ICESat-2-oriented ice surface elevation large-range extraction method

Abstract

The invention discloses an ICESat-2-oriented ice surface elevation large-range extraction method, which is characterized in that a surface trend line is constructed by using a photon weight local maximum value, and simple and complex form region division is carried out in combination with a roughness index. For a complex form region, determining a photon ellipse neighborhood direction by using an RANSAC algorithm, calculating a photon similarity density by using a Gaussian kernel function, and extracting signal photons through an along-track adaptive density threshold; for a simple form area, a model effective point obtained by an RANSAC algorithm is directly used as a signal photon point. According to the method, the problem of insufficient large-range signal photon extraction precision caused by factors such as surface form change and data quality along-rail difference can be better solved, the data processing efficiency can be effectively improved through simple and complex form region division and region self-adaptive ice surface elevation extraction, and then the method serves large-range ice surface elevation change monitoring.

Description

Technical Field

[0001] The present invention belongs to the field of spaceborne lidar photon point cloud processing, and particularly relates to a method for large-scale extraction of ice surface elevation for ICESat-2. Background Art

[0002] ICESat-2 (The Ice, Cloud and Land Elevation Satellite-2) is the world's first laser altimetry satellite using the photon counting mode. The Advanced Topography Laser Altimeter System (ATLAS) carried on the satellite uses a photon counting lidar with a wavelength of 532 nm, which has the characteristics of high repetition frequency and low pulse energy, and can quickly obtain high-precision and large-scale ground three-dimensional data. The 6-beam laser equipped with ATLAS is divided into 3 pairs according to the pulse energy strength, and can obtain a laser footprint with a diameter of 11 m, and the along-track resolution is as high as 0.7 m, and the spatial resolution has been significantly improved compared with ICESat, bringing opportunities for problems such as ice sheet elevation monitoring, glacier surface crack detection, and ice surface lake water depth inversion. At present, ICESat-2 ATL06 data is mainly used for large-scale monitoring of the glacier surface by spaceborne lidar. However, the along-track resolution of ICESat-2 ATL06 is 20 m, and its along-track resolution is too sparse to obtain fine glacier surface morphological features such as small cracks and ice surface lakes, thus affecting the measurement of the mass and dynamic balance of glaciers. Therefore, how to quickly and high-quality process the ICESat-2 raw data ATL03 is the key to large-scale ice surface elevation extraction and monitoring.

[0003] Photon counting lidar has high sensitivity and is easily affected by solar background noise, atmospheric scattering and instrument dark counting, resulting in a large number of noise photons in the ICESat-2 ATL03 data. At present, the main methods for photon point cloud denoising can be summarized into the following three categories: 1) denoising algorithms based on image processing; 2) denoising algorithms based on density; 3) denoising algorithms based on local statistical parameters. These methods have high quality under specific terrains, but when facing large-scale processing, the algorithm performance will be affected by terrain changes and data quality, that is, the accuracy is not high. In addition, the strategy of uniformly processing different terrain data by existing methods will also consume a lot of time, that is, the efficiency is not high. Summary of the Invention

[0004] The object of the present invention is to provide a method for large-scale extraction of ice surface elevation for ICESat-2, aiming at the problems of low efficiency and accuracy in processing large-scale ICESat-2 ATL03 data in the prior art. By using the present invention, data in different morphological regions can be quickly divided, and the accuracy after adaptive step-by-step processing can be guaranteed, which helps to provide high-resolution surface features in the research of ice surface change process.

[0005] To solve the above technical problems, the present invention adopts the following technical solutions: A method for large-scale extraction of ice surface elevation for ICESat-2, comprising the following steps: Step 1: Extract the photon point data of each photon point in the study area of ICESat-2 ATL03 data, where the photon point data includes photon longitude and latitude, photon elevation, photon confidence, and photon weight; Step 2: Calculate the along-track distance of each photon point; Step 3: Screen photon points based on photon confidence; Step 4: Construct a surface trend line based on the screened photon points; Step 5: Divide each along-track segment divided along the track into simple morphological regions and complex morphological regions, where the along-track segment includes multiple along-track sections; Step 6: Obtain the extraction result of the ice surface elevation in the simple morphological region; Step 7: Obtain the extraction result of the ice surface elevation in the complex morphological region; Step 8: Stitch and merge the extraction results of the ice surface elevation of the along-track sections in different morphological regions.

[0006] The along-track distance of the photon point is calculated based on the following formula: , In the formula: represents the along-track distance of the i-th photon point, ) is the UTM projection coordinate of the starting photon point, ) is the UTM projection coordinate of the i-th photon point.

[0007] In step 3, photons with low, medium, and high confidence are selected as the screened photon points.

[0008] The construction of the surface trend line in step 4 includes the following steps: Segment the photon points along the track to obtain along-track segments. For each along-track segment, divide it into along-track sections; for each along-track section , find the photon with the maximum weight in each along-track section according to the photon weight extracted in step 1; then The maximum weight photon points of each along-track section are set as a set ; Select the set The maximum elevation point as the trend point, and connect all the trend points to complete the construction of the trend line.

[0009] The division of the simple morphology area and the complex morphology area in the step 5 includes the following steps: Calculate the distance from the maximum weight photon point in the along-track section to the trend line; The distance values from the maximum weight photon points of each along-track section are set as a set ; Calculate the surface roughness based on the following formula: ,[[]] In the formula: is the surface roughness, is the distance from the maximum weight photon point of the first along-track section to the th along-track section in the along-track section to the surface trend line, is the total number of along-track sections in the along-track segment, If the surface roughness of the along-track segment is greater than or equal to the roughness threshold, then all the along-track sections of this along-track segment are complex morphology areas; if the surface roughness of the along-track segment is less than the roughness threshold, then all the along-track sections of this along-track segment are simple morphology areas.

[0010] The extraction result of the ice surface elevation of the simple morphology area in the step 6 includes the following steps: Use the RANSAC algorithm to perform linear fitting on the photon points in the along-track section to obtain the inlier set in the along-track section is the ice surface elevation of the simple morphology area.

[0011] The extraction result of the ice surface elevation of the complex morphology area in the step 7 includes the following steps: Step 7.1: Take each photon point in the along-track section as the current central photon point , search for the photon points in its circular neighborhood, and perform RANSAC linear fitting on the photon points in the neighborhood to obtain the direction of the current central photon point in the circular neighborhood, change the circular neighborhood to an elliptical neighborhood, the major axis size of the ellipse is the diameter of the circle, the major axis direction of the ellipse is the direction of the current central photon in the circular neighborhood, and the minor axis of the ellipse is set to a fixed value; Step 7.2: Calculate the K-nearest neighbor distance of each photon point in the along-track section; Step 7.3: Define the average value of the distances between the photon points in the elliptical neighborhood and their nearest photon points as the photon spacing , evaluate the photon points within the elliptical neighborhood and the current central photon point to find the difference in photon spacing between them. The photon points are from the current central photon point and the set of photon points within the elliptical neighborhood . Then, use the one-dimensional Gaussian kernel function to calculate the similarity density of the current central photon point : , wherein: is the similarity density of the current photon center point . The photon points are from the current central photon point and the set of photon points within the elliptical neighborhood . is the standard deviation of the Gaussian kernel function, is the spatial distribution difference between the photon point within the elliptical neighborhood and the current central photon point , that is, the difference in photon spacing; Step 7.4: Determine the photon points with similarity density greater than the density threshold in each along-track section as signal photons, and the photon points with similarity density less than the density threshold as noise photons. Finally, use the signal photons screened by the density threshold as the ice surface elevation extraction result of the complex morphology area.

[0012] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the method for large-scale extraction of ice surface elevation for ICESat-2 according to any one of claims 1 to 7.

[0013] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps of the method for large-scale extraction of ice surface elevation for ICESat-2 according to any one of claims 1 to 7.

[0014] A computer program product includes a computer program, and when the computer program is executed by a processor, it implements the steps of the method for large-scale extraction of ice surface elevation for ICESat-2 according to any one of claims 1 to 7.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) High precision: The surface roughness calculated by the present invention can reflect the geometric feature differences between the simple morphology area and the complex morphology area of the glacier surface, and then set a threshold to achieve high-precision division of the simple and complex morphology areas.

[0016] (2) High efficiency: The present invention divides the data of different morphological regions through the roughness index, and formulates corresponding denoising schemes according to the data characteristics of different morphological regions, effectively solving the problem of low data processing efficiency of ICESat-2 ATL03, and then serving the extraction and monitoring of ice surface elevation over a large area. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the study area and data, where (a) is the location of Jakobshavn Glacier in the Greenland Ice Sheet, (b) is the distribution of ICESat-2 data orbits, (c) is the monthly ICESat-2 data volume from 2019 to 2023, (d) is a schematic diagram of confidence photons, and (e) is a schematic diagram of photon weights; Figure 2 It is a schematic flow chart of the present invention; Figure 3 It is a schematic diagram for constructing the surface trend line. Among them, (a) is a schematic diagram of extracting the maximum weight photons at a 3-meter segment along the track in a complex morphological region, (b) is a schematic diagram of extracting the maximum elevation photons at a 150-meter segment along the track in a complex morphological region and calculating the trend line, (c) is a schematic diagram of extracting the maximum weight photons at a 3-meter segment along the track in a simple morphological region, (d) is a schematic diagram of extracting the maximum elevation photons at a 150-meter segment along the track in a simple morphological region and calculating the trend line, and (e) is Figure 3 A line description diagram of (a)-(d) in Figure 4 It is a schematic diagram for dividing simple and complex morphological regions. Among them, (a) is a roughness result diagram of data containing complex morphological regions, and (b) is a roughness result diagram of data with only simple morphological regions; Figure 5 It is a schematic diagram for extracting the ice surface elevation with regional adaptability. Among them, (a) is a result diagram of extracting the ice surface elevation in a simple morphological region, and (b) is a result diagram of extracting the ice surface elevation in a complex morphological region; Figure 6 It is a partial schematic diagram of the result of large-scale extraction of the ice surface elevation of ICESat-2 ATL03 data. Among them, (a)-(f) are result diagrams of extracting the ice surface elevation in different morphological regions and with different data qualities. The specific data descriptions are as follows: (a) is a simple morphological region with good data quality, (b) is a simple morphological region with poor data quality, (c) contains simple and complex morphological regions with good data quality, (d) contains simple and complex morphological regions with good data quality, (e) is a simple morphological region with poor data quality, and (f) is a simple morphological region with good data quality. DETAILED DESCRIPTION OF THE INVENTION

[0018] For the convenience of those of ordinary skill in the art to understand and implement the present invention, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0019] Embodiment 1

[0020] A method for large-scale extraction of ice surface elevation for ICESat-2 specifically includes the following steps: Step 1: Data reading and cropping. The ICESat-2-ATL03 data is stored in the HDF5 format. Use the Python h5py package to extract the photon point data of each photon point from the ICESat-2-ATL03 data in the HDF5 format. The photon point data includes photon longitude and latitude, photon elevation, photon confidence, and photon weight, and save it as a text file. According to the photon longitude and latitude, use the nearest neighbor algorithm to obtain the intersection points of the photon trajectory and the boundary of the study area, so as to obtain the photon point data within the study area, including photon longitude and latitude, photon elevation, photon confidence, and photon weight.

[0021] Step 2: Calculate the along-track distance of photon points. The photon point data obtained in Step 1 uses the geographic coordinate system as a reference, which is not convenient for distance calculation. Therefore, first convert the photon point cloud within the study area into the UTM projection coordinate system, and use the starting photon point within the study area as the origin to calculate the along-track distance of the photon points; Formula (1) In the formula: represents the along-track distance of the i-th photon point, ) is the UTM projection coordinate of the starting photon point, ) is the UTM projection coordinate of the i-th photon point.

[0022] Step 3: Screen photon points based on photon confidence. The photon point cloud contains a large amount of noise. The ICESat-2 ATL03 data evaluates the deviation degree of each photon point from the ground surface according to the reference DEM, and assigns confidence according to the deviation degree (the smaller the deviation of the photon point from the reference DEM, the higher the confidence). The confidence is divided into 5 levels, where: 0 represents noise photons; 1 represents extended signal photons, that is, a possible signal photon area is added outside the low, medium, and high confidence photon ranges; 2 represents low confidence photons; 3 represents medium confidence photons; 4 represents high confidence photons. According to the photon confidence of each photon point read in Step 1, select low, medium, and high confidence photons as the screened photon points to eliminate a large number of noise photons far from the ground surface, and all subsequent steps are processed based on the screened photon points.

[0023] Step 4: Construction of surface trend line. After screening the photon points, in order to reduce the influence brought by terrain changes, the photon points are segmented along the track to obtain segments along the track. For each segment along the track, it is divided into sections along the track. For each section along the track , the photon with the maximum weight within each section along the track is obtained according to the photon weights extracted in Step 1; then, the photon points with the maximum weight of the sections along the track are set as set ; finally, the highest elevation point of set is selected as the trend point. After calculating the trend points of each segment along the track, connecting all the trend points can complete the construction of the trend line.

[0024] In this embodiment, takes the value of 50.

[0025] Step 5: Roughness estimation. According to the trend line constructed in Step 4, first calculate the distance from the photon point with the maximum weight within the section along the track to the trend line; then, the distance values from the photon points with the maximum weight of the sections along the track to the trend line are set as set

[0026] Formula (2) In the formula: is the surface roughness, is the distance from the photon point with the maximum weight in the first section along the track to the surface trend line within the segment along the track to the th section along the track, is the total number of sections along the track within the segment along the track. In this embodiment, takes the value of 50.

[0027] Generally speaking, in the area with complex morphology, the terrain undulates greatly, and the distance from the photon point with the maximum weight within the section along the track to the trend line is relatively large. Therefore, the calculated roughness is larger than that in the area with simple morphology. According to this characteristic, a roughness threshold is set to divide the area with simple morphology and the area with complex morphology. If the surface roughness of the segment along the track is greater than or equal to the roughness threshold, all sections along the track of this segment along the track are areas with complex morphology; if the surface roughness of the segment along the track is less than the roughness threshold, all sections along the track of this segment along the track are areas with simple morphology; thus realizing the division of the area with simple morphology and the area with complex morphology, providing a basis for the rapid extraction of ice surface elevation.

[0028] Step 6: Extraction of ice surface elevation in the area with simple morphology.

[0029] Use the RANSAC (Random Sample Consensus) algorithm to perform linear fitting on the photon points within the along-track section, and obtain the set of inliers within the along-track section , and use the set of inliers as the extraction result of the ice surface elevation in the simple morphology area.

[0030] Step 7: Extraction of the ice surface elevation in the complex morphology area.

[0031] Step 7.1: Use the RANSAC algorithm to determine the elliptical neighborhood of the photons. Take each photon point within the along-track section as the current central photon point , search for the photon points within its circular neighborhood, and perform RANSAC linear fitting on the photon points within the neighborhood to obtain the direction of the current central photon point within the circular neighborhood. Generally speaking, the signal photon points are densely distributed along the photon direction. To suppress the influence of noise photons in other directions on the subsequent estimation of the signal photon density, after obtaining the photon direction, replace the circular neighborhood with an elliptical neighborhood. Among them, the major axis size of the ellipse is the diameter of the circle, the major axis direction of the ellipse is the direction of the current central photon within the circular neighborhood, and the minor axis of the ellipse is set to a fixed value.

[0032] Step 7.2: Calculate the K-nearest neighbor distance of each photon point.

[0033] Due to the large amount of photon point cloud data, use KdTree to construct a photon space index, and calculate the K-nearest neighbor distance of each photon point within the along-track section, as shown in formula (3): Formula (3) In the formula:[[]] represents the total distance from the th photon point within the along-track section to the nearest photons around it. Among them, the nearest th photon point's nearest photons are sorted according to the calculated Euclidean distance. and respectively represent the along-track distances of the th photon point and the jth photon point, and respectively represent the photon elevations of the th photon point and the jth photon point, is an important parameter for calculating , that is, the number of neighboring photons selected. Substantially, it is a smoothing factor, and its function is to smooth the spatial heterogeneity within the same category of photons. If is too small, the smoothing effect is not obvious, and the sum of the distances between signal photons and noise photons will be relatively similar, making it difficult to distinguish; if If it is too large, the difference between the noise photons and the signal photons will be over-smoothed, making the photon denoising incomplete.

[0034] Step 7.3: Calculate the photon similarity density using the Gaussian kernel function.

[0035] After calculating the K-nearest neighbor distance, define the average value of the distances between the photon points within the elliptical neighborhood and their K nearest photon points as the photon spacing (i.e., K-nearest neighbor distance / ), and evaluate the difference in photon spacing between the photon points within the elliptical neighborhood and the current central photon point . The photon point comes from the set of photon points within the elliptical neighborhood of the current central photon point . Then, calculate the similarity density of the current central photon point using the one-dimensional Gaussian kernel function, as shown in formula (4).

[0036] Formula (4) In the formula: is the similarity density of the current photon center point . The photon point comes from the set of photon points within the elliptical neighborhood of the current central photon point . is the standard deviation of the Gaussian kernel function, is the spatial distribution difference between the photon point within the elliptical neighborhood and the current central photon point , that is, the difference in photon spacing.

[0037] Step 7.4: Calculate the along-track adaptive density threshold and screen out the signal photons, and extract the ice surface elevation of the screened signal photons.

[0038] Generally speaking, there is a density difference between the signal photons and the noise photons in each along-track section. The density threshold is set as the average value of the similarity densities of all photon points within the along-track section. Determine the photon points with similarity density greater than the density threshold within each along-track section as signal photons, and the photon points with similarity density less than the density threshold as noise photons. Finally, use the signal photons screened by the density threshold as the extraction result of the ice surface elevation in the complex terrain area.

[0039] Step 8: Merging the ice surface elevation extraction results of different morphological regions. After the region-adaptive ice surface elevation extraction, to ensure the overall coherence of the data results, the ice surface elevation extraction results of the along-track sections in different morphological regions are spliced and merged according to the along-track distance, thus completing the large-scale ice surface elevation extraction method for ICESat-2.

[0040] To illustrate the effectiveness of the present invention, all the data of Jakobshavn Isbræ in Greenland from 2019 to 2023 were selected for testing. Figure 1 Fig. 1 is a schematic diagram of the study area and data. There were 2019 ICESat-2 ATL03 data in Jakobshavn Isbræ from 2019 to 2023, all of which were version V6. Table 1 is a parameter setting description table for the region-adaptive ice surface elevation extraction method.

[0041] Table 1

[0042] The present invention provides relatively accurate simple and complex morphological region divisions and ice surface elevation results. The data from January to March 2022 in Jakobshavn Isbræ were selected for visual inspection. There were 104 data passing through Jakobshavn Isbræ from January to March 2022. The overall evaluation of the simple and complex morphological region divisions and ice surface elevation extraction is shown in Table 2. Table 2 is the accuracy evaluation table for the complex morphological region detection results of the present invention.

[0043] Table 2

[0044] Figure 4 In (a), it is the roughness estimation of the data containing complex morphological regions. The 20 km data region only took 11 s; Figure 4 In (b), it is the roughness estimation of the data with only simple morphological regions. The 200 km data region only took 100 s. Table 3 is the elevation extraction time table for the data containing complex morphological regions ( Figure 4 In (a)) and the efficiency comparison of using and not using the region-adaptive ice surface elevation extraction method.

[0045] Table 3

[0046] From Table 3, Figure 6 it can be seen that after adopting the present invention, the efficiency is significantly improved. Thus, it can be seen that the present invention can meet the large-scale ice surface elevation extraction of ICESat-2.

[0047] In addition, according to Figure 4 in (a), Figure 4In (b) of the present invention, the selected roughness threshold has an obvious difference between the simple morphology area and the complex morphology area, and can accurately divide the simple and complex morphology areas.

[0048] Those of ordinary skill in the art can understand that all or part of the processes in the above-described method embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-described method embodiments.

[0049] Embodiment 2 In this embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above-described Embodiment 1 are implemented.

[0050] Embodiment 3 In this embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-described Embodiment 1 are implemented.

[0051] Embodiment 4 In this embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above-described Embodiment 1 are implemented.

[0052] It should be noted that the embodiments described in the present invention are only illustrative of the spirit of the present invention. Those skilled in the art to which the present invention pertains can make various modifications or supplements to the described embodiments or use similar ways to replace them, but will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A method for large-scale extraction of ice surface elevation for ICESat-2, characterized in that, It includes the following steps: Step 1: Extract the photon point data of each photon point in the study area from ICESat-2 ATL03 data. The photon point data includes photon longitude and latitude, photon elevation, photon confidence, and photon weight; Step 2: Calculate the along-track distance of each photon point; Step 3: Screen photon points based on photon confidence; Step 4: Construct a surface trend line based on the screened photon points; Step 5: Divide each along-track segment divided along the track into a simple morphology area and a complex morphology area. The along-track segment includes multiple along-track sections; Step 6: Obtain the extraction result of the ice surface elevation in the simple morphology area; Step 7: Obtain the extraction result of the ice surface elevation in the complex morphology area; Step 8: Stitch and merge the extraction results of the ice surface elevation of the along-track sections in different morphology areas.

2. The large-scale extraction method of ice surface elevation for ICESat-2 according to claim 1, characterized in that The along-track distance of the photon point is calculated based on the following formula: ， In the formula: represents the along-track distance of the i-th photon point, ) is the UTM projection coordinates of the starting photon point, ) is the UTM projection coordinates of the i-th photon point.

3. The large-scale ice surface elevation extraction method for ICESat-2 according to claim 1, characterized in that In Step 3, select low, medium, and high-confidence photons as the screened photon points.

4. A method for large-scale extraction of ice surface elevation for ICESat-2 according to claim 1, characterized in that, In Step 4, the construction of the surface trend line includes the following steps: Segment the photon points along the track to obtain segments along the track. For each segment along the track, divide it into sub - segments along the track; for each sub - segment along the track , obtain the photon with the maximum weight within each sub - segment along the track according to the photon weights extracted in step 1; then set the photon points with the maximum weight of sub - segments along the track as a set ; select the point with the maximum elevation in the set as the trend point, and connect all the trend points to complete the construction of the trend line.

5. A method for large-scale extraction of ice surface elevation for ICESat-2, characterized in that, In Step 5, the division of the simple morphology area and the complex morphology area includes the following steps: Calculate the distance from the photon point with the maximum weight within the along-track section to the trend line; set the distance values of the photon points with the maximum weight in along-track sections to the trend line as a set ; calculate the surface roughness based on the following formula: ， Wherein: is the surface roughness, is the distance from the maximum weighted photon point in the first along-track section to the th along-track section within the along-track section to the surface trend line, is the total number of along-track sections within the along-track section, If the surface roughness along the track segment is greater than or equal to the roughness threshold, then the track-aligned sections along this track segment are all complex morphology regions; if the surface roughness along the track segment is less than the roughness threshold, then the track-aligned sections along this track segment are all simple morphology regions.

6. The method for extracting large-scale ice surface elevation for ICESat-2 according to claim 1, wherein, In Step 6, obtaining the extraction result of the ice surface elevation in the simple morphology area includes the following steps: Use the RANSAC algorithm to perform linear fitting on the photon points within the along-track section to obtain the set of inliers within the model in the along-track section is the ice surface elevation of the simple morphological area.

7. A method for large-scale extraction of ice surface elevation for ICESat-2 according to claim 1, characterized in that In Step 7, obtaining the extraction result of the ice surface elevation in the complex morphology area includes the following steps: Step 7.1: Take each photon point within the along-track section as the current central photon point , search for the photon points within its circular neighborhood, and perform RANSAC linear fitting on the photon points within the neighborhood to obtain the current central photon point In the direction within the circular neighborhood, replace the circular neighborhood with an elliptical neighborhood. The major axis size of the ellipse is the diameter of the circle, and the major axis direction of the ellipse is the direction of the current central photon in the circular neighborhood, and set the minor axis of the ellipse to a fixed value; Step 7.2: Calculate the K-nearest neighbor distance of each photon point within the along-track section; Step 7.3: Define the average distance between a photon point within the elliptical neighborhood and its nearest photon points as the photon spacing , and evaluate the difference in photon spacing between the photon points within the elliptical neighborhood and the current central photon point . The photon points are from the set of photon points within the elliptical neighborhood of the current central photon point . Then, use a one-dimensional Gaussian kernel function to calculate the similarity density of the current central photon point : ， Wherein: is the current photon center point is the similarity density of the photon point from the current central photon point is the set of photons within the elliptical neighborhood , is the standard deviation of the Gaussian kernel function, is the photon point within the elliptical neighborhood and the current central photon point is the spatial distribution difference, that is, the difference in photon spacing; Step 7.4: Determine the photon points with a similarity density greater than the density threshold within each along-track section as signal photons, and the photon points with a similarity density less than the density threshold as noise photons. Finally, use the signal photons screened by the density threshold as the extraction result of the ice surface elevation in the complex morphology area.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for large-scale extraction of ice surface elevation for ICESat-2 described in any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for large-scale extraction of ice surface elevation for ICESat-2 described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for large-scale extraction of ice surface elevation for ICESat-2 described in any one of claims 1 to 7.

Images