing, performs cross-layer concatenation of the multi-scale features output by the discrete wavelet transform through 3×3 convolution, and combines channel attention and spatial attention for attention fusion; and an ecological riparian zone ratio quantification module, configured to quantify the ecological riparian zone ratio results for each river segment based on the riparian zone land use classification results.

[0019] Thirdly, some embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the methods described in any of the embodiments described in the first aspect.

[0020] Fourthly, some embodiments of this application provide an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, can implement the method described in any of the embodiments described in the first aspect.

[0021] Fifthly, some embodiments of this application provide a computer program product including computer program instructions, which, when read and executed by a processor, can implement the method described in any of the embodiments included in the first aspect above. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is an architecture diagram of a remote sensing monitoring system for river and lake riparian zones provided in an embodiment of this application.

[0024] Figure 2 A flowchart of a remote sensing monitoring method for river and lake shorelines provided in this application embodiment;

[0025] Figure 3 A schematic diagram of the architecture of the wavelet transform U-Net deep convolutional neural network model provided in the embodiments of this application;

[0026] Figure 4 A schematic diagram of the architecture of the attention fusion module provided in an embodiment of this application;

[0027] Figure 5 A block diagram of the composition of a remote sensing monitoring device for river and lake shorelines is provided for embodiments of this application;

[0028] Figure 6 A schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0029] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0030] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0031] River and lake riparian zones, as crucial transitional areas where land and water meet, play an irreplaceable ecological barrier role in maintaining biodiversity, water conservation, and intercepting non-point source pollution. However, the inventors of this application have discovered two key technical challenges in current remote sensing-based identification of riparian ecological land use:

[0032] Insufficient accuracy in identifying the boundaries of water level fluctuation zones: The spatial extent of the riparian zone typically relies on the definition of water level fluctuation zones, which is extrapolated from the inundated area between the multi-year average high and low water levels to the land buffer zone. Although long-term satellite remote sensing imagery can be used to invert the maximum inundation range and preliminarily delineate the dynamic boundaries of water bodies, it is limited by the low acquisition frequency and limited coverage of high-resolution satellite imagery, making long-term continuous monitoring difficult. In contrast, medium-resolution satellite imagery is acquired more frequently and has stronger coverage, but its spatial resolution is limited and it is susceptible to interference from mixed pixel effects, making it difficult to accurately locate water body boundaries at the pixel scale, thus restricting the accurate determination of the riparian zone extent.

[0033] Limited ability to classify and identify ecological land use: Existing land use classification methods based on remote sensing still rely primarily on manual visual interpretation when identifying ecological land (such as forest land and grassland) and non-ecological land (such as construction land, transportation land, and agricultural land) within the riparian zone. This method is inefficient, cannot meet the needs of large-scale rapid monitoring, and is also affected by the subjective judgment of the interpreters, resulting in poor consistency and low repeatability.

[0034] To address the background and aforementioned technical problems, embodiments of this application provide a remote sensing monitoring method for river and lake riparian zones. This method belongs to the remote sensing quantification method for the ecological riparian zone ratio based on hybrid pixel decomposition and deep learning. This method accurately extracts the boundaries of water level fluctuation zones through an improved hybrid pixel decomposition algorithm, and automatically identifies ecological land using a deep convolutional neural network (e.g., wavelet transform U-Net), solving the problems of blurred boundaries and low classification efficiency in traditional methods. For example, in some embodiments of this application, the method for obtaining the ecological riparian zone ratio of the river and lake riparian zone includes: defining the water level fluctuation zone based on medium-to-high resolution satellite imagery (medium-resolution remote sensing imagery data can be Sentinel-2, and high-resolution remote sensing imagery data can be GF-1 / GF-6) through spectral unmixing and flooding frequency analysis, and determining the riparian zone monitoring range by buffering several meters towards the land area; constructing a wavelet transform U-Net deep convolutional neural network model, and using this model to automatically classify ecological land using multi-temporal high-resolution images within the riparian zone monitoring range; and quantifying the ecological riparian zone ratio by statistically analyzing the proportion of ecological land by river segment. The method for obtaining the ecological shoreline ratio of river and lake riparian zones provided in this application significantly improves boundary accuracy and identification efficiency, supporting large-scale dynamic monitoring of riparian zones.

[0035] Please refer to Figure 1 , Figure 1The river and lake riparian zone remote sensing monitoring system provided in some embodiments of this application includes: a remote sensing image acquisition device 100 and a classification server 200. The remote sensing image acquisition device 100 is used to acquire medium-resolution remote sensing image data or high-resolution remote sensing image data, which includes both historical and current remote sensing image data. The classification server 200 executes the river and lake riparian zone ecological riparian zone ratio acquisition method provided in the embodiments of this application based on the data provided by the remote sensing image acquisition device 100.

[0036] It should be noted that, Figure 1 The system architecture described is merely a specific example; those skilled in the art can design different systems based on actual needs. Figure 1 The system used in this application is not limited to the specific architecture of the system for obtaining the ecological shoreline ratio of river and lake shorelines.

[0037] like Figure 2 As shown, embodiments of this application provide a remote sensing monitoring method for river and lake riparian zones, which includes:

[0038] S101, acquire historical medium-resolution remote sensing image data and historical high-resolution remote sensing image data.

[0039] S102, extract endmember spectra of typical land features from the historical medium-resolution remote sensing image data, wherein the typical land features include: water bodies, vegetation and bare land.

[0040] S103, the endmember spectrum is corrected based on the high-resolution remote sensing image data and shadow or cloud interference is removed to obtain the target endmember spectral matrix, wherein the target endmember spectral matrix is ​​used to provide a spectral decomposition reference.

[0041] S104, perform pixel unmixing on the medium-resolution remote sensing image data to be decomposed based on the target endmember spectral matrix to obtain the pixel unmixing result, wherein the medium-resolution remote sensing image data to be decomposed is used to provide the spectrum of the pixels to be decomposed. In the embodiments of this application, the medium-resolution remote sensing image data to be decomposed belongs to remote sensing image data from different periods in the same area as historical medium-resolution remote sensing image data and historical high-resolution remote sensing image data, wherein the medium-resolution remote sensing image data to be decomposed belongs to the remote sensing image data of the current period.

[0042] S105, determine the water level fluctuation zone based on the pixel demixing result and the flooding frequency.

[0043] S106, Determine the monitoring range of the coastal zone based on the water level fluctuation area.

[0044] S107 acquires image data corresponding to the coastal zone monitoring range, and obtains data for four high-resolution bands based on the image data, wherein the four high-resolution bands include red, green, blue and near-infrared bands;

[0045] S108 inputs the data from the four high-resolution bands into the wavelet transform U-Net deep convolutional neural network model to automatically identify the land use type within the monitoring range of the riparian zone and obtain the riparian zone land use classification result. The wavelet transform U-Net deep convolutional neural network model uses discrete wavelet transform for downsampling processing, and performs cross-layer concatenation of the multi-scale features output by the discrete wavelet transform through 3×3 convolution and combines channel attention and spatial attention for attention fusion.

[0046] S109, Based on the land use classification results of the riparian zone, quantify the ecological riparian zone ratio results for each river section. For example, S108 further includes the following steps: Combining the generated riparian zone monitoring range and land use classification identification results, classify and statistically analyze the land use according to ecological land and non-ecological land; according to the river section division, statistically analyze the proportion of ecological land area in the riparian zone, and calculate the ecological riparian zone ratio for each river section.

[0047] The inventors of this application discovered in their research that medium-resolution remote sensing images (such as Sentinel-2) are affected by the mixed pixel effect, resulting in blurred water body boundaries. Meanwhile, high-resolution remote sensing images have low coverage frequency, making it difficult to achieve long-term continuous monitoring. Therefore, the embodiments of this application use high-resolution remote sensing images to assist endmember correction and employ an improved demixing algorithm to improve boundary continuity, thereby obtaining a more accurate monitoring range and improving the accuracy of land use type identification in the region of interest.

[0048] In some embodiments of this application, the process of obtaining pixel unmixing results by performing pixel unmixing on the medium-resolution remote sensing image data to be decomposed based on the target endmember spectral matrix in step S104 includes, for example, obtaining a spatially continuous water abundance distribution map based on the target endmember spectral matrix and the medium-resolution remote sensing image data to be decomposed; correspondingly, the process of determining the water level fluctuation zone based on the pixel unmixing results and the flooding frequency in step S105 includes, for example, obtaining high-resolution remote sensing image data with a spatial range corresponding to the medium-resolution remote sensing image data to be decomposed, and obtaining a water body binary map based on the high-resolution remote sensing image data corresponding to the medium-resolution remote sensing image data to be decomposed; and obtaining the water level fluctuation zone based on the water body binary map and the water abundance distribution map.

[0049] Some embodiments of this application define the water level variation range by using spectral unmixing and flooding frequency analysis, thereby improving the accuracy of the obtained water level variation range results.

[0050] In some embodiments of this application, obtaining a spatially continuous water abundance distribution map based on the target endmember spectral matrix and the medium-resolution remote sensing image data to be decomposed includes:

[0051] The first step is to perform a linear spectral mixture model solution on the medium-resolution remote sensing image data to be decomposed pixel by pixel based on the target endmember spectral matrix to obtain the abundance vector of each pixel. The abundance vector is used to quantitatively characterize the abundance distribution of endmembers of each typical land cover in each pixel.

[0052] For example, this first step typically includes: spatially resampling the multi-phase abundance distribution map to obtain a spatially sampled water body abundance distribution map, ensuring that the resolution of the sampling result is consistent with the high-resolution remote sensing image data; interpolating the spatially sampled water body abundance distribution map using bilinear interpolation to obtain a target water body abundance distribution map; and constructing a long-term water body raster dataset based on the water body binary map and multiple target water body abundance distribution maps, wherein the long-term water body raster dataset is a spatiotemporal representation obtained from the multi-temporal target water body abundance distribution map. The cube-shaped structure stores the sub-pixel level water coverage ratio in the long-term water raster dataset, which is used to quantify the dynamic characteristics of the water level fluctuation area. The flooding frequency of each pixel is calculated based on the long-term water raster dataset to obtain a flooding frequency distribution map. The water level fluctuation area is obtained based on the flooding frequency distribution map. For example, obtaining the water level fluctuation area based on the flooding frequency distribution map includes: extracting all pixels satisfying 0.25≤WIF≤0.75 from the flooding frequency distribution map to obtain the water level fluctuation area.

[0053] It is easy to understand that the embodiments of this application spatially resample the water abundance distribution map (which is the result of water body identification from medium-resolution remote sensing image data) to make its resolution consistent with that of high-resolution remote sensing images (such as GF-1 / GF-6). Furthermore, the embodiments of this application use bilinear interpolation to maintain the continuity of abundance values ​​after sampling (avoiding pixel value jumps), thus ensuring accurate spatial matching during subsequent overlay analysis. The embodiments of this application determine sub-pixel flooding (i.e., through α...). 水体 By combining statistical frequency thresholds (i.e., flooding frequency), it accurately captures the gradual transition zone between land and water that is ignored by traditional methods, providing irreplaceable dynamic baseline data for coastal zone management.

[0054] The second step involves extracting the initial water body extent using the Normalized Differential Water Index (NDWI) thresholding method, and then obtaining the water abundance components in each pixel based on the initial water body extent and the abundance vector, thus obtaining the water endmember abundance value α. 水体 。

[0055] The third step is to determine the abundance value α of the water endmember.水体 Pixels with an NDWI greater than 0 and less than 0.1 are forcibly assigned a value of 0, while the remaining pixels retain the water endmember abundance value α. 水体 The spatially continuous water abundance distribution map is obtained.

[0056] The embodiments of this application are based on the pixel-level water abundance value α obtained by solving a linear mixture model. 水体 Furthermore, the calculated water abundance values ​​are corrected based on NDWI, with the correction principle being that values ​​satisfying α... 水体 Pixels with an NDWI value greater than 0 (indicating the presence of water components in the unmixed model) and an NDWI value less than 0.1 (indicating that the optical characteristics do not match the typical reflection characteristics of water) have their water abundance values ​​forcibly reset to 0. By correcting the NDWI, interference from "false water bodies" such as shadows / dark buildings can be eliminated, significantly improving the accuracy of water areas on the water abundance distribution map.

[0057] In some embodiments of this application, the step of performing a linear spectral mixing model to solve the medium-resolution remote sensing image data to be decomposed pixel by pixel based on the target endmember spectral matrix to obtain the abundance vector of each pixel includes: based on the linear spectral mixing model, under the dual constraints of "abundance is non-negative" and "abundance sum is 1", applying the least squares method to perform spectral unmixing on each pixel to obtain the abundance vector of each pixel.

[0058] Some embodiments of this application obtain the abundance vector of each pixel through two principles, thereby improving the accuracy of the obtained abundance vector.

[0059] In summary, the embodiments of this application relate to the field of environmental remote sensing monitoring technology, and particularly to an intelligent identification method for water level fluctuation zones and riparian land use in long-term satellite imagery. Specifically, it relates to an improved hybrid pixel decomposition algorithm and a deep convolutional neural network model, which enhances the accuracy of water level fluctuation zone boundary identification, enables rapid and high-precision identification of ecological land use types within a large riparian space, and ultimately establishes an ecological riparian zone ratio quantification method to provide data support for riparian zone ecological management.

[0060] The embodiments of this application aim to address the following technical problems existing in current riparian zone remote sensing monitoring: First, it is difficult to accurately extract the range of water level fluctuation zones based on medium- and high-resolution satellite remote sensing data, resulting in inaccurate delineation of riparian zone spatial boundaries; second, there is a lack of efficient and accurate ecological land identification methods, which cannot support the rapid calculation and dynamic monitoring of the riparian zone ecological-shore zone ratio over large-scale areas. Therefore, the embodiments of this application propose a method that comprehensively utilizes medium- and high-resolution satellite imagery, combined with an improved hybrid pixel decomposition algorithm and a deep convolutional neural network model, to achieve high-precision extraction of the water level fluctuation zone boundary, and on this basis, buffers the land area, thereby accurately determining the spatial range of the riparian zone; simultaneously, it realizes the automatic classification and identification of riparian zone ecological land, thereby supporting the dynamic monitoring of the spatiotemporal distribution characteristics of riparian zone ecological land at the watershed scale and the rapid calculation of the riparian zone ecological-shore zone ratio, solving the problems of low data acquisition accuracy and poor timeliness in current aquatic habitat protection and supervision.

[0061] The following uses specific medium-to-high resolution remote sensing image data as an example to illustrate the method for obtaining the ecological shoreline ratio of river and lake riparian zones provided in some embodiments of this application.

[0062] This application provides a method for obtaining the ecological shoreline ratio of river and lake riparian zones. The method includes: S1: acquiring and preprocessing medium- and high-resolution remote sensing images to collect basic geographic data such as the distribution of target water bodies; S2: comprehensively utilizing medium- and high-resolution remote sensing images, extracting water level variation zones and determining the riparian zone monitoring range through hybrid pixel decomposition and water body frequency analysis; S3: constructing a wavelet transform U-Net deep convolutional neural network model to automatically identify land use types within the riparian zone; S4: quantifying the ecological shoreline ratio of each river segment based on the land use classification results of the riparian zone in S3.

[0063] Better still, S1 further includes the following steps: S1.1: Acquire historical Sentinel2 MSI images of the target water body area over many years (as an example of medium-resolution remote sensing image data) and GF-1 / GF-6 PMS high-resolution remote sensing image data (as an example of high-resolution remote sensing image data), and perform geometric correction, spatial cropping, radiometric calibration, atmospheric correction, image registration, etc.; S1.2: Collect background geographical data such as water body distribution in the monitoring area to assist in analysis and model construction.

[0064] Better still, S2 further includes the following steps:

[0065] S2.1: The N-FINDR endmember extraction algorithm (i.e., N-dimensional spectral feature lookup method) is used to automatically identify the endmember spectral information of typical land cover types (such as water bodies, vegetation, bare land, roads, etc.) in Sentinel 2 MSI medium-resolution imagery. To improve the representativeness and purity of the endmember spectra, high-resolution remote sensing images from the same period are introduced for spectral correction and anomaly removal processing, effectively eliminating the interference caused by shadows and clouds.

[0066] S2.2: Based on the linear spectral mixing model, under the dual constraints of "abundance is non-negative" and "abundance summation is 1", the least squares method is applied to perform spectral unmixing on the target pixels, and the abundance distribution of each endmember in each pixel is quantitatively obtained (the specific calculation is as follows: formula (1)). At the same time, combined with the water body range initially extracted by the NDWI threshold method, the abundance components of water in the mixed pixels are further identified and analyzed, thereby accurately depicting the water-land transition zone.

[0067] Assuming each pixel in multispectral data is composed of a linear combination of several endmembers (pure spectra), the mathematical expression for the linear spectral mixture model is:

[0068] (1)

[0069] In the formula, Let be the pixel surface reflectance vector. For the endmember spectral matrix, For abundance vectors, For the measurement error vector, m The number of multispectral image bands, n This represents the number of endmembers.

[0070] In unmixing, the abundance vector is obtained using the least squares method under the following physical constraints:

[0071] 1) The sum of abundance equals 1.

[0072] 2) Abundance is a non-negative value.

[0073] S2.3: The water body identification results from the Sentinel 2 imagery are spatially resampled to maintain consistency with the high-resolution remote sensing imagery. The spectral unmixing results of each image are overlaid with long-term water raster data for analysis, and the flooding frequency (WIF) of each pixel is calculated. Based on the statistical results, spatial regions satisfying 0.25 ≤ WIF ≤ 0.75 are defined as typical water level fluctuation zones, used to characterize the dynamic changes in water body boundaries under seasonal or human-induced regulation.

[0074] S2.4: Based on the outer boundary of the water level fluctuation zone, a buffer of 150 meters is added to the land to fully cover the area affected by the interaction between water and land, and finally determine the monitoring range of the riparian zone, providing accurate spatial boundary support for subsequent dynamic land use assessment.

[0075] Better still, S3 further includes the following steps: constructing a wavelet transform U-Net deep convolutional neural network model to automatically identify land use types within the riparian zone, including:

[0076] S3.1: The geomedian method is used to fuse multi-temporal high-resolution remote sensing images, and combined with existing land use type information, spatiotemporal matching is carried out to complete the annotation of the training dataset. The land use classification system includes ecological land and non-ecological land, and ecological land and non-ecological land further include corresponding subcategories.

[0077] S3.2: Construct as follows Figure 3 The wavelet transform U-Net deep convolutional neural network model shown is designed with a loss function adapted to multi-class recognition (as shown below), and model training and accuracy verification are carried out.

[0078] Figure 3 The wavelet transform U-Net deep convolutional neural network model, by introducing discrete wavelet transform (DWT), dense cross-layer connection mechanism and attention fusion module AFF, can improve feature representation ability and improve the accuracy of coastal land use type interpretation results.

[0079] The input data for this model includes four high-resolution bands, which refer to the red, green, blue, and near-infrared reflectance data of the high-resolution imagery corresponding to the coastal zone monitoring area generated by combining S2 and S3. The model output data is the spatial distribution information of land use type identification results, including water areas, forest land, grassland, agricultural land, unused land, hardened embankments, and other land uses.

[0080] The specific improvements to this model include:

[0081] 1) Application of wavelet transform in encoders

[0082] This model employs Discrete Wavelet Transform (DWT) instead of traditional pooling for downsampling. Through orthogonal invertible filters such as Haar wavelets (including four sub-band filters: LL, LH, HL, and HH), the input image is decomposed into multi-resolution sub-band images, preserving local frequency domain features while reducing spatial resolution. The fixed filter and step size design of the wavelet transform avoids information loss, and high-resolution feature reconstruction is achieved through inverse wavelet transform (IWT), effectively recovering detailed information.

[0083] (Equation 1)

[0084] In the formula,

[0085] —Low-pass filter,

[0086] —High-pass filter.

[0087] After Haar transform (DWT), the low-resolution image result can be denoted as:

[0088] (Equation 2)

[0089] Typically, pooling layers in convolutional neural networks, which restore the original image resolution through transposed convolution or interpolation, suffer some information loss. The inverse wavelet transform (IWT) method can obtain image reconstruction results without information loss.

[0090]

[0091] 2) Dense cross-layer connection mechanism

[0092] Inspired by DenseNet, this model introduces a dense cross-level connection strategy. This strategy utilizes 3×3 convolutions (such as...) Figure 3 The Conv 3×3 method shown cascades multi-scale features from the DWT output across layers, enhancing feature reuse and mitigating the gradient vanishing problem. By fusing shallow detail features with deep semantic features, the accuracy of boundary segmentation is significantly improved.

[0093] 3) Attention Fusion Module (AFF)

[0094] To further optimize feature fusion, the model integrates an attention mechanism, and proposes, for example... Figure 3 The diagram shows multiple attention fusion modules (AFF). This module combines channel attention and spatial attention, dynamically weights important feature regions, and suppresses redundant information. It can effectively improve the model's ability to focus on non-ecological land in the riparian zone and achieve more accurate classification in complex land use types.

[0095] The attention fusion module (AFF) architecture of this application embodiment is as follows: Figure 4 As shown, it includes:

[0096] The system consists of several modules: low-level features (corresponding to spatial path output), high-level features (corresponding to channel attention), concatenate, convolution and activation (conv+bn+relu), global pooling, 1×1 convolution (conv 1×1), batch normalization, activation to generate channel attention weights (sigmoid), multiplication (mul), and addition (add).

[0097] Feature concatenation: Low-level features and high-level features are merged through the "concatenate" operation.

[0098] Convolution and Activation: After concatenation, the features are processed by "conv+relu" (convolution + ReLU activation) to extract preliminary fused features.

[0099] Attention mechanism module: Perform "global pooling" on the above features to compress the spatial dimension; learn channel correlation through "conv 1×1" (1×1 convolution); stabilize training with "batch normalization"; and activate and generate channel attention weights with "sigmoid".

[0100] Feature weighting and fusion: Attention weights and original features are weighted by "mul" (multiplication); the weighted features are then fused by "add".

[0101] To enhance the model's generalization ability, the images undergo data enhancement processing such as image cropping, rotation, brightness adjustment, and contrast adjustment before input, thereby improving the model's robustness under different imaging conditions.

[0102] Model training parameter settings include:

[0103] Optimizer: Adam, initial learning rate 1×10^-4, weight decay coefficient 1×10^-4.

[0104] Loss function: Combining Dice Loss and Focal Loss to alleviate the class imbalance problem.

[0105] Batch Size: 16.

[0106] Number of iterations (Epochs): 200.

[0107] Weight initialization: Kaiming Normal initialization.

[0108] Data split ratio: 70% training set, 15% validation set, and 15% test set.

[0109] Regularization strategies: Dropout (p=0.5) and L2 regularization to prevent overfitting.

[0110] To address the class imbalance problem, some embodiments of this application design the following loss function:

[0111] Total_loss=Dice_loss+(1×Focal_loss)

[0112] Among them, Dice loss effectively improves the model's segmentation performance on minority classes and is suitable for handling imbalanced pixel distributions; Focal loss, on the other hand, enhances the model's ability to recognize complex boundaries by reducing attention to easily classified samples and strengthening the learning of difficult-to-classify samples. By combining the two, the model can simultaneously optimize the overall segmentation accuracy and the ability to recognize difficult-to-classify regions during training, thereby improving the robustness and generalization ability of the performance.

[0113] Dice Loss, based on the Dice Coefficient, measures the degree of overlap between two sets (predicted values ​​and actual values). Its calculation method is as follows:

[0114]

[0115] Focal Loss is an improvement on Cross Entropy Loss by introducing an adjustable factor that makes the model pay more attention to samples that are difficult to classify, while reducing its attention to samples that are easy to classify.

[0116]

[0117] S3.3: Input the processed remote sensing image and output a distribution map of various land use types within the coastal zone.

[0118] S3.4 The model output results are manually reviewed and corrected to ensure the reliability and consistency of the recognition results.

[0119] Better still, S4 further includes the following steps: S4.1: Combining the riparian zone monitoring range and land use classification identification results generated by S2 and S3, classify and statistically analyze them according to ecological land and non-ecological land; S4.2 According to the river section division, statistically analyze the proportion of ecological land area in the riparian zone, and calculate the ecological riparian zone ratio of each river section.

[0120] It is easy to understand that the embodiments of this application innovatively integrate hybrid pixel decomposition technology and deep learning methods, achieving multiple breakthroughs in accuracy, efficiency, and application depth of coastal zone remote sensing monitoring, and has the following significant technical effects:

[0121] 1) Improve the accuracy of boundary extraction: By introducing an improved hybrid pixel decomposition algorithm, the boundary blurring problem caused by hybrid pixels in medium resolution images is overcome, which significantly improves the extraction accuracy and spatial continuity of the boundary of the water level fluctuation zone, and provides technical support for the scientific delineation of the spatial range of the coastal zone.

[0122] 2) Improve the efficiency and accuracy of ecological land identification: Construct a wavelet transform-based U-Net deep convolutional neural network model to achieve automated and high-precision identification of ecological land types within the coastal zone, solve the problems of low efficiency and strong subjectivity of traditional manual interpretation, and improve the level of intelligence in ecological land identification.

[0123] 3) Support for large-scale dynamic monitoring of riparian zones: A dynamic quantitative method for ecological riparian zone ratio is proposed, which can realize spatiotemporal change monitoring and health assessment of ecological land use in riparian zones on a large scale and in multiple time phases, providing data support for large-scale riparian zone ecological management in watersheds, regions and other areas.

[0124] Please refer to Figure 5 , Figure 5 This application illustrates a remote sensing monitoring device for river and lake riparian zones, which includes:

[0125] The historical data acquisition module 201 is configured to acquire historical medium-resolution remote sensing image data and historical high-resolution remote sensing image data.

[0126] The endmember spectrum acquisition module 202 is configured to extract endmember spectra of typical land features from the historical medium-resolution remote sensing image data, wherein the typical land features include: water bodies, vegetation or bare land.

[0127] The preprocessing module 203 is configured to correct the endmember spectrum and remove shadow or cloud interference based on the high-resolution remote sensing image data to obtain a target endmember spectral matrix, wherein the target endmember spectral matrix is ​​used to provide a spectral decomposition reference.

[0128] The pixel unmixing module 204 is configured to perform pixel unmixing on the medium-resolution remote sensing image data to be decomposed based on the target endmember spectral matrix to obtain the pixel unmixing result, wherein the medium-resolution remote sensing image data to be decomposed is used to provide the spectrum of the pixels to be decomposed.

[0129] The water level fluctuation zone determination module 205 is configured to determine the water level fluctuation zone based on the pixel demixing result and the flooding frequency.

[0130] The coastal zone monitoring range determination module 206 is configured to determine the coastal zone monitoring range based on the water level fluctuation area.

[0131] The land use type identification module 207 is configured to acquire image data corresponding to the monitoring range of the riparian zone, and obtain data of four high-resolution bands based on the image data, wherein the four high-resolution bands include red, green, blue and near-infrared bands; input the data of the four high-resolution bands into the wavelet decomposition U-Net deep convolutional neural network model to automatically identify the land use type of the riparian zone monitoring range and obtain the riparian zone land use classification result, wherein the wavelet transform U-Net deep convolutional neural network model uses discrete wavelet transform for downsampling processing, and performs cross-layer concatenation of the multi-scale features output by the discrete wavelet transform through 3×3 convolution and combines channel attention and spatial attention for attention fusion.

[0132] The ecological shoreline ratio quantification module 208 is configured to quantify the ecological shoreline ratio results of each river segment based on the shoreline land use classification results.

[0133] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the aforementioned method, and will not be elaborated further here.

[0134] Some embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can implement the non-ecological land use identification method for river and lake riparian areas as described in any of the above embodiments.

[0135] like Figure 6 As shown, some embodiments of this application provide an electronic device 500, which includes a memory 510, a processor 520, and a computer program stored in the memory 510 and executable on the processor 520. When the processor 520 executes the program, it can implement the remote sensing monitoring method for river and lake shorelines as described in any of the above embodiments.

[0136] Processor 520 can process digital signals and can include various computing architectures. For example, it can be a complex instruction set computer architecture, a reduced instruction set computer architecture, or an architecture that implements multiple instruction set combinations. In some examples, processor 520 can be a microprocessor.

[0137] The memory 510 can be used to store instructions executed by the processor 520 or data related to the execution of instructions. These instructions and / or data may include code used to implement some or all of the functions of one or more modules described in the embodiments of this application. The processor 520 of the embodiments of this disclosure can be used to execute the instructions in the memory 510 to implement… Figure 2 The method shown. Memory 510 includes dynamic random access memory, static random access memory, flash memory, optical memory, or other memory well known to those skilled in the art.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0139] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0140] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0142] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0143] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A remote sensing monitoring method for river and lake riparian zones, characterized in that, The remote sensing monitoring methods for river and lake shorelines include: Acquire historical medium-resolution remote sensing image data and historical high-resolution remote sensing image data; The endmember spectra of typical land features are extracted from the historical medium-resolution remote sensing image data, wherein the typical land features include: water bodies, vegetation and bare land; The endmember spectrum is corrected based on the high-resolution remote sensing image data and shadow or cloud interference is removed to obtain the target endmember spectral matrix, wherein the target endmember spectral matrix is ​​used to provide a spectral decomposition benchmark. The target endmember spectral matrix is ​​used to perform pixel unmixing on the medium-resolution remote sensing image data to be decomposed, and the pixel unmixing result is obtained. The medium-resolution remote sensing image data to be decomposed is used to provide the spectrum of the pixels to be decomposed. The water level fluctuation zone is determined based on the pixel unmixing results and the flooding frequency. The monitoring range of the coastal zone is determined based on the aforementioned water level fluctuation area; Image data corresponding to the coastal zone monitoring range is acquired, and data for four high-resolution bands are obtained based on the image data, wherein the four high-resolution bands include red, green, blue and near-infrared bands; The data from the four high-resolution bands are input into the wavelet transform U-Net deep convolutional neural network model to automatically identify the land use types within the riparian zone monitoring area and obtain the riparian zone land use classification results. The wavelet transform U-Net deep convolutional neural network model uses discrete wavelet transform for downsampling processing, and performs cross-layer concatenation of the multi-scale features output by the discrete wavelet transform through 3×3 convolution and combines channel attention and spatial attention for attention fusion. Based on the land use classification results of the riparian zone, the ecological riparian zone ratio of each river section was quantified.

2. The remote sensing monitoring method for river and lake shorelines as described in claim 1, characterized in that, The step of unmixing the medium-resolution remote sensing image data to be decomposed based on the target endmember spectral matrix to obtain the pixel unmixing result includes: A spatially continuous water abundance distribution map is obtained based on the target endmember spectral matrix and the medium-resolution remote sensing image data to be decomposed. The step of determining the water level fluctuation range based on the pixel demixing result and the flooding frequency includes: Acquire high-resolution remote sensing image data with a spatial range corresponding to the medium-resolution remote sensing image data to be decomposed, and obtain a water body binary map based on the high-resolution remote sensing image data corresponding to the medium-resolution remote sensing image data to be decomposed. The water level fluctuation zone is obtained based on the binary water body map and the water abundance distribution map.

3. The remote sensing monitoring method for river and lake shorelines as described in claim 2, characterized in that, The step of obtaining a spatially continuous water abundance distribution map based on the target endmember spectral matrix and the medium-resolution remote sensing image data to be decomposed includes: Based on the target endmember spectral matrix, the linear spectral mixture model is performed pixel by pixel on the medium-resolution remote sensing image data to be decomposed to obtain the abundance vector of each pixel. The abundance vector is used to quantitatively characterize the abundance distribution of endmembers of each typical land cover in each pixel. The initial water body extent was extracted using the Normalized Differential Water Index (NDWI) thresholding method. Based on the initial water body extent and the abundance vector, the abundance components of the water body in each pixel were obtained, resulting in the water body endmember abundance value α. 水体 ； The water body endmember abundance value α 水体 Pixels with an NDWI greater than 0 and less than 0.1 are forcibly assigned a value of 0, while the remaining pixels retain the water endmember abundance value α. 水体 The spatially continuous water abundance distribution map is obtained.

4. The remote sensing monitoring method for river and lake shorelines as described in claim 3, characterized in that, The step of performing a linear spectral mixture model solution on the medium-resolution remote sensing image data to be decomposed pixel by pixel based on the target endmember spectral matrix to obtain the abundance vector of each pixel includes: Based on the linear spectral mixing model, under the dual constraints of "abundance is non-negative" and "abundance summation is 1", the least squares method is applied to perform spectral unmixing on each pixel to obtain the abundance vector of each pixel.

5. The remote sensing monitoring method for river and lake shorelines as described in claim 4, characterized in that, The step of obtaining the water level fluctuation zone based on the binary water body map and the water abundance distribution map includes: Spatially resample the water abundance distribution map to obtain a spatially sampled water abundance distribution map, so that the resolution of the spatially sampled water abundance distribution map is consistent with that of the high-resolution remote sensing image data. The abundance distribution map of the spatially sampled water body was interpolated using the bilinear interpolation method to obtain the abundance distribution map of the target water body. A long-term water body raster dataset is constructed based on the binary water body map and multiple target water body abundance distribution maps. The long-term water body raster dataset is a spatiotemporal cube obtained from the multi-temporal target water body abundance distribution maps. The pixel values ​​in the long-term water body raster dataset are used to store the sub-pixel level water body coverage ratio and to quantify the dynamic characteristics of the water level fluctuation zone. The flooding frequency distribution map is obtained by calculating the flooding frequency of each pixel based on the long-time water body raster dataset. The water level fluctuation zone is obtained from the flooding frequency distribution map.

6. The remote sensing monitoring method for river and lake shorelines as described in claim 5, characterized in that, The method of obtaining the water level fluctuation zone based on the flooding frequency distribution map includes: The water level fluctuation zone is obtained by extracting all pixels that satisfy 0.25≤floodingfrequency≤0.75 from the flooding frequency distribution map.

7. A remote sensing monitoring device for river and lake shorelines, characterized in that, The remote sensing monitoring device for river and lake shorelines includes: The historical data acquisition module is configured to acquire historical medium-resolution remote sensing image data and historical high-resolution remote sensing image data. The endmember spectrum acquisition module is configured to extract endmember spectra of typical land features from the historical medium-resolution remote sensing image data, wherein the typical land features include: water bodies, vegetation, or bare land. The preprocessing module is configured to correct the endmember spectrum and remove shadow or cloud interference based on the high-resolution remote sensing image data to obtain the target endmember spectral matrix, wherein the target endmember spectral matrix is ​​used to provide a spectral decomposition reference. The pixel unmixing module is configured to perform pixel unmixing on the medium-resolution remote sensing image data to be decomposed based on the target endmember spectral matrix to obtain the pixel unmixing result, wherein the medium-resolution remote sensing image data to be decomposed is used to provide the pixel spectrum to be decomposed. The water level fluctuation zone determination module is configured to determine the water level fluctuation zone based on the pixel unmixing result and the flooding frequency; The coastal zone monitoring range determination module is configured to determine the coastal zone monitoring range based on the water level fluctuation zone; The land use type identification module is configured to: acquire image data corresponding to the monitoring range of the riparian zone, and obtain data for four high-resolution bands based on the image data, wherein the four high-resolution bands include red, green, blue, and near-infrared bands; input the data of the four high-resolution bands into a wavelet transform U-Net deep convolutional neural network model to automatically identify the land use type of the monitoring range of the riparian zone and obtain the land use classification result of the riparian zone, wherein the wavelet transform U-Net deep convolutional neural network model uses discrete wavelet transform for downsampling processing, performs cross-layer concatenation of the multi-scale features output by the discrete wavelet transform through 3×3 convolution, and combines channel attention and spatial attention for attention fusion; The ecological shoreline ratio quantification module is configured to quantify the ecological shoreline ratio results for each river segment based on the shoreline land use classification results.

8. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, can implement the remote sensing monitoring method for river and lake shorelines as described in any one of claims 1-6.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein, When the processor executes the computer program, it can implement the remote sensing monitoring method for river and lake shorelines as described in any one of claims 1-6.

10. A computer program product, characterized in that, It includes computer program instructions, which, when read and executed by a processor, can implement the remote sensing monitoring method for river and lake shorelines as described in any one of claims 1-6.

Images