Method and system for inverting forest terrain based on photon point cloud data using multi-level window detection

Abstract

The present invention provides a method and system for inverting forest terrain using photon point cloud data based on multi-level window detection. The method comprises: obtaining ground photon points under the forest based on the multi-level window terrain detection method; obtaining terrain elevation residuals under multi-level window sizes based on the ground photon points under the forest; diverting the terrain elevation residuals based on a preset residual threshold to obtain deviation points of different categories; correcting the deviation points of different categories and fusing the ground photon points under multi-level window sizes based on the correction results to obtain corrected ground photon points under a small window; and obtaining forest terrain inversion results based on the corrected ground photon points under the small window. The present invention proposes a method for inverting forest terrain using multi-level window detection. By fusing the forest terrain inverted under windows of different sizes, it resists residual noise photons and terrain errors in dense vegetation areas, thereby obtaining a terrain result with rich terrain details and high precision.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-precision forest understory terrain inversion, and specifically relates to a method and system for forest understory terrain inversion based on photon point cloud data of multi-level window detection. Background Art

[0002] Methods for inverting understory terrain from photon point cloud data are based on filtered photon point clouds. The filtering results directly affect the accuracy of understory terrain inversion. Existing filtering methods cannot completely separate signal photons from noise photons, and some noise photons adjacent to signal photons still remain in the filtered results. These residual noise photons are located near ground photons and photons at the top of the canopy, which makes it difficult to invert understory terrain and forest height. In addition, in areas with dense vegetation, the sparse sampling of ground signal photons makes it impossible to accurately identify ground photons. In summary, the accuracy of understory terrain inversion is primarily limited by two factors: residual noise near the ground and sparse ground photons beneath dense vegetation. Achieving high-precision understory terrain inversion requires overcoming these constraints.

[0003] Previous studies have shown that large-sized terrain detection windows can accurately obtain ground photon points in densely vegetated areas, but this reduces the terrain resolution, making the inverted terrain relatively rough and difficult to depict terrain details. Although small-sized terrain detection windows can describe more detailed terrain, it is difficult to obtain ground photon points in densely vegetated areas, and residual noise photons below the ground may also be mistakenly selected as ground photon points. In view of this, the present invention proposes a method and system for inverting forest terrain based on photon point cloud data using multi-level window detection. Summary of the Invention

[0004] To address the problems existing in the prior art, the present invention provides a method and system for inverting forest terrain using photon point cloud data based on multi-level window detection. First, ground photon points are acquired in a large window, which have higher accuracy. Second, terrain photon points are acquired in a small window. These ground photon points have better terrain details, but may contain noise photons near the ground and canopy photons above the ground. Therefore, the present invention verifies the terrain in the small window with the terrain acquired in the large window to generate the correct terrain.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for inverting forest terrain based on photon point cloud data using multi-level window detection, the method comprising:

[0007] Based on the multi-level window terrain detection method, the ground photon points under the forest are obtained; the multi-level window includes a large window and a small window; the large window is a window of preset large size, and the small window is a window of preset small size;

[0008] Based on the forest ground photon points, obtaining terrain elevation residuals in multiple window sizes;

[0009] Dividing the terrain elevation residual based on a preset residual threshold to obtain deviation points of different categories;

[0010] Correcting the deviation points of different categories and fusing the ground photon points under multiple window sizes based on the correction results to obtain the corrected ground photon points under the small window;

[0011] Based on the corrected ground photon points in the small window, an understory terrain inversion result is obtained.

[0012] Preferably, the method for obtaining the forest ground photon points includes:

[0013] The multi-level window forest ground photon point cloud obtained was filtered based on iterative median filtering to obtain the potential ground photon benchmark;

[0014] Based on the elevation histogram frequency statistics method, the peak value of the multi-level window forest ground photon point cloud is obtained;

[0015] The understory ground photon point is obtained based on the position of the maximum peak within the preset distance of the potential ground photon benchmark.

[0016] Preferably, the method for obtaining the terrain elevation residual under multiple window sizes includes:

[0017] The ground photon points obtained under the large window are used to generate continuous terrain through a cubic spline function, and the elevation of the ground photon points obtained under the small window and the terrain elevation residual at the same position in the continuous terrain generated by the large window are obtained; wherein, the terrain elevation residual represents the difference between the ground photon points obtained under the large window and the small window sizes.

[0018] Preferably, the different categories of deviation points include no deviation points, first category deviation points and second category deviation points.

[0019] Preferably, the method for obtaining the corrected ground photon point under the small window includes:

[0020] Obtaining a starting range of a buffer zone around the terrain of the first type of deviation point;

[0021] Based on the starting range of the buffer zone around the terrain, comparing the number of ground photons in the buffer zone around the terrain inverted by the large window and the small window;

[0022] When the number of ground photons in the buffer zone around the large window terrain is greater than the amount of ground photon data in the buffer zone around the small window terrain, the ground photon point obtained by the small window is determined to be a residual noise photon point, and the residual noise photon point is replaced by the elevation interpolation point under the large window; otherwise, the ground photon point obtained by the small window is determined to be correct;

[0023] Perform photon deviation point mutation detection on the second-type deviation points, and compare the number of ground photons in the buffer zone around the terrain inverted by the large window and the small window for the second-type deviation points that are not located in the preset mutation zone; otherwise, the second-type deviation points are determined to be non-ground photon points obtained by the small window, and the non-ground photon points are replaced by elevation interpolation points at the same position under the large window;

[0024] The correct ground photon points and the ground photon points replaced by the large window elevation interpolation points are fused to obtain the corrected ground photon points under the small window; among them, the correct ground photon points include the deviation-free points.

[0025] The present invention also provides a forest terrain inversion system based on photon point cloud data of multi-level window detection, which is used to implement the method. The system includes:

[0026] The photon point acquisition module is used to obtain photon points on the ground under the forest based on a multi-level window terrain detection method; wherein the multi-level window includes a large window and a small window; the large window is a window of a preset large size, and the small window is a window of a preset small size;

[0027] A residual acquisition module, used for obtaining terrain elevation residuals in multiple window sizes based on the forest ground photon points;

[0028] A residual diversion module is used to divert the terrain elevation residual based on a preset residual threshold to obtain deviation points of different categories;

[0029] A residual correction module is used to correct the deviation points of different categories and fuse the ground photon points under multiple window sizes based on the correction results to obtain the corrected ground photon points under the small window;

[0030] The inversion result acquisition module is used to obtain the forest understory terrain inversion result based on the corrected ground photon points in the small window.

[0031] Preferably, the photon point acquisition module includes:

[0032] A median filter unit is used to filter the acquired multi-level window forest ground photon point cloud based on iterative median filtering to obtain a potential ground photon benchmark;

[0033] A peak acquisition unit is used to obtain the peak value of the ground photon point cloud under the forest in a multi-level window based on the elevation histogram frequency statistics method;

[0034] A photon point acquisition unit is used to obtain the forest ground photon point based on the position of the maximum peak within a preset distance of the potential ground photon benchmark.

[0035] Preferably, in the residual acquisition module, the ground photon points obtained under the large window are used to generate continuous terrain through a cubic spline function to obtain the elevation of the ground photon points obtained under the small window and the terrain elevation residual of the same position in the continuous terrain generated by the large window; wherein, the terrain elevation residual represents the difference between the ground photon points obtained under the large window and the small window sizes.

[0036] Compared with existing technologies, the present invention offers the following advantages: It proposes a multi-level window detection method for inverting understory terrain. By fusing understory terrain inverted using windows of different sizes, it mitigates residual noise photons and terrain errors in dense vegetation areas, resulting in a highly detailed and accurate topographic result. This method not only captures detailed terrain but also achieves high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0038] Figure 1 This is a flowchart of obtaining ground photon points according to an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of the elevation residuals between the ground photon points obtained in the small window and the terrain in the large window according to an embodiment of the present invention;

[0040] Figure 3 Schematic diagram of ground photons, continuous terrain, and a buffer zone around the terrain obtained in two-level windows according to an embodiment of the present invention; (a) indicates that the small window obtains an erroneous ground photon point; (b) indicates that the small window obtains a correct ground photon point;

[0041] Figure 4 Schematic diagram of deviation point mutation detection according to an embodiment of the present invention; (a) shows an erroneous ground photon point acquired in a small window; (b) shows a deviation point with no elevation mutation in elevated terrain; (c) shows a deviation point with no elevation mutation in mountaintop terrain; and (d) shows a deviation point with no elevation mutation in terrain where flat land meets a steep slope.

[0042] Figure 5 This is a flow chart of a method for inverting forest terrain using photon point cloud data based on multi-level window detection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] Example 1

[0046] A method for inverting forest terrain based on photon point cloud data using multi-level window detection, including:

[0047] S1: Based on the multi-level window terrain detection method, the ground photon points under the forest are obtained; wherein the multi-level window includes a large window and a small window, and the window range is [100m, 50m, 30m, 20m]. In this embodiment, the large window is a large window with a preset large size, and the small window is a small window with a preset small size. The photon point cloud data in this embodiment is obtained by ICESat-2 (Ice, Cloud and land Elevation Satellite-2), that is, the Ice, Cloud and Land Elevation Satellite-2. ICESat-2 obtains photon point cloud data through its Advanced Topographic Laser Altimeter System (ATLAS) and has a specific along-track distance design. The along-track distance refers to the distance between the projection points of adjacent laser pulses on the surface. The ATLAS system of ICESat-2 is designed with an along-track spacing of approximately 0.7 meters, which means that the light spots formed by each pair of laser pulses on the surface are approximately 0.7 meters apart. This design enables ICESat-2 to capture the terrain changes on the earth's surface with high resolution.

[0048] A further embodiment is that the method for obtaining ground photon points under the forest includes:

[0049] S11: Filter the acquired multi-level window forest ground photon point cloud based on iterative median filtering to obtain the potential ground photon benchmark.

[0050] In this embodiment, the iterative median filtering method and the elevation histogram statistics method are introduced to jointly constrain the acquisition of accurate ground photon index. The ground photon benchmark is roughly determined by iterative median filtering, and the potential ground photon benchmark output by iterative median filtering is matched with the peak value of the elevation frequency histogram statistics, such as Figure 1 The general process of iterative median filtering in the present invention is shown in formula (1):

[0051]

[0052] Where Ω represents the photon point cloud set within each window, Z represents the elevation value, and median(·) represents the median value. For the forested area, after five iterations of median filtering, the median elevation value Y2 output during the second iteration of median filtering falls between the vegetation canopy and the ground. Therefore, the present invention uses the median elevation values ​​Y3, Y4, and Y5 output from the third to fifth iterations of median filtering as alternative references for the final output ground point.

[0053] S12: Based on the elevation histogram frequency statistics method, obtain the peak value of the multi-level window forest ground photon point cloud. Figure 1 As shown, the difference between Y3 and Y5 is first compared. If the difference is less than the threshold value of 3m, the ground peak index is selected according to formula (2).

[0054] Groundpeak=f_max{N(c_peak(Y3)),N(c_peak(Y4)),N(c_peak(Y5))} (2)

[0055] Where, Groundpeak represents the position of the ground peak, N(·) represents the number of photons, c_peak(Y) represents the peak closest to Y, and f_max{·} represents the position of the maximum value within {·}. In the present invention, it represents the position of the maximum peak within {·}. If the difference between Y3 and Y5 is greater than the threshold value 3m, then the difference between Y3 and Y4 and the difference between Y4 and Y5 are compared, and the group with the smaller difference is used as the alternative benchmark. Figure 1 As shown in Figure 2, if the difference between Y3 and Y4 is greater than the difference between Y4 and Y5, the ground peak index is selected according to formula (3); if the difference between Y3 and Y4 is less than the difference between Y4 and Y5, the ground peak index is selected according to formula (4). Finally, the average value of the photons where the searched ground peak is located is output as the ground photon point in the window, and the ground photon points output in each window are fitted to the ground trend through cubic spline interpolation.

[0056] Groundpeak=f_max{N(c_peak(Y4)),N(c_peak(Y5))} (3)

[0057] Groundpeak=f_max{N(c_peak(Y3)),N(c_peak(Y4))} (4)

[0058] The understory ground photon point is obtained based on the position of the maximum peak within a preset distance of the potential ground photon benchmark.

[0059] S2: Based on the ground photon points under the forest, the terrain elevation residuals with multiple window sizes are obtained.

[0060] Generating terrain using the above-mentioned method of acquiring ground photon points within a single-sized window is not rational. Larger window sizes can mitigate the sparseness of ground photons in densely vegetated areas and the residual noise photons near the ground, ensuring the accuracy of ground photon point selection. However, terrain generated from ground photons acquired with a large window size poorly reflects topographic details, resulting in large errors in the resulting terrain. While terrain acquired with a small window size can well reflect detailed terrain, it exhibits significant errors in the inverted terrain in densely vegetated areas and when there are a high number of residual noise photons. In densely vegetated areas, the sparseness of ground photons leads to canopy photons being mistaken for ground photons; in areas with a high number of residual noise photons, some noise photons are mistakenly classified as ground photons.

[0061] In view of this, the present invention fuses the forest terrain obtained under multiple levels of windows. The forest terrain obtained under a small-size window will be checked by the forest terrain obtained under a large-size window to eliminate errors in the terrain under the small window.

[0062] A further embodiment is that the method for obtaining terrain elevation residuals under multiple window sizes includes:

[0063] The ground photon points obtained under the large window are used to generate continuous terrain through a cubic spline function, and the elevation of the ground photon points obtained under the small window and the terrain elevation residual at the same position in the continuous terrain generated by the large window are obtained; among them, the terrain elevation residual represents the difference between the ground photon points obtained under the large window and the small window sizes.

[0064] Specifically, first, the ground photon points obtained under the large window are used to generate a continuous terrain through a cubic spline function, and the elevation of the ground photon points obtained in the next level window is compared with the residual of the continuous terrain in the large window, as shown in formula (5):

[0065]

[0066] Where, They represent the along-track distance and elevation of the ground photon point obtained under the large window; Respectively represent the along-track distance and elevation of the ground photon point obtained in the small window; D = spline (A, B; C) represents the cubic spline interpolation function, where A and B are the interpolation point vectors, C is the horizontal coordinate vector of the point to be found, and D is the vertical coordinate vector value of the point to be found; Indicates the residual difference in terrain elevation between the ground photon point obtained in the small window and the same position in the large window. It represents the difference of ground photon points obtained under two levels of window size, such as Figure 2 shown.

[0067] S3: Based on the preset residual threshold, the terrain elevation residual is divided into different categories of deviation points; a further embodiment is that the different categories of deviation points include no deviation points, first-class deviation points and second-class deviation points. In this embodiment, no deviation points are ground photon points with an absolute value less than 0.5m; first-class deviation points are elevation residuals. The ground photon points with elevation residuals less than -0.5m or greater than 0.5m and less than 2m. The second type of deviation points are ground photon points with elevation residuals greater than 2m.

[0068] S4: Correct the deviation points of different categories and fuse the ground photon points under multiple window sizes based on the correction results to obtain the corrected ground photon points under the small window.

[0069] A further embodiment is that the method for obtaining the corrected ground photon point in the small window includes:

[0070] For ground photon points whose absolute value of elevation residual is less than 0.5m, it is considered that these ground photon points are consistent with the terrain inverted under the large window, and the ground photon points at this location can be retained.

[0071] Get the starting range of the buffer zone around the terrain of the first type of deviation point; Specifically, in this embodiment, for the elevation residual Ground photon points with a distance less than -0.5m or greater than 0.5m and less than 2m are considered to have deviations from the terrain inverted under the large window. This type of deviation may be due to mistaking noise photon points near the ground for ground photon points under the small window, such as Figure 3 (a); It may also be that the terrain obtained under a large window cannot reflect the terrain details, such as Figure 3 (b), so it is necessary to determine the source of the deviation. Considering that the number of residual noise photons is smaller than that of ground signal photons, the source of the deviation is determined by comparing the number of photons in the buffer zone around the terrain inverted under different size windows. Before comparing the number of photon points in the buffer zone, the range of the buffer zone should be determined. For isolated deviation points (such as Figure 3 (a) Dashed box), the ground points with no deviation on its left and right sides are used as the starting range of the buffer zone. For consecutive and adjacent deviation ground points, they can be first clustered and merged into one class, and then the ground points with no deviation on its left and right sides are used as the starting range of the buffer zone.

[0072] Based on the starting range of the buffer zone around the terrain, the number of ground photons in the buffer zone around the terrain inverted by the large window and the small window is compared;

[0073] When the number of ground photons in the buffer zone around the large window terrain is greater than the amount of ground photon data in the buffer zone around the small window terrain, the ground photon points obtained in the small window are judged to be residual noise photon points, and the residual noise photon points are replaced with the elevation interpolation points under the large window; otherwise, the ground photon points obtained in the small window are judged to be correct, that is, the ground photon points obtained in the large window cannot reflect the terrain details, and these ground photon points obtained in the small window are considered to be correct.

[0074] The second type of deviation points are subjected to photon deviation point mutation detection, and the second type of deviation points that are not located in the preset mutation area are compared with the number of ground photons in the buffer zone around the terrain inverted by the large window and the small window; otherwise, the second type of deviation points are determined to be non-ground photon points obtained by the small window, and the elevation interpolation points at the same position under the large window are used to replace the non-ground photon points; specifically, for ground photon points with elevation residuals greater than 2m, these ground photons are also considered to have deviations from the terrain inverted under the large window. The deviation may come from mistaking canopy photon points for ground photon points under the small window, as shown in the figure. Figure 4 (a); It may also come from scenes with rapid changes in terrain, such as ground uplift, mountain tops, and the junction of flat land and steep slopes. When determining the source of deviation points, continuous and adjacent deviation ground points should be clustered into one category, such as Figure 4 (a)-(c). If the canopy photon points are mistaken for ground photon points, it will cause large elevation changes on both sides of these deviated ground points, resulting in large changes in the slopes of the correct ground points on both sides, such as Figure 4 (a). Figure 4 As shown in (c), although the flat ground uplift and the topography of the mountain will also lead to the existence of deviation points, this terrain changes slowly. Figure 4 As shown in (d), for the terrain where the flat land meets the steep slope, only the slope of the correct ground point at one end of the flat land changes greatly, such as Figure 4 (d). Therefore, the slope change before and after the correct ground point close to the deviation point can be detected to distinguish Figure 4 (a) and Figure 4 The sources of the deviation points in (b)-(d). If the slope change of at least one of the correct ground points on the left and right sides is less than the threshold, the deviation point is considered not to be a canopy photon and is misclassified as a ground point. Finally, the ground photon points in the two-level window are fused using the strategy for determining the number of photons in the buffer zone described above. Conversely, if the slope change of the correct ground points on the left and right sides is both greater than the threshold, the deviation point is likely to be an incorrect ground point. The deviated ground photon point can be assumed to originate from the canopy, and these non-ground photon points are replaced with elevation interpolation points at the same location in the large window. Figure 5 Flowchart for detecting terrain with multi-level windows.

[0075] The correct ground photon points and the ground photon points replaced by the large window elevation interpolation points are fused to obtain the corrected ground photon points under the small window; among them, the correct ground photon points include the deviation-free points.

[0076] S5: Based on the corrected ground photon points in the small window, the understory terrain inversion result is obtained.

[0077] This paper proposes a method for inverting forest terrain using multi-level window detection. By fusing inverted forest terrain using windows of different sizes, the method mitigates residual noise photons and terrain errors caused by dense vegetation, resulting in highly detailed and accurate terrain results. Residual noise photons after filtering and sparse ground photons caused by dense vegetation are two major factors that affect terrain accuracy. While large windows effectively mitigate these two major error factors, they significantly sacrifice terrain detail. While small windows can maintain topographic detail, the terrain results are significantly affected by these two error factors. This paper proposes a strategy for detecting terrain using multi-level windows, fusing terrain information from large windows to eliminate errors from small windows. The resulting terrain from small windows not only reflects detailed terrain but also exhibits high accuracy. Experiments were conducted in areas with varying forest types, slopes, and terrain gradients. The results show that the inverted terrain achieved accuracies of 0.97m, 0.81m, and 0.78m in flat areas, and 1.36m and 1.1m in steeper areas, respectively. By comparing with the terrain results of single window inversion, the proposed method is superior to the strategy of single window terrain inversion. By fusing the terrain results of multiple windows, high-precision and stable terrain results are obtained.

[0078] Example 2

[0079] The present invention also provides a forest topography inversion system based on photon point cloud data of multi-level window detection, which is used to implement the method. The system includes:

[0080] A photon point acquisition module is used to obtain photon points on the ground under the forest based on a multi-level window terrain detection method; wherein the multi-level window includes a large window and a small window;

[0081] The residual acquisition module is used to obtain the terrain elevation residuals in multiple window sizes based on the ground photon points under the forest;

[0082] The residual diversion module is used to divert the terrain elevation residual based on the preset residual threshold to obtain deviation points of different categories;

[0083] The residual correction module is used to correct the deviation points of different categories and fuse the ground photon points under multiple window sizes based on the correction results to obtain the corrected ground photon points under a small window;

[0084] The inversion result acquisition module is used to obtain the inversion result of the forest terrain based on the corrected ground photon points in the small window.

[0085] In a further embodiment, the photon point acquisition module includes:

[0086] A median filter unit is used to filter the acquired multi-level window forest ground photon point cloud based on iterative median filtering to obtain a potential ground photon benchmark;

[0087] A peak acquisition unit is used to obtain the peak value of the ground photon point cloud under the forest in a multi-level window based on the elevation histogram frequency statistics method;

[0088] The photon point acquisition unit is used to obtain the ground photon point under the forest based on the position of the maximum peak within a preset distance of the potential ground photon benchmark.

[0089] A further implementation method is that, in the residual acquisition module, the ground photon points obtained under the large window are used to generate continuous terrain through a cubic spline function, and the elevation of the ground photon points obtained under the small window and the terrain elevation residual of the same position in the continuous terrain generated by the large window are obtained; wherein, the terrain elevation residual represents the difference between the ground photon points obtained under the large window and the small window sizes.

[0090] The embodiment of the present invention inverts the forest terrain in five experimental areas with different terrains and vegetation coverage, and compares the terrain accuracy with that inverted under a single window size (100m, 50m, 20m). The results are shown in Table 1. The root mean square error and standard deviation of the terrain inverted by the multi-level window proposed by the present invention are lower than those of the terrain inverted under a single size. This shows that the terrain detection method based on the multi-level window proposed by the present invention is superior to the terrain inverted under a single size in terms of both accuracy and stability. The terrain accuracy of the multi-level window inversion is higher than that of the terrain inverted under a single large / medium (100m, 50m) window. The main reason is that the terrain inverted under a single large window is not sufficient to reflect the changes in terrain details, while the multi-level window joint inversion method can well express these terrain details, mainly because the strategy of multi-level window joint inversion terrain integrates the sensitivity of small windows to terrain detail information. The terrain accuracy of the multi-level window joint inversion is higher than that of the terrain inverted using a single small (20m) window. This is mainly due to the fact that the terrain results inverted using a single small window contain more errors and are significantly affected by canopy photons or residual noise photons. Therefore, the multi-level window terrain detection method proposed in this paper not only combines the ability of a small window to depict detailed terrain, but also corrects the errors in the small window by inverting the terrain using a large window, resulting in better and more stable performance. (See Table 1)

[0091] Table 1 Comparison of terrain accuracy between inverted terrain using multi-level windows and inverted terrain using a single window size

[0092]

[0093] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A method for inverting forest terrain based on photon point cloud data using multi-level window detection, characterized in that: The method comprises: Based on the multi-level window terrain detection method, the ground photon points under the forest are obtained; the multi-level window includes a large window and a small window; the large window is a window of preset large size, and the small window is a window of preset small size; Based on the forest ground photon points, obtaining terrain elevation residuals in multiple window sizes; Dividing the terrain elevation residual based on a preset residual threshold to obtain deviation points of different categories; Correcting the deviation points of different categories and fusing the ground photon points under multiple window sizes based on the correction results to obtain the corrected ground photon points under the small window; the deviation points of different categories include no deviation points, first-category deviation points, and second-category deviation points; The method for obtaining the corrected ground photon point in the small window includes: Obtaining a starting range of a buffer zone around the terrain of the first type of deviation point; Based on the starting range of the buffer zone around the terrain, comparing the number of ground photons in the buffer zone around the terrain inverted by the large window and the small window; When the number of ground photons in the buffer zone around the large window terrain is greater than the amount of ground photon data in the buffer zone around the small window terrain, the ground photon point obtained by the small window is determined to be a residual noise photon point, and the residual noise photon point is replaced by the elevation interpolation point under the large window; otherwise, the ground photon point obtained by the small window is determined to be correct; Perform photon deviation point mutation detection on the second-type deviation points, and compare the number of ground photons in the buffer zone around the terrain inverted by the large window and the small window for the second-type deviation points that are not located in the preset mutation zone; otherwise, the second-type deviation points are determined to be non-ground photon points obtained by the small window, and the non-ground photon points are replaced by elevation interpolation points at the same position under the large window; The correct ground photon points and the ground photon points replaced by the large window elevation interpolation points are fused to obtain the corrected ground photon points under the small window; wherein, the correct ground photon points include the non-deviation points; Based on the corrected ground photon points in the small window, an understory terrain inversion result is obtained.

2. The method according to claim 1, characterized in that The method for obtaining the forest ground photon point includes: The multi-level window forest ground photon point cloud obtained was filtered based on iterative median filtering to obtain the potential ground photon benchmark; Based on the elevation histogram frequency statistics method, the peak value of the multi-level window forest ground photon point cloud is obtained; The understory ground photon point is obtained based on the position of the maximum peak within the preset distance of the potential ground photon benchmark.

3. The method according to claim 1, characterized in that Methods for obtaining the terrain elevation residuals under multiple window sizes include: The ground photon points obtained under the large window are used to generate continuous terrain through a cubic spline function, and the elevation of the ground photon points obtained under the small window and the terrain elevation residual at the same position in the continuous terrain generated by the large window are obtained; wherein, the terrain elevation residual represents the difference between the ground photon points obtained under the large window and the small window sizes.

4. A forest topography inversion system based on photon point cloud data of multi-level window detection, used to implement the method described in any one of claims 1 to 3, characterized in that: The system comprises: The photon point acquisition module is used to obtain photon points on the ground under the forest based on a multi-level window terrain detection method; wherein the multi-level window includes a large window and a small window; the large window is a window of a preset large size, and the small window is a window of a preset small size; A residual acquisition module, used for obtaining terrain elevation residuals in multiple window sizes based on the forest ground photon points; A residual diversion module is used to divert the terrain elevation residual based on a preset residual threshold to obtain deviation points of different categories; A residual correction module is used to correct the deviation points of different categories and fuse the ground photon points under multiple window sizes based on the correction results to obtain the corrected ground photon points under the small window; The inversion result acquisition module is used to obtain the forest understory terrain inversion result based on the corrected ground photon points in the small window.

5. The system according to claim 4, characterized in that The photon point acquisition module includes: A median filter unit is used to filter the acquired multi-level window forest ground photon point cloud based on iterative median filtering to obtain a potential ground photon benchmark; A peak acquisition unit is used to obtain the peak value of the ground photon point cloud under the forest in a multi-level window based on the elevation histogram frequency statistics method; A photon point acquisition unit is used to obtain the forest ground photon point based on the position of the maximum peak within a preset distance of the potential ground photon benchmark.

6. The system according to claim 4, characterized in that In the residual acquisition module, the ground photon points obtained under the large window are used to generate continuous terrain through a cubic spline function, and the elevation of the ground photon points obtained under the small window and the terrain elevation residual at the same position in the continuous terrain generated by the large window are obtained; wherein, the terrain elevation residual represents the difference between the ground photon points obtained under the large window and the small window sizes.

Images