A method and system for evaluating forest volume using feature-level point cloud fusion

Abstract

The present invention discloses a method and system for evaluating forest volume by using feature-level point cloud fusion, the method comprising the following steps: assigning geographic coordinates to ALS and TLS point cloud data in a sample plot respectively, and establishing an associated matching relationship between the two according to the common geographic coordinates; extracting single tree feature parameters from ALS and TLS point cloud data respectively, and performing feature-level fusion based on the geographic coordinates of the trunk position, using the crown width and tree height in the ALS point cloud data of the sample plot and the diameter at breast height data in the TLS point cloud data, obtaining a diameter at breast height estimation model through regression analysis, using the crown width and tree height in the fused data set, according to the diameter at breast height estimation model, estimating the diameter at breast height of each tree in the target area, and inputting the tree height together with the tree height into a binary volume model to estimate the forest volume in the target area. The present invention avoids the time-consuming and laborious process of traditional point cloud fusion, quickly realizes the measurement of each tree on the sample plot, and has strong operability in obtaining the forest volume of a large area.

Description

Technical Field

[0001] The present invention relates to the technical field of forest resource management, and in particular to a method and system for evaluating forest volume by using feature-level point cloud fusion. Background Art

[0002] Forests are an important part of terrestrial ecosystems and play an important role in maintaining ecological security and responding to climate change.

[0003] Forest volume assessment is of great significance for forest management and ecological monitoring. The basic unit of forest volume assessment is each standing tree. The traditional method is manual field measurement, using tools such as girth rulers and height gauges to measure each tree (tree height, breast diameter), and then bringing the results of each tree measurement into a binary volume model. The traditional method is time-consuming and labor-intensive, and can only obtain data for a small area. In recent years, remote sensing technology, such as LiDAR, which has been widely used in forestry, can quickly obtain the status of large areas of forests, effectively shortening operation time and improving work efficiency.

[0004] The application of LiDAR technology mainly includes airborne laser scanning (ALS) and terrestrial laser scanning (TLS). Among them, ALS technology can extract more accurate tree height (TH) parameters, but the extraction of diameter at breast height (DBH) is unsatisfactory; TLS can extract more accurate DBH, but the extraction of tree height has great uncertainty.

[0005] To this end, Chinese invention patent CN 104867180 B discloses a forest stand feature inversion method integrating UAV and LIDAR. It extracts typical stand features by combining UAV and LiDAR data, generates a three-dimensional point cloud of the study area as a digital surface model using feature point matching algorithm and aerial triangulation, calculates the canopy model by combining the digital terrain model extracted by LiDAR, and then constructs a multivariate regression estimation model by combining the ground-measured stand survey data with the jointly extracted point cloud variables and verifies its accuracy. The combination of the two, taking advantage of each other, is a cheap and flexible way to monitor forest status. However, this method uses multivariate regression analysis to construct a joint extraction estimation model, and the algorithm for high-precision matching of the two point clouds is very complex and inefficient.

[0006] In view of this, it is necessary to improve the existing point cloud fusion method for assessing forest volume in order to simplify the algorithm and improve efficiency. Summary of the invention

[0007] In view of the above-mentioned defects, the technical problem to be solved by the present invention is to provide a method and system for evaluating forest volume by using feature-level point cloud fusion, so as to solve the problems of complex algorithms and low efficiency in the prior art.

[0008] To this end, the present invention provides a method for evaluating forest volume by using feature-level point cloud fusion, comprising the following steps:

[0009] Collect ALS point cloud data of the target area and TLS point cloud data of the sample plots arranged in the target area, and assign geographic coordinates to the ALS point cloud data and TLS point cloud data of the sample plots respectively;

[0010] Based on the common geographic coordinates of the ALS point cloud data and the TLS point cloud data in the sample plot, an association and matching relationship between the two is established;

[0011] The ALS point cloud data and TLS point cloud data of the sample plot are processed respectively, and the single tree characteristic parameters of each tree in the sample plot are extracted to obtain the ALS characteristic data set and the TLS characteristic data set; wherein the ALS characteristic data set includes the trunk geographical coordinates, tree height and crown width of each tree in the sample plot, and the TLS characteristic data set includes the trunk geographical coordinates and breast diameter of each tree in the sample plot;

[0012] Based on the geographical coordinates of the tree trunks, the ALS feature data and the TLS feature data set were fused at the feature level to obtain a fused data set, including the geographical coordinates of the trunks, tree height, breast diameter and crown width of each tree in the sample plot, where the tree height and crown width were taken from the ALS feature data set, and the breast diameter was taken from the TLS feature data set;

[0013] The crown width, tree height and DBH data in the fused data set were used to obtain a DBH estimation model through regression analysis. The DBH estimation model was then used to estimate the DBH of each tree in the target area using the crown width of each tree in the ALS point cloud data.

[0014] The DBH and tree height of each tree in the target area are input into the binary volume model to estimate the forest volume in the target area.

[0015] In the above technical method, preferably, a TLS base station and a target sphere for positioning the TLS base station are set at the center of the sample site, and the geographical coordinates of each point in the TLS point cloud data are obtained by measuring the position of each point with the base station and the target sphere, and by three-dimensional coordinate conversion and least squares adjustment, the geographical coordinates of each point are assigned to the TLS point cloud data;

[0016] Through the onboard global satellite navigation system, the ALS point cloud data is directly assigned geographic coordinates while being acquired.

[0017] In the above technical method, preferably, before establishing the associated matching relationship between the TLS point cloud data and the ALS point cloud data of the sample site, the TLS point cloud data and the ALS point cloud data of the sample site are firstly processed by position coordinate conversion using a consistent projection and coordinate system.

[0018] In the above technical method, preferably, the method for obtaining the TLS feature data set and the ALS feature data set is as follows:

[0019] The TLS point cloud data and ALS point cloud data of the sample site are segmented based on the region growing segmentation algorithm to remove the gross errors in the point cloud data;

[0020] Use the progressive irregular triangulated network encryption algorithm to perform ground filtering on the point cloud and identify ground points;

[0021] Generate normalized digital surface model nDSM using ground points and original point cloud data;

[0022] Utilizing nDSM data, a point cloud segmentation method for linear entity extraction is used to extract the single tree feature parameters of each tree, thereby obtaining the TLS feature dataset and the ALS feature dataset.

[0023] In the above technical method, preferably, the DBH estimation model function used in the regression analysis is as follows:

[0024]

[0025] Where f(x) is the regression function, DBH TLS is the DBH of the corresponding trees in the TLS feature dataset, H ALS is the tree height of the corresponding tree in the ALS feature dataset, CD ALS is the crown width of the corresponding tree in the ALS feature data set, and a, b, and c are the regression coefficients respectively.

[0026] In the above technical method, preferably, the forest volume V of the target area is estimated using the following formula:

[0027] V=p×DBH q ×H r ;

[0028] Among them, p, q, r are constant coefficients respectively, DBH is the tree's diameter at breast height, H is the tree's height, and p = 0.0000942941, q = 1.832223553, r = 1.832223553.

[0029] The present invention also provides a system for evaluating forest volume by using feature-level point cloud fusion, comprising:

[0030] A data acquisition device, used for acquiring ALS point cloud data of a target area and TLS point cloud data of a sample plot arranged in the target area, and assigning geographic coordinates to the ALS point cloud data and the TLS point cloud data respectively;

[0031] A matching device, based on the geographic coordinates of the ALS point cloud data and the TLS point cloud data of the sample site, establishes an associated matching relationship between the two;

[0032] A feature extraction device is used to process the ALS point cloud data and the TLS point cloud data of the sample plot respectively, and extract the single tree feature parameters of each tree in the sample plot to obtain an ALS feature data set and a TLS feature data set; wherein the ALS feature data set includes the trunk geographic coordinates, tree height and crown width of each tree in the sample plot, and the TLS feature data set includes the trunk geographic coordinates and breast diameter of each tree in the sample plot;

[0033] A feature fusion device is used to perform feature-level fusion of the ALS feature data and the TLS feature data set based on the trunk geographic coordinates to obtain a fused data set, including the trunk geographic coordinates, tree height, breast diameter and crown width of each tree in the sample plot, wherein the tree height and crown width are taken from the ALS feature data set, and the breast diameter is taken from the TLS feature data set;

[0034] The diameter at breast height estimation device is used to obtain a diameter at breast height estimation model by using the crown width, tree height and diameter at breast height data in the fused data set through regression analysis, and then use the crown width of each tree in the ALS point cloud data of the target area to estimate the diameter at breast height of each tree in the target area according to the diameter at breast height estimation model;

[0035] The forest volume calculation device is used to input the breast diameter and tree height of each tree in the target area into a binary volume model to estimate the forest volume in the target area.

[0036] In the system, preferably, the data acquisition device comprises:

[0037] The TLS point cloud acquisition device includes a three-dimensional laser scanner, a TLS base station set at the center of the sample site, and a target ball used for positioning the TLS base station. The geographical coordinates of each point in the TLS point cloud data are obtained by measuring the position of each point with the base station and the target ball, and by three-dimensional coordinate conversion and least square adjustment, the geographical coordinates of each point are assigned to the TLS point cloud data.

[0038] The ALS point cloud acquisition device, including an airborne laser ranging sensor and a global satellite navigation system, directly assigns geographic coordinates to the ALS point cloud data while acquiring the ALS point cloud data.

[0039] In the above system, preferably, the feature extraction device comprises:

[0040] A noise removal unit is used to segment the TLS point cloud data and ALS point cloud data of the sample site based on the region growing segmentation algorithm to remove gross errors in the point cloud data;

[0041] A ground point recognition unit is used to perform ground filtering of point clouds using a progressive irregular triangulated network encryption algorithm to identify ground points;

[0042] A normalization unit, used for generating a normalized digital surface model nDSM using ground points and original point cloud data;

[0043] The feature extraction unit is used to extract the single tree feature parameters of each tree by using the nDSM data and a point cloud segmentation method for linear entity extraction.

[0044] In the above system, preferably, the DBH estimation model function used in the regression analysis is as follows:

[0045]

[0046] Where f(x) is the regression function, DBH TLs is the DBH of the corresponding trees in the TLS feature dataset, H ALS is the tree height of the corresponding tree in the ALS feature dataset, CD ALS is the crown width of the corresponding tree in the ALS feature data set, and a, b, and c are the regression coefficients respectively.

[0047] It can be seen from the above technical solutions that the method and system for evaluating forest volume by using feature-level point cloud fusion provided by the present invention solves the problem of complex high-precision matching algorithm and low efficiency of two point clouds in the prior art. Compared with the prior art, the present invention has the following beneficial effects:

[0048] (1) Geographic coordinates are assigned to ALS point cloud data and TLS point cloud data respectively. Based on their common geographic coordinates, an associated matching relationship is established between the two. Feature-level fusion is performed based on the geographic coordinates of the tree trunks in the two. The fusion speed is fast, avoiding the time-consuming and laborious process of traditional point cloud fusion methods.

[0049] (2) Based on the geographic coordinates of the tree trunks in the ALS point cloud data and TLS point cloud data of the sample plot, the tree height and crown width were extracted from the ALS point cloud data, and the breast diameter was extracted from the TLS point cloud data, respectively, thus quickly measuring each tree in the sample plot.

[0050] (3) The ALS point cloud data and TLS point cloud data of the sample plot were used to obtain the DBH estimation model through regression analysis. The crown width of each tree in the ALS point cloud data of the target area was then used to estimate the DBH of each tree in the target area according to the DBH estimation model. The DBH and the tree height were input into the binary volume model to estimate the forest volume in the target area. This method has strong operability in obtaining the forest volume of a large area. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces and describes the drawings required for use in the embodiments of the present invention or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 A flow chart of a method for evaluating forest volume using feature-level point cloud fusion provided by the present invention;

[0053] Figure 2 A schematic diagram of target area and sample plot selection in an embodiment of the present invention;

[0054] Figure 3 It is an enlarged view of a sample plot in an embodiment of the present invention;

[0055] Figure 4 Schematic diagram of the matching results of ALS point cloud and TLS point cloud in an embodiment of the present invention;

[0056] Figure 5 This is a schematic diagram of the comparison of DBH obtained by three DBH acquisition methods;

[0057] Figure 6 This is a schematic diagram of the crown and trunk positions extracted from ALS point cloud data;

[0058] Figure 7 Schematic diagram of the tree trunk position extracted from ALS point cloud data and TLS data;

[0059] Figure 8 The following is a comparison chart of the timber volume of each tree obtained by the four methods;

[0060] Fig. 9 The diagram is a schematic diagram of the results of timber volume calculation in the sample plot using four methods;

[0061] Fig.10 This is a schematic diagram of the full-area timber volume results calculated by the method of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the accompanying drawings of the embodiments of the present invention to clearly and completely describe the technical solutions of the embodiments of the present invention. Obviously, the embodiments described below 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 work are within the scope of protection of the present invention.

[0063] In order to more clearly explain and illustrate the technical solution and implementation of the present invention, several preferred specific embodiments for implementing the technical solution of the present invention are introduced below.

[0064] It should be noted that the directional words such as "inside, outside", "front, back" and "left, right" in this article are expressed based on the product usage status. Obviously, the use of the corresponding directional words does not constitute a limitation on the protection scope of this scheme.

[0065] The terms used in this application are defined as follows:

[0066] DGNSS, Differential Global Navigation and Positioning System, differential global satellite navigation system.

[0067] TLS, Terrestrial Laser Scanner, 3D laser scanning.

[0068] ALS, Airborne Laser Scanner, airborne laser scanning.

[0069] DBH, Diameter at Breast Height, diameter at breast height.

[0070] TH, Tree Height, tree height.

[0071] CD, Crown Diameter, crown diameter, in this application, crown diameter refers to the crown diameter.

[0072] See also Figure 1 , Figure 1 A flow chart of a method for evaluating forest volume using feature-level point cloud fusion provided by the present invention.

[0073] like Figure 1 As shown, the present invention provides a method for evaluating forest volume by using feature-level point cloud fusion, comprising the following steps:

[0074] Step 110, collect ALS point cloud data of the target area and TLS point cloud data of the sample plot arranged in the target area, and assign geographic coordinates to the ALS point cloud data and TLS point cloud data of the sample plot, respectively. The sample plot is an area in the target area, and a TLS base station is arranged at the center of the sample plot, and a target ball for TLS base station positioning is arranged in the sample plot.

[0075] ALS point cloud data is obtained by drone-mounted laser rangefinders, and geographic coordinates are directly assigned to ALS point cloud data through airborne DGNSS. TLS point cloud data is obtained by 3D laser scanners, and geographic coordinates are assigned to TLS point cloud data using the "base station position + single target sphere position" method.

[0076] In this example, the target area is selected as the Saihanba Mechanical Forest Farm in Chengde City, Hebei Province (42°02′~42°36′N, 116°51′~117°39′E, 1100-1940m above sea level), with an area of ​​76,600 hm 2 The sample plot was selected in the southwest corner of the target area, with an area of ​​160m×420m and flat terrain (slope less than 5°). Figure 2 As shown in the figure, a TLS base station is deployed in the center of the sample site (☆ in the figure), and a target ball for TLS base station positioning is deployed about 13m away from the TLS base station (● in the figure). The visibility between the TLS base station and the target ball is good, and both have open airspace to ensure good DGNSS signals.

[0077] The TLS point cloud data of the sample site was obtained using the Stonex X300 3D laser scanner. The point cloud density is about 5339.49 pts / m 2 The geographic coordinates of the TLS base station and the target sphere are measured respectively, and these two coordinates are used to assign geographic coordinates to the TLS point cloud data. This process is called orientation.

[0078] A DJI M300 multi-rotor drone equipped with a laser rangefinder and DGNSS was used to obtain the ALS point cloud data and geographic coordinates in the target area, including the ALS point cloud data and geographic coordinates of the sample plot. The drone was flown at an altitude of 80m, a speed of 3.5m / s, a side overlap rate of 80%, and three echoes were used to obtain the ALS point cloud data, with a point cloud density of 2031.0lpts / m 2 .

[0079] Step 120 , based on the common geographic coordinates of the ALS point cloud data and the TLS point cloud data of the sample site, establish an association and matching relationship between the ALS point cloud data and the TLS point cloud data of the sample site.

[0080] In this step, the TLS point cloud data and ALS point cloud data of the sample site are first converted using a consistent projection and coordinate system, and then the TLS point cloud data and ALS point cloud data are associated and matched based on the common geographic coordinates.

[0081] In this embodiment, DJI Zhitu v3.5 is used to process the ALS point cloud data, the coordinate system is China Geographic Coordinate System 2000 (CGCS2000), the projection mode is Gauss Kruger three-zone projection, the central meridian is 117°E, and the horizontal accuracy of the processed ALS point cloud data is 10cm and the vertical accuracy is 5cm. Stonex Siscan v3.0 point cloud processing software is used to process the TLS point cloud data, and the projection and coordinate system consistent with ALS are selected. The horizontal accuracy of the processed TLS point cloud data is 8mm and the vertical accuracy is 15mm.

[0082] In this example, Figure 4 The matching results of the four directions of front, back, left, and right of a tree in the ALS point cloud and the TLS point cloud (based on the superposition of common geographic coordinates), where dark colors represent TLS point clouds and light colors represent ALS point clouds. Figure 4 It can be seen that the two point clouds have a good degree of overlap.

[0083] Step 130, respectively process the ALS point cloud data and TLS point cloud data of the sample plot, and extract the single tree characteristic parameters of each tree in the sample plot to obtain the ALS characteristic data set and the TLS characteristic data set; wherein the ALS characteristic data set includes the trunk geographic coordinates LOC of each tree in the sample plot. ALS 、Tree height H ALS CD ALS The TLS feature dataset includes the geographic coordinates of each tree trunk in the plot. TLS , DBH TLS .

[0084] This step includes point cloud denoising, ground point recognition, point cloud normalization, single tree segmentation, and single tree feature parameter extraction, specifically including:

[0085] Step 131 , segmenting the TLS point cloud data and the ALS point cloud data respectively based on a region growing segmentation algorithm, and removing gross errors in the point cloud data.

[0086] Step 132, using a progressive irregular triangulated network encryption algorithm to perform point cloud ground filtering on the TLS point cloud data and the ALS point cloud data respectively, and identify ground points;

[0087] Step 133, generating two normalized digital surface models nDSM, TLS and ALS, using the identified ground points and the original TLS point cloud data and ALS point cloud data respectively;

[0088] Step 134 , using the two nDSM data respectively and using a point cloud segmentation method for linear entity extraction, extract the single tree feature parameters of each tree in the TLS and ALS point cloud data to obtain an ALS feature data set and a TLS feature data set.

[0089] Among them, the ALS feature dataset includes the geographic coordinates LOC of each tree trunk in the sample plot. ALS 、Tree height H ALS CD ALs The TLS feature dataset includes the geographic coordinates of each tree trunk in the plot. TLS , DBH TLs . And, the geographical coordinates LOC of the trunk of each tree in the plot TLS =LOC ALS In the following applications, the trunk geographic coordinates of each tree are uniformly used as the trunk geographic coordinates LOC in the ALS feature dataset. ALS .

[0090] Since it is difficult to obtain a tree trunk point cloud with sufficient density using the ALS point cloud, the LOC is calculated based on the position of the highest point of the tree crown in the method of the present invention. ALS , to facilitate calculation.

[0091] Step 140, based on the geographical coordinates of each tree trunk, the ALS feature dataset and the TLS feature dataset are fused at the feature level to obtain a fused dataset {LOC ALS , H ALs , CD ALs , DBH TLs}, including the geographical coordinates LOC of each tree trunk in the plot ALs 、Tree height H ALs , Crown CD ALS and DBH TLS The tree height H ALS CD ALS Taken from the ALS feature dataset, DBH TLS Taken from the TLS features dataset.

[0092] In the above steps, we use the fact that the trunk positions extracted from the ALS point cloud and the TLS point cloud in the sample plot are very close to each other, and associate all other features extracted from the two point cloud data to finally obtain the tree height H of each tree. ALs , Crown CD ALs and DBH TLS , obtain the fused data set {LOCALS , H ALS , CD ALS , DBH TLs}, thus achieving the feature-level fusion of the two point cloud data.

[0093] For two point cloud data in a plot, the fusion process of determining whether they are the same tree can be expressed by the following formula:

[0094] c={(a1, b1)|a1∈A and b1∈B, L(a1, b1)≤r}.

[0095] Among them, A represents the TLS feature dataset, B represents the ALS feature dataset, a1 and b1 represent the trunk positions obtained by the point cloud respectively, and a1∈A, b1∈B, L(a, b) represents the distance between the trunk positions a1 and b1. When the distance between the trunk positions obtained by the two point cloud data exceeds the search distance r, they are considered not to be the same tree (c is false), otherwise they are considered to be the same tree (c is true).

[0096] Step 150, using the tree height H in the fused data set ALS , Crown CD ALS and DBH TLS The DBH estimation model was obtained through regression analysis, and the crown width CD of each tree in the ALS point cloud data of the target area was used. ALs , according to the DBH estimation model, estimate the DBH of each tree in the target area ALS opt .

[0097] Since the DBH of trees outside the sample plot cannot be directly obtained through TLS point cloud data or ALS point cloud data, the method of the present invention uses the tree height H obtained from the ALS point cloud data through regression analysis. ALS , Crown CD ALs And the DBH obtained from TLS point cloud data TLs , a DBH estimation model was constructed, and then the crown width CD in the ALS point cloud data of the target area was used ALs Estimate the DBH of each tree in the entire target area ALS opt , including the DBH of each tree inside and outside the sample plot.

[0098] Specifically, the DBH estimation model function is first established as follows:

[0099]

[0100] Where f(x) is the regression function, DBH TLS is the DBH of the corresponding tree in the TLS feature dataset, H ALSis the tree height of the corresponding tree in the ALS feature dataset, CD ALS is the crown width of the corresponding trees in the ALS feature data set, and a, b, and c are the regression coefficients respectively.

[0101] Then, using the crown width CD in the fused dataset ALS 、Tree height H ALS and DBH TLs The regression coefficients a, b, and c are obtained by regression analysis.

[0102] Then, the regression coefficients a, b, and c are substituted into the above regression function to obtain the DBH estimation model. After a large amount of data optimization, the regression coefficients finally determined by the present invention are: a=0.174, b=0.455, and c=8.871.

[0103] Finally, the DBH estimation model is used to calculate the tree height H of each tree in the target area. ALs CD ALS Substitute into the DBH estimation model to obtain the DBH of each tree in the target area. ALs opt .

[0104] Step 160: The DBH of each tree in the target area ALs opt He Shugao H ALS Input the binary volume model to estimate the forest volume in the target area.

[0105] The volume of each tree V = p × DBH q ×H r ;

[0106] Among them, p, q, r are constant coefficients, DBH is the tree's diameter at breast height, and H is the tree's height.

[0107] After a lot of data optimization and verification, it was finally determined that p=0.0000942941, q=1.832223553, 1=1.832223553.

[0108] The effectiveness of the method of the present invention is verified by using specific examples below.

[0109] First, verify the accuracy of the DBH estimation model.

[0110] The verification method is: extract the DBH from TLS ref As a reference, compare the DBH extracted from the ALS point cloud of the sample plot ALs And the DBH estimated by optimizing the DBH estimation model ALS opt , the comparison results are as follows Figure 5 And as shown in Table 1.

[0111] Figure 5 In the figure, the horizontal axis represents the ID of a single tree, and the vertical axis represents the DBH value, where the black value is DBH. ref , light color is DBH ALs , dark color is DBH ALs opt As can be seen from Table 1, the optimized estimated DBH ALS opt The average value is 23.553cm, which is very different from the reference average value (23.616cm). DBH extracted from ALS point cloud ALS The mean value is 25.513cm, which is nearly 2cm larger than the reference value (the error exceeds 8%). In terms of standard deviation, the standard deviation of the reference extracted from TLS is the largest (4.307), and the standard deviations of the other two methods are relatively low, especially the result of the optimization estimation is the lowest, which is only 1.4304. Although the STD of the ALS method (3.164) seems to be closer to the reference value from the statistical indicator standard deviation STD, the maximum and minimum deviations are large, so when it comes to each tree, it will affect the accuracy of the "measurement of each tree".

[0112] Table 1.

[0113]

[0114] Second, the evaluation of feature fusion effect.

[0115] The evaluation mainly adopted the method of combining visual comparison and quantitative analysis. The crown position extracted from ALS and the trunk position extracted from ALS and TLS were superimposed and displayed, and the difference in trunk position within the sample plot between the two point cloud data was analyzed using indicators such as average error and maximum / minimum error.

[0116] like Figure 6 A schematic diagram of extracting the trunk position and crown outline is shown, where the vector is the crown outline extracted by ALS, and ● is the trunk position extracted by TLS; Figure 7 Schematic diagram of the tree trunk position extracted using ALS point cloud data and TLS data, ● is the tree trunk position extracted by TLS, ▲ is the tree trunk position extracted by ALS.

[0117] It can be seen that there is little overall difference in the trunk positions extracted by TLS and ALS, and both are within the coverage of the tree crown; from the point cloud detail images of the three selected individual trees, it can be seen that the trunk positions extracted by TLS are almost completely consistent with the trunk positions obtained by the actual ALS scan.

[0118] Table 2.

[0119]

[0120] The statistical results are basically consistent with those shown in the figure. The mean deviation of the trunk positions extracted by TLS and ALS is 0.401m, and the mean deviation accounts for 4.962% of the mean crown width extracted by ALS. It can be seen that the overall difference in the trunk position deviation is not very large, and within the crown coverage range, it meets the requirements of feature-level fusion.

[0121] Third, the estimation accuracy of forest volume was evaluated.

[0122] Since it is generally believed that the DBH of TLS and the H of ALS are closest to the true value, the DBH in the sample plot is used as the TLS and H ALS As the input parameter of the binary volume model, the calculated forest volume V TLS+ALS For reference, the volume V calculated using only TLS was compared and analyzed. TL , S , the volume V calculated by ALS without DBH optimization ALS 、Volume V calculated by ALS after optimizing DBH ALSopt .

[0123] Figure 8 The comparison diagrams of the timber volume of each tree obtained by the four methods are shown in Table 3 , and the estimated timber volume, error, error percentage and root mean square error of the four methods are given in Table 3 .

[0124] Table 3.

[0125]

[0126] Figure 8 The horizontal axis represents the ID of a single tree, and the vertical axis represents the volume V value. The lightest color at the bottom is the volume V calculated only using TLS. TLS , most of which are located in the upper part. The lighter color is the volume V calculated using the uncorrected parameters. ALs The second darkest color is the volume V calculated by the method of the present invention. ALS opt , black is the reference V TLS+ALS .

[0127] It can be seen that the standard deviation of the reference value is O.128, with the largest range of variation. ALS 、V TLS 、V ALSopt The standard deviations of the wood volume decreased successively, namely 0.108, 0.063 and 0.054. From the total volume calculation value, the reference value is 40.748m 3 , V TLS 、V ALS and V ALS opt The difference is 29.630m 3(72.053%, percentage error), 5.840m 3 (14.332%, percent error) and 1.256m 3 (3.082%, percentage error), it can be seen that TLS has the worst estimation accuracy for timber volume, which is mainly due to the inaccurate tree height measured by TLS, resulting in a serious underestimation of timber volume; the method of the present invention has the highest estimation accuracy for timber volume, which is similar to the timber volume V calculated using uncorrected parameters. ALS In comparison, the error percentage is reduced by 11.25%. It can be seen that the method of the present invention can greatly improve the estimation accuracy of timber volume.

[0128] Fig. 9 This is a comparison of the volume calculation results of the four methods in the sample plot. In the figure, a represents V ALS+TLS The volume calculated by the method is used as a reference. b, c and d represent V TLs 、V ALS and V ALS opt The volume and V calculated by the method ALS+TLS The difference between the volumes calculated by the TLS method, the darker the color, the greater the difference. TLS The difference with the reference is the largest. The volume V calculated by the method of the present invention is ALS opt Closest.

[0129] Fig.10 This is a graph of the timber volume results for the entire area calculated using the method of the present invention.

[0130] Based on the description of the above specific embodiments, the method and system for evaluating forest volume by using feature-level point cloud fusion provided by the present invention have the following advantages compared with the prior art:

[0131] First, by assigning geographic coordinates to ALS point cloud data and TLS point cloud data, the feature-level point cloud fusion of the two is realized. The breast diameter and tree height of each tree in the fused data set are input into the binary volume model to estimate the forest volume of the target area. The algorithm is simple, efficient and fast.

[0132] Second, using the crown width, tree height and breast diameter data in the ALS point cloud data, a breast diameter estimation model was obtained through regression analysis. Using the crown width in the fused data set, the breast diameter of each tree outside the sample plot was estimated according to the breast diameter estimation model, which reduced the amount of data collection, improved efficiency, and was highly operational for obtaining large-area forest volume.

[0133] Third, the results are highly accurate.

[0134] Finally, it should be noted that the terms "comprises", "includes" or any other variations thereof used in this article are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements that are inherent to such process, method, article or device. In the absence of more restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0135] The present invention is not limited to the above-mentioned optimal implementation mode. Anyone should be aware that any structural changes made under the inspiration of the present invention, and any technical solutions that are the same or similar to the present invention, fall within the protection scope of the present invention.

Claims

1. A method for evaluating forest volume using feature-level point cloud fusion, characterized in that: The following steps are involved: Collect ALS point cloud data of the target area and TLS point cloud data of the sample plots arranged in the target area, and assign geographic coordinates to the ALS point cloud data and TLS point cloud data of the sample plots respectively; Based on the common geographic coordinates of the ALS point cloud data and the TLS point cloud data of the sample site, an association and matching relationship between the two is established; The ALS point cloud data and TLS point cloud data of the sample plot are processed respectively, and the single tree characteristic parameters of each tree in the sample plot are extracted to obtain the ALS characteristic data set and the TLS characteristic data set; wherein the ALS characteristic data set includes the trunk geographical coordinates, tree height and crown width of each tree in the sample plot, and the TLS characteristic data set includes the trunk geographical coordinates and breast diameter of each tree in the sample plot; Based on the geographical coordinates of the tree trunks, the ALS feature data and the TLS feature data set were fused at the feature level to obtain a fused data set, including the geographical coordinates of the tree trunks, tree height, breast diameter and crown width of each tree in the sample plot, where the tree height and crown width were taken from the ALS feature data set, and the breast diameter was taken from the TLS feature data set; The crown width, tree height and DBH data in the fused data set were used to obtain a DBH estimation model through regression analysis. The DBH estimation model was then used to estimate the DBH of each tree in the target area using the tree height and crown width of each tree in the ALS point cloud data. The DBH and tree height of each tree in the target area are input into the binary volume model to estimate the forest volume in the target area.

2. The method according to claim 1, characterized in that A TLS base station and a target ball for TLS base station positioning are set at the center of the sample plot. The geographical coordinates of each point in the TLS point cloud data are obtained by measuring the position of each point in the TLS point cloud data relative to the base station and the target ball, and by three-dimensional coordinate transformation and least squares adjustment, the geographical coordinates of each point are assigned to the TLS point cloud data. Through the onboard global satellite navigation system, the ALS point cloud data is directly assigned geographic coordinates while being acquired.

3. The method according to claim 1, characterized in that Before establishing the correlation and matching relationship between the TLS point cloud data and the ALS point cloud data of the sample site, the position coordinates of the TLS point cloud data and the ALS point cloud data of the sample site are first converted using a consistent projection and coordinate system.

4. The method according to claim 1, characterized in that: The method to obtain the TLS feature dataset and the ALS feature dataset is as follows: The TLS point cloud data and ALS point cloud data of the sample site are segmented based on the region growing segmentation algorithm to remove the gross errors in the point cloud data; Use the progressive irregular triangulated network encryption algorithm to perform ground filtering on the point cloud and identify ground points; Generate normalized digital surface model nDSM using ground points and original point cloud data; Utilizing nDSM data, a point cloud segmentation method for linear entity extraction is used to extract the single tree feature parameters of each tree, and then the TLS feature dataset and ALS feature dataset are obtained.

5. The method according to claim 1, characterized in that The DBH estimation model function used in regression analysis is as follows: Where f(x) is the regression function, DBH TLS is the DBH of the corresponding trees in the TLS feature dataset, H ALS is the tree height of the corresponding tree in the ALS feature dataset, CD ALS is the crown width of the corresponding tree in the ALS feature data set, and a, b, and c are the regression coefficients respectively.

6. The method according to claim 5, characterized in that The following formula is used to estimate the forest volume V in the target area: V=p×DBH q ×H r ; Among them, p, q, r are constant coefficients respectively, DBH is the tree's diameter at breast height, H is the tree's height, and p = 0.0000942941, q = 1.832223553, r = 1.832223553.

7. A system for evaluating forest volume using feature-level point cloud fusion, characterized in that: include: A data acquisition device, used for acquiring ALS point cloud data of a target area and TLS point cloud data of a sample plot arranged in the target area, and assigning geographic coordinates to the ALS point cloud data and the TLS point cloud data respectively; A matching device, based on the geographic coordinates of the ALS point cloud data and the TLS point cloud data of the sample site, establishes an associated matching relationship between the two; A feature extraction device is used to process the ALS point cloud data and the TLS point cloud data of the sample plot respectively, and extract the single tree feature parameters of each tree in the sample plot to obtain an ALS feature data set and a TLS feature data set; wherein the ALS feature data set includes the trunk geographic coordinates, tree height and crown width of each tree in the sample plot, and the TLS feature data set includes the trunk geographic coordinates and breast diameter of each tree in the sample plot; A feature fusion device is used to perform feature-level fusion of the ALS feature data and the TLS feature data set based on the trunk geographic coordinates to obtain a fused data set, including the trunk geographic coordinates, tree height, breast diameter and crown width of each tree in the sample plot, wherein the tree height and crown width are taken from the ALS feature data set, and the breast diameter is taken from the TLS feature data set; The diameter at breast height estimation device is used to obtain a diameter at breast height estimation model by using the crown width, tree height and diameter at breast height data in the fused data set through regression analysis, and then use the crown width of each tree in the ALS point cloud data of the target area to estimate the diameter at breast height of each tree in the target area according to the diameter at breast height estimation model; The forest volume calculation device is used to input the breast diameter and tree height of each tree in the target area into a binary volume model to estimate the forest volume in the target area.

8. The system according to claim 7, characterized in that The data acquisition device comprises: The TLS point cloud acquisition device includes a three-dimensional laser scanner, a TLS base station set at the center of the sample site, and a target ball used for positioning the TLS base station. The geographical coordinates of each point in the TLS point cloud data are obtained by measuring the position of each point with the base station and the target ball, and by three-dimensional coordinate conversion and least square adjustment, the geographical coordinates of each point are assigned to the TLS point cloud data. The ALS point cloud acquisition device, including an airborne laser ranging sensor and a global satellite navigation system, directly assigns geographic coordinates to the ALS point cloud data while acquiring the ALS point cloud data.

9. The system according to claim 7, characterized in that The feature extraction device comprises: A noise removal unit is used to segment the TLS point cloud data and ALS point cloud data of the sample site based on the region growing segmentation algorithm to remove gross errors in the point cloud data; A ground point recognition unit is used to perform ground filtering of point clouds using a progressive irregular triangulated network encryption algorithm to identify ground points; A normalization unit, used for generating a normalized digital surface model nDSM using ground points and original point cloud data; The feature extraction unit is used to extract the single tree feature parameters of each tree by using the nDSM data and a point cloud segmentation method for linear entity extraction.

10. The system according to claim 7, characterized in that The DBH estimation model function used in the regression analysis is as follows: Where f(x) is the regression function, DBH TLS is the DBH of the corresponding trees in the TLS feature dataset, H ALS is the tree height of the corresponding tree in the ALS feature dataset, CD ALS is the crown width of the corresponding tree in the ALS feature data set, and a, b, and c are the regression coefficients respectively.

Images