CN114119863A - Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data - Google Patents
Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data Download PDFInfo
- Publication number
- CN114119863A CN114119863A CN202111317369.1A CN202111317369A CN114119863A CN 114119863 A CN114119863 A CN 114119863A CN 202111317369 A CN202111317369 A CN 202111317369A CN 114119863 A CN114119863 A CN 114119863A
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- crown
- calculating
- tree
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/89—Lidar systems specially adapted for specific applications for mapping or imaging
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
Abstract
The invention discloses a method for automatically extracting a street tree target and forest attributes thereof based on vehicle-mounted laser radar data, which comprises the following steps: preprocessing the original point cloud data by using commercial software; determining the road driving direction according to the track points scanned on the vehicle; establishing a three-dimensional regular grid index for the preprocessed point cloud and extracting a rod-shaped point set by combining rod-shaped characteristics; distinguishing a street tree and an artificial rod target according to the morphological difference of the point cloud on the upper part of the rod-shaped ground object; extracting and dividing the street trees connected with the crown by adopting a layered crown width fitting algorithm; and calculating the attribute of the street tree in combination with the information of the Digital Elevation Model (DEM) of the target area. The invention adopts a full-automatic mode in the data processing process, does not need any human intervention after setting relevant parameters, and can efficiently and automatically extract the three-dimensional information of the street tree under the urban road environment: the position, the tree height, the breast height, the crown width, the branch height and the like of the tree improve the production efficiency of urban part collection and mapping large-scale topographic map.
Description
Technical Field
The invention relates to the technical field of surveying and mapping science, in particular to a method for automatically extracting street tree targets and forest attributes thereof based on vehicle-mounted laser radar data, which is particularly suitable for automatically acquiring three-dimensional information of garden street trees in urban fine management by utilizing a mobile measurement system.
Background
The street trees play a significant role in urban landscaping, ecological environment improvement and the like. At present, ecological smart city construction, unmanned system development and city component collection and updating are actively promoted, street tree three-dimensional information is taken as one of important basic data in the fields, and how to efficiently and accurately extract a street tree target becomes a great hotspot of current research. Therefore, it is imperative to develop a method for automatically extracting the street tree target so as to improve the surveying and mapping production efficiency and the intelligence level. The three-dimensional laser scanning technology is a great technical innovation in the field of surveying and mapping, and greatly promotes the development of the surveying and mapping industry. The vehicle-mounted mobile measurement technology is widely applied to point cloud data acquisition by domestic and foreign scholars due to the characteristics of high data acquisition precision, wide range and high speed.
The vehicle-mounted mobile measurement system is provided with physical devices such as a laser scanner, a Global Navigation Satellite System (GNSS), an Inertial Measurement Unit (IMU), a CCD camera and the like, and an automobile is used as a main carrier, so that high-precision true-color three-dimensional point clouds in a measurement area can be rapidly collected on the premise of no loss of mobility, the collection efficiency of surveying and mapping data is greatly improved, and a high-quality data source is also provided for the identification of various ground object targets.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a method for automatically extracting street tree targets and forest attributes thereof based on vehicle-mounted laser radar data.
The purpose of the invention is realized as follows:
a method for automatically extracting a street tree target and forest attributes thereof based on vehicle-mounted laser radar data comprises the following steps:
firstly, preprocessing original point cloud data: the method comprises point cloud partitioning, ground filtering and DEM generation; dividing the wide-range urban road point cloud into a plurality of sub-blocks in a Terrasolid software according to a self-defined mode; sequentially performing ground filtering operation on the block point clouds by using terraSoid software to obtain ground points and non-ground points, and then generating a DEM (digital elevation model) based on the ground point clouds;
secondly, point cloud coordinate conversion after pretreatment: reading a Terrasolid engineering file to obtain boundary information of the block point cloud, calculating a horizontal rotation angle by combining the GpsTime of the track point and the slope of a connecting line between adjacent track points under the constraint of a block boundary point set, and then rotating the point cloud until the road driving direction is parallel to the X axis;
thirdly, identifying and extracting the rod-column-shaped point set: establishing a three-dimensional regular grid index for the target point cloud, and performing spatial neighborhood analysis by taking a non-space grid as a research unit and combining with ground object rod-shaped characteristics to obtain a rod-shaped point set;
fourthly, identifying and filtering the artificial rod target: the rod-column-shaped point set acquired in the third step also comprises a part of artificial rod point cloud, and the artificial rod and the street tree target are distinguished mainly according to the morphological difference of the point cloud on the upper parts of different rod-shaped ground objects;
and fifthly, finely dividing connected crowns: extracting complete crown point cloud by using a spatial clustering algorithm based on a regular grid, directly clustering if only one street tree exists, and dividing the crown point cloud by using a layered crown width fitting algorithm after clustering is finished if the crowns of a plurality of street trees are connected to each other to realize the refined extraction of the street tree monomer;
sixthly, automatically calculating the attributes of the street tree and outputting the result: on the basis of correctly extracting the single point cloud of the street tree, a related operator is designed to automatically calculate the tree attributes of the street tree, such as the tree height, the crown width, the under-branch height and the like, and the attributes are output as a text file and a vector file.
The second step includes the following substeps:
(1) reading and parsing a Terrasolid engineering file: analyzing prj format project text file, wherein the file contains peripheral horizontal convex hulls (marked as C) of all block point clouds1，C2，……，Cn) Each block corresponds to a horizontal convex hull, and the vertex of the convex hull is marked as Q1，Q2，……，Qn；
(2) And (3) calculating the horizontal rotation angle according to the track point information: sequentially taking convex hull vertexes Q according to clockwise or anticlockwiseiForm a vector v with a certain track point G1And Qi+1Form a vector v2Calculating v1And v2The obtained value is of the same sign, which indicates that the current track point G is in the convex hull; traversing all track points, and determining the track points in the current block point cloud area according to the method; reordering the track points according to the GpsTime ascending mode, calculating the time difference delta t of the adjacent track points and the connection line slope difference delta k of the front track point and the rear track point, if the conditions are met: Δ t < tThreshold value&&Δk＜kThreshold valueThen, considering that the adjacent track points are distributed on the same track line, and successfully separating the track points on different track lines according to the distribution; taking track points P of the head and the tail of any one track line1(x1,y1)、P2(x2,y2) According to the formula:
calculating a horizontal rotation angle theta;
(3) point cloud coordinate conversion: for computer processing, the original point cloud is first centered around the geometric center P of the boxc(xc,yc,zc) The fixed point is rotated counterclockwise by an angle theta around the Z axis, and then the x, y, Z coordinates of the rotated point cloud are subtracted by fixed offsets dx, dy, dz, respectively. Setting an arbitrary point cloud P (x, y, z), according to a formula: x ═ xc)cosθ+(y-yc)sinθ+xc-dx，y'＝-(y-yc)sinθ+(y-yc)cosθ+yc-dy, z ' -z-dz calculating the coordinate transformation result P ' (x ', y ', z ');
the third step comprises the following substeps:
(1) establishing a three-dimensional regular grid index: traversing the target point set to obtain the minimum value (x) of the point cloud coordinate in the three-axis directionmin,ymin,zmin) Setting the side length of the grid as delta d, and aiming at any point cloud Pi(xi,yi,zi) Can be expressed according to the formula:
calculating row (R), column (C) and layer (L) indexes of the grid unit where the three-dimensional regular grid index is located, and sequentially calculating the grid unit index numbers corresponding to all point clouds in the target point set by using the rules, thereby completing the construction of the three-dimensional regular grid index;
(2) grid horizontal neighborhood analysis: firstly, selecting any non-blank net in a horizontal grid of the same layer as an initial seed grid S, carrying out horizontal eight-neighborhood clustering by taking S as a center, indicating that the single body clustering is finished when a new non-empty grid cannot be added into a clustering body, then selecting any grid in the rest non-empty grids again as a new seed grid S, and repeating the steps until all non-blank nets in the layer are processed. Traversing all clustering bodies, and if the clustering bodies meet the geometric condition: number of clustering grids<a1And horizontal span<a2And cluster the number of point clouds<a3. Then judging the point cloud to be a rod-shaped section point cloud, reserving the point cloud, and discarding the clustering object which does not meet the conditions. Performing horizontal neighborhood analysis on the non-blank nets in each layer from bottom to top by adopting the method to finally obtain a plurality of cross section point sets of the suspected rod-shaped objects;
(3) grid vertical neighborhood analysis: analyzing the distribution of the clustering bodies in the grid network of the nth layer and the (n + 1) th layer, and if the clustering bodies S in the nth layernWith n +1 intralayer cluster Sn+1If there is a grid with the same row and column indices, S will benAnd Sn+1Are combined into a whole; clustering and merging are started by taking the clusters in the upper and lower adjacent layers of grids as research objects according to the mode, and finally a three-dimensional space clustering point set is obtained; using the formula: hc＝Zmax-ZminCalculating the height difference (H) of the cluster point setc) If H isc<HThreshold valueThen the current clustering point set is considered as a pole-column-shaped point set;
the fourth step includes the substeps of: taking the grid in which the barycentric coordinates of the point clouds of the rod-shaped part are located as the center, intercepting a grid window with the size of m multiplied by n from top to bottom from the highest point of the rod-shaped object, projecting the point clouds in the window into a horizontal plane, and calculating the azimuth coverage value of the projection point clouds under the annular neighborhoodW; defining 8 directions of east, south, west, north, northeast, southeast, southwest and northwest, completely scanning the point cloud of the crown on the inner side of the traffic lane, but only scanning partial point cloud on the outer side, so that the coverage exists on at least 4 directions for the crown, but the projection point cloud extends in a single direction for the street lamp head, the traffic sign board and the signal lamp arm, and the coverage exists on at most 2 directions; the following discriminant rules are therefore designed: according to the formula:
calculating the coverage degree of the ith position in the k layer annular neighborhood grid; secondly, according to the formula:
calculating the coverage of the k-th layer annular neighborhood grid; thirdly, according to the formula:
calculating the integral annular neighborhood coverage W (W is more than or equal to 0 and less than or equal to 8) of the projection point cloud; fourthly, the theoretical coverage of the known crown point cloud is 4, the theoretical coverage of the point cloud at the top of the artificial rod is 1, the average value of the two is 2.5, and according to the formula:
to calculate the discrimination threshold WTIf W is>WTThen the rod target is determined as a street tree, if W<WTJudging the rod target as an artificial rod and filtering the artificial rod;
the fifth step includes the substeps of:
(1) clustering crown point clouds: carrying out European clustering on a non-space net in a three-dimensional space to extract complete crown point cloud, and setting a clustering rule: if the number of point clouds in the grid network is more than 5, the point clouds participate in clustering, and all the non-grid networks with common surfaces are clustered into a whole; clustering termination conditions: new grids cannot be searched and added into the current clustering monomer; if the adjacent crowns are crossed and overlapped, the adjacent crowns can be directly clustered into a crown monomer;
(2) rough segmentation of the point clouds of the connected crowns: firstly, connecting crown points of treesCloud projection onto XOZ plane according to trunk horizontal centroid X coordinate (X)1，x2，x3，……，xn) Numbering the street trees in an ascending manner: 1, 2, 3, … …, n. ② setting step length delta x to pair interval [ x1，xn]Performing equal interval division, and counting the number of sub-intervals [ x ]1+k*Δx，x1+(k+1)*Δx]Point clouds inside; traversing all the intervals, respectively finding out a point p with the maximum elevation in each interval and storing the point p into a linked list L; taking out the interval [ x ] for determining the dividing point between the ith street tree and the (i + 1) th street treei，xi+1]The point set in the corresponding linked list L with the minimum height in the point set is the initial segmentation point SegPoint; fourthly, sequentially carrying out pairwise division on the adjacent street trees from left to right according to the mode until all the street trees are processed;
(3) finely dividing the point clouds of the connected crowns: firstly, setting fixed intervals along the Z-axis direction for the roughly divided crown point cloud to perform layered projection; then extracting a contour point set P of the left and right canopy point clouds in the same layer by combining a convex hull algorithmL、PRFitting the set of points P according to the least squares principleLAnd PRPlane circular model O ofL、OR(ii) a Finally, analyzing the canopy point cloud P and the fitting circle OL、OR(radii are each rL、rR) The relative position relationship between the P points and the fitting circle to determine the segmentation range of the connected crown, and respectively calculate the horizontal distance d between the P points and the fitting circleL、dRThe crown point cloud is segmented according to the following geometrical conditions: satisfy dL<rL&&dR<rRIf the P point belongs to the left tree crown and the right tree crown simultaneously; ② satisfy dL<rL&&dR>rRIf the P point belongs to the left crown but not the right crown, judging that the P point belongs to the left crown but not the right crown; satisfy dL>rL&&dR<rRThen point P is considered to belong to the right crown but not to the left crown. Traversing all the layered crown point clouds, and sequentially segmenting every two adjacent street trees according to the mode;
the sixth step includes the substeps of: (1) performing coordinate inverse transformation in the substep (3) of the second step on the extracted street tree monomer point cloud, and restoring the real coordinates of the street tree point cloud;
(2) calculating horizontal positioning coordinates
Traversing the trunk point cloud of the single-trunk street tree, and according to a formula:
calculating a trunk point cloud horizontal positioning coordinate;
(3) calculating the tree height H: acquiring a Digital Elevation Model (DEM) in a measuring area range, and calculating the row and column index numbers of corresponding pixels V in the DEM and the elevation value Z of the pixels V by using the horizontal positioning coordinates of the street treesdemThe height of the ground at the bottom of the trunk is obtained; traversing crown point cloud to obtain elevation Z of highest pointmaxAccording to the formula: h ═ Zmax-ZdemCalculating the tree height H;
(4) calculating the breast diameter DBH: from Z in substep (2)demFirstly, upwards respectively intercepting a point set P at three height positions of the trunk, which are 1.2m, 1.3m and 1.4m away from the ground1、P2、P3Estimating the radius r of a target point set fitting circle by using a RANSAC plane circle fitting model1、r2、r3According to the formula:
calculating the breast diameter DBH;
(5) calculating the height h below the branch: and layering the street tree point cloud along the Z-axis direction at equal intervals, and determining fitting circle parameters for the point clouds in each layer by using an RANSAC algorithm. Recording the area of a point cloud fitting circle in the kth layer as SkAccording to the formula: Δ S ═ Sk+1-SkCalculating the area difference delta S of the point cloud fitting circles in the two adjacent layers, if delta S>The threshold value indicates that the trunk cross-sectional area is mutated and the first branch point is detected. Recording the maximum value of the k layer point cloud elevation as ZkAccording to the formula: h ═ Zk-ZdemCalculating the height under the branch;
(6) calculating the crown width CW: projecting the crown point cloud into an XOY plane and calculating the maximum X of the horizontal coordinatesmin、Xmax、Ymin、YmaxDefining the extension of the crown in the north-south direction as: d1＝Ymax-Ymin(ii) a The east-west extension is: d2＝Xmax-XminAccording to the formula:
calculating the crown width CW;
(7) outputting the attribute result of the street tree: outputting the attribute of the shade forest tree as a format required by the topographic map software, such as: text files, SHP files, DWG files.
Has the positive and beneficial effects that: the invention adopts a full-automatic mode in the data processing process, does not need any human intervention after setting relevant parameters, and can efficiently and automatically extract the three-dimensional information of the street tree under the urban road environment: the position, the tree height, the breast height, the crown width, the branch height and the like of the tree improve the production efficiency of urban part collection and mapping large-scale topographic map.
Drawings
FIG. 1 is a flow chart;
FIG. 2 is a diagram of the effect of non-ground point clouds remaining after ground filtering by Terrasolide software according to the present invention;
FIG. 3 is a schematic diagram of the transformation of point cloud coordinates based on the driving direction of a road according to the present invention, wherein the point cloud to be processed is transformed from an XOY coordinate system to an X ' O ' Y ' coordinate system according to a track point in the middle of the road;
FIG. 4 is a drawing of the result of the extraction of the pole-column-shaped point set of the present invention, wherein the street light pole, the trunk and the traffic sign pole are all completely extracted;
FIG. 5 is a graph showing the filtering effect of the artificial pole target, the street lamp and the traffic sign are inserted into the crown, and the red point cloud is the complete artificial pole point cloud identified by the algorithm;
FIG. 6 is a schematic diagram of a layered crown width fitting algorithm for segmenting connected crowns according to the present invention, fitting a planar circular model according to a set of horizontal contour points of point clouds in different left and right crowns;
FIG. 7 is a comparison graph of the prior and post segmentation results of the connected crown point clouds of the present invention.
Detailed Description
The present invention is further illustrated by the following figures and specific examples, which are to be understood as illustrative only and not as limiting the scope of the invention, which is to be given the full breadth of the appended claims and any and all equivalent modifications thereof which may occur to those skilled in the art upon reading the present specification.
According to the method flow shown in fig. 1, the method further clarifies that the road tree target in the vehicle-mounted point cloud data of a certain road section in Nanjing is automatically extracted as an application example:
firstly, preprocessing original point cloud data: partitioning and ground filtering processing are carried out on the original point cloud in Terrasolide software, the obtained non-ground point cloud of the measuring area is shown in figure 2, and then a DEM (digital elevation model) of the measuring area is generated based on the ground point cloud obtained through extraction;
secondly, point cloud coordinate conversion after pretreatment: determining track points positioned in the current blocks according to the block point cloud boundary information, automatically calculating a horizontal rotation angle theta by using the GPSTIme attribute of the track points and the slope of a connecting line between two adjacent points, rotating the point cloud until the road driving direction is parallel to the X axis, and performing coordinate conversion as shown in FIG. 3;
thirdly, identifying and extracting the rod-column-shaped point set: establishing a three-dimensional regular grid index for the target point cloud, performing spatial neighborhood analysis by taking a non-space grid as a research object and combining with ground object rod-shaped characteristics to obtain a rod column-shaped point set, and taking a block point cloud with the number of 6 as an example in the implementation to obtain the rod column-shaped point set shown in fig. 4;
fourthly, identifying and filtering the artificial rod target: taking the grid in which the barycentric coordinates of the point clouds of the rod-shaped part are located as the center, intercepting a grid window with the size of 13 multiplied by 5 from top to bottom from the highest point of the rod-shaped object, projecting the point clouds in the window into a horizontal plane, and calculating the azimuth coverage W of the projection point clouds under the annular neighborhood. Calculating W values of the annular neighborhood in 8 azimuths according to a formula (I) mentioned in the fourth step of the specification and setting an azimuth coverage threshold MTWill satisfy M>MTThe pole target of (1) is judged as a street tree, otherwise, the pole target of (1) is an artificial pole. The discrimination results of all the rod-shaped ground objects in the block point cloud No. 6 are shown in the following table (selection), and taking the local point cloud in the survey area as an example, the separation result of the street tree and the artificial rod point cloud is shown in FIG. 5, wherein the red point cloud belongs to the artificial rod target.
And fifthly, finely dividing connected crowns: primarily extracting crown point clouds by using an European clustering algorithm based on a three-dimensional regular grid, directly clustering if isolated street trees exist, and dividing the crown point clouds by using a layered crown width fitting algorithm after clustering is finished if a plurality of street tree crowns are connected to each other to realize the refined extraction of single street trees; the method comprises the following substeps:
(1) and carrying out European clustering on the non-space nets in the three-dimensional space to preliminarily extract crown point clouds.
(2) And projecting the connected crown point cloud into an XOZ plane, determining an initial segmentation point SegPoint by combining the distribution characteristics of the crown surface contour point set, and dividing the target crown point cloud into a left part and a right part again by taking the SegPoint as a reference.
(3) Dividing the crown point cloud into a plurality of crown layers from top to bottom along the horizontal direction, determining a crown layer fitting circle model by using a layered crown breadth fitting algorithm, setting distribution rules of left and right crown point clouds, and dividing all the crown layer point clouds layer by layer;
(4) and pairwise segmenting the connected crown point clouds from left to right according to the mode to realize refined extraction.
Taking a street tree with two connected crowns as an example, fig. 6 is a schematic diagram of a model of a layered crown breadth fitting circle, the left side of fig. 7 is an effect before algorithm segmentation, and the right side of fig. 7 is an effect after algorithm segmentation.
And sixthly, automatically calculating the attributes of the street tree and outputting results. Designing a correlation operator on the basis of correctly extracting the single point cloud of the street tree to automatically calculate the tree attributes of the street tree, such as the tree height, the crown breadth, the under-branch height and the like, and outputting the tree attributes as a text file (excerpt) shown in the following table:
| ID | X(m) | Y(m) | tree height (m) | Chest diameter (m) | Crown width (m) | Height below branch (m) |
| tree_15 | 377290.325 | 3539754.834 | 7.920 | 0.224 | 5.883 | 2.288 |
| tree_16 | 377248.908 | 3539807.809 | 6.984 | 0.188 | 4.056 | 1.825 |
| tree_17 | 377250.328 | 3539807.675 | 6.379 | 0.199 | 3.674 | 2.065 |
| tree_18 | 377254.305 | 3539801.088 | 6.941 | 0.184 | 3.348 | 2.479 |
| tree_19 | 377241.976 | 3539789.676 | 5.724 | 0.195 | 4.746 | 1.220 |
| tree_20 | 377233.512 | 3539768.800 | 6.998 | 0.162 | 4.781 | 2.256 |
| tree_21 | 377262.775 | 3539786.665 | 7.448 | 0.187 | 5.121 | 2.061 |
| tree_22 | 377267.790 | 3539777.998 | 6.624 | 0.171 | 3.286 | 2.437 |
| ... | ... | ... | ... | ... | ... | ... |
| tree_52 | 377267.820 | 3539750.598 | 8.367 | 0.208 | 5.728 | 2.286 |
| tree_53 | 377264.061 | 3539748.275 | 9.166 | 0.215 | 5.886 | 2.278 |
| tree_54 | 377266.125 | 3539744.293 | 8.924 | 0.184 | 5.926 | 2.255 |
| tree_55 | 377270.538 | 3539746.334 | 8.565 | 0.196 | 5.848 | 2.296 |
| tree_56 | 377219.411 | 3539788.578 | 6.705 | 0.303 | 4.340 | 2.681 |
The invention adopts a full-automatic mode in the data processing process, does not need any human intervention after setting relevant parameters, and can efficiently and automatically extract the three-dimensional information of the street tree under the urban road environment: the position, the tree height, the breast height, the crown width, the branch height and the like of the tree improve the production efficiency of urban part collection and mapping large-scale topographic map.
Claims (6)
1. A method for automatically extracting a street tree target and forest attributes thereof based on vehicle-mounted laser radar data is characterized by comprising the following steps:
firstly, preprocessing original point cloud data: the method comprises point cloud partitioning, ground filtering and DEM generation; dividing the wide-range urban road point cloud into a plurality of sub-blocks in a Terrasolid software according to a self-defined mode; sequentially performing ground filtering operation on the block point clouds by using terraSoid software to obtain ground points and non-ground points, and then generating a DEM (digital elevation model) based on the ground point clouds;
secondly, point cloud coordinate conversion after pretreatment: reading a Terrasolid engineering file to obtain block point cloud boundary information, calculating a horizontal rotation angle by combining the GPSTIme of the track points and the slope of a connecting line between adjacent track points under the constraint of a block boundary point set, and then rotating the point cloud until the road driving direction is parallel to the X axis;
thirdly, identifying and extracting the rod-column-shaped point set: establishing a three-dimensional regular grid index for the target point cloud, and performing spatial neighborhood analysis by taking a non-space grid as a research unit and combining with ground object rod-shaped characteristics to obtain a rod-shaped point set;
fourthly, identifying and filtering the artificial rod target: the rod-column-shaped point set acquired in the third step also comprises a part of artificial rod point cloud, and the artificial rod and the street tree target are distinguished mainly according to the morphological difference of the point cloud on the upper parts of different rod-shaped ground objects;
and fifthly, finely dividing connected crowns: extracting a complete crown point cloud by using a spatial clustering algorithm based on a regular grid; if the street trees are isolated, the clustering is directly carried out; if the crowns of a plurality of street trees are connected, dividing the crown point cloud by using a layered crown width fitting algorithm after clustering is finished, and realizing the refined extraction of the single street trees;
sixthly, automatically calculating the attributes of the street tree and outputting the result: on the basis of correctly extracting the single point cloud of the street tree, a related operator is designed to automatically calculate the tree attributes of the street tree, such as the tree height, the crown width, the under-branch height and the like, and the attributes are output as a text file and a vector file.
2. The method for automatically extracting the street tree target and the forest attributes thereof based on the vehicle-mounted laser radar data according to the claim 1, wherein the second step comprises the following substeps:
(1) reading and parsing a Terrasolid engineering file: parsing prj format engineering documentThe peripheral horizontal convex hull (marked as C) containing all the block point clouds1，C2，……，Cn) Each block corresponds to a horizontal convex hull, and the vertex of the convex hull is marked as Q1，Q2，……，Qn；
(2) And (3) calculating the horizontal rotation angle according to the track point information: sequentially taking convex hull vertexes Q according to clockwise or anticlockwiseiForm a vector v with a certain track point G1And Qi+1Form a vector v2Calculating v1And v2The obtained value is of the same sign, which indicates that the current track point G is in the convex hull; traversing all track points, and determining the track points in the current block point cloud area according to the method; reordering the track points according to the GpsTime ascending mode, calculating the time difference delta t of the adjacent track points and the connection line slope difference delta k of the front track point and the rear track point, if the conditions are met: Δ t < tThreshold value&&Δk＜kThreshold valueThen, considering that the adjacent track points are distributed on the same track line, and successfully separating the track points on different track lines according to the distribution; taking track points P of the head and the tail of any one track line1(x1,y1)、P2(x2,y2) According to the formula:
calculating a horizontal rotation angle theta;
(3) point cloud coordinate conversion: for computer processing, the original point cloud is first centered around the geometric center P of the boxc(xc,yc,zc) Rotating the fixed point counterclockwise around the Z axis by an angle theta, and then respectively subtracting fixed offset dx, dy and dz from x, y and Z coordinates of the rotated point cloud; setting an arbitrary point cloud P (x, y, z), according to a formula: x ═ xc)cosθ+(y-yc)sinθ+xc-dx，y'＝-(y-yc)sinθ+(y-yc)cosθ+yc-dy, z ' -z-dz calculating the coordinate transformation result P ' (x ', y ', z ').
3. The method for automatically extracting the street tree target and the forest attribute thereof based on the vehicle-mounted laser radar data as claimed in claim 1, wherein the third step comprises the following sub-steps:
(1) establishing a three-dimensional regular grid index: traversing the target point set to obtain the minimum value (x) of the point cloud coordinate in the three-axis directionmin,ymin,zmin) Setting the side length of the grid as delta d, and aiming at any point cloud Pi(xi,yi,zi) Can be expressed according to the formula:
calculating row (R), column (C) and layer (L) indexes of the grid unit where the three-dimensional regular grid index is located, and sequentially calculating the grid unit index numbers corresponding to all point clouds in the target point set by using the rules, thereby completing the construction of the three-dimensional regular grid index;
(2) grid horizontal neighborhood analysis: firstly, selecting any non-blank net as an initial seed grid S in a horizontal grid of the same layer, carrying out horizontal eight-neighborhood clustering by taking S as a center, indicating that the single body clustering is finished when a new non-empty grid cannot be added into a clustering body, then selecting any grid as a new seed grid S in the rest non-empty grids again, and repeating the steps until all the non-blank nets in the layer are processed; traversing all clustering bodies, and if the clustering bodies meet the geometric condition: number of clustering grids<a1And horizontal span<a2And clustering the number of the body point clouds<a3(ii) a Judging that the point cloud is a rod-shaped section point cloud, reserving the point cloud, and discarding the clustering object which does not meet the conditions; performing horizontal neighborhood analysis on the non-blank nets in each layer from bottom to top by adopting the method to finally obtain a plurality of cross section point sets of the suspected rod-shaped objects;
(3) grid vertical neighborhood analysis: analyzing the distribution of the clustering bodies in the grid network of the nth layer and the (n + 1) th layer, and if the clustering bodies S in the nth layernWith n +1 intralayer cluster Sn+1If there is a grid with the same row and column indices, S will benAnd Sn+1Are combined into a whole; clustering and merging are started by taking the clusters in the upper and lower adjacent layers of grids as research objects according to the mode, and finally a three-dimensional space clustering point set is obtained; using the formula: hc＝Zmax-ZminCalculating the height difference (H) of the cluster point setc) If H isc<HThreshold valueThe current set of cluster points is considered to be a set of rodlike points.
4. The method for automatically extracting the street tree target and the forest attribute thereof based on the vehicle-mounted laser radar data as claimed in claim 1, wherein the fourth step comprises the following substeps:
taking a grid in which the barycentric coordinates of the point clouds of the rod-shaped part are located as a center, intercepting a grid window of m multiplied by n from top to bottom from the highest point of the rod-shaped object, projecting the point clouds in the window into a horizontal plane, and calculating the azimuth coverage value W of the projection point clouds under the annular neighborhood; defining 8 directions of east, south, west, north, northeast, southeast, southwest and northwest, completely scanning the point cloud of the crown on the inner side of the traffic lane, but only scanning partial point cloud on the outer side, so that the coverage exists on at least 4 directions for the crown, but the projection point cloud extends in a single direction for the street lamp head, the traffic sign board and the signal lamp arm, and the coverage exists on at most 2 directions; the following discriminant rules are therefore designed: according to the formula:
calculating the coverage degree of the ith position in the k layer annular neighborhood grid; secondly, according to the formula:
calculating the coverage of the k-th layer annular neighborhood grid; thirdly, according to the formula:
calculating the integral annular neighborhood coverage W of the projection point cloud; fourthly, the theoretical coverage of the known crown point cloud is 4, the theoretical coverage of the point cloud at the top of the artificial rod is 1, and the average value of the two is 2.5; according to the formula:
to calculate a discrimination thresholdValue WTIf W is>WTThen the rod target is determined as a street tree, if W<WTThe pole target is determined to be an artificial pole and needs to be filtered out.
5. The method for automatically extracting the street tree target and the forest attributes thereof based on the vehicle-mounted laser radar data as claimed in claim 1, wherein the fifth step comprises the following substeps:
(1) clustering crown point clouds: carrying out European clustering on a non-space net in a three-dimensional space to extract complete crown point cloud, and setting a clustering rule: if the number of point clouds in the grid network is more than 5, the point clouds participate in clustering, and all the non-grid networks with common surfaces are clustered into a whole; clustering termination conditions: new grids cannot be searched and added into the current clustering monomer; if the adjacent crowns are crossed and overlapped, the adjacent crowns can be directly clustered into a crown monomer;
(2) rough segmentation of the point clouds of the connected crowns: firstly, projecting the point clouds of the connected tree crowns on an XOZ plane according to the X coordinate (X) of the horizontal mass center of the tree trunk1，x2，x3，……，xn) Numbering the street trees in an ascending manner: 1, 2, 3, … …, n; ② setting step length delta x to pair interval [ x1，xn]Performing equal interval division, and counting the number of sub-intervals [ x ]1+k*Δx，x1+(k+1)*Δx]Point clouds inside; traversing all the intervals, respectively finding out a point p with the maximum elevation in each interval and storing the point p into a linked list L; taking out the interval [ x ] for determining the dividing point between the ith street tree and the (i + 1) th street treei，xi+1]The point set in the corresponding linked list L with the minimum height in the point set is the initial segmentation point SegPoint; fourthly, sequentially carrying out pairwise division on the adjacent street trees from left to right according to the mode until all the street trees are processed;
(3) finely dividing the point clouds of the connected crowns: firstly, setting fixed intervals along the Z-axis direction for the roughly divided crown point cloud to perform layered projection; then extracting a contour point set P of the left and right canopy point clouds in the same layer by combining a convex hull algorithmL、PRFitting the set of points P according to the least squares principleLAnd PRPlane circular model O ofL、OR(ii) a Finally, analyzing the canopy point cloud P and the fitting circle OL、OR(radii are each rL、rR) The relative position relationship between the P points and the fitting circle to determine the segmentation range of the connected crown, and respectively calculate the horizontal distance d between the P points and the fitting circleL、dRThe crown point cloud is segmented according to the following geometrical conditions: satisfy dL<rL&&dR<rRIf the P point belongs to the left tree crown and the right tree crown simultaneously; ② satisfy dL<rL&&dR>rRIf the P point belongs to the left crown but not the right crown, judging that the P point belongs to the left crown but not the right crown; satisfy dL>rL&&dR<rRIf the P point belongs to the right crown but not the left crown, judging that the P point belongs to the right crown but not the left crown; traversing all the layered crown point clouds, and sequentially segmenting every two adjacent street trees according to the method.
6. The method for automatically extracting the street tree target and the forest attributes thereof based on the vehicle-mounted radar laser data as claimed in claim 1, wherein the sixth step comprises the following sub-steps:
(1) performing coordinate inverse transformation in the substep (3) of the second step on the extracted street tree monomer point cloud, and restoring the real coordinates of the street tree point cloud;
(2) calculating horizontal positioning coordinates
Traversing the trunk point cloud of the single-trunk street tree, and according to a formula:
calculating a trunk point cloud horizontal positioning coordinate;
(3) calculating the tree height H: acquiring a Digital Elevation Model (DEM) in a measuring area range, and calculating the row and column index numbers of corresponding pixels V in the DEM and the elevation value Z of the pixels V by using the horizontal positioning coordinates of the street treesdemThe height of the ground at the bottom of the trunk is obtained; traversing crown point cloud to obtain elevation Z of highest pointmaxAccording to the formula：H＝Zmax-ZdemCalculating the tree height H;
(4) calculating the breast diameter DBH: from Z in substep (2)demFirstly, upwards respectively intercepting a point set P at three height positions of the trunk, which are 1.2m, 1.3m and 1.4m away from the ground1、P2、P3Estimating the radius r of a target point set fitting circle by using a RANSAC plane circle fitting model1、r2、r3According to the formula:
calculating the breast diameter DBH;
(5) calculating the height h under the branch, namely layering the point cloud of the street tree at equal intervals along the Z-axis direction, and determining a fitting circle parameter for the point cloud in each layer by using an RANSAC algorithm; recording the area of a point cloud fitting circle in the kth layer as SkAccording to the formula: Δ S ═ Sk+1-SkCalculating the area difference delta S of the point cloud fitting circles in the two adjacent layers, if delta S>The threshold value indicates that the cross section area of the trunk is mutated, and a first branch point is detected; recording the maximum value of the k layer point cloud elevation as ZkAccording to the formula: h ═ Zk-ZdemCalculating the height under the branch;
(6) calculating the crown width CW: projecting the crown point cloud into an XOY plane and calculating the maximum X of the horizontal coordinatesmin、Xmax、Ymin、YmaxDefining the extension of the crown in the north-south direction as: d1＝Ymax-Ymin(ii) a The east-west extension is: d2＝Xmax-XminAccording to the formula:
calculating the crown width CW;
(7) outputting the attribute result of the street tree: outputting the attribute of the shade forest tree as a format required by the topographic map software, such as: text files, SHP files, DWG files.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111317369.1A CN114119863A (en) | 2021-11-09 | 2021-11-09 | Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111317369.1A CN114119863A (en) | 2021-11-09 | 2021-11-09 | Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114119863A true CN114119863A (en) | 2022-03-01 |
Family
ID=80377600
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111317369.1A Pending CN114119863A (en) | 2021-11-09 | 2021-11-09 | Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114119863A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114782925A (en) * | 2022-06-17 | 2022-07-22 | 四川省公路规划勘察设计研究院有限公司 | Vehicle-mounted LIDAR data-based highway guardrail vectorization method and device |
| CN115018897A (en) * | 2022-06-24 | 2022-09-06 | 滁州学院 | Method for extracting typical surface feature elements of laser point cloud city |
| CN115205690A (en) * | 2022-09-14 | 2022-10-18 | 深圳大学 | Method and device for extracting street tree in monomer mode based on MLS point cloud data |
| CN116129391A (en) * | 2023-04-18 | 2023-05-16 | 山东省国土测绘院 | Method and system for extracting pavement tree from vehicle-mounted laser point cloud |
| CN116309566A (en) * | 2023-05-17 | 2023-06-23 | 深圳大学 | Point cloud-based adhesive artificial rod-shaped object monomer extraction method and related equipment |
-
2021
- 2021-11-09 CN CN202111317369.1A patent/CN114119863A/en active Pending
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114782925A (en) * | 2022-06-17 | 2022-07-22 | 四川省公路规划勘察设计研究院有限公司 | Vehicle-mounted LIDAR data-based highway guardrail vectorization method and device |
| CN114782925B (en) * | 2022-06-17 | 2022-09-02 | 四川省公路规划勘察设计研究院有限公司 | Highway guardrail vectorization method and device based on vehicle-mounted LIDAR data |
| CN115018897A (en) * | 2022-06-24 | 2022-09-06 | 滁州学院 | Method for extracting typical surface feature elements of laser point cloud city |
| CN115205690A (en) * | 2022-09-14 | 2022-10-18 | 深圳大学 | Method and device for extracting street tree in monomer mode based on MLS point cloud data |
| CN115205690B (en) * | 2022-09-14 | 2022-12-06 | 深圳大学 | Method and device for extracting street tree in monomer mode based on MLS point cloud data |
| CN116129391A (en) * | 2023-04-18 | 2023-05-16 | 山东省国土测绘院 | Method and system for extracting pavement tree from vehicle-mounted laser point cloud |
| CN116309566A (en) * | 2023-05-17 | 2023-06-23 | 深圳大学 | Point cloud-based adhesive artificial rod-shaped object monomer extraction method and related equipment |
| CN116309566B (en) * | 2023-05-17 | 2023-09-12 | 深圳大学 | Point cloud-based adhesive artificial rod-shaped object monomer extraction method and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114119863A (en) | Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data | |
| CN113034689B (en) | Laser point cloud-based terrain three-dimensional model, terrain map construction method and system, and storage medium | |
| CN111598823B (en) | Multisource mobile measurement point cloud data space-ground integration method and storage medium | |
| CN110570428B (en) | Method and system for dividing building roof sheet from large-scale image dense matching point cloud | |
| CN107516077B (en) | Traffic sign information extraction method based on fusion of laser point cloud and image data | |
| CN109270544A (en) | Mobile robot self-localization system based on shaft identification | |
| CN106529469A (en) | Unmanned aerial vehicle airborne LiDAR point cloud filtering method based on adaptive gradient | |
| CN112070769A (en) | Layered point cloud segmentation method based on DBSCAN | |
| CN108564650B (en) | Lane tree target identification method based on vehicle-mounted 2D LiDAR point cloud data | |
| CN114764871B (en) | Urban building attribute extraction method based on airborne laser point cloud | |
| CN112132840B (en) | Vehicle-mounted street tree point cloud classification and feature information extraction method | |
| CN112729130A (en) | Method for measuring height of tree canopy by satellite remote sensing | |
| WO2023060632A1 (en) | Street view ground object multi-dimensional extraction method and system based on point cloud data | |
| CN111091079A (en) | TLS-based method for measuring dominant single plant structural parameters of vegetation in alpine and fragile regions | |
| CN112241661A (en) | Urban ground feature fine classification method combining airborne LiDAR point cloud data and aerial image | |
| CN115063555A (en) | Method for extracting vehicle-mounted LiDAR point cloud street tree growing in Gaussian distribution area | |
| CN111487643A (en) | Building detection method based on laser radar point cloud and near-infrared image | |
| CN107993242B (en) | Method for extracting boundary of missing area based on airborne LiDAR point cloud data | |
| CN114136208A (en) | Low-cost tree structure automatic reconstruction method for lossless estimation of stumpage volume | |
| CN112906719A (en) | Standing tree factor measuring method based on consumption-level depth camera | |
| Cong et al. | Research on a point cloud registration method of mobile laser scanning and terrestrial laser scanning | |
| CN116977580A (en) | Method for manufacturing mountain area large scale DEM based on airborne LiDAR | |
| CN115984721A (en) | Method for realizing country landscape management based on oblique photography and image recognition technology | |
| CN115690138A (en) | Road boundary extraction and vectorization method fusing vehicle-mounted image and point cloud | |
| CN116977593A (en) | Single wood segmentation method based on super-voxel concave-convex segmentation and color region growth |
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 |