CN112150479A - Single tree segmentation and tree height and crown width extraction method based on Gaussian clustering - Google Patents
Single tree segmentation and tree height and crown width extraction method based on Gaussian clustering Download PDFInfo
- Publication number
- CN112150479A CN112150479A CN202010912849.1A CN202010912849A CN112150479A CN 112150479 A CN112150479 A CN 112150479A CN 202010912849 A CN202010912849 A CN 202010912849A CN 112150479 A CN112150479 A CN 112150479A
- Authority
- CN
- China
- Prior art keywords
- gaussian
- tree
- crown
- height
- point cloud
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 31
- 230000011218 segmentation Effects 0.000 title claims abstract description 29
- 238000000034 method Methods 0.000 claims abstract description 59
- 238000002945 steepest descent method Methods 0.000 claims abstract description 15
- 238000001914 filtration Methods 0.000 claims abstract description 11
- 238000011430 maximum method Methods 0.000 claims abstract description 11
- 230000000877 morphologic effect Effects 0.000 claims abstract description 11
- 239000002023 wood Substances 0.000 claims description 10
- 238000012417 linear regression Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 3
- 238000011160 research Methods 0.000 description 7
- 238000001514 detection method Methods 0.000 description 6
- 239000000203 mixture Substances 0.000 description 5
- 241000196324 Embryophyta Species 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 241000218631 Coniferophyta Species 0.000 description 2
- 241000218638 Thuja plicata Species 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 241000894007 species Species 0.000 description 2
- 239000002028 Biomass Substances 0.000 description 1
- 235000014466 Douglas bleu Nutrition 0.000 description 1
- 241000218683 Pseudotsuga Species 0.000 description 1
- 240000001416 Pseudotsuga menziesii Species 0.000 description 1
- 235000008572 Pseudotsuga menziesii Nutrition 0.000 description 1
- 235000005386 Pseudotsuga menziesii var menziesii Nutrition 0.000 description 1
- 241000218685 Tsuga Species 0.000 description 1
- 240000003021 Tsuga heterophylla Species 0.000 description 1
- 235000008554 Tsuga heterophylla Nutrition 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Geometry (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a single tree segmentation and tree height and crown width extraction method based on Gaussian clustering, which comprises the steps of firstly generating a crown height model for acquired point cloud data by using an interpolation method, and obtaining a Gaussian model through morphological open operation and Gaussian filtering; then, obtaining a crown vertex by using a local maximum method, and fitting the Gaussian curved surface by using a steepest descent method to obtain a single tree position and a preliminary crown amplitude range; and finally, clustering the normalized point cloud data by adopting the minimum two-dimensional Euclidean distance, and extracting and verifying the obtained tree height and crown width by utilizing the actually measured data, so that the extraction precision of the tree height and the crown width is improved.
Description
Technical Field
The invention relates to the technical field of single tree information extraction, in particular to a method for single tree segmentation and extraction of tree height and crown based on Gaussian clustering.
Background
The tree height and the crown width not only reflect important indexes of forest stand structural characteristics, but also are the most commonly used observation variables in the researches such as production potential estimation, biomass prediction, forest growth prediction and the like. Therefore, the single tree segmentation and the high crown extraction of the trees have important significance in forest resource investigation. The airborne laser radar has strong penetrating power as an active remote sensing technology, is not easily influenced by weather and illumination conditions, and can obtain high-precision vertical structure information of the earth surface and ground objects. By using an airborne laser radar technology, parameter information such as height and crown of the single trees and the trees can be effectively extracted; at present, methods for single-tree segmentation and tree high crown extraction can be divided into two types: a gridding-based Canopy Height Model (CHM) and a point cloud clustering-based segmentation method; but all have certain limitations, which lead to inaccurate extraction of tree height and crown width.
Disclosure of Invention
The invention aims to provide a method for single tree segmentation and tree height and crown breadth extraction based on Gaussian clustering, and the extraction precision of the tree height and the crown breadth is improved.
In order to achieve the purpose, the invention provides a single tree segmentation and tree height and crown width extraction method based on Gaussian clustering, which comprises the following steps:
generating a canopy height model for the acquired point cloud data by using an interpolation method, and obtaining a Gaussian model through morphological open operation and Gaussian filtering;
fitting the Gaussian curved surface by using a local maximum method and a steepest descent method to obtain a single wood position and a preliminary crown width range;
and clustering the normalized point cloud data by adopting the minimum two-dimensional Euclidean distance, and extracting and verifying the obtained tree height and crown width.
The method comprises the following steps of generating a canopy height model for acquired point cloud data by using an interpolation method, and obtaining a Gaussian model through morphological open operation and Gaussian filtering, wherein the method comprises the following steps:
and replacing the black or gray image in the obtained canopy height model with a point cloud height threshold value in a specified range by using morphological opening operation, and smoothing by using Gaussian filtering to obtain the Gaussian model.
Wherein, utilize local maximum method and steepest descent method to carry out the fitting to the gaussian surface, obtain single wood position and preliminary crown width scope, include:
and identifying a local threshold point of the Gaussian model through a set window, and calculating the parameters of the Gaussian model by using a steepest descent method.
The method comprises the following steps of clustering the normalized point cloud data by adopting a minimum two-dimensional Euclidean distance, and extracting and verifying the obtained tree height and crown breadth, and comprises the following steps:
and extracting a point cloud cluster of which the distance between the sample and the center line is less than or equal to a set threshold value according to the center line of the set cluster, dividing the corresponding point cloud into designated clusters after calculating the Euclidean distance of the point cloud in the horizontal direction by using a two-dimensional Euclidean formula until all the point cloud is divided.
Wherein, adopt minimum two-dimentional Euclidean distance to carry out the clustering with normalization point cloud data to extracting and verifying obtaining tree height and crown width, still include:
and measuring the single tree segmentation precision by using the recall rate, the accuracy and the harmonic value, and evaluating the precision of the tree high crown and performing linear regression analysis.
The invention relates to a single tree segmentation and tree height and crown width extraction method based on Gaussian clustering, which comprises the steps of firstly generating a crown height model for acquired point cloud data by using an interpolation method, and obtaining a Gaussian model through morphological open operation and Gaussian filtering; then, obtaining a crown vertex by using a local maximum method, and fitting the Gaussian curved surface by using a steepest descent method to obtain a single tree position and a preliminary crown amplitude range; and finally, clustering the normalized point cloud data by adopting the minimum two-dimensional Euclidean distance, and extracting and verifying the obtained tree height and crown width by utilizing the actually measured data, so that the extraction precision of the tree height and the crown width is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic step diagram of a method for splitting a single tree and extracting a tree height and a crown based on Gaussian clustering according to the present invention.
FIG. 2 is a schematic flow chart of a method for splitting a single tree and extracting a tree height and a crown based on Gaussian clustering according to the present invention.
FIG. 3 is a comparison diagram of the segmentation accuracy of the Gaussian model clustering method and the watershed method provided by the present invention.
FIG. 4 is a linear regression relationship between the crown width extracted by correctly segmenting the three-dimensional single wood and the actually measured crown width by the two methods provided by the invention.
FIG. 5 is a linear regression relationship between the height of the single-tree extracted by correctly segmenting the three-dimensional single tree and the height of the actually measured single tree according to the two methods provided by the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
In the description of the present invention, "a plurality" means two or more unless specifically defined otherwise.
Referring to fig. 1 and 2, the present invention provides a method for splitting a single tree and extracting a tree height and a crown based on gaussian clustering, comprising:
s101, generating a canopy height model for the acquired point cloud data by using an interpolation method, and obtaining a Gaussian model through morphological open operation and Gaussian filtering.
Specifically, the original point cloud is divided into non-ground points and ground points, interpolation operation is respectively carried out to obtain a digital ground model (DSM) and a Digital Elevation Model (DEM), and the difference operation is carried out on the two to obtain CHM; the detection of tree vertices is affected by the presence of black or gray "hole" holes in the original CHM. Performing morphological on operation on the surface of the CHM model, replacing a hole in the CHM model with a point cloud height maximum point (height threshold) in a specified range or neighborhood to form a CMM model, namely:
The CMM model can enhance the information of the edge of the crown to make the surface of the crown smoother, and a Gaussian filtering method is used for carrying out linear smoothing on the CMM model to obtain a Gaussian model GCMM (Gaussian filter CMM), namely:
GCMM=G5×5×CMM
s102, fitting the Gaussian curved surface by using a local maximum method and a steepest descent method to obtain the single-tree position and the initial crown range.
Specifically, the fixed window can accurately identify the top point of the tree, a set window with the size of 5 multiplied by 5 is selected through experimental analysis to find the local maximum point of the GCMM model, the GCMM model can be used for eliminating the pseudo local maximum, and error detection in treetop identification is reduced; and then, calculating parameters of the Gaussian model by using a steepest descent method, specifically:
performing Gaussian surface fitting through grid points of a GCMM model, wherein the gray level image matrix of the model is EM×N(1. ltoreq. i.ltoreq.M, 1. ltoreq. j.ltoreq.N), z ═ e when x ═ i and y ≦ j in a three-dimensional spaceijZ generationThe gray scale value of the table image satisfies the condition eijThe expression of the Gaussian surface model is as follows:
the method is simplified and can be obtained:
let B be lnA, then we get:
the above problem translates into solving the parameters B, x0,y0α (parameters B, x)0,y0And α are all arrays of 1 × n) such that the objective function is:
due to the fact that directly orderSolving parameters is complex, so the research adopts the steepest descent method to solve the optimal parameters of the Gaussian mixture model, and X is made to be (B, X)0,y0,α)TThen F (B, x)0,y0α) the first derivative at point X is:
the experiment adopts the algorithm of the steepest descent method to calculate X ═ B, X0,y0,α)TThe process is as follows:
given initial value X0,>0, k is 0, calculate:
in the formula: g (x) ═ g1,g2,g3,g4)T;
Calculating:
Xk+1=Xk-tkg(Xk)
FK+1=F(Xk+1),gk+1=g(Xk+1)
Solving the fitted Gaussian mixture model by using a steepest descent method as follows:
in the formula: (x)n,yn) For the single-wood position after fitting of the Gaussian mixture surface, AnIs (x)n,yn) The gray value of the location, alphanAnd fitting the crown range for the Gaussian mixture curved surface.
And S103, clustering the normalized point cloud data by adopting the minimum two-dimensional Euclidean distance, and extracting and verifying the obtained tree height and crown breadth.
Specifically, the minimum two-dimensional Euclidean distance clustering based on the Gaussian mixture model is to perform distance iteration judgment on the elevation normalized point cloud, divide the attribution condition of the adjacent point cloud, and realize accurate segmentation of the three-dimensional single tree. The minimum two-dimensional Euclidean distance clustering algorithm is used as follows:
(x) definitionn,yn) The position of the space is the central line of the cluster, and the distance between the extracted sample and the central line is less than or equal to 4 alphanThe point cloud cluster of (1);
② according to two-dimensional Euclidean distance formulaJudging that the distance center line is more than 4 alphanSearching a cluster with the shortest Euclidean distance in the horizontal direction of the point cloud, and dividing the point cloud into the clusters;
and thirdly, repeating the step II until all the sample points are divided into corresponding clusters, and realizing the three-dimensional segmentation of the single wood.
The method comprises the following steps of evaluating the capabilities of extracting singles and tree high crown width by airborne radar point cloud from two aspects, and obtaining verification of the singles segmentation precision by carrying out visual interpretation on actually measured ground data and high spatial resolution pictures, wherein the specific method comprises the following steps:
and measuring the single tree segmentation precision by adopting three indexes of recall rate r, accuracy rate p and harmonic value F:
r=TP/(TP+FN)
p=TP/(TP+FP)
F=2×(r×p)/(r+p)
the method further comprises the following steps:
and point cloud data required by research are acquired, and region discrete processing is performed.
Specifically, the research area is located in an area with main landform mainly on hills, the elevation is 260-. Coniferous forests are the main species in the research area, and forest vegetation tree species are mainly douglas fir (Pseudotsuga menziesii), western hemlock (Tsuga herbaphylla) and western red cedar (Thuja plicata). The basic statistics of the single-wood structure parameters actually measured by the topographic survey of 347 field survey points in the region are shown in table 1.
TABLE 1 basic statistics of measured Single Wood structural parameters
And acquiring small-spot airborne radar point cloud data by adopting a Saab TopEye LiDAR system. The flying height of the remote sensing platform is 200m, the flying inclination angle is 8 degrees, the sampling density is 4 pulse echoes per square meter, the laser pulse speed is 7000 points/second, the maximum echo is recorded for 4 times, the light spot diameter is 40cm, the flying speed of the helicopter is 25m/s, and the scanning width is 70 m. 6 round pure forest sample plots (actually measured single trees total 893 strains) with the radius of 30m are selected in the experiment, the sample plot area is 1.69 hectare, and the average point density is about 4.86pts/m2Labeled as Plot1, Plot2, Plot3, Plot4, Plot5, Plot 6. The color aerial image of the research area uses a Pseudo-crown Surface Model (Pseudo-canopy Surface Model) derived from the area through an autocorrelation technology to finish positive shooting correction by using a soft-copy photogrammetry system, and the final image resolution is 0.3 m.
A circular sample plot with the radius of 10m is selected from 6 sample plots, 12 actually measured trees are shared in the circular sample plot, and when the crown top is detected based on a local maximum method, 12 pseudo local maxima can be eliminated by adopting a GCMM model, and the crown top identification is not influenced.
Local maximum detection is performed on the GCMM models of 6 plots using windows of different sizes to find the preliminary crown vertices as shown in table 2. The actual single trees of the 6 blocks are 893 strains, 260 crown vertexes are detected more than the actual single tree value in a 3 × 3 window, and 356 unidentified crown vertexes exist in a 7 × 7 window, so that a 5 × 5 window is selected for an experiment to find a local maximum point in the GCMM model.
TABLE 2 detection of the Effect of a Window on crown vertices
Detection window size | W=3 | W=5 | W=7 |
Crown vertex identification number | 1153 | 860 | 537 |
The method is characterized in that a Gaussian model clustering method and a watershed method are adopted to respectively carry out single-tree segmentation on 6 sample place clouds, the point clouds of Plot plots 5 and Plot6 are high in density and large in number of trees, the single-tree segmentation carried out by the watershed method has a serious over-segmentation phenomenon, and the closed edge of a crown extracted has a defect phenomenon. As the overlooking shape of the conifer canopy is similar to a circle, the single-tree point cloud accords with Gaussian distribution in the GCMM model, and the crown range preliminarily extracted by the Gaussian model clustering method is closer to the shape of a real tree. The Gaussian model clustering method utilizes a local maximum method to detect the position of a preliminary tree on the GCMM model, so that the number of single tree over-segmentation is reduced, and the segmentation accuracy is higher by comparing with the actually measured tree position.
As shown in table 3, 893 strains were shared in 6 blocks, and 860 strains were divided by the gaussian model clustering method, of which 790 strains were correctly divided, 103 strains were over-divided, and 70 strains were under-divided. 867 plants are segmented based on the watershed method, and the algorithm is accurately segmented into 709 plants, 184 plants are over-segmented, and 158 plants are under-segmented. Compared with a watershed method, the Gaussian model clustering method is based on the local maximum detection of the GCMM model to perform single tree segmentation, and the accuracy is improved by 11.3%.
TABLE 3 Gaussian model clustering method and watershed Algorithm accuracy evaluation
As shown in fig. 3, the mean harmonic value of the watershed algorithm is only 0.80, and pixels with similar spatial positions and gray values are connected to form a closed contour based on the CHM model, so that an over-segmentation phenomenon is easily caused when a single-wood canopy is formed. And the Gaussian model clustering method is to obtain a crown vertex by adopting a local maximum method based on a GCMM model, and to fit the real shape of the tree at the crown vertex by using a Gaussian surface fitting of a steepest descent method to the maximum extent, so as to ensure the accuracy of single tree segmentation, but as the canopy density of the plots 5 and the plot6 is higher, the laser signal cannot identify smaller trees under the forest, and the average F value is 0.89. Compared with a watershed algorithm, the Gaussian model clustering method obtains the average r value of 0.87 and the average p value of 0.91 in 6 blocks respectively, and the superiority of the method is proved.
Single tree height and crown width extraction are carried out on 6 sample lands by using a Gaussian model clustering method and a watershed method, wherein the tree height extraction precision is 95% and 90%, and the crown width extraction precision is 91% and 86%, respectively. The linear regression relationship between the crown width extracted by correctly segmenting the three-dimensional single wood and the actually measured crown width by the two methods is shown in (a) and (b) of FIG. 4, and the Gaussian model clustering method R20.84, but the watershed method crown extraction result is weak, R20.73; the linear regression relationship of the height of the single tree is shown in FIG. 5(a) (b), and the watershed method R2Compared with the Gaussian model clustering method, the method has better tree height extraction result and R is 0.842=0.92。
The average error of the tree height extraction of the Gaussian model clustering method and the watershed method is-0.83 m and-1.41 m respectively, and the average error is a negative value, which indicates that the tree height is underestimated, and probably the system cannot acquire all tree information due to insufficient sampling density of laser points; the average errors of the crown amplitudes extracted by using the two algorithms are-0.42 m and 1.05m respectively, and analysis shows that: (1) the tree in the research sample plot is of a conifer type, the overlooking shape of the crown is approximate to a circle, the crown extracted by the Gaussian curve surface fitting method is closer to the real condition than the crown extracted by the watershed algorithm, and the crown amplitude extraction average error is reduced by 0.63 m; (2) compared with a watershed method, the Gaussian model clustering method directly clusters the original point cloud, tree information loss caused by the adoption of a maximum value method in a grid by the CHM model is avoided, integrity of point cloud information is guaranteed, and the average error of extracting the tree height is reduced by 0.58 m. The Gaussian model clustering algorithm has good performance, compared with a watershed algorithm, the single tree segmentation recall rate r is 0.87, the accuracy rate p is 0.91, the harmonic value F of the comprehensive r and p is 0.89, the accuracy of crown width extraction and tree height extraction of the algorithm is 95% and 91%, the average error is-0.42 m and-0.83 m, and the extraction accuracy of the tree height and the crown width is improved.
The invention relates to a single tree segmentation and tree height and crown width extraction method based on Gaussian clustering, which comprises the steps of firstly generating a crown height model for acquired point cloud data by using an interpolation method, and obtaining a Gaussian model through morphological open operation and Gaussian filtering; then, obtaining a crown vertex by using a local maximum method, and fitting the Gaussian curved surface by using a steepest descent method to obtain a single tree position and a preliminary crown amplitude range; and finally, clustering the normalized point cloud data by adopting the minimum two-dimensional Euclidean distance, and extracting and verifying the obtained tree height and crown width by utilizing the actually measured data, so that the extraction precision of the tree height and the crown width is improved.
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (5)
1. A single tree segmentation and tree height and crown width extraction method based on Gaussian clustering is characterized by comprising the following steps:
generating a canopy height model for the acquired point cloud data by using an interpolation method, and obtaining a Gaussian model through morphological open operation and Gaussian filtering;
fitting the Gaussian curved surface by using a local maximum method and a steepest descent method to obtain a single wood position and a preliminary crown width range;
and clustering the normalized point cloud data by adopting the minimum two-dimensional Euclidean distance, and extracting and verifying the obtained tree height and crown width.
2. The method for single tree segmentation and tree height and crown width extraction based on gaussian clustering of claim 1, wherein the step of generating a crown height model for the acquired point cloud data by interpolation and obtaining a gaussian model through morphological open operation and gaussian filtering comprises:
and replacing the black or gray image in the obtained canopy height model with a point cloud height threshold value in a specified range by using morphological opening operation, and smoothing by using Gaussian filtering to obtain the Gaussian model.
3. The method for single-tree segmentation and tree height and crown width extraction based on gaussian clustering of claim 2, wherein the step of fitting the gaussian surface by using the local maximum method and the steepest descent method to obtain the single-tree position and the preliminary crown width range comprises the following steps:
and identifying a local threshold point of the Gaussian model through a set window, and calculating the parameters of the Gaussian model by using a steepest descent method.
4. The method for single tree segmentation and tree height and crown width extraction based on Gaussian clustering of claim 3, wherein the step of clustering the normalized point cloud data by using the minimum two-dimensional Euclidean distance and extracting and verifying the obtained tree height and crown width comprises the following steps:
and extracting a point cloud cluster of which the distance between the sample and the center line is less than or equal to a set threshold value according to the center line of the set cluster, dividing the corresponding point cloud into designated clusters after calculating the Euclidean distance of the point cloud in the horizontal direction by using a two-dimensional Euclidean formula until all the point cloud is divided.
5. The method for single tree segmentation and crown width extraction based on gaussian clustering of claim 4, wherein the normalized point cloud data is clustered by using the minimum two-dimensional euclidean distance, and the obtained tree height and crown width are extracted and verified, further comprising:
and measuring the single tree segmentation precision by using the recall rate, the accuracy and the harmonic value, and evaluating the precision of the tree high crown and performing linear regression analysis.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010912849.1A CN112150479A (en) | 2020-09-03 | 2020-09-03 | Single tree segmentation and tree height and crown width extraction method based on Gaussian clustering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010912849.1A CN112150479A (en) | 2020-09-03 | 2020-09-03 | Single tree segmentation and tree height and crown width extraction method based on Gaussian clustering |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112150479A true CN112150479A (en) | 2020-12-29 |
Family
ID=73889229
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010912849.1A Withdrawn CN112150479A (en) | 2020-09-03 | 2020-09-03 | Single tree segmentation and tree height and crown width extraction method based on Gaussian clustering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112150479A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112396699A (en) * | 2020-11-30 | 2021-02-23 | 常州市星图测绘科技有限公司 | Method for automatically sketching land parcel based on unmanned aerial vehicle image |
-
2020
- 2020-09-03 CN CN202010912849.1A patent/CN112150479A/en not_active Withdrawn
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112396699A (en) * | 2020-11-30 | 2021-02-23 | 常州市星图测绘科技有限公司 | Method for automatically sketching land parcel based on unmanned aerial vehicle image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11615606B2 (en) | Tree crown extraction method based on unmanned aerial vehicle multi-source remote sensing | |
CN105488770B (en) | A kind of airborne laser radar point cloud filtering method of object-oriented | |
CN107274417B (en) | Single tree segmentation method based on airborne laser point cloud aggregation relation | |
CN112381861B (en) | Forest land point cloud data registration and segmentation method based on foundation laser radar | |
Wu et al. | Individual tree crown delineation using localized contour tree method and airborne LiDAR data in coniferous forests | |
CN106845399B (en) | Method for extracting single-tree information from LiDAR point cloud by using hierarchical clustering mode | |
Chen et al. | Isolating individual trees in a savanna woodland using small footprint lidar data | |
Ferraz et al. | 3-D mapping of a multi-layered Mediterranean forest using ALS data | |
CN104155638B (en) | A kind of seeds sorting technique based on the pseudo-vertical waveform model of LiDAR | |
CN113591766B (en) | Multi-source remote sensing tree species identification method for unmanned aerial vehicle | |
CN114463403A (en) | Tree carbon sink amount calculation method based on point cloud data and image recognition technology | |
CN108074232B (en) | Voxel segmentation-based airborne LIDAR building detection method | |
CN114764871B (en) | Urban building attribute extraction method based on airborne laser point cloud | |
CN112669333B (en) | Single wood information extraction method | |
Pirotti et al. | A comparison of tree segmentation methods using very high density airborne laser scanner data | |
Su et al. | Extracting wood point cloud of individual trees based on geometric features | |
CN110988909A (en) | TLS-based vegetation coverage determination method for sandy land vegetation in alpine and fragile areas | |
CN111091079A (en) | TLS-based method for measuring dominant single plant structural parameters of vegetation in alpine and fragile regions | |
Putman et al. | Automated estimation of standing dead tree volume using voxelized terrestrial lidar data | |
CN107705344A (en) | Plant canopy model extracting method in laser scanning environment cloud data | |
CN112150479A (en) | Single tree segmentation and tree height and crown width extraction method based on Gaussian clustering | |
CN117765006A (en) | Multi-level dense crown segmentation method based on unmanned aerial vehicle image and laser point cloud | |
Lin et al. | Noise point detection from airborne lidar point cloud based on spatial hierarchical directional relationship | |
CN114862872B (en) | Mangrove single wood segmentation method based on Faster R-CNN | |
CN112381029B (en) | Method for extracting airborne LiDAR data building based on Euclidean distance |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20201229 |