A Ground Photon Extraction Method for Spaceborne Photon Counting Lidar Data

Abstract

The present invention proposes a method for extracting ground photons from spaceborne photon counting lidar data, belonging to the technical field of ground photon extraction. First, the spaceborne photon counting lidar data of the target area is segmented into multiple long segments and corresponding several short segments. The terrain index of each short segment is calculated by using the elevation difference of the long segments and the elevation difference of the short segments. The rebound distance between the fabric particles in the original fabric simulation is adjusted according to the calculated terrain index, and the adjusted fabric simulation algorithm is used to extract the ground photons from the acquired data. Considering that the spaceborne photon counting lidar data covers a wide area and the terrain differences within the area are large, the present invention adopts a segmented method, and uses the terrain indices under different short segments to accurately characterize the terrain under the corresponding short segments, extracts the ground photons of each segment, ensures the accuracy of the extracted ground photons, and improves the extraction accuracy of the ground photons.

Description

Technical Field

[0001] The present invention relates to a method for extracting ground photons from spaceborne photon counting lidar data, belonging to the technical field of ground photon extraction. Background Art

[0002] High-precision ground elevation is crucial for understanding unprecedented environmental changes. Based on spaceborne photon counting lidar data, the ground elevation over a large area can be retrieved, which is beneficial for environmental monitoring. However, this highly sensitive lidar is easily triggered by solar noise, atmospheric noise, and dark current, resulting in a large number of random noise photons, making the extracted ground photons contain noise photons, which affects the retrieval of ground elevation.

[0003] In the algorithm for extracting ground photons, there is an algorithm based on the minimum value that marks the photons in the lowest height or a lower height range as the ground. Although this method is feasible in ideal situations, vegetation coverage, terrain undulation, and residual noise photons will all lead to extraction errors. The algorithm based on the minimum value lacks consideration of various influencing factors when extracting ground photons, resulting in low extraction accuracy. On this basis, empirical mode decomposition (EMD) is introduced for precise classification. However, the ground photons extracted by EMD are uneven, with more photons distributed on flat ground than on steep slopes. Therefore, a triangulated irregular network (TIN) is used for ground photon densification. Since the triangle threshold will change in different scenarios, it is difficult to accurately give the threshold without prior knowledge, and the threshold setting also lacks a theoretical basis, unable to guarantee the accuracy of ground photon extraction, resulting in low accuracy of the extracted ground photons and unable to provide good data support for subsequent research.

[0004] In the existing processing of airborne photon counting lidar point cloud data, a cloth simulation algorithm is introduced to achieve automated processing, and the accuracy of the extracted ground photons is relatively good. Cloth simulation is a surface-based classification algorithm, where the cloth hardness directly affects the simulation result, and the cloth hardness is provided by the user. Since the area covered by airborne photon counting lidar data is small and the terrain undulation change in the whole area is not too large, the parameter values set by the user will not cause too much deviation to the whole area. Therefore, cloth simulation can meet the accuracy requirements for extracting ground photons from airborne photon counting lidar data. However, the area covered by spaceborne photon counting lidar data is wide, the photon data volume is too large, and the terrain undulation deviation in the area is too large. Still using the original cloth simulation method of user-defined cloth hardness for ground photon extraction, the set parameters cannot represent the terrain of the whole area, and the accuracy of ground photon extraction cannot be guaranteed. Summary of the Invention

[0005] The object of the present invention is to provide a method for extracting ground photons from spaceborne photon counting lidar data, so as to solve the problem of low accuracy in extracting ground photons from existing spaceborne photon counting lidar data.

[0006] The present invention proposes a method for extracting ground photons from spaceborne photon counting lidar data, and the method comprises the following steps:

[0007] 1) Obtain spaceborne photon counting lidar data of a target area;

[0008] 2) Divide the obtained spaceborne photon counting lidar data into multiple long segments and corresponding several short segments respectively according to a set first segmentation threshold and a second segmentation threshold, and calculate the terrain index of each short segment according to the elevation difference of the long segment and the elevation difference of the corresponding short segment;

[0009] 3) Calculate the rebound distance between fabric particles within the distance of the corresponding short segment according to the terrain index of each short segment;

[0010] 4) Adopt a fabric simulation algorithm according to the rebound distance between fabric particles to extract ground photons from the obtained spaceborne photon counting lidar data;

[0011] Wherein, the calculation formula of the terrain index of each short segment is:

[0012]

[0013] In the formula, TI i is the terrain index of the i-th short segment, ED si is the elevation difference of the i-th short segment, and ED li is the elevation difference of the long segment corresponding to the i-th short segment;

[0014] The calculation formula of the rebound distance of each short segment of fabric particles is:

[0015] RD i = ED i-particle * TI i

[0016] In the formula, RD i is the rebound distance of the fabric particles of the i-th short segment, ED i-particle is the maximum height difference between fabric particles in the i-th short segment, and TI i is the terrain index of the i-th short segment;

[0017] The first segmentation threshold and the second segmentation threshold are determined according to the correlation between the terrain index calculated from the corresponding long segment and short segment combination and the normalized digital terrain model, and the long segment distance and the short segment distance corresponding to the highest correlation are respectively selected as the first segmentation threshold and the second segmentation threshold.

[0018] The present invention first divides the spaceborne photon counting lidar data in the target area into long segments and short segments, calculates the terrain index of each short segment using the elevation difference of the long segment and the elevation difference of the short segment, adjusts the rebound distance between the fabric particles in the original fabric simulation according to the calculated terrain index, and extracts the ground photons from the acquired data using the adjusted fabric simulation algorithm. The present invention uses a fabric simulation algorithm based on the terrain index to extract ground photons, and uses the terrain index calculated from the actually acquired data to replace the fabric hardness in the fabric simulation algorithm. Considering that the area covered by the spaceborne photon counting lidar data is relatively large and the terrain differences within the area are significant, the present invention adopts a segmented method, calculates different terrain indices under different short segments to accurately characterize the terrain under the corresponding short segments, extracts the ground photons of each segment, ensures the accuracy of ground photon extraction, and improves the ground photon extraction accuracy; at the same time, considering that the original fabric hardness needs to be set manually, which affects the processing efficiency, the present invention uses the terrain index to replace the fabric hardness to achieve automated processing, improving the ground photon extraction efficiency.

[0019] The present invention calculates the terrain index of each short segment according to the elevation difference of the divided long segment and the corresponding elevation difference of the short segment, and uses the actually calculated terrain indices of different short segments to replace the fabric hardness in the fabric simulation. Considering the large terrain undulation factor within the area, the terrain indices under different short segments are used to accurately characterize the terrain under the corresponding short segments, ensuring the accuracy of extracting the ground photons of each short segment and improving the ground photon extraction accuracy.

[0020] Automatically adjust the rebound distance of the fabric particles using the calculated terrain index to ensure the fabric simulation result.

[0021] Furthermore, step 4) also includes noise detection and filtering of the ground photons: when the elevation difference between particles does not meet the set condition, the fabric will be damaged, and the photons at the damaged part are removed as noise.

[0022] Furthermore, the set condition is determined according to the 3σ principle, that is, when the elevation difference between particles exceeds μ + 3σ, the fabric will be damaged, where μ is the average value between particles and σ is the variance between particles.

[0023] Since there will still be residual noise in the ground photons extracted by the fabric simulation algorithm, resulting in a sudden height change of the fabric particles after simulation, which affects the accuracy of ground photon extraction, noise detection and filtering are still required. When the elevation difference between particles exceeds μ + 3σ, it proves that there is a height mutation, resulting in fabric damage, and the photons at the damaged part are noise and should be removed.

[0024] Furthermore, when performing fabric simulation, the required parameters include the spacing between fabric particles, the moving distance of particles under the action of gravity in each iteration, and the maximum height difference between fabric particles and the corresponding photons.

[0025] Furthermore, the spacing between the fabric particles is determined according to the number of ground photons extracted. When the number of ground photons extracted is larger, the spacing between the fabric particles is smaller.

[0026] Furthermore, in each iteration, the moving distance of the particle under the action of gravity is 9.8 m.

[0027] Furthermore, the maximum height difference between the fabric particle and the corresponding photon is determined according to the number of ground photons extracted. When the number of ground photons extracted is larger, the maximum height difference between the fabric particle and the corresponding photon is smaller.

[0028] When performing fabric simulation, other parameters also need to be input, and the fabric simulation is carried out according to these parameters. The spacing between the fabric particles and the maximum height difference between the fabric particle and the corresponding photon in the parameters can both be determined according to the actual accuracy requirements for the extracted ground photons to meet the actual data requirements of users.

[0029] Since different combinations of long segments and short segments will directly change the terrain index, thereby affecting the extraction accuracy of subsequent ground photons, the optimal segment combination of the target area is determined according to the above process, so that the correlation between the terrain index calculated by the combination of long segments and short segments and the normalized digital terrain model is the highest, ensuring the accuracy of the calculated terrain index, and further ensuring the extraction accuracy of subsequent ground photons. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is the flowchart of ground photon extraction from spaceborne photon counting lidar data of the present invention;

[0031] Fig. 2(a) is a typical measurement scene of a flat area of spaceborne photon counting lidar;

[0032] Fig. 2(b) is a typical measurement scene of a steep area of spaceborne photon counting lidar;

[0033] Figure 3 is the schematic diagram of particle noise in the fabric simulation result;

[0034] Fig. 4(a) is the ground photon extraction result of bare land in a flat terrain in the experiment;

[0035] Fig. 4(b) is the ground photon extraction result of bare land in a steep terrain in the experiment;

[0036] Fig. 4(c) is the ground photon extraction result of a vegetation area in a flat terrain in the experiment;

[0037] Fig. 4(d) is the ground photon extraction result of a vegetation area in a steep terrain in the experiment;

[0038] Figure 5(a) is a scatter plot comparing the ground photon height extracted by the method of the present invention in the experiment with the G-LiHT DTM height;

[0039] Figure 5(b) is a scatter plot comparing the ground photon height extracted by the minimum value algorithm in the experiment with the G-LiHT DTM height;

[0040] Figure 5(c) is a scatter plot comparing the ground photon height extracted by the minimum percentile algorithm in the experiment with the G-LiHT DTM height;

[0041] Figure 5(d) is a scatter plot comparing the ground photon height extracted by the EMD algorithm in the experiment with the G-LiHT DTM height;

[0042] Figure 6 is the RMSE box plot under different vegetation heights obtained by the method of the present invention in the experiment;

[0043] Figure 7 is the RMSE box plot under different slopes obtained by the method of the present invention in the experiment. Detailed implementation manners

[0044] The following further explains the detailed implementation manners of the present invention with reference to the accompanying drawings.

[0045] The present invention proposes a method for extracting ground photons from spaceborne photon counting lidar data, which uses a cloth simulation method improved based on terrain index to extract ground photons. Since the cloth hardness in the original cloth simulation method needs to be set by the user, and the cloth hardness directly affects the simulation result. For the problem that the spaceborne data covers a wide area, the custom cloth hardness cannot accurately characterize the terrain undulation in the area, which directly affects the extraction accuracy of ground photons. Therefore, the present invention uses terrain index to replace the cloth hardness to improve the extraction accuracy of ground photons. The specific process of this method is as Figure 1 shown. First, the spaceborne photon counting lidar data of the target area is segmented into multiple long segments and corresponding several short segments. The terrain index of each short segment is calculated using the elevation difference of the long segment and the elevation difference of the short segment. The rebound distance between cloth particles in the original cloth simulation is adjusted according to the calculated terrain index, and the adjusted cloth simulation algorithm is used to extract ground photons from the acquired data.

[0046] Step 1. Obtain data

[0047] The present invention needs to obtain spaceborne photon counting lidar data of the target area. Since the ICE-sat satellite (Ice, Cloud, and Land Elevation Satellite) can obtain the elevations of ice sheets, forests, oceans, etc., providing a reliable basis for monitoring global climate change, the present invention uses the ICE-sat satellite to obtain photon data of the target area. In this embodiment, the ICESat-2 ATL03 data is used. ICESat-2 provides 21 data products (ATL00 - ATL21). ATL03 (Global Geolocated Photons) provides the time, longitude, latitude, altitude, and signal confidence tags for each photon, and is the only source of photon information required for land vegetation along-track products and other advanced products. In this embodiment, 16 data for May and June 2019 are collected, and photons with a signal confidence label greater than 2 are marked as signals, obtaining the signal photons within the target area.

[0048] Step 2. Data segmentation

[0049] The present invention divides the signal photon data into multiple long segments according to a set first segmentation threshold, and divides the signal photon data into multiple short segments according to a set second segmentation threshold. One long segment corresponds to multiple short segments. Since different combinations of long segments and short segments will directly affect the value of the terrain index, in order to ensure the consistency of the terrain index obtained from the long segments and short segments after segmentation with the real ground, it is necessary to ensure reasonable segmentation of the long segments and short segments. Therefore, it can be determined according to the correlation between the terrain index and the normalized digital terrain model (DTM). By calculating the correlation between the terrain index under different combinations and the normalized DTM within the corresponding area, and selecting the combination with the highest correlation, the segmentation threshold for the long segments and the segmentation threshold for the short segments can be determined. For example, the long segment segmentation range is 500 - 1000m (interval is 100m), and the short segment segmentation range is 50 - 100m (interval is 10m). Calculate the correlation between the terrain index and the DTM in different combination methods. The results are shown in Table 1. It is obtained that when the short segment is 70m and the long segment is 800m, the correlation is the highest (95.28%), proving that this line segment combination can best describe the signal terrain of the research area. Therefore, the first segmentation threshold is set to 800m, and the second segmentation threshold is set to 70m.

[0050] Table 1:

[0051]

[0052] Step 3. Terrain index calculation

[0053] The core idea of cloth simulation is that the shape of the cloth is related to the object it covers. When the cloth covers the point cloud, its shape is the Digital Surface Model (DSM); when the cloth covers the inverted point cloud, its shape is the Digital Terrain Model (DTM). The cloth is composed of a set of cloth particles and the springs between them, and the simulation is achieved by iteratively positioning the particles. In each iteration, the forces acting on the cloth particles are decomposed into gravity and spring forces in the vertical direction. The cloth particles first fall under the action of gravity and then bounce back under the action of elasticity. Therefore, the rebound rate of the cloth particles affects the simulation results.

[0054] The rebound rate of the cloth particles is determined by the cloth hardness provided by the user, as shown in formula (1), and this formula decreases as the terrain undulation increases.

[0055]

[0056] Where RD is the rebound distance, and ED particle is the height difference between cloth particles. Hardness is the cloth hardness. When the cloth hardness is greater, the particle rebound is smaller and the deformation of the cloth is smaller. When the hardness values are 1, 2, and 3, the moving distances are 1 / 2, 3 / 4, and 7 / 8 of the height difference between cloth grains respectively.

[0057] To make the cloth simulation applicable to spaceborne photon counting lidar data and calculate the rebound distance of cloth particles more accurately, the terrain index is used instead of the cloth hardness, where the terrain index is designed based on the observed values of typical measurements. The height difference is composed of terrain and ground objects, as Figure 2(a)-2(b) shown below, specifically:

[0058] 1) The height difference in flat areas is mainly contributed by ground objects, and the height difference in steep areas is mainly caused by the terrain;

[0059] 2) The height difference at close range is mainly contributed by ground objects, and the height difference at long range is mainly caused by the terrain.

[0060] Therefore, comparing the height differences of signal photons at different distances can reflect the terrain undulation, although it is difficult to obtain the height difference of ground objects in advance. Calculate the terrain index of each short segment according to the segmented long segment and short segment, and its formula is as follows:

[0061]

[0062] In the formula, TI i is the terrain index of the i-th short segment, ED si is the elevation difference of the i-th short segment, and ED li is the elevation difference of the long segment corresponding to the i-th short segment.

[0063] Calculate the rebound distance of the fabric particles for each short segment based on the terrain index obtained, using the following formula:

[0064]

[0065] In the formula, RD i is the rebound distance of the fabric particles in the i-th short segment, ED i-particle is the maximum height difference between the fabric particles in the i-th short segment, and TI i is the terrain index of the i-th short segment.

[0066] Step 4. Fabric simulation

[0067] When performing fabric simulation, in addition to the rebound distance calculated based on the terrain index, the parameters required also include the distance between fabric particles, the moving distance of the particles under the action of gravity in each iteration, and the maximum height difference between the fabric particles and the corresponding photons. Among them, the distance between fabric particles refers to the length of the interval between fabric particles. Since the distance between fabric particles determines the distance between the extracted ground photons, this parameter can be set according to actual requirements. When the user needs high-precision ground photon data, the distance between fabric particles can be set smaller, so that more photons are extracted. The moving distance of the particles under the action of gravity in each iteration is 9.8 m. The maximum height difference between the fabric particles and the corresponding photons is also related to the actual needs of the user. When the user needs high-precision ground photon data, the maximum height difference can be set smaller. In this embodiment, the distance between fabric particles is set to 10 m, and the maximum height difference is set to 0.3 m.

[0068] Step 5. Output ground photons

[0069] There will be certain noise remaining in the initial ground photons extracted from the fabric simulation results. As Figure 3 shown, the remaining noise will cause sudden height changes in the fabric particles after simulation. Since the noise is far from the signal, the spring force at the position of the residual noise is much greater than the average value. Since the remaining noise will cause sudden changes in the elevation of the fabric particles, therefore, in the present invention, by calculating the elevation difference between the particles and their average value μ and variance σ, according to the 3σ principle, when the elevation difference between the particles exceeds μ + 3σ, the fabric will be damaged. Therefore, the photons at the damaged position are marked as noise photons and removed, and the positions of the fabric particles are recalculated.

[0070] To further verify the accuracy of the ground photons extracted by the method of the present invention, verification is carried out from two aspects: the extraction accuracy and the influencing factors of the adaptive fabric simulation.

[0071] A. Accuracy verification

[0072] Photons with a signal confidence label greater than 2 are regarded as signals, and the ground photons are extracted from them using the method of the present invention. The cloth particle spacing is set to 10 m, the moving distance of the particles under the action of gravity in each iteration is set to 9.8 m, and the maximum height difference is set to 0.3 m.

[0073] The results are as Figure 4(a)-4(d) shown. The extracted ground photons visually conform to the actual ground. Specifically, from Figure 4(a)-4(d) it can be found that vegetation cover and terrain have little effect on the cloth simulation performance, indicating that the cloth simulation based on the terrain index has terrain perception and adaptive capabilities. In addition, since the cloth particle spacing is preset, the extracted ground photons are evenly distributed even on steep mountaintops or valleys. Although there are many noise photons remaining underground in the original signal photons, the ground photons extracted by the method of the present invention do not contain any noise photons, and the cloth breaks at the positions of the noise photons and is recalculated to the positions of the real ground photons, indicating that the rag (segmentation) is effective.

[0074] To quantitatively evaluate the adaptive compressive sensing performance of the cloth simulation, the heights of the extracted ground photons and the DTM are compared. The other three algorithms (the minimum value algorithm, the minimum percentile algorithm, and the EMD algorithm) are used for comparison with the method of the present invention. Using formulas (4), (5), and (6), the R 2 , mean absolute error (MAE), and root mean square error (RMSE) of each algorithm are calculated.

[0075]

[0076] where n is the number of ground photons, e i is the elevation of the i-th ground photon, and r i is the elevation of the digital terrain model with a resolution of 1 m generated by the corresponding thermal imager (G-LiHT).

[0077] The ground photons are extracted using the algorithm of the present invention and the other three algorithms, and the elevations of the extracted photons are compared with the reference data (G-LiHT DTM). The scatter plot is as Figure 5(a)-5(d) shown. Since the R 2 is greater than 0.9997, there is a strong consistency between the extracted photons and the reference ground height. The fitting curve is close to 1:1, indicating that the method of the present invention and the other three algorithms are all effective. Specifically, the two minimum value-based algorithms (Fig. 5(b) and Fig. 5(c)) have the lowest accuracy, and these two algorithms prove that the lack of terrain adaptability and residual noise detection ability will reduce the accuracy.

[0078] The accuracy of the EDM-based algorithm is only slightly worse than that of the method of the present invention, indicating that both algorithms have terrain perception capabilities. However, when comparing the average spacing of ground photons (i.e., the data length divided by the number of ground photons), it is found that the method of the present invention (10 m / pts) is much better than the EDM-based algorithm (12.9 m / pts). Probably due to EMD, fewer ground photons are extracted in steep areas. Obviously, the method of the present invention can not only extract sufficient ground photons but also has higher accuracy.

[0079] B. Influence factor verification of adaptive cloth simulation

[0080] Since the canopy height and terrain undulation are factors affecting the extraction of ground photons, their performance under different canopy heights and slopes is determined through experiments. Therefore, the RMSE between the ground photons extracted by the method of the present invention and the reference data in different scenarios is calculated, and box plots are drawn (as Figure 6 and Figure 7 shown).

[0081] Among them, the box plot of RMSE under different canopy heights is as Figure 6 shown. The canopy height is divided into 0 - 5 m, 5 - 10 m, 10 - 15 m, 15 - 20 m, and >20 m. As the canopy height increases, the median RMSE increases from 3.24 m to 3.64 m. In addition, as the canopy height increases, the range of RMSE becomes more concentrated. Specifically, when the canopy height is greater than 20 m, the RMSE range is less than 0.8 m. After analysis, the method of the present invention is hardly affected by the canopy because the extraction idea of cloth simulation is to invert the signal photons and place the simulated cloth on top of the inverted photons. Therefore, the decrease in accuracy is more likely due to the absence of ground photons under high canopies, and the method of the present invention is robust under different canopy heights.

[0082] Among them, the box plot of RMSE under different slopes is as Figure 7 shown. The slopes are divided into 0 - 2, 2 - 5, 5 - 15, 15 - 30, and >30. Figure 7 Among them, the accuracy is the highest when the slope is 0 - 2, and the median RMSE is 3.37 m. The accuracy is the lowest when the slope >30°, and the median RMSE is 3.50 m. As the slope increases, the accuracy of the method of the present invention decreases slightly, indicating that the cloth simulation based on the terrain index is effective and can make the simulated cloth fully fit the actual ground surface. In addition, as the slope increases, the RMSE range also shrinks, proving that the broken cloth can effectively identify and eliminate the residual noise photons in steep areas.

[0083] Through the above two verification methods, it can be seen that compared with the ground photon extraction methods of other spaceborne photon counting lidars, the extraction accuracy of the present invention is the highest. Moreover, due to the excessive slope in the steep area, in the existing methods, many photons in the original data are considered non-ground photons in places with a large slope, resulting in a small number of ground photons extracted in the steep area. However, the terrain index is introduced in the cloth simulation algorithm of the present invention, fully considering the influence of terrain undulation, and sufficient ground photons can be extracted in the steep area, and the extraction accuracy is higher. At the same time, the method of the present invention is hardly affected by the canopy height and terrain undulation when extracting ground photons, and can ensure the extraction accuracy under different influencing factors.

Claims

1. A method for extracting ground photons from spaceborne photon counting lidar data, characterized in that, the method comprises the following steps: 1) Obtain spaceborne photon counting lidar data of the target area; 2) Divide the obtained spaceborne photon counting lidar data into multiple long segments and corresponding several short segments according to a set first segmentation threshold and a second segmentation threshold respectively, and calculate the terrain index of each short segment according to the elevation difference of the long segment and the elevation difference of the corresponding short segment; 3) Calculate the rebound distance between fabric particles within the distance of the corresponding short segment according to the terrain index of each short segment; 4) Adopt a fabric simulation algorithm according to the rebound distance between fabric particles to extract ground photons from the obtained spaceborne photon counting lidar data; wherein, the calculation formula of the terrain index of each short segment is: where TI i is the topographic index of the i-th short segment, and ED si is the elevation difference of the i-th short segment, and ED li is the elevation difference of the long segment corresponding to the i-th short segment; The calculation formula of the rebound distance of each short segment fabric particle is: RD i = ED i-particle * TI i wherein, RD i is the rebound distance of the fabric particles in the i-th short segment, ED i-particle is the maximum height difference between the fabric particles in the i-th short segment, TI i is the terrain index of the i-th short segment; The first segmentation threshold and the second segmentation threshold are determined according to the correlation between the terrain index calculated by the combination of the corresponding long segment and short segment and the normalized digital terrain model, and the corresponding long segment distance and short segment distance when the correlation is the highest are respectively selected as the first segmentation threshold and the second segmentation threshold.

2. The method for extracting ground photons from spaceborne photon counting lidar data according to claim 1, characterized in that, Step 4) further includes noise detection and filtering of ground photons: when the elevation difference between particles does not meet the set conditions, the fabric will be damaged, and the photons at the damaged part are removed as noise.

3. The method for extracting ground photons from spaceborne photon counting lidar data according to claim 2, characterized in that, The set conditions are determined according to the 3σ principle, that is, when the elevation difference between particles exceeds μ + 3σ, the fabric will be damaged, where μ is the average value between particles and σ is the variance between particles.

4. The method for extracting ground photons from spaceborne photon counting lidar data according to claim 1, characterized in that, When performing fabric simulation, the required parameters include the fabric particle spacing, the moving distance of particles under the action of gravity in each iteration, and the maximum height difference between fabric particles and corresponding photons.

5. The method for extracting ground photons from spaceborne photon counting lidar data according to claim 4, characterized in that, The fabric particle spacing is determined according to the number of ground photons extracted. When the number of ground photons extracted is more, the fabric particle spacing is smaller.

6. The method for extracting ground photons from spaceborne photon counting lidar data according to claim 4, characterized in that, The moving distance of particles under the action of gravity in each iteration is 9.8 m.

7. The method for extracting ground photons from spaceborne photon counting lidar data according to claim 4, characterized in that, The maximum height difference between fabric particles and corresponding photons is determined according to the number of ground photons extracted. When the number of ground photons extracted is more, the maximum height difference between fabric particles and corresponding photons is smaller.

Images