A method for remote sensing monitoring of ice lakes

Abstract

The application discloses a kind of ice lake remote sensing monitoring methods, comprising:1) remote sensing data download and pre-processing;2) based on NDWI and DNSI composite remote sensing index CI is constructed;3) composite remote sensing index CI pre-processing;4) DEM data scale reduction and pixel matching processing;5) based on DEM dynamic rate CI threshold value;6) ice water classification and area extraction.The application aims at the problem that single remote sensing index and fixed threshold method is low in precision of ice lake area extraction and ice water classification under complex topography conditions in plateau, poor in universality, composite remote sensing index is constructed, and topographic factor is introduced to dynamically adjust classification threshold value, the fitting precision of CI threshold value and elevation relationship is improved using support vector regression method, and the precision of ice lake area extraction and ice water classification discrimination and the like is effectively improved.The method significantly improves the spatial adaptability and monitoring precision of ice lake classification result, provides technical support for ice lake evolution analysis and disaster warning in plateau area.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing monitoring of ice lakes, and belongs to the technical field of remote sensing. Specifically, it is an ice lake remote sensing monitoring method, which can realize accurate identification and area extraction of ice water boundaries of ice lakes in plateau areas, and is suitable for scenarios such as ice lake dynamic change analysis and disaster risk monitoring. BACKGROUND

[0002] As a unique water body in plateau areas, the boundary change of ice lakes directly reflects glacier ablation, water quantity change and climate change. Ice lake boundary extraction is an important basis for evaluating ice lake flood risk, formulating disaster prevention and mitigation measures, and carrying out water resource management. With the continuous development of remote sensing technology, automatic extraction of ice lake boundaries based on remote sensing has gradually become a research hotspot, and accurate and efficient extraction of ice lake boundaries has become a key link in plateau remote sensing monitoring.

[0003] At present, the remote sensing index method has been widely used in ice lake extraction and water body and ice surface classification due to its simple structure, efficient calculation and other advantages. Common water body identification indexes include normalized water index (NDWI), modified normalized water index (MNDWI), automatic water index (AWEI), etc. These indexes mainly use the difference in reflection characteristics of water body in visible light and near-infrared bands for discrimination; and common indexes for ice and snow identification include normalized snow index (NDSI), normalized difference ice index (NDII), snow cover index (SCI), etc., which mainly reflect ice and snow coverage through the combination of green light and short-wave infrared. In practical application, different indexes have different sensitivity to targets. NDWI is sensitive to water body but unstable in ice and snow mixed areas, NDSI can identify snow surface but is easy to be confused with clear water body, MNDWI and AWEI have certain advantages in high-reflectance background, in addition, some researches introduce spectral angle method, principal component analysis (PCA) and other improvement means to improve the stability of the index in complex background. However, most of these methods still rely on fixed threshold, and lack of linkage adjustment mechanism with terrain and time change factors.

[0004] In current remote sensing monitoring of ice lakes, the identification and classification of ice lake boundaries are mostly dependent on a single remote sensing index (such as NDWI, NDSI, etc.) combined with a fixed threshold. Although this method has certain applicability under certain conditions, it has significant limitations when faced with complex and variable topographic environments and seasonal climate changes. On the one hand, this method generally lacks the ability to respond to factors such as topographic relief and time-varying water body state, making it difficult to adapt to the dynamic changes of ice lake boundaries in different time periods, different elevation regions, and even different remote sensing images. On the other hand, traditional index thresholds are mostly set by artificial experience, lacking a scientific and reasonable dynamic calibration mechanism and regional adaptation strategy, and are easily disturbed by image lighting conditions, sensor differences, and surface coverage, resulting in large fluctuations and low precision in ice-water classification results, making it difficult to meet the actual needs of rapid warning and high-precision monitoring of ice lake disasters. Therefore, there is an urgent need to develop a new type of ice lake monitoring method that integrates multi-source remote sensing index information, has dynamic calibration capability, and can fully adapt to complex topographic conditions, to achieve automation, high precision, and strong robustness in the process of ice lake boundary extraction, and to provide more scientific and reliable data support for ice lake outburst risk warning, water resource management, and climate change research. SUMMARY

[0005] In view of the low ice-water boundary identification precision, poor adaptability of fixed threshold, and lack of topographic adjustment capability in existing ice lake remote sensing monitoring methods, the present application provides an ice lake remote sensing monitoring method. The purpose of the present application is achieved by the following technical solutions:

[0006] An ice lake remote sensing monitoring method, which is based on optical remote sensing images and digital elevation model (DEM), constructs a composite remote sensing index through NDWI and NDSI, and dynamically calibrates the threshold based on the SVR method in combination with the downscaled DEM, simultaneously achieving accurate extraction of ice lake area and fine classification of ice-water, comprising the following steps:

[0007] An ice lake remote sensing monitoring method, comprising the following steps:

[0008] 1) Data download and preprocessing:

[0009] Obtain multi-source optical remote sensing images of different time periods covering the target ice lake area, and perform atmospheric correction and geometric correction. Obtain the original DEM data of the target ice lake area.

[0010] The method of the present application is mainly based on Landsat, Sentinel 2, etc. optical remote sensing data. First, the remote sensing data platform is used to download remote sensing image data of the target area within a specific time period. The downloaded data should include different time series of images of the ice lake area. After downloading the image data, necessary preprocessing is performed, including atmospheric correction, geometric correction, etc. to ensure the spatial accuracy and spectral consistency of the remote sensing images.

[0011] 2) Composite remote sensing index construction:

[0012] The composite remote sensing index (CI) is established by the ratio of the two remote sensing indexes NDWI and NDSI. NDWI can be used to enhance the recognition ability of water bodies, and NDSI can help to enhance the differentiation between ice and water. CI reflects the comprehensive spectral characteristics of the ice-water mixing area, and is used to enhance the recognition ability of the ice lake area.

[0013] 3) Ice lake area extraction based on CI index:

[0014] The CI layer is standardized and preprocessed, and the Otsu maximum inter-class variance method is used for threshold segmentation of the standardized CI image to obtain the optimal ice lake boundary threshold t, and the overall ice lake area containing ice and water is extracted.

[0015] This step aims to extract the overall range of the ice lake area based on the composite remote sensing index (CI) layer, including the comprehensive coverage area of water and ice. The preprocessed CI layer has a unified numerical scale and significantly improves the differentiation of the index for ice-water areas. On this basis, the Otsu maximum inter-class variance method is used for segmentation of the standardized CI layer to automatically determine an optimal global threshold, thereby realizing the extraction of the overall ice lake area (containing water and ice pixels). The extraction result forms an ice lake mask layer, which is used as the input for the subsequent ice-water classification step based on the dynamic rate of elevation, and lays a data foundation for further distinguishing ice and water.

[0016] 4) DEM data downscaling and pixel matching processing:

[0017] The cubic interpolation method is used to resample the DEM data to realize the spatial alignment of the CI layer at the pixel level. To realize the pixel-level matching of 30m resolution DEM data and 10m resolution CI layer, the present application proposes to use the cubic interpolation (Cubic Interpolation) method to downscale the DEM data to improve its spatial resolution to 10m consistent with CI, to ensure the consistency of spatial position and the feasibility of data fusion.

[0018] 5) Dynamic calibration of ice-water classification threshold based on DEM:

[0019] Based on the downscaled DEM layer, the natural break method is used to segment the regional elevation, divide several elevation intervals, and extract the corresponding CI value set in each interval. The K-means clustering algorithm is used to divide the CI values into ice and water two categories, and the optimal ice-water classification CI threshold value is determined for each elevation interval. A sample set is constructed between the center elevation of each interval and the CI threshold value, and a support vector regression (SVR) is used to establish the mapping function of the elevation-CI threshold value. The SVR model is constructed by using a radial basis kernel function (RBF), and the model parameters include the penalty factor C and the kernel function parameter γ, which are automatically determined by K-fold cross-validation.

[0020] The specific operation is as follows:

[0021] First, the elevation interval of the downscaled DEM layer is divided, and the Jenks Natural Breaks method is used to perform optimal partitioning on the DEM data to ensure that the elevation difference within the interval is minimized and the elevation difference between the intervals is maximized. The optimization objective function of the natural break method is:

[0022]

[0023] Where h i is the elevation value, μ j is the mean value of the jth elevation interval, H j represents the jth interval, and k represents the total number of elevation intervals. The value of k (total number of elevation intervals) is usually flexibly set according to the terrain complexity and sample distribution of the study area, and is an adjustable parameter in the natural break method. This method ensures that the elevation interval division can best reflect the changes in the terrain, and can most reasonably partition the elevation characteristics of different regions, thereby providing a clear elevation structure basis for the subsequent ice-water classification.

[0024] In each elevation interval, the corresponding CI value set is extracted. Then, the K-means clustering algorithm is used to classify these CI values into ice and water two categories, thereby determining the optimal ice-water classification CI threshold value for each elevation interval. The K-means clustering algorithm performs optimal clustering through the following objective function:

[0025]

[0026] Where r ij is an indicator function, indicating whether the sample x i belongs to the jth class, μ j is the center of the jth class, n represents the total number of samples, i.e. the total number of CI layer pixels obtained in step 4), and k represents the total number of elevation intervals after optimal partitioning of the DEM data;‖x i -μ j ‖ represents the sample x iDistance to the cluster center μ j The algorithm iteratively optimizes the class labels of each data point such that all data points minimize the distance to the center of the class they belong to, thus achieving an effective classification of CI values. The goal of K-means clustering is to separate CI values into two classes of ice and water, and generate a representative ice-water classification threshold for each elevation interval through the clustering results.

[0027] Finally, based on the center elevation of each elevation interval and the corresponding CI threshold sample, a support vector regression (SVR) method is used to establish an elevation-threshold mapping function. SVR is a widely used nonlinear method for regression problems, which can map the original data to a high-dimensional space, thus improving the fitting ability of the model to complex nonlinear relationships. By minimizing the structural risk, the SVR method can effectively avoid overfitting while ensuring prediction accuracy, thus providing an accurate ice-water classification threshold for each elevation value.

[0028] The objective function of SVR is as follows:

[0029]

[0030] The constraint condition of regression is:

[0031] T i <ω,Φ(h i )>-b≤∈+ξ i ,

[0032] In the above formula: ‖ω 2 ‖ is the square norm of the weight vector, representing the complexity of the regression function; k represents the total number of elevation intervals after optimal partitioning of DEM data; i represents the interval number; ξ i and are slack variables used to allow a certain prediction error; C is the penalty coefficient used to control the degree of error penalty; T i represents the CI threshold corresponding to each elevation interval; <ω,φ(h i )> is the inner product of the elevation value mapped by the nonlinear kernel function and the regression function, b is the bias term, and ∈ is the regression tolerance, allowing a certain range of errors.

[0033] In the application of the present application, h i is the center elevation of each elevation interval, representing the typical representative value of different elevation intervals. When performing SVR modeling, the CI threshold T i obtained from step 5) is used to construct an elevation-CI threshold sample set with the corresponding elevation value h i . These sample sets include elevations and corresponding ice-water classification thresholds as input and target output data for SVR.

[0034] The SVR trains a regression model with these samples to fit the relationship between elevation and ice-water classification threshold, forming an elevation-CI threshold mapping function:

[0035] T = f(h)

[0036] Where f(h) is a function calculated by the SVR model, representing the ice-water classification threshold T corresponding to a given elevation value h. In the regression process, the SVR method uses a radial basis function (RBF) kernel for non-linear projection, and the kernel function form is:

[0037] K(h i ,h j ) = exp(-γ‖h i -h j ‖ 2 )

[0038] Where γ is the parameter of the RBF kernel function, controlling the width of the kernel function, and ‖h i -h j ‖ is the Euclidean distance between two elevation points. Through this non-linear mapping, SVR can effectively capture the complex relationship between elevation and ice-water classification threshold.

[0039] Finally, using the elevation-threshold mapping function obtained by SVR, the dynamic classification threshold can be calculated according to the elevation value of each pixel. The calculated dynamic threshold T i is combined with the CI layer to perform accurate ice-water classification. This method significantly improves the accuracy of ice lake remote sensing classification through dynamic calibration of elevation features, and adapts to changes under different terrain conditions.

[0040] 6) Fine classification of ice surface and water surface:

[0041] Based on the elevation-CI threshold mapping function, for each pixel in the ice lake area, its elevation value is extracted and substituted into the function to calculate the corresponding CI threshold, which is compared with the pixel CI value to realize ice-water classification. Finally, a high-resolution per-pixel ice-water distribution map is obtained, realizing fine remote sensing division of ice lake ice surface and water body.

[0042] Combined with the elevation-CI threshold mapping relationship established in step 5), the fine classification of ice surface and water surface of the standardized CI layer obtained in step 3) is realized. Specifically, for each image pixel to be classified, first extract its corresponding elevation value h i , and input it into the support vector regression model to calculate its matched dynamic classification threshold T i Then, the dynamic threshold T i is compared with the CI value C iIn contrast, the following criteria are used for discrimination:

[0043]

[0044] Finally, based on the above dynamic classification method, the ice-water classification result of each pixel in the entire remote sensing image is obtained, that is, the high-resolution, pixel-by-pixel ice lake and ice water distribution map.

[0045] In a further aspect, the formula for constructing CI in step 2) is as follows:

[0046]

[0047]

[0048] Where Green, NIR, and SWIR represent the reflectivity of the green band, near-infrared band, and short-wave infrared band of the remote sensing data, respectively. Reflectivity is a unitless quantity with a value ranging from 0 to 1, and is extracted from the remote sensing data processing results obtained in step 1).

[0049] In a further aspect, in step 3), the standardization preprocessing step is as follows: first, the CI layer constructed and calculated in the previous step is subjected to effective pixel screening, and pixels with CI values less than 0 are removed to exclude the influence of non-water body and ice surface interference factors such as cloud shadows, bare land, and snow. Subsequently, the remaining effective pixels are normalized by linearly stretching the CI values to the interval of 0 to 1 to enhance the numerical consistency and comparability between different time series and different sensor images.

[0050] In a further aspect, in step 3), the Otsu maximum between-class variance method formula is as follows:

[0051]

[0052] Where t is the ice lake boundary threshold value, i.e., the dividing value between the ice lake area (ice + water) and the non-ice lake area after standardization of the CI value; is the between-class variance, which is a measure of the difference between classes, used to measure the effectiveness of the ice lake and non-ice lake area under the current threshold, and the larger the value, the better the effect; ω0(t) and ω1(t) respectively represent the proportion of pixels belonging to and not belonging to the ice lake area (ice and water) in the standardized CI layer; μ0(t) and μ1(t) respectively represent the average CI value of pixels belonging to and not belonging to the ice lake area (ice + water). This method maximizes the between-class variance between the ice lake area and the background, effectively avoiding the bias caused by artificial experience setting, and improving the objectivity and robustness of ice lake boundary extraction.

[0053] In a further scheme, the cubic interpolation method in step 3) is: constructing a cubic fitting function at the pixel to be interpolated using the surrounding DEM values, completing the spatial refinement of the elevation data and ensuring the spatial consistency with the remote sensing data.

[0054] Specifically, at the position of each DEM target pixel to be interpolated, four adjacent known pixel values are selected, and a cubic interpolation function is constructed for fitting. The interpolation function is in the form of:

[0055] f(x) = a3(x-x0) 3 + a2(x-x0) 2 + a1(x-x0) + a0

[0056] Where x0 is the reference point, and a0 to a3 are coefficients obtained by least squares fitting. In two-dimensional space, the interpolation operation is performed on the x direction and the y direction respectively, and then the elevation value of the target pixel is obtained. After completing the interpolation processing, all DEM data are resampled to 10m resolution, and are spatially aligned with the CI layer to realize accurate pixel-level matching processing. This step provides key topographic auxiliary information support for subsequent dynamic CI threshold setting and ice lake area extraction based on DEM.

[0057] The beneficial effects of the present application are:

[0058] The method of the present application is based on Landsat, Sentinel 2 and other optical remote sensing data, and by introducing digital elevation model (DEM) information, the threshold value of the composite remote sensing index (CI) constructed based on NDWI and NDSI is dynamically adjusted, which can automatically identify the ice lake boundary under different terrain conditions and time stages, and improve the accuracy and stability of ice water classification.

[0059] The present application effectively solves the problems of traditional methods, such as inability to adaptively adjust the threshold value, dependence on experience for identification effect, and boundary extraction susceptible to environmental interference, and has the advantages of high automation, strong adaptability, and strong generalizability, and is especially suitable for ice lake area dynamic monitoring tasks under complex terrain conditions in plateau areas. BRIEF DESCRIPTION OF DRAWINGS

[0060] The present application will be further described below in conjunction with the drawings and examples.

[0061] Figure 1 The technical flowchart of the ice lake remote sensing monitoring method of the present application.

[0062] Figure 2 The ice lake boundary extraction and ice water classification results in Example 1. DETAILED DESCRIPTION

[0063] Example 1:

[0064] A remote sensing monitoring method for ice lakes, in this embodiment, the monitoring area is the Selin Co ice lake, which is a large ice lake located on the Qinghai-Tibet Plateau and belongs to a typical plateau lake. The ecological environment around the lake is fragile and sensitive, and is significantly affected by climate change, glacier retreat, and water level changes.

[0065] Remote sensing data introduction:

[0066] In embodiment 1, the optical remote sensing image data used is Sentinel-2, and the data source is the public remote sensing image released by the European Space Agency (ESA), which can be obtained and processed for free. The imaging dates are November 15, 2023, December 21, 2023, January 14, 2024, and January 21, 2024. The Sentinel-2 satellite carries a multi-spectral imager (MSI) that can obtain 13-band remote sensing data, covering a wide range of wavelengths from visible light, near-infrared to short-wave infrared, with a spatial resolution of 10 meters, which can accurately capture the lake boundary and its changes, especially during the winter and spring seasons, providing detailed change information of ice lakes. The specific data source is the public remote sensing image released by the European Space Agency (ESA), which can be obtained and processed for free. This allows researchers to use these image data for multi-temporal analysis in the Selin Co ice lake monitoring, to track real-time changes in ice and water, area changes, and other key information of the lake.

[0067] Digital Elevation Model (DEM) is a model that digitally represents the height of the ground, widely used in Geographic Information System (GIS). In embodiment 1, the DEM data comes from ASTER GDEM (Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Digital Elevation Model). ASTER GDEM is a global elevation data set released by NASA and the University of Tokyo in Japan in cooperation with the use of ASTER satellite remote sensing image data. The spatial resolution of this data set is 30 meters, covering the global range, including polar regions, with high data accuracy.

[0068] The following are the specific steps of embodiment 1:

[0069] 1) Data acquisition and preprocessing:

[0070] Obtain multi-source optical remote sensing images of different time periods covering the target ice lake area, and perform atmospheric correction and geometric correction. Obtain DEM data of the target ice lake area.

[0071] First, download four Sentinel-2 Level-1C images covering Silingcuo Glacier Lake from the Copernicus Open Access Hub platform. The images were taken on November 15, 2023, December 21, 2023, January 14, 2024, and January 21, 2024, respectively.

[0072] The downloaded Sentinel-2 Level-1C imagery was first atmospherically corrected using the Sen2Cor tool to convert it into a surface reflectance product (Level-2A). Sen2Cor, based on the physical radiative transfer model, can effectively remove interference from atmospheric aerosols and water vapor, improving the accuracy of spectral indicators such as the CI index.

[0073] Subsequently, geometric fine corrections were performed on the four images to reduce spatial deviations between them. To ensure spatial consistency among the multi-temporal images, a ensemble control point (GCP) registration method based on Google Earth high-resolution imagery was adopted. Each image underwent secondary geometric correction and sub-pixel-level registration to ensure perfect alignment at the pixel level, thereby improving the spatial accuracy of subsequent glacial lake boundary extraction and change detection.

[0074] 2) Construction of Composite Remote Sensing Index (CI):

[0075] The Normalized Difference Water Index (NDWI) and Normalized Difference Snow Index (NDSI) are combined by their ratio to enhance the water body's response while reducing the confusion effect between snow cover and ice. In this step, CI construction first extracts three bands (Band 3, Band 8, and Band 11) from four preprocessed Sentinel-2 images, calculating the NDWI and NDSI values ​​for each image. Then, CI values ​​are calculated sequentially at the pixel level. To further enhance the sensitivity of CI to the water-ice transition characteristics of typical glacial lake areas, the Selin Co Lake area is divided into five typical sub-regions during processing: the southwest shore mudflat zone, the northeast bay, the northern part of the main lake area, the glacial lake inlet, and the nearshore transition zone. 500–800 representative pixels are selected from each sub-region for subsequent extraction of the typical CI range and verification of the ice-water response effect.

[0076] 3) Glacial lake area extraction based on CI index:

[0077] The CI layer is preprocessed by standardization. The Otsu method is used to perform threshold segmentation on the standardized CI image to obtain the optimal ice lake boundary threshold t, and the overall ice lake region including the ice surface and water body is extracted.

[0078] Based on the CI layer generated in step 2), the overall coverage of ice lakes in the Cele Lake region is extracted, including the combined area of water bodies and ice surfaces. First, the CI layer is subjected to effective pixel screening to remove pixels with CI values less than 0, excluding invalid information interference caused by cloud shadows, bare land, snow, and other high reflectance interference. Subsequently, the CI values of all effective pixels are linearly normalized to fall within the value range of 0 to 1. Then, the Otsu maximum inter-class variance method is used to binarize the CI layer to automatically determine the optimal threshold t for the ice lake boundary. The final binary mask layer identifies all pixels with CI greater than the threshold t as ice lake areas, forming an ice lake extraction mask layer covering the entire Cele Lake region.

[0079] 4) DEM data downscaling and pixel matching processing:

[0080] The original 30-meter DEM is down-scaled using the Cubic Interpolation method. For each target pixel of 10m resolution to be generated, based on its 16 surrounding 30m pixels, a cubic interpolation function is constructed in the x and y directions to obtain the elevation value at the target location.

[0081] To ensure the spatial registration accuracy of the resampled data, the projection information of the Sentinel-2 image is used as a standard reference to perform a collective correction on the interpolated DEM, making it completely consistent with the CI layer in the projection coordinate system and spatial distribution. The final output of the 10m resolution DEM is aligned with the CI layer at the pixel level, with a spatial deviation controlled within one pixel, providing a high-precision fusion basis.

[0082] 5) Dynamic calibration of ice-water classification threshold based on DEM

[0083] First, the Jenks Natural Breaks method is used to automatically partition the elevation data in the Cele Lake region. According to prior knowledge of the Cele Lake, the elevation interval is divided into 12 levels to maximize the homogeneity of the terrain within the interval and enhance the expression of the elevation difference between intervals. Subsequently, the corresponding pixel set in the CI layer is extracted within each elevation interval. To improve classification accuracy, at least 200 effective pixels are selected in each interval. Considering the regional distribution and spectral stability, the K-means clustering algorithm is used to divide the CI values into two categories, representing ice and water pixels, respectively, and determine the optimal ice-water classification CI threshold for each elevation interval.

[0084] The support vector regression (SVR) model is used to construct a mapping function between the ice-water classification threshold and the elevation. The center elevation value of the elevation interval and the corresponding CI threshold are input into the model as training samples, and the radial basis function (RBF) kernel function is used for nonlinear fitting to capture the complex relationship between the ice-water reflectivity under different elevation conditions. During the model training process, the kernel width σ is 0.3, γ is 5.56, and the penalty coefficient C is set to 100. The model performs stably in five-fold cross-validation, and the fitting accuracy R 2 is 0.92, indicating that there is a good mapping relationship between the elevation and the threshold.

[0085] 6) Fine classification of ice surface and water surface

[0086] Through the elevation- CI threshold mapping relationship established in step 5), the fine classification of ice surface and water surface for each pixel in the ice lake image is realized, and the results are shown in the accompanying Figure 2 Figures (a)-(d) represent the extraction results of Selinco Lake on November 15, 2023, December 21, 2023, January 14, 2024, and January 21, 2024, respectively. It can be seen that the ice lake outlines of different dates are relatively clear, and the extraction results of different periods are analyzed: Figure (a) shows the ice-water classification extraction in the non-freezing period, when the lake surface has not yet frozen, and the water extraction result is accurate; Figure (b) shows the ice-water classification of the lake surface in the early freezing period, and it can be seen that the lake has begun to freeze, and the ice-water boundary is clear and the classification is accurate in the relatively low-lying areas such as the north, south and east coasts of the lake; Figure (c) shows the ice-water classification of the lake surface in the freezing period, and it can be seen that only the lake surface in the west coast and other high-lying areas still has some water surface, and the ice-water boundary is clear and the classification is accurate; Figure (d) shows the lake surface in the complete freezing stage. From the above analysis, it can be seen that the overall monitoring of the ice lake monitoring work of the present application for Example 1 is relatively accurate, and reflects the ice-water change of Selinco Lake in the freezing period.

[0087] Finally, it should be noted that the above is only used to illustrate the technical solutions of the present application and is not limiting. Although the present application has been described in detail with reference to the preferred arrangement, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for remote sensing monitoring of glacial lakes, characterized in that: Includes the following steps: 1) Data acquisition and preprocessing: Acquire multi-source optical remote sensing images covering the target glacial lake area at different time phases, and perform atmospheric and geometric corrections; and acquire DEM data of the target glacial lake area; 2) Construction of composite remote sensing index: The composite remote sensing index CI is constructed based on the ratio of normalized water index NDWI to normalized snow difference index NDSI; the CI layer can be obtained by calculating the pixels of the image based on the formula of remote sensing index CI. 3) Ice lake area extraction based on CI index: The CI layer is standardized and preprocessed. The Otsu's maximum inter-class variance method is used to perform threshold segmentation on the standardized CI image to obtain the optimal ice lake boundary threshold and extract the overall ice lake area including the ice surface and water body. 4) DEM data downscaling and cell matching processing: The cubic interpolation method is used to resample and downscale the DEM data to achieve cell-level spatial alignment with the CI layer; 5) Dynamic calibration of ice-water classification threshold based on DEM: Based on the downscaled DEM layer, the regional elevation is segmented using the natural breakpoint method, dividing it into several elevation intervals. The corresponding CI value set is extracted in each interval, and the CI values ​​are divided into ice and water using the K-means clustering algorithm. The optimal ice-water classification CI threshold is determined for each elevation interval. A sample set between the center elevation of each interval and the CI threshold is constructed, and the elevation-CI threshold mapping function is established using the support vector regression (SVR) method. 6) Fine classification of ice and water surfaces: Based on the elevation-CI threshold mapping function, for each pixel in the ice lake area, its elevation value is extracted and substituted into the function to calculate the corresponding CI threshold, which is then compared with the pixel CI value to achieve ice and water classification.

2. The method for remote sensing monitoring of glacial lakes according to claim 1, characterized in that: The formula for the CI value in step 2) is as follows: In the formula: Green, NIR, and SWIR represent the reflectance of the green band, near-infrared band, and short-wave infrared band of remote sensing data, respectively.

3. The method for remote sensing monitoring of glacial lakes according to claim 1, characterized in that: In step 3), the steps for standardizing CI include: removing pixels with CI values ​​less than 0, retaining valid data and then performing normalization processing to stretch the CI values ​​to the range of 0 to 1.

4. The method for remote sensing monitoring of glacial lakes according to claim 1, characterized in that: In step 4), the 30m resolution DEM data is specifically matched with the 10m resolution CI layer at the pixel level.

5. The method for remote sensing monitoring of glacial lakes according to claim 1, characterized in that: In step 4), the cubic interpolation method is as follows: a cubic fitting function is constructed at the pixel to be interpolated using the surrounding DEM values ​​to complete the spatial refinement of the elevation data and maintain the spatial consistency of the remote sensing data.

6. The method for remote sensing monitoring of glacial lakes according to claim 1, characterized in that: In step 5), the SVR model is constructed using the radial basis function (RBF). The model parameters, including the penalty factor C and the kernel function parameter γ, are automatically determined through K-fold cross-validation.

7. The method for remote sensing monitoring of glacial lakes according to claim 1, characterized in that: In step 6), for each image pixel to be classified, its corresponding elevation value h is extracted. i Based on the elevation-CI threshold mapping function, the dynamic classification CI threshold T that it matches is calculated. i ; set the threshold T i CI value C of the current pixel i Make a comparison and make a judgment based on the following judgment principles: Based on the above dynamic classification, the ice and water classification results of each pixel in the entire remote sensing image are obtained, and the distribution map of ice lake ice and water is obtained pixel by pixel.

Images