CN114511546A - A method for obtaining diameter at breast height of trees from laser point cloud based on DBSCAN clustering and four quadrants - Google Patents

Abstract

The invention discloses a method for obtaining a diameter at breast height of a laser point cloud based on DBSCAN clustering and four quadrants, comprising the following steps: 1) collecting laser point cloud data of trees in a survey area; Generate tree diameter at breast height point cloud files; 3) Use DBSCAN clustering method to group the DBH point clouds of different trees, and obtain the point cloud collection of the diameter at breast height of a single tree and its supporting rods; 4) Use DBSCAN clustering method to classify and eliminate The noise points of the support rods around a single tree are obtained to obtain the point cloud set at the DBH of a single tree; 5) filter the noise points to obtain the optimal point cloud set of the DBH of a single tree; 6) use the least squares calculation method to calculate the optimal circle, Get tree trails and tree centers. The invention can automatically and quickly separate the DBH point cloud of each forest tree, and segment the optimal DBH point set of a single tree tree, and then perform the calculation, with accurate calculation, high intelligence and high efficiency.

Description

A method for obtaining diameter at breast height of trees from laser point cloud based on DBSCAN clustering and four quadrants

technical field

The invention relates to the technical field of survey and measurement of urban greening forest trees, in particular to a method for obtaining the diameter at breast height of a laser point cloud based on DBSCAN clustering and four quadrants.

Background technique

With the development of society, people's requirements for living environment are getting higher and higher, and the functions of urban greening are gradually known by people. These functions are closely related to people's physical and mental health and the improvement and improvement of living environment quality. Urban greening has played an irreplaceable and measurable role in the city. Forests play an important role in urban greening. In order to maintain urban greening resources, it is necessary to conduct regular surveys of urban greening resources. In the investigation of forest resources, it is most critical to measure the diameter of the trees. In urban greening projects, the tree diameter also directly affects the project cost.

With the development of 3D laser scanning technology, 3D laser scanners have been widely used in the acquisition of tree diameter at breast height. At present, the extraction work mainly uses the computer and its graphics equipment (CAD) to determine the location of the DBH point cloud, manually remove the noise points, and then calculate the individual tree [Diameter at breast height extraction method based on single-station 3D laser scanning data, Ke. Min et al., Zhejiang Forestry Science and Technology, 2019, 39(2), 87-91], this method requires a lot of manual participation, filtering noise points, filtering point cloud data, and the efficiency is low.

SUMMARY OF THE INVENTION

In order to solve the technical problems existing in the known technology, the present invention provides a method for obtaining the DBH of a laser point cloud based on DBSCAN clustering and four quadrants. The method can automatically and quickly separate the DBH point cloud of each tree, and segment a single tree The optimal DBH point set of forest trees is then calculated. The calculation is accurate, the degree of intelligence is high, and the efficiency is high.

The technical scheme adopted by the present invention to solve the technical problems existing in the known technology is: a method for obtaining the diameter at breast height of a laser point cloud based on DBSCAN clustering and four quadrants, comprising the following steps:

1) Using a station-type 3D laser scanner to collect laser point cloud data of forest trees in the survey area;

2) Splicing, cutting and intercepting the laser point cloud data of trees, obtaining the point cloud of the specified thickness at the DBH of the tree in the survey area, and generating the DBH point cloud file;

3) The DBSCAN clustering method is used to group the DBH point clouds of different trees, and obtain the point cloud set of DBH of a single tree and its support rods;

4) The DBSCAN clustering method is used for classification, and the noise points of the support sticks around the single tree are removed to obtain the point cloud set at the DBH of the single tree;

5) Filter the noise points to obtain the cloud set of the optimal point at breast height of a single tree;

6) Taking the optimal point set as the point set and the circle equation as the basic equation, the least squares calculation method is used to calculate the optimal circle of the point set, the radius of the optimal circle is the diameter of the tree, and the center of the optimal circle is Forest Center.

The method of described step 4) is, to the point cloud collection of individual forest trees and forest support sticks, according to the DBSCAN clustering method, the setting scanning radius is 0.04 meters, and the minimum number of scanning points is 7 points. Support stick point cloud data for classification, when the point cloud is divided into one category, the point set is a single tree point set; when the point set is divided into two categories, the point set with more points is a single tree tree point set ; When the point set is divided into three categories, first determine whether there is one type of point set with more than 5 times the number of points of the other two types of point sets, if so, the point set is a single tree point set; if not, calculate three The center point of the classified point set forms a triangle, and the point set with the largest angle is the single tree point set; when the number of point sets is more than four, calculate the center point of each category point set and the center point of all category point sets, close to The point set of the center point of all category point sets is the single tree tree point set.

The method of the step 5) is to obtain the mean value of the point cloud set of a single tree tree, obtain the mean value point of the point cloud set of a single tree tree tree, and take the mean value point as a center point, and divide the point cloud data of a single tree tree tree into four quadrants; Calculate the point that is farthest from the center point in each quadrant, and construct a triangle with the point farthest from the center point in each quadrant and the center point in order; The triangle where the center of the tree diameter at breast height is located;

Calculate the circumcircle of the triangle with the least number of points per unit area, and obtain the center and radius of the circumcircle; take the center of the circumcircle as the center, expand the radius of the circumcircle by 0.03 meters as the new radius, form a new circumscribed circle, and calculate the points that fall into the new circumscribed circle , and then judge whether the ratio of the point set to the DBH point set of a single tree is greater than 80%, if so, the point set is the optimal point set; , calculate the points that fall into the center circle and the circumscribed new circle at the same time in a single tree point set, and take this point set as the new tree point set;

Taking the center of the circumscribed circle as the initial vibration center, and taking the connecting line between the center of the circumcircle and the center point as the diagonal line to construct a rectangular area, move 0.003 meters in the X/Y directions respectively to form a set of vibration centers in the rectangular area; construct a set of vibration centers When , deduct the point 0.01 meters away from the center point; take the vibration center as the center, construct the vibration circle with the distance between the vibration center and the center point as the radius, and construct the vibration ring with the radius of the vibration circle 0.01 meters inside and outside;

Move the vibration center in the vibration center set, build different vibration rings, and calculate the points that fall into each vibration ring in the new tree point set. When the whole vibration center set is cycled once, the vibration ring with the most points is the optimal vibration ring , the point set in the vibration ring is the optimal point set.

The method of described step 6) is to calculate the median value of the optimal point set, obtain the difference between the coordinates of all points and the median value, and then according to the difference value, according to the least squares method, calculate the optimal fitting circle of the point set. , so as to obtain the position and DBH of a single tree.

The method of the described step 3) is to load the tree diameter point cloud file, adopt the DBSCAN clustering method, set the scanning radius to be 0.4 meters, and the minimum number of scanning points is 30 points, and the forest tree data in the entire area is divided, and the number of filtered points is less than 30 points. 30 points of noise points, so as to segment the point cloud data of each tree and tree support sticks.

To sum up, the present invention proposes a method for obtaining diameter at breast height of forest trees based on DBSCAN clustering and four-quadrant laser point cloud. First, the three-dimensional laser point cloud data of trees is collected, and the point cloud data at breast height of trees is cut out. Based on the DBSCAN clustering , perform single tree segmentation on the tree data, separate the single tree tree data and the tree support stick data, obtain the point cloud data of single tree tree diameter at breast height, and then filter the noise points based on the four-quadrant method to obtain the optimal point set of tree diameter at breast height. Finally, Use the least squares method to calculate the tree position and DBH.

The advantages and positive effects of the present invention are: by applying the DBSCAN clustering method to the classification of the DBH laser point cloud data of trees, and the acquisition of the DBH laser point cloud data of individual trees, the DBH point cloud of each tree can be automatically and quickly separated, Helps to improve the calculation efficiency of tree diameter at breast height. By using the four-quadrant method to filter the noise points, the optimal diameter at breast height laser point cloud of a single tree is obtained, and then the calculation is carried out. The calculation is accurate, the degree of intelligence is high, the use is convenient, and the efficiency is high.

Description of drawings

Fig. 1 is the flow chart of the present invention;

Fig. 2 is a schematic diagram of forest tree laser point cloud data scanned by a three-dimensional laser scanner;

Figure 3 is a schematic diagram of the intercepted tree diameter at breast height laser point cloud;

Figure 4 is a schematic diagram of the point cloud collection of different individual trees and their support rods after the initial classification of forest trees;

Figure 5 is a schematic diagram of the point cloud collection of a single tree tree after fine classification;

Figure 6 is a schematic diagram of the DBH point set after the four-quadrant method has removed noise points;

FIG. 7 is a schematic diagram of the optimal circle fitted according to the least squares method.

Detailed ways

In order to further understand the content of the invention, features and effects of the present invention, the following embodiments are exemplified and described in detail with the accompanying drawings as follows:

Please refer to FIGS. 1 to 7. In order to solve the problem of low efficiency in manually interpreting the DBH point cloud of trees in CAD, the present invention provides a DBSCAN clustering and four-quadrant laser point cloud DBH acquisition method, which includes the following steps:

1) Using a station-type 3D laser scanner to collect laser point cloud data of forest trees in the survey area

The station laser scanner is used to collect the laser point cloud of the forest tree, so as to obtain the laser point cloud of the forest tree from the ground to the canopy in the entire survey area.

Before use, place the scanner in the observation environment for more than 30 minutes, then connect and power on the test, and start the operation after confirming that it is correct. During the scanning operation, set the distance between the stations to 50m, and set up the stations one by one to scan to obtain data.

2) Splicing, cutting, and intercepting the tree laser point cloud data, obtaining the point cloud with the specified thickness at the DBH of the tree, and generating the DBH point cloud file

The vertical line connecting the two adjacent stations is the boundary line, and the closed area formed by the borders on both sides of the urban greening trees is the single-site cloud data cropping range, and the rectangular single-site cloud data is cut out. The cut point cloud data is spliced with the coordinates of the set site to obtain the point cloud data of the entire survey area; according to the ground laser point cloud data of the entire survey area, the DTM model of the survey area is established, and the point cloud data of the specified thickness at the DBH of the tree is intercepted. Generate a tree DBH point cloud data file.

3) Using the DBSCAN clustering method, the DBH point clouds of different trees are grouped, and the point cloud collection of the DBH of a single tree and its supporting rods is obtained.

Load the tree diameter at breast height point cloud file, use the DBSCAN clustering method, set the scanning radius to 0.4 meters, and set the minimum number of scanning points to 30 points. Point cloud data for each tree and tree support.

4) The DBSCAN clustering method is used for classification, and the noise points of the support sticks around a single tree are removed to obtain the point cloud set at the DBH of a single tree.

For the point cloud collection of a single tree and tree support rods, according to the DBSCAN clustering method, set the scanning radius to 0.04 meters and the minimum number of scanning points to 7 points. A collection of point clouds at DBH of a single tree.

The point cloud of a single tree and its support rod has the following characteristics. The point cloud of the DBH of a single tree is generally in the middle, the point cloud of the DBH of a single tree is denser and has more points; the point cloud of the support rod is at the edge and has fewer points. , the support rods are generally three, symmetrically distributed;

According to the point cloud characteristics of a single tree and its supporting rods, when the point cloud is divided into one category, the point set is a single tree point set; when the point set is divided into two categories, the point set with more points is Single tree point set; when the point set is divided into three categories, first determine whether there is one type of point set with more than 5 times the number of points of the other two types of point sets, if so, the point set is a single tree tree point set; if No, calculate the center point of three classification point sets to form a triangle, the point set with the largest angle is the single tree point set; when the number of point sets is more than four, calculate the center point of each category point set and all category point sets The center point of , and the point set close to the center point of all category point sets is the single tree point set.

5) Filter the noise points to obtain the optimal point cloud set for the DBH of a single tree

Take the mean value of the point cloud set of a single tree tree to obtain the mean point of the point cloud set of a single tree tree, take the mean point as the center point, and divide the point cloud data of a single tree tree into four quadrants; calculate the distance between each quadrant and the center point. For the farthest point, construct a triangle with the points farthest from the center point in each quadrant and the center point in sequence; calculate the number of points per unit area of each triangle, and determine the triangle with the least number of points per unit area as the triangle with the center of the diameter at breast height of a single tree;

Calculate the circumcircle of the triangle with the least number of points per unit area, and obtain the center and radius of the circumcircle; take the center of the circumcircle as the center, expand the radius of the circumcircle by 0.03 meters as the new radius, form a new circumscribed circle, and calculate the points that fall into the new circumscribed circle , and then judge whether the ratio of the point set to the DBH point set of a single tree is greater than 80%, if so, the point set is the optimal point set; , calculate the points that fall into the center circle and the circumscribed new circle at the same time in a single tree point set, and take this point set as the new tree point set;

Taking the center of the circumscribed circle as the initial vibration center, and taking the connecting line between the center of the circumcircle and the center point as the diagonal line to construct a rectangular area, move 0.003 meters in the X/Y directions respectively to form a set of vibration centers in the rectangular area; construct a set of vibration centers When , deduct the point 0.01 meters away from the center point; take the vibration center as the center, construct the vibration circle with the distance between the vibration center and the center point as the radius, and construct the vibration ring with the radius of the vibration circle 0.01 meters inside and outside;

Move the vibration center in the vibration center set, build different vibration rings, and calculate the points that fall into each vibration ring in the new tree point set. When the whole vibration center set is cycled once, the vibration ring with the most points is the optimal vibration ring , the point set in the vibration ring is the optimal point set.

The four-quadrant method was used to screen out the position of the center of a single tree, and then the proportion method, intersection method and moving center method were used to filter noise points to obtain the optimal point set at DBH of a single tree.

6) Taking the optimal point set as the point set and the circle equation as the basic equation, the least squares calculation method is used to calculate the optimal circle of the point set, the radius of the optimal circle is the diameter of the tree, and the center of the optimal circle is Forest Center.

In order to prevent singular matrices from appearing in the calculation process, calculate the mean value of the optimal point set, find the difference between the coordinates of all points and the median value, and then calculate the optimal fitting circle of the point set according to the difference according to the least square method, Thus, the position and DBH of individual trees were obtained.

The survey of urban greening trees has the characteristics of large number, high density, poor visibility and high survey accuracy. The stand-type 3D laser scanner has the characteristics of high scanning accuracy and long scanning distance. The station-type 3D laser scanner can quickly obtain forest tree laser point cloud data.

According to the scanned laser point cloud data, the corresponding cutting, splicing and interception are performed to obtain the required forest tree laser point cloud data.

Preliminary classification, this step relies on computer program processing. For the laser point cloud data of tree diameter at breast height, according to the DBSCAN clustering method, the scanning radius is set to 0.4 meters, and the minimum number of scanning points is 30 points.

For fine-grained classification, this step relies on computer program processing. For the point cloud set of the segmented single tree and tree support rods, based on the DBSCAN clustering method, the scanning radius is set to 0.04 meters, and the minimum number of scanning points is 7 points, and the data of the single tree tree and the surrounding support rods are segmented; and According to the point cloud characteristics of individual trees and support rods, the DBH point cloud data of individual trees were screened out.

For the finely classified DBH point cloud, a four-quadrant method is proposed to filter out the noise points and determine the optimal point set. Calculate the mean as the center point; divide the point cloud into four quadrants according to the center point; then construct triangles in order according to the farthest point from the center point to the four quadrants, determine the triangle with the smallest point density as the location of the tree center, and then follow a certain The radius forms a new circumscribed circle, and then uses proportion judgment, intersection calculation, and moving the center of the circle to filter out noise points to obtain the optimal point set.

For the obtained optimal point set, calculate the mean value, calculate the difference between the optimal point set and the mean value, and then use the least squares method to obtain the center and diameter of the circle.

Examples incorporating numerical values:

Take a road greening area in Tianjin as an example. The area is located in Tianjin Binhai New Area. Among them, there are trees on the west side and shrubs on the east side. Only trees are collected this time. The arbor spacing is about 2.5m×2.5m, and the age of the seedlings is about 3a.

Step 1, use Maptek I-Site 8820 3D laser scanner to scan and measure, and use Trimble R8GPS dual-frequency receiver to collect data. Place the scanner in the observation environment for more than 30 minutes, then connect and power on the test, and start the operation after confirming that it is correct. The steps of the scanning operation are as follows:

Use GPS-RTK to obtain the three-dimensional coordinate information of the scanning site and the backsight point. Set up the 3D laser scanner at the design station, level the instrument, then locate the backsight point of the 3D laser scanner, and enter the scanning site and the coordinate information of the backsight point into the 3D laser scanner control handbook, and scan the area. The angle is set to 360°, the scan quality is set to medium, and the scan measurement is started.

After the scan is completed, check the data integrity, supplement the scan if it is missing, and switch stations if not, until the scan of the entire survey area is completed. The tree point cloud effect in the survey area after scanning is shown in Figure 2.

Step 2: The vertical line connecting the two adjacent sites is the boundary line, and the closed area formed by the borders on both sides of the urban greening trees is the single-site cloud data cutting width, and the cutting length is set to 50 meters. Rectangular single-site cloud data, and then splicing the cut point cloud data of each station using the site coordinates to obtain the point cloud data of the entire area; The point cloud data of the thickness of 0.04 meters at a distance of 13 meters from the ground is used to generate the point cloud data file of tree diameter at breast height. As shown in Figure 3, for the sake of clarity and beauty, the schematic diagram only shows the point cloud data of 4 trees.

Step 3: Preliminary classification, using DBSCAN clustering algorithm, according to the distance between trees is greater than 0.4 meters, set the classification radius to 0.4 meters, the minimum number of points is 30 points, segment the forest tree data in the entire area, and filter out the number of points less than 30 points. The noise points of the points are divided into the point cloud data of each tree and tree support rods. As shown in Figure 4, the data of 4 forest trees after preliminary classification are shown.

DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm. Unlike partitioning and hierarchical clustering methods, it defines a cluster as the largest set of densely connected points, can divide regions with high enough density into clusters, and can find clusters of arbitrary shapes in noisy spatial databases.

Step 4, fine classification, set the scanning radius to 0.04 meters, and the minimum number of scanning points to 7 points, classify the data of a single tree and the surrounding support sticks of the tree; filter out the point cloud data of each tree. As shown in Figure 5, the point cloud data of the support rods were filtered out, and the point cloud data of a single tree was obtained.

The point cloud of a single tree and support rod has the following characteristics. The point cloud of the DBH of a single tree is generally in the middle, the point cloud of the DBH of a single tree is denser and more numerous, and the point cloud of the support rod is located at the edge and has a small number. There are generally three or four supporting rods, symmetrically distributed; according to the point cloud characteristics of a single tree and supporting rods, when the point cloud is divided into one category, the point set is a single tree point set; when the point set is divided into In the case of two types, the number of point sets is the single tree point set; when the point set is divided into three types, first determine whether there is one type of point set more than 5 times the number of the other two types, if yes, then the point set is a single tree point set; otherwise, calculate the center points of three classification point sets to form a triangle, and the point set with the largest angle is the single tree tree point set; when the number of point sets is more than four, calculate the center of each category point set point and the center point of all category point sets, and the point set close to the center point of all point sets is a single tree tree point set.

When calculating the obtuse angle of a triangle, you need to use the formula for the distance between two points and the law of cosines.

The formula for the distance between two points is as follows:

Cosine Theorem:

Step 5: Calculate the average value of the single tree point set obtained in step 4 as the center point (X ₀ , Y ₀ ), and according to the distance between the X/Y coordinates of each point in the point set and the center point (X ₀ , Y ₀ ) The size relationship divides the point set into four quadrant point sets.

Find the point farthest from the center point (X ₀ , Y ₀ ) in each quadrant, and construct a triangle by ordering the center point and the farthest point in each quadrant. For example, the center point and the farthest point in the first quadrant and the farthest point in the second quadrant construct a triangle, and the center point and the farthest point in the second quadrant and the farthest point in the third quadrant construct a triangle, and so on. If there is no point in a quadrant, skip this quadrant. For example, if there is no point in the second quadrant, skip this quadrant, and the center point and the farthest point in the first quadrant and the third quadrant construct a triangle.

则三点共线，不构成三角形，否则，则构成三角形。According to the slope method, judge whether the three points are collinear, when x2=x1, x3=x1, or

If the three points are collinear, they do not form a triangle; otherwise, they form a triangle.

Calculate the area of each triangle according to Heron's formula. Heron's formula is as follows: Assuming that the side lengths of the triangle are a, b, and c, respectively, the area S of the triangle can be obtained by the following formula:

According to the positional relationship between the point and the triangle, the number of points that fall into each triangle is calculated.

The method for judging the positional relationship between a point and a triangle is as follows: Assume that the three points of the triangle are A, B, and C in a clockwise (or counterclockwise) order. For a certain point P, find three vectors PA, PB, PC, and then calculate the following three cross products (^ represents the symbol of cross product): t1=PA^PB, t2=PB^PC, t3=PC^PA, if t1, t2, t3 have the same sign (the same positive or the same negative), then P is inside the triangle, otherwise it is outside.

According to the area of each triangle and the number of points falling into the triangle, the point density of each triangle is calculated.

According to the circumcircle formula, calculate the circumcircle of the triangle with the smallest point density.

The formula for calculating the circumcircle is as follows: Let the three vertices of the triangle be A(X0,Y0), B(X1,Y1), C(X2,Y2),

Let a1=X1-X0, b1=Y1-Y0, c1=(a1*a1+b1*b1)/2,

a2=X2-X0, b2=Y2-Y0, c2=(a2*a2+b2*b2)/2,

d=a1*b2-a2*b1

Then: the coordinates of the center of the circle (XC, YC) are XC=x0+(c1*b2-c2*b1)/d

YC=y0+(a1*c2-a2*c1)/d

Taking the center of the circumcircle (X _C , Y _C ) as the center, and expanding the radius r of the circumcircle by 0.03 meters as a new radius, construct concentric circles.

According to the distance between the point and the center of the circle, calculate the points where the points of a single tree fall into the concentric circles, and determine the proportion of the points in the circumscribed new circle to the number of points in a single tree. If the proportion exceeds 80%, then the circumscribed new circle The points inside are the optimal point set.

If the proportion is less than 80%, take the center point (X ₀ , Y ₀ ) as the center of the circle and the radius of the circumscribed circle r as the radius to construct the central circle, and calculate the points that fall into the central circle and the circumscribed new circle at the same time for the points of a single tree. , take this point set as the new tree point set.

Taking the center of the circumcircle (X _C , Y _C ) as the initial vibration center, and taking the connecting line between the center of the circumcircle (X _C , Y _C ) and the center point (X ₀ , Y ₀ ) as the diagonal line, a rectangular area is constructed, and a rectangular area is constructed with X Move 0.003 meters in the /Y direction respectively to form a vibration center set in a rectangular area; when constructing a vibration center set, deduct points 0.01 meters away from the center point (X ₀ , Y ₀ ).

Taking the vibration center as the center, the vibration circle is constructed with the distance between the vibration center and the center point (X ₀ , Y ₀ ) as the radius, and the vibration ring is constructed with 0.01 meters inside and outside the radius of the vibration circle.

Move the vibration center in the vibration center set, build different vibration rings, and calculate the points that fall into each vibration ring in the new tree point set. When the entire vibration center set is cycled once, the vibration ring with the most points is the optimal vibration ring, and the point set in the vibration ring is the optimal point set.

Step 6, according to the optimal point set of a single tree calculated in step 5, calculate the mean value of the optimal point set, calculate the difference between the optimal point concentration point and the mean value, and then calculate the optimal circle according to the least square method, and add the coordinates of the center of the circle to the mean value. The position of the tree is obtained, and the diameter is the tree diameter at breast height. As shown in Figure 7.

The principle of least squares is as follows. Assuming that the coordinates of the center of the circle are (x ₀ , y ₀ ), the radius is R, n two-dimensional coordinate points are distributed on the circle, and the coordinate points are (x _i , y _i ), i=1,2,..., n;

The basic equation of a circular curve is:

R ² =(xx ₀ ) ² +(yy ₀ ) ² (1)

Expand to get:

make

which is

Then (2) is transformed into

x ² +y ² -ax-by+c=0 (5)

After the formula is extended to n points, it can be written in matrix form:

make:

but

AX=L (8)

According to the least squares algorithm, we can get:

X=(A ^T A) ^-1 A ^T L (9)

According to formula (9), a, b, c can be obtained, and they can be substituted into formula (4) to obtain the coordinates of the center of the circle and the radius.

To sum up, the present invention relates to a method for obtaining the diameter of trees in a laser point cloud based on DBSCAN clustering and four quadrants. A three-dimensional laser scanner is used to scan the urban greening trees, and then intercepted to obtain the laser point cloud data of the diameter at breast height of the trees. An improved rapid acquisition method of tree diameter at breast height is proposed, especially involving DBSCAN clustering and four-quadrant filtering noise points. After preliminary classification, fine classification, four-quadrant filtering noise points, and least squares calculation, it includes the following steps: 1) Using a standing 3D laser scanner to collect the laser point cloud data of trees; 2) Splicing, cutting and intercepting the laser point cloud data of trees, obtaining the point cloud of the specified thickness at the DBH of the tree, and generating the DBH point cloud file; 3 ) based on the DBSCAN clustering method, grouped the DBH point clouds of different trees, and obtained the point cloud set of DBH of a single tree and tree support rods. 4) Based on the DBSCAN clustering method, the noise points of the support rods around the single tree are removed, and the point cloud set of the single tree is obtained; 5) Based on the four-quadrant method, the noise points are filtered to obtain the optimal point cloud set of the single tree diameter at breast height; 6 ) using the least squares method to calculate the tree diameter at breast height.

Although the preferred embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the above-mentioned specific embodiments. Under the inspiration of the present invention, personnel can also make many forms without departing from the scope of the present invention and the protection scope of the claims, which all belong to the protection scope of the present invention.

Claims (5)

1. a laser point cloud tree diameter at breast height acquisition method based on DBSCAN clustering and four quadrants, is characterized in that, comprises the following steps: 1) Using a station-type 3D laser scanner to collect laser point cloud data of forest trees in the survey area; 2) Splicing, cutting and intercepting the laser point cloud data of trees, obtaining the point cloud of the specified thickness at the DBH of the tree in the survey area, and generating the DBH point cloud file; 3) The DBSCAN clustering method is used to group the DBH point clouds of different trees, and obtain the point cloud set of DBH of a single tree and its support rods; 4) The DBSCAN clustering method is used for classification, and the noise points of the support sticks around the single tree are removed to obtain the point cloud set at the DBH of the single tree; 5) Filter the noise points to obtain the cloud set of the optimal point at breast height of a single tree; 6) Taking the optimal point set as the point set and the circle equation as the basic equation, the least squares calculation method is used to calculate the optimal circle of the point set, the radius of the optimal circle is the diameter of the tree, and the center of the optimal circle is Forest Center. 2. the laser point cloud forest tree diameter at breast height acquisition method based on DBSCAN clustering and four quadrants according to claim 1, is characterized in that, the method for described step 4) is, to the point cloud collection of individual tree tree and forest support stick , according to the DBSCAN clustering method, the scanning radius is set to 0.04 meters, and the minimum number of scanning points is 7 points. It is a single tree point set; when the point set is divided into two types, the point set with more points is a single tree point set; when the point set is divided into three types, first determine whether there is one type of point set whose points are other More than 5 times of the two types of point sets, if yes, the point set is a single tree point set; if no, calculate the center points of the three classification point sets to form a triangle, and the point set with the largest angle is the single tree point set ; When the number of point sets is more than four, calculate the center point of each category point set and the center point of all category point sets, and the point set close to the center point of all category point sets is a single tree tree point set. 3. the laser point cloud tree diameter at breast height acquisition method based on DBSCAN clustering and four quadrants according to claim 1, is characterized in that, the method for described step 5) is, to obtain mean value to individual tree tree point cloud collection, obtain Set the mean point of the point cloud of a single tree, and divide the point cloud data of a single tree into four quadrants with the mean point as the center point; The points and center points of the triangle are constructed in sequence; the number of points per unit area of each triangle is calculated, and the triangle with the least number of points per unit area is determined as the triangle with the center of the diameter at breast height of the single tree; Calculate the circumcircle of the triangle with the least number of points per unit area, and obtain the center and radius of the circumcircle; take the center of the circumcircle as the center, expand the radius of the circumcircle by 0.03 meters as the new radius, form a new circumscribed circle, and calculate the points that fall into the new circumscribed circle , and then judge whether the ratio of the point set to the DBH point set of a single tree is greater than 80%, if so, the point set is the optimal point set; , calculate the points that fall into the center circle and the circumscribed new circle at the same time in a single tree point set, and take this point set as the new tree point set; Taking the center of the circumscribed circle as the initial vibration center, and taking the connecting line between the center of the circumcircle and the center point as the diagonal line to construct a rectangular area, move 0.003 meters in the X/Y directions respectively to form a set of vibration centers in the rectangular area; construct a set of vibration centers When , deduct the point 0.01 meters away from the center point; take the vibration center as the center, construct the vibration circle with the distance between the vibration center and the center point as the radius, and construct the vibration ring with the radius of the vibration circle 0.01 meters inside and outside; Move the vibration center in the vibration center set, build different vibration rings, and calculate the points that fall into each vibration ring in the new tree point set. When the whole vibration center set is cycled once, the vibration ring with the most points is the optimal vibration ring , the point set in the vibration ring is the optimal point set. 4. the laser point cloud tree diameter at breast height acquisition method based on DBSCAN clustering and four quadrants according to claim 1, is characterized in that, the method for described step 6) is, calculates the median value of optimal point set, seeks all points The difference between the coordinates and the median value, and then according to the difference, according to the least squares method, the optimal fitting circle of the point set is calculated, so as to obtain the position and DBH of a single tree. 5. the laser point cloud tree diameter at breast height acquisition method based on DBSCAN clustering and four quadrants according to claim 1, is characterized in that, the method for described step 3) is, load tree diameter at breast height point cloud file, adopt DBSCAN clustering method , set the scanning radius to 0.4 meters, and the minimum number of scanning points to be 30 points, segment the forest data in the entire area, filter out the noise points with less than 30 points, and then segment the point cloud data of each forest tree and forest support stick.

Images