An ice surface crack depth automatic extraction method and system based on linear cloth simulation

Abstract

This invention discloses an automatic method and system for extracting ice surface crack depth based on linear cloth simulation. The method includes: acquiring photon positioning data and photon classification data of the target area; generating target photon data and particle data through data preprocessing and meshing; simulating cloth drop based on particle data using a preset dynamic model combined with gravity to obtain a cloth height curve; classifying the denoised photon data based on the cloth height curve to determine the photon category and crack segment; and quantizing the crack segment depth information based on elevation-aided data. This invention adaptively fits the terrain through cloth simulation of the physical process, overcoming the limitations of traditional methods that rely on fixed resolution, manual thresholds, and empirical parameters. It enhances the detection stability and automation level in areas with weak or dense cracks, providing more reliable three-dimensional structural data for ice shelf stability assessment, and can be widely applied in the field of laser echo data detection technology.

Description

Technical Field

[0001] This invention relates to the field of laser echo data detection technology, and in particular to an automatic method and system for extracting the depth of ice surface cracks based on linear cloth simulation. Background Technology

[0002] The Antarctic ice sheet, the world's largest land ice mass, extends into the ocean, forming a vast floating ice shelf structure. Ice shelves play a crucial role in supporting land-based ice flows, regulating the mass balance of the ice sheet, and maintaining global sea-level stability. The integrity of the ice shelf directly affects its "supporting effect" on the inland ice sheet; a large-scale fracturing or collapse of the ice shelf often leads to accelerated inland ice flows into the sea, thus significantly impacting global sea-level rise.

[0003] Against the backdrop of global warming, cracks forming on and within ice shelves are developing at an accelerated pace. As one of the direct precursors to the eventual disintegration of ice shelves, the three-dimensional structural information, such as the depth and width of cracks, is crucial for assessing ice shelf stability. In particular, when cracks penetrate the entire thickness of the ice shelf, they significantly reduce the mechanical integrity of the ice shelf and may trigger the rapid separation of large icebergs. Therefore, obtaining the true depth parameters of cracks is a core requirement in current polar science, glacier dynamics, and ice shelf stability research.

[0004] However, traditional remote sensing methods still have fundamental limitations in accurately depicting the vertical structure of cracks. While optical and radar sensors can effectively depict the location and orientation of cracks, they cannot directly obtain the vertical morphology of cracks because optical images cannot penetrate the ice and radar echoes are strongly affected by crack width, geometry, and scattering characteristics. Therefore, they are difficult to use for accurate crack depth inversion. Early laser altimetry systems had higher spatial resolution than radar, but they could only identify large-scale crack structures and were still insufficient for the continuous vertical features of most ice shelf cracks.

[0005] Research based on this type of data has been able to obtain spatial distribution, temporal evolution, and some depth information of fractures. However, existing methods often rely on optical imagery, fixed-scale segmented elevation products, or threshold adjustments using multiple empirical parameters. These methods generally suffer from insufficient versatility: they are difficult to adapt to variations in fracture scale across different ice shelf regions, and they struggle to operate stably in situations where fracture development is subtle or photon distribution is sparse. Furthermore, some automated detection methods based on segmented elevation data are limited by fixed spatial resolution, often failing to capture subtle vertical changes in small-scale or densely populated fracture zones. While some methods have attempted to improve the precision of fracture depth extraction using photon-level data in recent years, they still rely heavily on manually set threshold parameters, requiring repeated adjustments for different regions and fracture types, making it difficult to achieve universal adaptability to the diverse fracture environments of Antarctic ice shelves. Overall, existing methods largely depend on direct detection of surface depressions or empirical rules to define fracture boundaries and depths, and their robustness significantly decreases when fracture morphology is atypical, development is weak, or environmental conditions are complex. Summary of the Invention

[0006] The main objective of this invention is to propose an automatic extraction method, system, electronic device, storage medium, and program product for ice surface crack depth based on linear cloth simulation, aiming to solve at least one problem of the prior art.

[0007] To achieve the above objectives, one aspect of this invention proposes an automatic method for extracting ice surface crack depth based on linear cloth simulation, the method comprising:

[0008] The system acquires photon positioning and classification data for the target area, and generates target photon data and particle data through data preprocessing and meshing. The target photon data includes denoised photon data and elevation-aided data, while the particle data includes initialized cloth particles and terrain height mask particles corresponding to each mesh cell.

[0009] Based on particle data, the fabric drop simulation was performed by combining a preset dynamic model with the effect of gravity to obtain the fabric height curve.

[0010] Based on the fabric height curve, the denoised photon data is classified to obtain the photon category and determine the crack segment.

[0011] The depth information of the fracture segment was obtained by quantification based on elevation-aided data.

[0012] In some embodiments, photon positioning data includes the time, latitude, longitude, and elevation information of photon events. Target photon data and particle data are generated through data preprocessing and gridding, including the following steps:

[0013] The photon classification labels in the photon classification data are used to filter out noisy photons from the photon localization data to obtain denoised photon data.

[0014] Based on the denoised photon data, scaled photon data is obtained by scaling the distance along the trace direction;

[0015] Based on photon positioning data, elevation auxiliary data is obtained by processing it using preset tools;

[0016] The scaled photon data is divided into grids at fixed intervals to obtain multiple grid cells;

[0017] Initialize the cloth particles corresponding to each grid cell based on the preset height;

[0018] Mask particles are initialized in each grid cell based on elevation-aided data.

[0019] In some embodiments, initializing mask particles in each grid cell based on elevation-aided data includes the following steps:

[0020] The elevation information corresponding to each grid cell is determined based on elevation auxiliary data;

[0021] Initialize the elevation of the mask particles corresponding to the corresponding grid cells based on the elevation information;

[0022] When there is a single mask particle with an empty elevation, the elevation of the corresponding mask particle is determined by linear interpolation based on the elevations of two adjacent mask particles.

[0023] If multiple mask particles have empty elevations, the elevation of the corresponding mask particles will be assigned a value of negative infinity.

[0024] In some embodiments, based on particle data, a fabric drop simulation is performed using a preset dynamic model combined with gravity to obtain a fabric height curve, including the following steps:

[0025] The cloth dynamic model is initialized using traction springs and bending springs; in the cloth dynamic model, the initial state of cloth particles in the moving state is movable.

[0026] Based on the fabric dynamic model combined with gravity to simulate the displacement of the falling fabric, the position of the fabric particles is iteratively updated every moment through an integral model;

[0027] Collision detection is performed on the result of each position update. If the height of the cloth particle after the update is less than or equal to the height of the mask particle, the height of the cloth particle is reset to the height of the mask particle, and the movement state of the cloth particle is updated to immovable.

[0028] The position iteration update stops when the height change of all cloth particles during the position iteration update is less than the first threshold, or when the number of iterations of the position iteration update reaches the second threshold.

[0029] The fabric height curve is constructed based on the spatial shape corresponding to the height of the fabric particles after the last position iteration update.

[0030] In some embodiments, photon classification is performed on the denoised photon data based on the fabric height curve to obtain photon categories and determine the crack segment, including the following steps:

[0031] Based on the height of fabric particles in the fabric height curve, a continuous fabric height model along the trajectory direction is generated by linear interpolation.

[0032] The first photon in the denoised photon data is taken as the candidate photon;

[0033] Based on the position of the candidate photon, extract the fabric particles from the fabric height model as candidate fabric;

[0034] If the height difference between the candidate photon and the candidate fabric is less than or equal to the third threshold, the candidate photon is classified as an ice surface photon; otherwise, the candidate photon is classified as a rift photon.

[0035] The next photon in the denoised photon data is taken as a candidate photon. The process of extracting fabric particles from the fabric height model based on the latitude and longitude of the candidate photon as candidate fabric is repeated until all photons in the denoised photon data have been traversed.

[0036] When the number of consecutive slit photons is greater than the fourth threshold, the corresponding consecutive segment is regarded as a slit segment.

[0037] The endpoints of the fractured segment are determined by the start and end positions of the continuous section.

[0038] In some embodiments, the fracture segment includes a continuous sequence of fracture photons and their corresponding location ranges. The depth information of the fracture segment is obtained by quantization based on elevation-aided data, including the following steps:

[0039] Gaussian filtering is applied to the elevation auxiliary data to smooth the elevation data.

[0040] Based on the location range corresponding to the fracture segment, the elevation sequence corresponding to the fracture segment is extracted from the elevation smoothing data;

[0041] Extreme value determination is performed on the high-order sequence. Based on the result of the extreme value determination, the fracture segment is divided into final fracture segments, and the final high-order sequence corresponding to each final fracture segment is determined according to the high-order sequence.

[0042] Obtain the endpoint elevations and minimum elevation within the segment from the final elevation sequence. Determine the depth of the final fracture segment based on the difference between the minimum endpoint elevation and the minimum elevation within the segment.

[0043] Using the endpoint of the final fracture segment as the base point, search for the corresponding point with the same elevation as the base point in the opposite direction of the base point in the final elevation sequence. Determine the width of the final fracture segment based on the positional distance between the base point and the corresponding point.

[0044] In some embodiments, extreme value determination is performed on the high-program sequence, and the fracture segment is divided into final fracture segments based on the result of the extreme value determination, including the following steps:

[0045] Iterate through the sequence of high-order segments to obtain the minimum and maximum values;

[0046] When the number of minimum values ​​is 0, the fracture segment is marked as an invalid fracture;

[0047] When the number of minimum values ​​is 1, the fracture segment is taken as the final fracture segment;

[0048] When the number of minima is greater than 1, the potential peak depth of the fracture is determined based on the difference between the maximum value and the minimum elevation within the fracture segment, and the potential fracture segment depth is determined based on the difference between the minimum elevation of the endpoints at both ends of the elevation sequence and the minimum elevation within the fracture segment.

[0049] When the ratio of the peak potential fracture depth to the potential fracture segment depth is less than the fifth threshold, the fracture segment is taken as the final fracture segment.

[0050] When the ratio of the potential fracture peak depth to the potential fracture segment depth is greater than or equal to the fifth threshold, all photons with height values ​​greater than the maximum value are deleted to update the target photon data and particle data. Then, the process returns to the step of simulating cloth falling based on particle data, using a preset dynamic model combined with gravity, until all fracture segments are split and the final fracture segment corresponding to each fracture segment is obtained.

[0051] To achieve the above objectives, another aspect of this invention proposes an automatic ice surface crack depth extraction system based on linear cloth simulation, the system comprising:

[0052] The first module is used to acquire photon positioning data and photon classification data of the target area, and generate target photon data and particle data through data preprocessing and gridding. The target photon data includes denoised photon data and elevation-aided data, and the particle data includes the initialized cloth particles and the mask particles of terrain height corresponding to each grid cell.

[0053] The second module is used to simulate the fabric falling based on particle data, using a preset dynamic model combined with gravity to obtain the fabric height curve.

[0054] The third module is used to classify the denoised photon data based on the fabric height curve to obtain the photon category and determine the crack segment.

[0055] The fourth module is used to quantify the depth information of the fracture segment based on elevation-aided data.

[0056] To achieve the above objectives, another aspect of the present invention provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned method.

[0057] To achieve the above objectives, another aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method.

[0058] To achieve the above objectives, another aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method.

[0059] The embodiments of this invention include at least the following beneficial effects: This invention provides a method, system, electronic device, storage medium, and program product for automatic extraction of ice surface crack depth based on linear cloth simulation. This scheme acquires photon positioning data and photon classification data of the target area, and generates target photon data and particle data through data preprocessing and meshing. The target photon data includes denoised photon data and elevation-aided data, and the particle data includes initialized cloth particles and terrain height mask particles corresponding to each mesh cell. Based on the particle data, cloth falling is simulated using a preset dynamic model combined with gravity to obtain a cloth surface height curve. Based on the cloth surface height curve, photon classification is performed on the denoised photon data to obtain photon categories and determine crack segments. The depth information of the crack segments is quantized based on the elevation-aided data. The embodiments of this invention adaptively fit the terrain through cloth simulation of the physical process, overcoming the limitations of traditional methods that rely on fixed resolution, manual thresholds, and empirical parameters. The beneficial effects of the embodiments of the present invention are as follows: it significantly improves the universality of depth extraction of cracks of different scales, morphologies and developmental degrees; through physical-driven adaptive simulation, it reduces the dependence on human intervention and empirical parameters, thereby enhancing the detection stability and automation level in areas with weak or dense cracks, and providing more reliable three-dimensional structural data for ice shelf stability assessment. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of an implementation environment for the automatic extraction method of ice surface crack depth based on linear cloth simulation provided in this embodiment of the invention;

[0061] Figure 2 This is a flowchart illustrating an automatic method for extracting ice surface crack depth based on linear cloth simulation, provided in an embodiment of the present invention.

[0062] Figure 3 This is a schematic diagram of the complete execution flow of the post-processing stage provided in the embodiments of the present invention;

[0063] Figure 4 This is a schematic diagram of the complete processing flow of the automatic extraction method for ice surface crack depth based on linear cloth simulation provided in the embodiments of the present invention;

[0064] Figure 5 This is a schematic diagram illustrating a conceptual example of the core working mechanism provided in the embodiments of the present invention;

[0065] Figure 6 This is a schematic diagram of an automatic ice surface crack depth extraction system based on linear cloth simulation provided in an embodiment of the present invention;

[0066] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of the present invention; they are merely examples of systems and methods consistent with some aspects of the embodiments of the present invention as detailed in the appended claims.

[0068] It is understood that the terms “first,” “second,” etc., used in this invention may be used herein to describe various concepts, but unless specifically stated otherwise, these concepts are not limited by these terms. These terms are used only to distinguish one concept from another. For example, first information may also be referred to as second information without departing from the scope of embodiments of the invention, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to determination” as used herein may be interpreted as “when…” or “when…” or “in response to determination.”

[0069] The terms “at least one,” “multiple,” “each,” “any,” etc., used in this invention, “at least one” includes one, two, or more than two; “multiple” includes two or more than two; “each” refers to each of the corresponding multiple; and “any” refers to any one of the multiple.

[0070] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein is for the purpose of describing embodiments of the invention only and is not intended to limit the invention.

[0071] To facilitate understanding of the technical solution of this invention, the meanings of the technical terms that may be involved in the technical solution of this invention will be explained first:

[0072] Antarctic ice shelves are floating ice masses that extend above the sea surface after a land ice sheet (or glacier) flows into the ocean. Ice shelves provide support to inland ice sheets through coupling with the inland ice mass, playing a crucial role in controlling ice flows, regulating ice sheet mass balance, and influencing global sea-level changes.

[0073] Ice-shelf fracture: This refers to the fracturing and formation of a crack in an ice shelf when the tensile or compressive stress in the ice exceeds its fracture strength and corresponding plastic deformation capacity. Depending on the depth, direction, and location of penetration, these fracture characteristics can be categorized as surface crevasses, basal crevasses (manifested on the surface of the ice shelf), and rifts that penetrate the entire ice shelf.

[0074] ICESat-2 Photon-Counting LiDAR Data: ICESat-2 is a laser altimeter satellite launched by NASA in 2018. It carries the ATLAS (Advanced Topographic Laser Altimeter System) and employs photon-counting ranging technology. At high orbital altitudes, it emits green light pulses at a high repetition rate and records the returning single-photon events, thereby obtaining dense photon-level echo points along the orbital direction. Key data products and highlights include:

[0075] ATL03 (Global Geolocated Photon Data): Raw geolocated photon data, containing the time, latitude and longitude, and ellipsoidal elevation of each received photon, is the main input source for LCSF-ice.

[0076] ATL08 (Land and Vegetation Height / Photon Classification): Classifies ATL03 photons (ground / vegetation / noise, etc.) to remove non-surface echoes and noise, thereby improving signal quality. The ATL08 classification results are then used for preprocessing noise reduction.

[0077] AE-10m (Average Elevation 10m): AE-10m is an average elevation product along the track generated by the ICESat-2 data processing workflow (e.g., it can be obtained using the SlideRule tool). Its core idea is to statistically calculate the effective photon heights falling within a fixed distance interval of 10 meters, using the raw photon observations of ICESat-2 laser altimetry data as the spatial resolution. The average value is typically used to obtain a smoother, more stable, and continuous ice surface elevation curve. It is a key auxiliary elevation data used in this invention for constructing the fabric simulation mask and calculating crack depth.

[0078] LCSF-ice (Linear Cloth Simulation Filtering for Ice Surface): LCSF-ice is an algorithmic framework for automatically separating ice surfaces from cracks and retrieving the 3D geometry (depth, width) of cracks from satellite photon-level laser altimetry data (ICESat-2ATL03). Its core idea leverages the physical property of "cloth falling under gravity and hanging from the edge of a depression." A layer of virtual cloth (composed of discrete cloth particles) falls from above and collides with "mask particles" (fixed particles based on actual elevation). The cloth eventually forms a curved surface that conforms to the ice surface and spans over the crack. By comparing the photon elevation with this cloth surface, ice surface photons and crack photons can be distinguished, thereby locating the crack region and calculating its depth and width. This method aims to reduce reliance on empirical thresholds, enhance adaptability to complex crack fields, and operate automatically at various scales (from large cracks to shallower / narrower cracks).

[0079] In related technologies, existing methods often rely on optical imagery, fixed-scale segmented elevation products, or threshold adjustments using multiple empirical parameters. These methods generally suffer from insufficient versatility: they struggle to adapt to variations in fracture scale across different ice shelf regions and operate stably in situations with subtle fracture development or sparse photon distribution. Furthermore, some automated detection methods based on segmented elevation data are limited by fixed spatial resolution, often failing to capture subtle vertical changes in small-scale or densely populated fracture zones. While some methods have attempted to improve the precision of fracture depth extraction using photon-level data in recent years, they still rely heavily on manually set threshold parameters, requiring repeated adjustments for different regions and fracture types, making it difficult to achieve universal adaptability to the diverse fracture environments of Antarctic ice shelves. Overall, existing methods largely depend on direct detection of surface depressions or empirical rules to define fracture boundaries and depths; their robustness significantly decreases when fracture morphology is atypical, development is weak, or environmental conditions are complex.

[0080] In view of this, this invention provides a method and system for automatically extracting ice surface crack depth based on linear cloth simulation. This method acquires photon positioning data and photon classification data of the target area, and generates target photon data and particle data through data preprocessing and meshing. The target photon data includes denoised photon data and elevation-aided data, while the particle data includes initialized cloth particles and terrain height mask particles corresponding to each mesh cell. Based on the particle data, a cloth drop simulation is performed using a preset dynamic model combined with gravity to obtain a cloth height curve. Based on the cloth height curve, photon classification is performed on the denoised photon data to obtain photon categories and determine crack segments. The depth information of the crack segments is quantized based on the elevation-aided data. This invention overcomes the limitations of traditional methods that rely on fixed resolution, manual thresholds, and empirical parameters by adaptively fitting the terrain through cloth simulation of the physical process. The beneficial effects of the embodiments of the present invention are as follows: it significantly improves the universality of depth extraction of cracks of different scales, morphologies and developmental degrees; through physical-driven adaptive simulation, it reduces the dependence on human intervention and empirical parameters, thereby enhancing the detection stability and automation level in areas with weak or dense cracks, and providing more reliable three-dimensional structural data for ice shelf stability assessment.

[0081] It is understood that the automatic ice surface crack depth extraction method based on linear cloth simulation provided by this invention can be applied to any computer device with data processing and computing capabilities, and this computer device can be various terminals or servers. When the computer device in the embodiment is a server, the server is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. Optionally, the terminal can be a smartphone, tablet, laptop, or desktop computer, but it is not limited to these.

[0082] like Figure 1 The diagram shown is a schematic representation of an implementation environment provided by an embodiment of the present invention. (Refer to...) Figure 1 The implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be connected via a network, either wirelessly or via a wired connection, to complete data transmission and exchange.

[0083] Server 101 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0084] Additionally, server 101 can also be a node server in a blockchain network. Blockchain is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms.

[0085] Terminal 102 can be a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. Terminal 102 and server 101 can be directly or indirectly connected via wired or wireless communication, and this embodiment of the invention does not impose any limitations.

[0086] For example, based on Figure 1 The implementation environment shown in this embodiment of the invention provides an automatic method for extracting ice surface crack depth based on linear cloth simulation. The following description uses the application of this automatic method for extracting ice surface crack depth based on linear cloth simulation in server 101 as an example. It can be understood that this automatic method for extracting ice surface crack depth based on linear cloth simulation can also be applied in terminal 102.

[0087] Reference Figure 2 , Figure 2 This is an optional flowchart of an automatic ice surface crack depth extraction method based on linear cloth simulation provided in an embodiment of the present invention. The execution subject of this automatic ice surface crack depth extraction method based on linear cloth simulation can be any of the aforementioned computer devices (including servers or terminals). Figure 2 The method may include, but is not limited to, steps S100 to S400.

[0088] Step S100: Obtain photon localization data and photon classification data of the target area, and generate target photon data and particle data through data preprocessing and gridding.

[0089] The target photon data includes denoised photon data and elevation-aided data, and the particle data includes the initialized cloth particles and the mask particles for terrain height corresponding to each grid cell.

[0090] It should be noted that the photon positioning data includes the time, latitude, longitude, and elevation information of photon events. In some embodiments, target photon data and particle data are generated through data preprocessing and gridding, which may include the following steps: using photon classification labels in photon classification data to filter noisy photons from the photon positioning data to obtain denoised photon data; based on the denoised photon data, scaling the distance along the track direction to obtain scaled photon data; based on the photon positioning data, processing it using a preset tool to obtain elevation auxiliary data; dividing the scaled photon data into grids at fixed intervals to obtain multiple grid cells; initializing the cloth particles corresponding to each grid cell based on a preset height; and initializing mask particles in each grid cell based on the elevation auxiliary data.

[0091] For example, in some specific implementations, data preprocessing and grid generation can be achieved as follows:

[0092] This invention is used to perform noise removal, scale conversion, meshing, and initialization of particle structures required for cloth simulation on ICESat-2 ATL03 photon counting data. Specifically, it includes the following steps:

[0093] Step 1: Obtaining raw data:

[0094] Acquire global photon positioning data from ICESat-2 ATL03 and simultaneously acquire the ICESat-2ATL08 product for photon classification. ATL03 provides time, latitude, longitude, and elevation information for photon events; ATL08 is used to identify noisy photons.

[0095] Step 2: Noise photon filtering (data cleaning):

[0096] Noise removal was performed on the ATL03 data using photon classification labels from ATL08:

[0097] Delete photons marked as "noise" or "unclassified";

[0098] Retain all valid surface photons.

[0099] This step ensures that the data used for crack detection is clean and reliable, reducing errors at the source.

[0100] Step 3: Scaling the distance along the track direction:

[0101] To enhance the vertical representation of the fracture geometry in the coordinate system, the distance along the trace direction of ICESat-2 is scaled by 1 / 1000.

[0102] This scaling enhances the vertical abrupt change at the top of the crack, making it easier for cloth simulations to distinguish the ice surface from the crack.

[0103] Step 4: Organization of Elevation Auxiliary Data

[0104] The SlideRule tool was used to generate the ICESat-2 average elevation data product with a distance of 10m along the track: AE-10m. Its features include: average elevation within a 10m segment length, 5m step size, and processing within the same frame as ATL06 but with higher resolution.

[0105] This data is used for calculating the "mask particle height" and crack depth in subsequent cloth simulations.

[0106] Step 5: Photon data meshing (particle initialization):

[0107] The scaled photon data is divided into grids at fixed intervals, and two types of particles are initialized in each grid cell:

[0108] Cloth particles: Simulate cloth nodes;

[0109] Masking particles: represent the height of the ice surface terrain.

[0110] It should be noted that, in some embodiments, initializing mask particles in each grid cell based on elevation auxiliary data may include the following steps: determining the elevation information corresponding to each grid cell based on the elevation auxiliary data; initializing the elevation of the mask particles corresponding to the corresponding grid cell based on the elevation information; when there is a single mask particle with an empty elevation, determining the elevation of the corresponding mask particle by linear interpolation based on the elevations of two adjacent mask particles; when there are multiple consecutive mask particles with empty elevations, assigning the elevation of the corresponding mask particles to negative infinity.

[0111] For example, in some specific implementations, the mask particle elevation is provided by a continuous ice surface elevation profile obtained through linear interpolation of 5m resolution AE-10m data. The mask particle elevation of an empty grid is obtained by linear interpolation of the elevations of two adjacent mask particles. If there are two or more consecutive empty grids, the elevation of the mask particles within the empty grid is assigned a value of negative infinity to avoid generating false surfaces. The cloth particles and mask particles use an 18m spatial resolution to lay a unified computational grid foundation for subsequent simulations.

[0112] Step S200: Based on particle data, simulate the fabric falling by combining a preset dynamic model with the effect of gravity to obtain the fabric height curve.

[0113] It should be noted that in some embodiments, step S200 may include the following steps: initializing the fabric dynamic model using traction springs and bending springs; wherein, the fabric dynamic model initially has fabric particles in a movable state; simulating fabric falling displacement based on the fabric dynamic model combined with gravity, and iteratively updating the position of the fabric particles at each time step using an integral model; performing collision detection judgment on the result of each position iteration update, and when the height of the updated fabric particle is less than or equal to the height of the mask particle, resetting the height of the fabric particle to the height of the mask particle, and updating the corresponding movement state of the fabric particle to immovable; stopping the position iteration update when the height change of all fabric particles in the position iteration update is less than a first threshold, or when the number of iterations of the position iteration update reaches a second threshold; and constructing a fabric height curve based on the spatial shape corresponding to the height of the fabric particles after the last position iteration update.

[0114] For example, in some specific embodiments, the fabric falling simulation can be implemented as follows:

[0115] Fabric simulation is the key innovation of this invention. It simulates "fabric" adhering to the ice surface under the influence of gravity, thus naturally crossing cracks and forming a "fabric height" that can be used for classification. Specifically, it includes the following steps:

[0116] Step 1: Initialize the cloth dynamics model:

[0117] The cloth mesh uses a physical model consisting of traction springs and flexion springs to ensure the stability of the cloth structure under stress. Initially, all cloth particles are set to "movable".

[0118] Step 2: Update cloth particle positions (Verlet integral):

[0119] In each iteration cycle, the cloth particles are displaced due to gravity and internal forces. The particle height H(t) is updated according to the following Verlet integral model:

[0120] H(t+Δt)=2H(t)-H(t-Δt)+(G / m)Δt 2

[0121] Where H(t) represents the particle height at the current time point t, H(t+Δt) represents the particle height at the next time point t+Δt, H(t-Δt) represents the particle height at the previous time point t-Δt, Δt is the time step, G represents gravity, and m represents the particle mass. This method is numerically stable, computationally efficient, and suitable for large-scale ice shelf simulation.

[0122] Step 3: Collision Detection and Stopping Mechanism

[0123] A collision is considered to have occurred when the height of the cloth particle is less than or equal to the height of the mask particle. This particle is:

[0124] The height is reset to the mask height;

[0125] The status has been updated to "immovable".

[0126] This rule ensures that the fabric fits the ice surface perfectly.

[0127] Step 4: Assessing fabric stability:

[0128] The cloth simulation continues to iterate until any of the following conditions are met:

[0129] The height variation of all fabric particles is <0.005m;

[0130] The number of iterations reached 200;

[0131] At this point, the fabric is considered to have reached a state of "physical equilibrium," and its spatial shape constitutes the "fabric height curve" used for subsequent classification.

[0132] In some specific application scenarios, the ice shelf crack identification process based on fabric simulation in this invention is implemented through the following steps. This process uses laser altimeter data as input and achieves automatic differentiation between the ice surface and cracks through fabric drop simulation under physical constraints. Specifically, it includes the following steps:

[0133] First, the spatial extent of the input photon point cloud data is analyzed. The system calculates the minimum bounding area of ​​the photon point cloud on a two-dimensional plane and constructs a regular cloth mesh above this area. This cloth mesh slightly extends beyond the coverage area of ​​the photon point cloud in the horizontal direction and is set at a certain height in the vertical direction, so that the cloth is initially suspended above the terrain, avoiding unnecessary contact in the initial stage.

[0134] Subsequently, based on the set spatial resolution, the fabric is divided into several uniformly distributed fabric particles, and corresponding mask particles are constructed below the fabric. The height of the mask particles is determined by the average elevation data along the track (AE-10m) and photon point cloud data, which is used to describe the real ice surface morphology, thereby providing stable terrain constraints for the fabric descent process.

[0135] After initializing the cloth and terrain, the system enters the cloth falling simulation phase. A constant downward gravitational force is applied to the cloth, causing it to gradually move downwards under gravity. In each simulation iteration, the position of the cloth particles is updated based on the previous and subsequent states, and collision detection is performed with the mask particles below. When the cloth particles fall to the ice surface, their movement is restricted, causing the cloth to gradually conform to the ice surface shape; in the crack areas, due to the lack of continuous terrain support below, the cloth will naturally cross the cracks, forming a suspended structure.

[0136] To improve computational efficiency, the system continuously monitors the overall displacement of the cloth during the cloth simulation. When the maximum displacement of the cloth particles falls below a preset threshold between two consecutive iterations, the cloth is considered to have reached a stable state, and the simulation terminates prematurely. This mechanism ensures that the cloth adheres fully to the ice surface while avoiding unnecessary recalculation.

[0137] After the fabric stabilizes, the system can smooth the fabric as needed to eliminate local irregular deformations, improve the continuity and overall stability of the fabric surface, and thus provide a more reliable reference surface for subsequent photon classification.

[0138] Finally, using the stabilized fabric surface as a reference, the system compares the distance between each photon point and the corresponding fabric height. When a photon is near the fabric surface, it is classified as an ice surface photon; when a photon is significantly below the fabric, it is classified as a rift photon. This method achieves automatic classification of photon points and ultimately outputs sets of ice surface photons, rift photons, and their corresponding classification labels.

[0139] Step S300: Based on the fabric height curve, perform photon classification on the denoised photon data to obtain photon categories and determine the crack segment;

[0140] It should be noted that in some embodiments, step S300 may include the following steps: based on the height of the fabric particles in the fabric height curve, a continuous fabric height model along the trajectory direction is generated by linear interpolation; the first photon in the denoised photon data is taken as a candidate photon; fabric particles are extracted from the fabric height model as candidate fabric according to the position corresponding to the candidate photon; when the height difference between the candidate photon and the candidate fabric is less than or equal to a third threshold, the candidate photon is classified as an ice surface photon; otherwise, the candidate photon is classified as a rift photon; the next photon in the denoised photon data is taken as a candidate photon, and the step of extracting fabric particles from the fabric height model as candidate fabric according to the latitude and longitude corresponding to the candidate photon is returned to be executed until all photons in the denoised photon data have been traversed; when the number of consecutive rift photons is greater than a fourth threshold, the corresponding continuous segment is taken as a rift segment; wherein, the endpoints of the two ends of the rift segment are determined by the start and end positions of the continuous segment.

[0141] For example, in some specific implementations, photon classification and preliminary crack extraction can be achieved as follows:

[0142] Step 1: Fabric Continuity Processing:

[0143] By linearly interpolating the height of the fabric particles, a continuous fabric height model along the trajectory direction is generated.

[0144] The fabric adheres closely to the ice surface but remains suspended in the cracked areas, providing natural boundaries for classification.

[0145] Step 2: Photon category determination:

[0146] For each photon, calculate the difference between its elevation and the height of the fabric surface:

[0147] If |H 光子 –H 布料 If |≤0.65m, it is classified as ice surface photon;

[0148] If |H 光子 –H 布料 If the value is greater than 0.65m, it is classified as a slit photon;

[0149] This is the key step in achieving automatic crack identification in this invention, eliminating the need for manual threshold adjustment.

[0150] Step 3: Extraction of potential fracture segments:

[0151] Traverse the photon sequence and mark all segments where "slit photons" appear consecutively. The start and end positions of these segments are the left and right endpoints of the slit segment.

[0152] Each crack segment constitutes the basic unit for subsequent geometric measurements in this invention.

[0153] Step S400: Quantize the depth information of the fracture segment based on elevation-aided data;

[0154] It should be noted that the fracture segment includes a continuous sequence of fracture photons and their corresponding position range. In some embodiments, step S300 may include the following steps: smoothing the elevation auxiliary data using Gaussian filtering to obtain elevation smoothed data; extracting the elevation sequence corresponding to the fracture segment from the elevation smoothed data based on the position range corresponding to the fracture segment; performing extreme value determination on the elevation sequence, splitting the fracture segment into final fracture segments based on the result of the extreme value determination, and determining the final elevation sequence corresponding to each final fracture segment based on the elevation sequence; obtaining the endpoint elevations at both ends of the final fracture segment and the minimum elevation within the segment from the final elevation sequence, and determining the depth value of the final fracture segment based on the difference between the minimum endpoint elevation and the minimum elevation within the segment; taking one endpoint of the final fracture segment as a base point, searching for a corresponding point with the same elevation as the base point in the opposite direction of the base point in the final elevation sequence, and determining the width value of the final fracture segment based on the positional distance between the base point and the corresponding point.

[0155] For example, in some specific embodiments, the present invention achieves the final extraction of fracture information through fracture depth and width calculation and post-processing optimization:

[0156] This invention utilizes AE-10m data to precisely quantify crack depth and width, and decomposes composite cracks to ensure that the obtained crack parameters are independent and accurate. Specifically, it includes the following steps:

[0157] Step 1: Calculation of Potential Fracture Segment Depth (PFSD):

[0158] For each potential fracture segment:

[0159] Read the elevation H of the left and right endpoints left H right ;

[0160] Obtain the minimum elevation H within the segment min

[0161] Calculate the potential fracture depth: PFSD = min(H) left H right ) - H min ;

[0162] Step 2: Calculation of Potential Fracture Segment Width (PFSW):

[0163] Taking the left end point X1 of the fracture segment as the base point;

[0164] Search in the opposite direction for point X2 with the same elevation as the left endpoint;

[0165] The calculated width is: PFSW = |X1–X2|;

[0166] This process utilizes the uniform 10m resolution of AE-10m to ensure consistency between elevation and horizontal distance.

[0167] Step 3: Local Extremum Analysis (Complex Fracturation Decomposition):

[0168] This is used to address the problem of "multiple fractures being misidentified as a single fracture". A decision split is performed on each individual potential fracture segment to determine the final fracture segment.

[0169] Recalculate the depth and width of all split crack segments and output the final independent crack set.

[0170] It should be noted that in some embodiments, extreme value determination is performed on the high-degree sequence, and the fracture segment is divided into final fracture segments based on the result of the extreme value determination. This may include the following steps: traversing and obtaining the minimum and maximum values ​​of the high-degree sequence; when the number of minimum values ​​is 0, the fracture segment is marked as an invalid fracture; when the number of minimum values ​​is 1, the fracture segment is taken as the final fracture segment; when the number of minimum values ​​is greater than 1, the potential fracture peak depth is determined based on the difference between the largest maximum value and the minimum elevation within the fracture segment, and the potential fracture peak depth is determined based on the minimum elevation of the endpoints at both ends of the high-degree sequence and the minimum elevation within the fracture segment. The difference in minimum elevation determines the potential fracture segment depth; when the ratio of the potential fracture peak depth to the potential fracture segment depth is less than the fifth threshold, the fracture segment is taken as the final fracture segment; when the ratio of the potential fracture peak depth to the potential fracture segment depth is greater than or equal to the fifth threshold, all photons with height values ​​greater than the maximum value are deleted to update the target photon data and particle data, and the process returns to the step of simulating cloth falling based on particle data, using a preset dynamic model combined with gravity, until all fracture segments are decomposed and the final fracture segment corresponding to each fracture segment is obtained.

[0171] For example, in some specific implementations, performing a decision split for each individual potential fracture segment specifically includes the following operations:

[0172] (1) Extreme value detection:

[0173] Extracting the AE-10m elevation curve after Gaussian smoothing:

[0174] Minimum local elevation values ​​within the potential fracture segment;

[0175] Maximum local elevation values ​​within the potential fracture segment.

[0176] (2) Preliminary determination of the number of cracks:

[0177] If it contains only one minimum value → single fracture;

[0178] If there is no minimum value, it is judged as an invalid crack and discarded.

[0179] If it contains multiple local minima, further analysis is required.

[0180] (3) Geometric proportion judgment:

[0181] Calculate the potential fracture peak depth (PFPD) corresponding to the highest maximum value within the potential fracture segment:

[0182] PFPD = H peak –H min ;

[0183] If the following conditions are met:

[0184] PFPD / PFSD < 1 / 2;

[0185] If no split is needed, proceed to the next step; otherwise.

[0186] (4) Data cutting and fabric reapplication simulation:

[0187] When it is necessary to split the fracture, for each potential fracture segment:

[0188] Remove all photons within the potential fracture segment whose height value is higher than the highest elevation maximum.

[0189] Update the left and right endpoints of the fracture segment;

[0190] The continuous photon data of the cut composite rift segment were re-applied to the repeating cloth simulation and reclassified.

[0191] This process can be performed iteratively until the composite fracture is completely separated.

[0192] It should be noted that in the above steps, the maximum elevation value above the maximum elevation value within each individual potential fracture segment is specifically cut off. After cutting, the data still represents the continuous photon data corresponding to that fracture segment. Furthermore, in this embodiment of the invention, the data of the cut fracture segments are directly reapplied to the same method as before for fabric simulation.

[0193] In some specific application scenarios, such as Figure 3 The diagram shows the complete execution flow of the post-processing stage in the LCSF-ice algorithm. This stage primarily processes the "potential crack segments" obtained from cloth simulation and photon classification to determine whether they contain multiple independent cracks and to separate cases of multiple crack stacking, making the depth and width estimates of each crack more accurate and unique. The process unfolds in a modular fashion from top to bottom, with each module connected by arrows to indicate the sequential data transfer logic.

[0194] The process begins with the potential fracture segment input module. This module receives all potential fracture segments extracted in the previous step (cloth simulation and photon classification). Each fracture segment contains a continuous sequence of fracture photons and its corresponding coordinate range along the track. Since photon morphology under certain wide fracture or noisy conditions may cause multiple fractures to be incorrectly merged, the goal of the subsequent processing is to identify whether merging occurs and to further separate them.

[0195] Next, the system proceeds to the local extremum extraction module. First, it smooths the AE-10m elevation data using a Gaussian filter to eliminate spurious peaks caused by local noise. Then, it searches for all local minima and maxima in the elevation sequence covered by the fracture segment: minima correspond to the bottom of the fracture, and maxima correspond to the ridge positions at fracture boundaries. Ideally, a fracture segment contains only one minima; if multiple minima are detected, it indicates that the fracture segment may be a composite fracture.

[0196] After outputting from the local extremum module, the process enters the crack quantity determination module. The system first checks the number of minimum values:

[0197] If the number of minimum values ​​is 0, the segment is judged as a meaningless fracture segment and is directly removed.

[0198] If it contains only one minimum value, the segment is considered a single crack and does not need to be split. It can proceed directly to the final parameter calculation stage.

[0199] If multiple minimum values ​​are included, proceed to the next step of geometric ratio determination.

[0200] The process then proceeds to the fracture peak depth calculation module. In this module, the system calculates the "fracture peak depth (PFPD)" between each maximum and the global minimum, and compares this ratio with the potential fracture segment depth (PFSD). If the ratio of all fracture peak depths to the total depth is less than 1 / 2, the system considers the local maxima in the potential fracture segment insufficient to separate fractures, and the fracture segment should be retained as a single-fracture structure.

[0201] If the peak depth of certain maxima exceeds the aforementioned threshold, the process proceeds to the maxima trimming module. In this stage, the system completely removes all photon points above the highest local maximum to eliminate "false top" height structures that could connect multiple maxima. After trimming, new boundaries are generated for the maxima.

[0202] After the trimming operation is completed, the data enters the re-layout simulation module. The system re-executes the cloth drop simulation on the trimmed photon data and performs photon classification and potential fracture segment extraction again. If the system still detects multiple minima in the new data, the process will repeat the steps of extreme value detection, peak depth determination, and data trimming. This process is iterative and can be performed in multiple rounds for complex fracture structures.

[0203] After several rounds of iterative decomposition, when each fracture segment has only a minimum value remaining, the process enters the final fracture parameter calculation module. This module recalculates the final fracture segment depth, final fracture segment width, and fracture location based on AE-10m elevation data, ensuring that the geometric parameters of each fracture correspond to an independent and unique fracture structure. The output data is the final fracture depth and width result of this invention.

[0204] In some specific application scenarios, the implementation logic of the post-processing steps is as follows:

[0205] After completing the initial crack identification based on fabric simulation, this invention further includes a post-processing step to address the problem of multiple cracks being misidentified as a single crack segment in complex ice shelf regions. This post-processing step refines and breaks down the crack segment through peak analysis and iterative reconstruction, ensuring that each crack segment corresponds to a real and unique crack structure.

[0206] First, the system performs peak value detection analysis on all initially extracted fracture segments. This analysis, based on the average elevation data along the track (AE-10m), identifies whether each fracture segment contains multiple obvious local minimum points or elevation fluctuations. Through this analysis, the system classifies fracture segments into two categories: one is "problem-free fracture segments" with simple internal structures and containing only a single fracture; the other is "problem fracture segments" that may contain multiple fractures or have complex structures. Problem-free fracture segments are directly retained as valid results, while problem fracture segments proceed to subsequent fine processing.

[0207] For segments identified as problematic, the system extracts corresponding local data intervals from the original ICESat-2 ATL03 photon point cloud data based on their left and right boundary positions, and saves these local point cloud data separately. In this way, each problematic segment is transformed into an independent subset of data, thus avoiding mutual interference between it and other segments.

[0208] Subsequently, the system enters the iterative processing phase. For each problematic fracture segment corresponding to a subset of data, the system re-executes the cloth simulation, photon classification, and fracture segment extraction processes. By reconstructing the interaction between the cloth and the terrain, fracture structures that might have been merged are re-analyzed within a smaller spatial range. After completing a new round of fracture segment extraction, the system performs peak analysis again and reclassifies the results into non-problematic fracture segments and problematic fracture segments.

[0209] In each iteration, newly identified problem-free crack segments are continuously accumulated and saved, while problem crack segments that still have complex structural features are processed again in the next round. This process can be repeated multiple times, but the system has a maximum iteration limit to prevent infinite loops and ensure that the algorithm completes the computation within a reasonable time.

[0210] To explain in detail the principle of the technical solution of the present invention, the overall process of the present invention will be described below with reference to some specific embodiments. It is easy to understand that the following is an explanation of the technical principle of the present invention and should not be regarded as a limitation of the present invention.

[0211] First, it should be noted that while existing technologies have made some progress in the three-dimensional structural identification and depth extraction of Antarctic ice shelf fissures using multi-source data such as optical images, radar images, and laser altimeters, they still have the following significant shortcomings:

[0212] (1) The crack depth cannot be automatically and quantitatively extracted, relying on manual methods and auxiliary images:

[0213] Existing research largely focuses on identifying the planar distribution of fractures based on optical or radar remote sensing images. However, image observation can only characterize the morphology and orientation, and cannot directly obtain information on the vertical morphology and depth of fractures. Some methods utilize ICESat-2 laser height data to study fractures, but still require optical images such as Landsat for auxiliary localization or verification, and lack the ability to extract fracture depth fully automatically and reliant on a single data source.

[0214] (2) It relies on a large number of thresholds and empirical parameters, resulting in poor generalization:

[0215] Existing ICESat-2 fracture detection methods based on ATL06 or ATL03 mostly employ a "depression identification + threshold determination" approach, involving numerous region-dependent empirical parameters (such as height difference threshold, photon density threshold, slope determination, etc.), making them difficult to apply to ice shelf regions with varying fracture scales, background noise, and topographical changes. For areas with complex fracture morphology, sparse signals, or abrupt slope changes, these threshold methods are prone to false positives or false negatives.

[0216] (3) Limited spatial resolution makes it difficult to capture small cracks or complex structures:

[0217] For example, the 40 m segmentation scale of the ICESat-2 ATL06 product makes it difficult to identify narrow, dense, or irregularly shaped fracture structures. Some subsequent methods using ATL03 data have improved the resolution, but they are still limited by photon sparsity, noise, etc., and have difficulty distinguishing multi-fracture composite structures, resulting in the merging or missegmentation of fracture depths.

[0218] (4) Lack of robustness verification in a wide range of fracture scenarios on Antarctic ice shelves:

[0219] While some algorithms have achieved good results in regions such as Greenland, they lack sufficient validation for the complex fracture systems of Antarctic ice shelves. Existing methods are generally only applicable to fracture types with single or distinct local characteristics, and it is difficult to maintain stable performance in regions with significant differences in fracture width, depth, and spacing.

[0220] To address the shortcomings of existing methods, this invention proposes an automatic crack depth extraction method based on linear cloth simulation filtering (LCSF-ice) using ICESat-2 photon data. This method draws inspiration from the physical process of cloth naturally drooping under gravity. By observing the state of the cloth lying flat on the ice surface and suspended above the crack, it can automatically identify crack boundaries and extract depth without the need for empirical thresholds. The main objective of this invention is:

[0221] The method automates the extraction of crack depth and width, and enables adaptive parameter extraction, reducing manual intervention and dependence on threshold tuning, thereby improving the universality and robustness of the method.

[0222] Improving the accuracy and spatial resolution of crack identification can effectively distinguish multi-crack composite structures and enhance the detection capability of narrow cracks and cracks with complex shapes.

[0223] It has achieved stable application on representative ice shelves in Antarctica, adapting to different fracture scales, noise levels, and terrain features, and has the potential for wide-ranging application.

[0224] We provide a scheme for extracting the three-dimensional structure of cracks based on physical process simulation, which provides more reliable basic data for ice shelf stability analysis, crack propagation mechanism research and ice shelf collapse risk assessment.

[0225] In specific application scenarios, this invention proposes an automatic ice shelf crack depth extraction method based on linear cloth simulation (LCSF-ice algorithm). Its core idea is to utilize ICESat-2 ATL03 photon-counting laser altimetry data to achieve automatic separation of the ice surface and crack region and analytical measurement of crack depth through a physical simulation process of cloth falling and conforming to the ice surface morphology. The algorithm is designed based on the physical principles of existing cloth simulation filtering (CSF) algorithms, but it incorporates systematic adaptive improvements in handling sparse photon data and complex ice shelf crack structures, achieving high-precision, fully automatic, and widely applicable crack depth extraction capabilities.

[0226] The entire technical solution consists of four main stages: data preprocessing and meshing, fabric drop simulation, photon classification and preliminary crack extraction, crack depth and width calculation, supplemented by post-processing steps based on local extremum features to further decompose composite cracks, correct crack boundaries and improve the reliability of depth estimation.

[0227] like Figure 4 The diagram illustrates the complete processing flow of the LCSF-ice algorithm. Starting with the input of raw ICESat-2 photon data, it goes through data cleaning, fabric simulation, photon classification, fracture segment extraction, and composite fracture decomposition, ultimately generating geometric parameters such as fracture depth and width. The entire flowchart unfolds sequentially from left to right and from top to bottom, with arrows connecting the various processing modules to indicate the sequential transfer and processing logic of data between each step.

[0228] The workflow begins with raw data input, containing two types of data: ICESat-2 ATL03 photon-level data, providing the spatial coordinates and elevation of each photon; and AE-10m elevation data along the track generated by SlideRule, used to supplement surface morphology information in areas with sparse photons. The workflow then proceeds to the data preprocessing module, where noisy photons in the ATL03 data are first removed using the ATL08 photon classification results, retaining only valid photons that may belong to the ice surface or cracks. Next, the track distance is scaled down by 1 / 1000, making the vertical morphology of cracks more prominent in subsequent cloth simulations. Afterwards, the photon and AE-10m data are re-meshd to generate equally spaced cloth particles and mask particles. The elevation of the mask particles is entirely supplemented by the AE-10m data, serving as "surface constraints" for the cloth simulation.

[0229] Next, we move to the core part of the flowchart—the cloth falling simulation module. This module includes the initialization of cloth particles, the establishment of spring connections, and gravity loading. The cloth falls gradually under gravity, and the particle positions are continuously updated using the Verlet integral method. In each iteration, the system checks whether the cloth particles collide with the mask particles below; once the height of a cloth particle is lower than the height of a mask particle, the height of the cloth particle is reset to the mask elevation and marked as "immovable". The entire cloth simulation process continues until the overall height change of the cloth particles is less than 0.005 meters, or the set upper limit of 200 iterations is reached, indicating that the cloth has stably adhered to the ice surface, while forming a natural suspended shape above the cracks. After the cloth stabilizes, the process enters the photon classification module. By linearly interpolating the height of the cloth particles, a continuous "cloth surface height curve" is obtained. The system calculates the difference between the height of each photon and the height of the corresponding cloth surface: if the difference is less than or equal to 0.65 meters, the photon is identified as an ice surface photon; if the difference is greater than 0.65 meters, it is identified as a rift photon.

[0230] The process then proceeds to the fracture geometry calculation module. Using AE-10m elevation data, the system calculates the elevation values ​​of the left and right endpoints of the fracture, as well as the elevation of the lowest point inside the fracture, thus obtaining the potential fracture segment depth (PFSD). The potential fracture segment width (PFSW) is determined by searching outwards from the left endpoint of the fracture to a position consistent with its height. This calculation process is entirely based on the uniform resolution of AE-10m, thus ensuring the geometric accuracy of both depth and width.

[0231] To address the possibility of multiple independent fractures within a fracture segment, the process proceeds to the composite fracture segmentation module. In this module, local maxima and minima are extracted after Gaussian smoothing of the AE-10m elevation to determine the number of fractures within a potential fracture segment. When multiple minima exist, the system determines whether segmentation is necessary based on the ratio of fracture peak depth to the overall fracture depth. If segmentation is deemed necessary, all photon data above the highest maxima are removed, and the cloth simulation, photon classification, and fracture segment identification are re-executed. This process may be repeated multiple times until all composite fractures are successfully separated and form independent fracture segments. The flowchart ends with the output module, which ultimately outputs the depth, width, location, and morphological characteristics of each fracture, constituting the final result of the automatic fracture extraction of this invention. This process, through cloth simulation, photon classification, and multi-round segmentation mechanisms, achieves automatic, robust, and high-precision three-dimensional structural analysis of both regular and complex fractures.

[0232] like Figure 5 The diagram shown illustrates a conceptual example of the core working mechanism of the LCSF-ice algorithm. The entire diagram is divided into four parts, corresponding to the four key stages of algorithm execution: data preprocessing, fabric initialization, photon classification, and calculation of crack depth and width. These four parts, arranged from left to right, collectively demonstrate how this invention automatically identifies ice shelf cracks and obtains their geometric features from ICESat-2 photon-level altimetry data. Figure 5 As shown:

[0233] Part 1(a) Data Preprocessing illustrates the state of the raw photon data after noise filtering. At this stage, the input ATL03 photon data contains a large number of valid photons and noisy photons. The noise may originate from atmospheric scattering, instrument false alarms, or non-surface reflections. By referencing the photon classification information from the ICESat-2 ATL08 product, noisy photons are identified and removed, retaining only true photon points related to the surface or fissures. The preprocessing effect shown in the figure aims to demonstrate that the filtered photon sequence more closely approximates the actual ice surface morphology, reducing interference points and allowing subsequent fabric modeling to be simulated on a cleaner data basis.

[0234] Part 2(b) Cloth Initialization demonstrates the process of photon data meshing and the creation of cloth particles and mask particles. First, the along-track photon data is divided into a regular grid according to a set spatial resolution; each grid cell corresponds to a cloth particle, serving as a node in the cloth model. Simultaneously, mask particles are generated at the same grid locations, their elevations provided by the AE-10m elevation product, used as ice surface constraints in the cloth simulation. In the schematic diagram, cloth particles and mask particles appear in a regular array, representing the uniform distribution of the cloth grid in space; while the curves formed by the mask particles represent the morphology of the ice surface along the track. This structure provides realistic terrain support for the cloth's descent, collision, and adhesion.

[0235] Part 3(c) Photon Classification illustrates the relative positional relationship between the fabric surface shape and photon points after the fabric simulation is completed. After multiple iterations and stable adhesion to the ice surface, the fabric model naturally forms a suspended shape in the crack region. Therefore, by comparing the elevation of each photon with the fabric surface height, photons can be automatically classified into "ice surface photons" and "crack photons". If the difference between the photon height and the fabric surface height is small, the photon is identified as being located on the ice surface; if the difference is large, the photon is located below the fabric surface and is identified as a crack photon. In the figure, the crack region is represented by a clearly concave photon point, while the fabric surface in this region has an arch-like suspended structure. After photon classification, by detecting areas where crack photons appear continuously, "potential crack segments" are extracted, and the left and right boundaries of the cracks are determined.

[0236] Part 4(d) Calculation of Fractured Depth and Width demonstrates how to calculate the geometric parameters of a potential fracture segment based on its boundary. In this stage, the algorithm uses AE-10m elevation data to obtain the elevation values ​​of the left and right endpoints of the fracture and determines the location of the lowest point inside the fracture. The fracture depth is defined as the difference between the smaller elevation of the left and right endpoints and the elevation of the lowest point; the fracture width is obtained by measuring the horizontal distance between the two endpoints. The figure illustrates the one-dimensional geometric measurement process of depth and width by showing the relationship between the fracture bottom location, the left and right endpoint locations, and their elevations. These parameters are ultimately used to describe the three-dimensional structure of each fracture.

[0237] In summary, this invention is the first to introduce a linear cloth simulation physical model into ICESat-2 photon counting laser altimetry data to calculate the depth of cracks in Antarctic ice shelves, achieving automatic, threshold-free separation between the ice surface and cracks through the collision of falling cloth with the ice surface. Specifically, this invention generates a cloth surface capable of crossing cracks by utilizing the falling, colliding, and stopping mechanism of cloth particles under gravity; and automatically identifies the location and morphology of cracks by determining the photons on the ice surface and cracks based on the difference between the photon height and the cloth surface height.

[0238] Compared to existing crack detection methods based on ICESat-2 ATL06 segmented elevation, ATL03 photon statistical features, or a large number of empirical thresholds, this invention achieves fully automatic and highly robust crack identification on ice shelves by introducing a mechanically meaningful cloth simulation mechanism. Traditional methods often rely on fixed thresholds, local slope variations, or abrupt changes in photon density, leading to false positives and false negatives in areas with sparse photons, complex crack morphologies, or multiple overlapping cracks. This invention utilizes the natural adhesion and traversal characteristics of cloth in a gravitational field, enabling the cloth surface to stably distinguish between the ice surface and crack locations, eliminating reliance on regional empirical parameters and significantly improving the universality and reliability of crack identification. Furthermore, this invention uses AE-10m high-resolution elevation data as cloth collision constraints, ensuring the model maintains a strong constraint fit to the ice surface even in photon-deficient regions, thus maintaining consistent performance across different ice shelf environments in Antarctica.

[0239] Compared to the common problem of insufficient ability to identify complex fractures in existing technologies, this invention proposes a fracture segmentation strategy of "fabric simulation—extreme value analysis—secondary simulation," which can effectively distinguish overlapping, multi-bottom, or complex fracture structures, ensuring the stability and uniqueness of fracture depth estimation. This adaptive segmentation method based on geometric proportions and local morphology enables the invention to maintain high accuracy in fracture-dense areas, regions difficult to identify in optical images, and environments with high photon point cloud noise. In summary, this invention significantly outperforms existing methods in terms of automation, adaptability, robustness, and fracture depth calculation accuracy, providing a more reliable technical means for Antarctic ice shelf stability research.

[0240] like Figure 6 As shown, this embodiment of the invention also provides an automatic ice surface crack depth extraction system 900 based on linear cloth simulation, which can implement the above-mentioned method. This system may include:

[0241] The first module 910 is used to acquire photon positioning data and photon classification data of the target area, and generate target photon data and particle data through data preprocessing and gridding; wherein, the target photon data includes denoised photon data and elevation-aided data, and the particle data includes the initialized cloth particles and the mask particles of terrain height corresponding to each grid cell;

[0242] The second module 920 is used to simulate the fall of fabric based on particle data, by combining a preset dynamic model with the effect of gravity, and to obtain the fabric height curve.

[0243] The third module 930 is used to classify the denoised photon data based on the fabric height curve to obtain the photon category and determine the crack segment.

[0244] The fourth module 940 is used to quantify the depth information of the fracture segment based on elevation-aided data.

[0245] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0246] This invention also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0247] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0248] like Figure 7 As shown, Figure 7 The hardware structure of an electronic device 1000 according to another embodiment is illustrated. The electronic device 1000 includes:

[0249] The processor 1001 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (aSIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention.

[0250] The memory 1002 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RaM). The memory 1002 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1002 and is called and executed by the processor 1001.

[0251] Input / output interface 1003 is used to implement information input and output;

[0252] The communication interface 1004 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0253] Bus 1005 transmits information between various components of the device (e.g., processor 1001, memory 1002, input / output interface 1003, and communication interface 1004);

[0254] The processor 1001, memory 1002, input / output interface 1003 and communication interface 1004 are connected to each other within the device via bus 1005.

[0255] The electronic device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0256] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0257] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0258] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0259] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0260] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0261] This invention provides an automatic ice surface crack depth extraction method, system, electronic device, storage medium, and program product based on linear cloth simulation. It acquires photon positioning and classification data of the target area, and generates target photon data and particle data through data preprocessing and meshing. The target photon data includes denoised photon data and elevation-aided data, while the particle data includes initialized cloth particles and terrain height mask particles corresponding to each mesh cell. Based on the particle data, a cloth drop simulation is performed using a preset dynamic model combined with gravity to obtain a cloth height curve. Based on the cloth height curve, photon classification is performed on the denoised photon data to determine the photon category and identify crack segments. The depth information of the crack segments is quantized based on the elevation-aided data. This invention overcomes the limitations of traditional methods that rely on fixed resolution, manual thresholds, and empirical parameters by adaptively fitting the terrain through cloth simulation of the physical process. The beneficial effects of the embodiments of the present invention are as follows: it significantly improves the universality of depth extraction of cracks of different scales, morphologies and developmental degrees; through physical-driven adaptive simulation, it reduces the dependence on human intervention and empirical parameters, thereby enhancing the detection stability and automation level in areas with weak or dense cracks, and providing more reliable three-dimensional structural data for ice shelf stability assessment.

[0262] The preferred embodiments of the present invention have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and spirit of the present invention should be within the scope of the claims of the present invention.

Claims

1. A method for automatically extracting the depth of ice surface cracks based on linear cloth simulation, characterized in that, The method includes the following steps: Photon positioning data and photon classification data of the target area are acquired, and target photon data and particle data are generated through data preprocessing and gridding. The target photon data includes denoised photon data and elevation-aided data, and the particle data includes initialized cloth particles and terrain height mask particles corresponding to each grid cell. Based on the particle data, the fabric falling is simulated by combining a preset dynamic model with the effect of gravity, and the fabric height curve is obtained. Based on the fabric height curve, the denoised photon data is classified to obtain the photon category and determine the crack segment. The depth information of the fracture segment is obtained by quantization based on the elevation-aided data; The step of classifying the denoised photon data based on the fabric height curve to obtain photon categories and determine the crack segment includes the following steps: Based on the height of the fabric particles in the fabric height curve, a continuous fabric height model along the trajectory direction is generated by linear interpolation. The first photon in the denoised photon data is taken as the candidate photon; Based on the position corresponding to the candidate photon, the fabric particles are extracted from the fabric height model as candidate fabric; If the height difference between the candidate photon and the candidate fabric is less than or equal to the third threshold, the candidate photon is classified as an ice surface photon; otherwise, the candidate photon is classified as a rift photon. The next photon in the denoised photon data is taken as the candidate photon, and the process of extracting the fabric particles from the fabric height model based on the latitude and longitude corresponding to the candidate photon is repeated until all photons in the denoised photon data have been traversed. When the number of consecutive slit photons is greater than the fourth threshold, the corresponding consecutive segment is taken as the slit segment; The endpoints of the fracture segment are determined by the start and end positions of the continuous section.

2. The method according to claim 1, characterized in that, The photon positioning data includes the time, latitude, longitude, and elevation information of the photon event. The generation of target photon data and particle data through data preprocessing and gridding includes the following steps: The photon localization data is filtered for noise using the photon classification labels in the photon classification data to obtain the denoised photon data. Based on the denoised photon data, scaled photon data is obtained by scaling the distance along the track direction; Based on the photon positioning data, the elevation auxiliary data is obtained by processing it using a preset tool. The scaled photon data is divided into grids at fixed intervals to obtain multiple grid cells; The fabric particles corresponding to each grid cell are initialized based on a preset height. The mask particles are initialized in each of the grid cells based on the elevation-aided data.

3. The method according to claim 2, characterized in that, The initialization of the mask particles in each of the grid cells based on the elevation-aided data includes the following steps: The elevation information corresponding to each grid cell is determined based on the elevation auxiliary data; The elevation of the mask particle corresponding to the grid cell is initialized based on the elevation information. When the elevation of a single mask particle is empty, the elevation of the corresponding mask particle is determined by linear interpolation based on the elevations of two adjacent mask particles. When there are multiple consecutive mask particles with empty elevations, the elevation of the corresponding mask particle is assigned a value of negative infinity.

4. The method according to claim 1, characterized in that, The process of simulating fabric fall based on the particle data, using a preset dynamic model combined with gravity, to obtain the fabric height curve includes the following steps: The fabric dynamic model is initialized using traction springs and bending springs; wherein, in the fabric dynamic model, the initial movement state of the fabric particles is movable; Based on the fabric dynamic model combined with the effect of gravity to simulate the displacement of the falling fabric, the position of the fabric particles is iteratively updated at each time step by the integral model. For each position iteration update result, a collision detection judgment is performed. If the height of the cloth particle is less than or equal to the height of the mask particle after the update, the height of the cloth particle is reset to the height of the mask particle, and the movement state corresponding to the cloth particle is updated to immovable. The position iteration update is stopped when the height change of all the fabric particles in the position iteration update is less than a first threshold, or when the number of iterations of the position iteration update reaches a second threshold. The fabric height curve is formed by the spatial shape corresponding to the height of the fabric particles after the last position iteration update.

5. The method according to claim 1, characterized in that, The fracture segment includes a continuous sequence of fracture photons and their corresponding location ranges. The process of quantizing the depth information of the fracture segment based on the elevation-aided data includes the following steps: The elevation auxiliary data is smoothed by Gaussian filtering to obtain smoothed elevation data; Based on the location range corresponding to the fracture segment, the elevation sequence corresponding to the fracture segment is extracted from the elevation smoothing data; The high-order sequence is subjected to extreme value determination. Based on the result of the extreme value determination, the fracture segment is divided into final fracture segments, and the final high-order sequence corresponding to each final fracture segment is determined according to the high-order sequence. The endpoint elevations and minimum elevations within the segment at both ends of the final fracture segment are obtained from the final elevation sequence. The depth value of the final fracture segment is determined based on the difference between the minimum endpoint elevation and the minimum elevation within the segment. Using the endpoint of the final crack segment as the base point, a corresponding point with the same elevation as the base point is searched in the opposite direction of the base point in the final elevation sequence. The width value of the final crack segment is determined based on the positional distance between the base point and the corresponding point.

6. The method according to claim 5, characterized in that, The process of determining the extreme values ​​of the high-order sequence and then dividing the fracture segment into final fracture segments based on the result of the extreme value determination includes the following steps: Iterate through the high-order sequence to obtain the minimum and maximum values; When the number of minimum values ​​is 0, the fracture segment is marked as an invalid fracture; When the number of minimum values ​​is 1, the fracture segment is taken as the final fracture segment; When the number of minimum values ​​is greater than 1, the potential fracture peak depth is determined based on the difference between the largest maximum value and the minimum elevation within the fracture segment, and the potential fracture segment depth is determined based on the difference between the minimum value of the endpoint elevations at both ends of the elevation sequence and the minimum elevation within the fracture segment. When the ratio of the peak depth of the potential fracture to the depth of the potential fracture segment is less than a fifth threshold, the fracture segment is taken as the final fracture segment. When the ratio of the potential fracture peak depth to the potential fracture segment depth is greater than or equal to the fifth threshold, all photons with height values ​​greater than the maximum value are deleted to update the target photon data and the particle data. Then, the step of simulating cloth falling based on the particle data and using a preset dynamic model combined with gravity is returned to be executed until all the fracture segments are split and the final fracture segment corresponding to each fracture segment is obtained.

7. An automatic ice surface crack depth extraction system based on linear cloth simulation, characterized in that, The system includes: The first module is used to acquire photon positioning data and photon classification data of the target area, and generate target photon data and particle data through data preprocessing and gridding; wherein, the target photon data includes denoised photon data and elevation-aided data, and the particle data includes initialized cloth particles and terrain height mask particles corresponding to each grid cell; The second module is used to simulate the fabric falling based on the particle data by combining a preset dynamic model with the effect of gravity, and to obtain the fabric height curve. The third module is used to perform photon classification on the denoised photon data based on the fabric height curve to obtain the photon category and determine the crack segment. The fourth module is used to quantize the depth information of the fracture segment based on the elevation auxiliary data; The step of classifying the denoised photon data based on the fabric height curve to obtain photon categories and determine the crack segment includes the following steps: Based on the height of the fabric particles in the fabric height curve, a continuous fabric height model along the trajectory direction is generated by linear interpolation. The first photon in the denoised photon data is taken as the candidate photon; Based on the position corresponding to the candidate photon, the fabric particles are extracted from the fabric height model as candidate fabric; If the height difference between the candidate photon and the candidate fabric is less than or equal to the third threshold, the candidate photon is classified as an ice surface photon; otherwise, the candidate photon is classified as a rift photon. The next photon in the denoised photon data is taken as the candidate photon, and the process of extracting the fabric particles from the fabric height model based on the latitude and longitude corresponding to the candidate photon is repeated until all photons in the denoised photon data have been traversed. When the number of consecutive slit photons is greater than the fourth threshold, the corresponding consecutive segment is taken as the slit segment; The endpoints of the fracture segment are determined by the start and end positions of the continuous section.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 6.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 6.

Images