CN110428438B - Single-tree modeling method and device and storage medium - Google Patents
Single-tree modeling method and device and storage medium Download PDFInfo
- Publication number
- CN110428438B CN110428438B CN201910625385.3A CN201910625385A CN110428438B CN 110428438 B CN110428438 B CN 110428438B CN 201910625385 A CN201910625385 A CN 201910625385A CN 110428438 B CN110428438 B CN 110428438B
- Authority
- CN
- China
- Prior art keywords
- tree
- point cloud
- cloud data
- target
- data set
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/64—Analysis of geometric attributes of convexity or concavity
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a single-tree modeling method, a single-tree modeling device and a storage medium. The method comprises the following steps: performing single tree clipping on forest point cloud data of a target forest area, and extracting partial single tree point cloud data as a sample data set; training the sample data set by utilizing a pre-designed 2.5D Manhattan distance model to obtain a target forest region model; respectively inputting all the single-tree point cloud data into the target forest region model to obtain a segmentation threshold value of the target forest region; and extracting the target tree data sets one by one according to the segmentation threshold. By utilizing the single-wood model constructed by the invention, all real single-wood sample point clouds can be input into the single-wood model to obtain the reference forms of the tree species, and the parameter threshold value required by the algorithm is extracted from the forms to achieve de-empirical thresholding.
Description
Technical Field
The invention relates to the field of data processing, in particular to a method and a device for modeling a single tree and a storage medium.
Background
Over the past decade, a number of methods have been proposed to measure the number of individuals through laser-like big data, which directly indicates the possibility of implementing the segmentation of the singles from the lidar data, and also highlights the challenges of implementing the segmentation of the singles. One common feature of early single-tree segmentation algorithms was the use of Canopy Height Models (CHMs), such as the local maximum focus filter algorithm proposed by Popescu and the watershed segmentation algorithm applied by Mei, Koch, etc. However, the generation of CHM requires: 1. firstly, separating out ground points and vegetation points; 2. generating a Digital Elevation Model (DEM) and a Digital Surface Model (DSM) by interpolation; 3. the difference between the DSM and the DEM generates the CHM. In the above process, many errors are often introduced due to various factors, such as a small density of dots, and the accuracy of the algorithm greatly depends on the fineness of the CHM. For the problems of the early algorithms, some more advanced solutions appear later, such as the method of smoothing CHM proposed by Persson, and the later algorithms that segment directly from the original point cloud: 1) li et al segment each tree by setting a spacing threshold between trees, and this algorithm has certain defects in forests with greater tree density and the optimal threshold will change with regional variation; 2) wei et al find iterative algorithms to locate the canopy using a non-parametric model such as Mean Shift, but Mean Shift algorithm requires a core parameter, wideband, and it is difficult to get a true forest distribution in the vertical direction.
It can be found from the previous research that the average accuracy of the CHM segmentation-based method is lower than that of the method of directly segmenting the point cloud. However, in the algorithm for directly performing the single-tree segmentation on the point cloud, an artificial experience threshold is inevitably added, for example, in the algorithm proposed by Li et al, it is considered that the distance between trees is larger than 1.5m in the region of more than 15 m. In areas below 15 meters, the tree-to-tree spacing can be greater than 1 m. The fixed threshold may result in the accuracy of the algorithm proposed by Li et al in performing the singulation in other forest areas not being maintained at a high level. In the algorithm proposed by Ferraz et al, a small number of reliable single-wood models are extracted by the CHM and used as training samples for statistics, so that a broadband threshold suitable for local application is obtained, but errors exist in the single-wood models obtained by the CHM, and the fitting fineness and difficulty are increased.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a single-tree modeling method, a single-tree modeling device and a storage medium, so that all real point clouds of single-tree samples are input into a single-tree model to obtain a reference form of the tree species, and a parameter threshold value required by an algorithm is extracted from the form to achieve de-empirical thresholding.
In order to solve the technical problem, the invention provides a single-tree modeling method, which comprises the following steps: performing single tree clipping on forest point cloud data of a target forest area, and extracting partial single tree point cloud data as a sample data set; training the sample data set by utilizing a pre-designed 2.5D Manhattan distance model to obtain a target forest region model; respectively inputting all the single-tree point cloud data into the target forest region model to obtain a segmentation threshold value of the target forest region; and extracting the target tree data sets one by one according to the segmentation threshold.
Further, the single-tree clipping is performed on the forest point cloud data of the target forest area, and part of the single-tree point cloud data is extracted as a sample data set, and the method further includes: dividing the forest point cloud data into ground points and vegetation point clouds; interpolating the ground points and generating a digital elevation model; performing normalization processing on the vegetation point cloud by using the digital elevation model; wherein the normalized value of the vegetation point cloud is an elevation value; and when the point corresponding to the vegetation point cloud is a tree top point, the tree top elevation value is obtained.
Further, training the sample data set by utilizing a pre-designed 2.5D-based Manhattan distance model to obtain a target forest region model, specifically, calculating the Manhattan distance between the single-tree point cloud data with the maximum elevation value in the sample data set and the rest single-tree point cloud data to obtain the Manhattan distance model; performing convex hull calculation by using the Manhattan distance model to obtain a convex hull point set; wherein the convex hull point set is a point set sorted from the single-wood point cloud data of the maximum elevation value in a reverse-time-oriented manner; sequentially traversing the convex hull points until the Manhattan distance of the traversed convex hull points is the maximum value; if the Manhattan distance of the traversed convex hull point is not the maximum value, adding the convex hull point to an upper convex hull point set; performing least square fitting on the upper convex hull point set to obtain k and b characteristic values of a linear model of the target tree; and obtaining the outmost Manhattan distance, the average outmost point distance and the maximum outmost distance of the single-wood point cloud data by utilizing the linear model.
Further, all the single-tree point cloud data are respectively input into the target forest zone model, specifically, all the single-tree point cloud data are arranged in a reverse order according to the height values of the single-tree point cloud data, and a reverse order data set is obtained; acquiring the single-tree point cloud data of the maximum elevation value in the reverse data set; transferring the single-tree point cloud data with the maximum elevation in the inverted data set and all the single-tree point cloud data in the maximum radius of the crown thereof from the inverted data set to the target tree data set according to the maximum radius parameter of the crown of the single-tree point cloud data; and circularly executing until all the single wood point cloud data in the reverse data set are traversed.
Further, the target tree data sets are extracted one by one according to the segmentation threshold, and specifically, the distances between all the single-tree point cloud data and the single-tree point cloud data with the maximum height difference are calculated; taking all the single-wood point cloud data in the 2-time maximum peripheral range as a data set to be processed; traversing the single-tree point cloud data in the data set to be processed in a reverse order according to the elevation values; the single-tree point cloud data with the maximum elevation value in the data set to be processed are divided into a target tree data set, and the single-tree point cloud data at infinity are divided into a non-target tree data set.
Further, the method for modeling the single trees further comprises the steps of detecting the target tree data set, specifically, judging whether target tree point cloud data exist in the target tree data set or not, and whether the difference value of the tree top elevation value of any single tree point cloud data in the target tree data set and the tree top elevation value of the target tree point cloud data is smaller than 1.5m or not; if the point cloud data exists, the single-tree point cloud data is classified into the target tree point cloud data set, and the detection operation is finished; if not, continuing to execute the detection operation; judging whether the number of elements of the target tree point cloud data set is less than 5; if yes, marking as an error tree and deleting; if not, marking as the target tree.
Further, the single tree modeling method further includes performing single tree modeling fitting on the target tree data sets, specifically performing least square linear fitting and standard gaussian function fitting on all the target tree data sets to obtain a fitting value of each target tree data set; and marking the target tree data set with the fitting value less than 0.05 as an abnormal tree and deleting the abnormal tree.
Further, the method for modeling the single-tree model further comprises the step of performing tree state correction on the target tree, specifically, the single-tree point cloud data with the characteristic line distance less than 0.1m between the non-target tree data set and the target tree is attributed to the target tree data set by traversing the target tree data set.
The embodiment of the invention has the following beneficial effects:
by utilizing the single-wood model constructed by the embodiment of the invention, all real point clouds of the single-wood sample can be input into the single-wood model to obtain the reference form of the tree species, and the parameter threshold value required by the algorithm is extracted from the form to achieve the de-empirical thresholding.
The invention also provides a single-wood modeling device, which comprises: the processing module is used for performing single-tree clipping on the forest point cloud data of the target forest area and extracting part of the single-tree point cloud data as a sample data set; the modeling module is used for training the sample data set by utilizing a pre-designed 2.5D Manhattan distance model to obtain a target forest region model; the calculation module is used for respectively inputting all the single-tree point cloud data into the target forest region model to obtain a segmentation threshold value of the target forest region; and the extraction module is used for extracting the target tree data sets one by one according to the segmentation threshold.
The embodiment of the invention has the following beneficial effects:
by utilizing the single-wood model constructed by the embodiment of the invention, all real point clouds of the single-wood sample can be input into the single-wood model to obtain the reference form of the tree species, and the parameter threshold value required by the algorithm is extracted from the form to achieve the de-empirical thresholding.
The invention also provides a computer readable storage medium, which includes a stored computer program, wherein when the computer program runs, the apparatus on which the computer readable storage medium is located is controlled to execute the method for modeling single trees as described above.
Drawings
FIG. 1 is a schematic flow chart diagram of a single wood modeling method according to a first embodiment of the present invention;
fig. 2 is a schematic structural diagram of a single-wood modeling apparatus according to a second embodiment of the present invention.
Detailed Description
The technical solutions in the present invention will be described clearly and completely with reference to the accompanying drawings, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
It should be noted that, the step numbers in the text are only for convenience of explanation of the specific embodiments, and do not serve to limit the execution sequence of the steps. The method provided by the embodiment can be executed by the relevant server, and the server is taken as an example for explanation below.
In a first embodiment, please refer to fig. 1.
As shown in fig. 1, a single wood modeling method provided by the first embodiment includes steps S1 to S4:
and S1, performing single-tree clipping on the forest point cloud data of the target forest area, and extracting part of the single-tree point cloud data as a sample data set.
And S2, training the sample data set by utilizing a pre-designed 2.5D Manhattan distance-based model to obtain a target forest region model.
And S3, respectively inputting all the single-tree point cloud data into the target forest region model to obtain a segmentation threshold value of the target forest region.
And S4, extracting the target tree data sets one by one according to the segmentation threshold.
In a specific embodiment, the step S1 further includes: dividing the forest point cloud data into ground points and vegetation point clouds; interpolating the ground points and generating a digital elevation model; performing normalization processing on the vegetation point cloud by using the digital elevation model; wherein the normalized value of the vegetation point cloud is an elevation value; and when the point corresponding to the vegetation point cloud is a tree top point, the tree top elevation value is obtained.
It should be noted that when performing single-tree clipping on the forest point cloud data of the target forest area, a relatively independent tree should be selected as much as possible, that is, no other trees exist around the tree, and ground points are clipped out at the same time, so as to ensure that the real state of the single-tree point cloud data can be completely and correctly reflected.
In a specific embodiment, in the step S2, specifically, a manhattan distance between the single-wood point cloud data with the maximum elevation value in the sample data set and the remaining single-wood point cloud data is calculated to obtain the manhattan distance model; performing convex hull calculation by using the Manhattan distance model to obtain a convex hull point set; wherein the convex hull point set is a point set sorted from the single-wood point cloud data of the maximum elevation value in a reverse-time-oriented manner; sequentially traversing the convex hull points until the Manhattan distance of the traversed convex hull points is the maximum value; if the Manhattan distance of the traversed convex hull point is not the maximum value, adding the convex hull point to an upper convex hull point set; performing least square fitting on the upper convex hull point set to obtain k and b characteristic values of a linear model of the target tree; and obtaining the outmost Manhattan distance, the average outmost point distance and the maximum outmost distance of the single-wood point cloud data by utilizing the linear model.
The calculation formula of the Manhattan distance is shown as formula 1:
wherein the content of the first and second substances,
point i 2.5D Manhattan distance, X, from the tree vertex point 、Y point 、Z point 、X top 、Y top 、Z top Pointing to the X, Y, Z values of i and the X, Y, Z values of the tree vertices, respectively.
It can be understood that by least square linear fitting, the k and b characteristic values of the linear model of the tree can be obtained, so that the tree top elevation value (i.e., z characteristic value) of the single-tree point cloud data is input in formula 2, the outmost manhattan distance of the elevation of the single-tree point cloud data can be obtained, and further, the average outmost point distance and the maximum outmost distance of the single-tree point cloud data are obtained by calculating formula 3 and formula 4.
Wherein, formula 2, formula 3, formula 4 are as follows respectively:
in a specific embodiment, in the step S3, specifically, all the single-wood point cloud data are arranged in a reverse order according to their elevation values to obtain a reverse-order data set; acquiring the single-tree point cloud data of the maximum elevation value in the reverse data set; transferring the single-tree point cloud data with the maximum elevation in the inverted data set and all the single-tree point cloud data in the maximum radius of the crown thereof from the inverted data set to the target tree data set according to the maximum radius parameter of the crown of the single-tree point cloud data; and circularly executing until all the single-tree point cloud data in the reverse data set are traversed.
In a specific embodiment, in the step S4, specifically, the distances between all the single-tree point cloud data and the single-tree point cloud data with the maximum height difference are calculated; taking all the single-wood point cloud data in the 2-time maximum peripheral range as a data set to be processed; traversing the single-tree point cloud data in the data set to be processed in a reverse order according to the elevation values; the single-tree point cloud data with the maximum elevation value in the data set to be processed are divided into a target tree data set, and the single-tree point cloud data at infinity are divided into a non-target tree data set.
It can be understood that the range selection is performed on all the single-tree point cloud data, that is, all the point cloud data are not traversed, so that the processing efficiency of the algorithm is effectively improved. Meanwhile, traversing the selected single tree point cloud data in a reverse order according to the elevation values, and dividing the single tree point cloud data at infinity corresponding to the currently traversed target tree point cloud data into non-target tree data sets, so that two data are processed at one time, and the processing efficiency of the algorithm is further improved.
In a specific embodiment, the method for modeling the single trees further includes step S5, detecting the target tree data set, specifically, determining whether target tree point cloud data exists in the target tree data set, and whether a difference between a tree top elevation value of any single tree point cloud data in the target tree data set and a tree top elevation value of the target tree point cloud data is less than 1.5 m; if the point cloud data exists, the single-tree point cloud data is classified into the target tree point cloud data set, and the detection operation is finished; if not, continuing to execute the detection operation; judging whether the number of elements of the target tree point cloud data set is less than 5; if yes, marking as an error tree and deleting; if not, marking as the target tree.
It can be understood that each time the division is completed, there is a case that there is some error, and therefore, it is necessary to detect each divided tree and determine whether it is a correct tree.
In a specific embodiment, the single tree modeling method further includes step S6, performing single tree modeling fitting on the target tree data sets, specifically, performing least square linear fitting and standard gaussian function fitting on all the target tree data sets to obtain fitting values of each target tree data set; and marking the target tree data set with the fitting value less than 0.05 as an abnormal tree and deleting the abnormal tree.
It can be understood that after the above steps are completed, some trees with a part of wrong segmentation due to too long outer branches of the trees or noisy points existing in the data center exist, but the point cloud structure of the trees deviates from the normal tree characteristics, so that single-tree modeling fitting needs to be performed on all the trees.
In a specific embodiment, the method for modeling single-tree further includes step S7, performing tree-form modification on the target tree, specifically, by traversing the target tree dataset, attributing the single-tree point cloud data in the non-target tree dataset, whose characteristic line distance from the target tree is less than 0.1m, to the target tree dataset.
It can be understood that, in the segmentation result, since the algorithm performs segmentation from high to low, some point cloud data that originally belongs to the tree is segmented first by the tree higher next to the tree, so that the shape of the tree is missing, and therefore, the tree shape correction is required.
Traversing each tree, calculating the distance of the point clouds not corresponding to the tree, and classifying the point clouds with the distance less than 0.1m from the characteristic line of the tree into the point clouds corresponding to the tree, wherein the formula is shown as formula 5:
the embodiment of the invention has the following beneficial effects:
by utilizing the single-wood model constructed by the embodiment of the invention, all real point clouds of the single-wood sample can be input into the single-wood model to obtain the reference form of the tree species, and the parameter threshold value required by the algorithm is extracted from the form to achieve the de-empirical thresholding.
Please refer to fig. 2 for a second embodiment.
As shown in fig. 2, a second embodiment provides a single-wood modeling apparatus, including: the processing module 21 is used for performing single-tree clipping on the forest point cloud data of the target forest area and extracting part of the single-tree point cloud data as a sample data set; the modeling module 22 is used for training the sample data set by utilizing a pre-designed 2.5D Manhattan distance-based model to obtain a target forest region model; the calculation module 23 is configured to input all the single-tree point cloud data into the target forest region model, so as to obtain a segmentation threshold of the target forest region; and the extracting module 24 is configured to extract the target tree data sets one by one according to the segmentation threshold.
In a specific embodiment, the processing module 21 further includes: dividing the forest point cloud data into ground points and vegetation point clouds; interpolating the ground points and generating a digital elevation model; performing normalization processing on the vegetation point cloud by using the digital elevation model; wherein the normalized value of the vegetation point cloud is an elevation value; and when the point corresponding to the vegetation point cloud is a tree top point, the tree top elevation value is obtained.
It should be noted that when performing single-tree clipping on the forest point cloud data of the target forest area, a relatively independent tree should be selected as much as possible, that is, no other trees exist around the tree, and the ground points are clipped out to ensure that the real state of the single-tree point cloud data can be completely and correctly reflected.
In a specific embodiment, the modeling module 22 specifically calculates a manhattan distance between the single-wood point cloud data with the maximum elevation value in the sample data set and the remaining single-wood point cloud data to obtain the manhattan distance model; performing convex hull calculation by using the Manhattan distance model to obtain a convex hull point set; wherein the convex hull point set is a point set sorted from the single-wood point cloud data of the maximum elevation value in a reverse-time-oriented manner; sequentially traversing the convex hull points until the Manhattan distance of the traversed convex hull points is the maximum value; if the Manhattan distance of the traversed convex hull point is not the maximum value, adding the convex hull point to an upper convex hull point set; performing least square fitting on the upper convex hull point set to obtain k and b characteristic values of a linear model of the target tree; and obtaining the outmost Manhattan distance, the average outmost point distance and the maximum outmost distance of the single-wood point cloud data by utilizing the linear model.
The calculation formula of the Manhattan distance is shown as formula 1:
wherein the content of the first and second substances,
point i 2.5D Manhattan distance, X, from the tree vertex point 、Y point 、Z point 、X top 、Y top 、Z top Pointing to the X, Y, Z values of i and the X, Y, Z values of the tree vertices, respectively.
It can be understood that by means of least square linear fitting, k and b characteristic values of a linear model of the tree can be obtained, so that the tree top elevation value (i.e., z characteristic value) of the single-tree point cloud data is input in formula 2, and the outmost Manhattan distance of the elevation where the single-tree point cloud data is located can be obtained, and then the average outmost point distance and the maximum outmost distance of the single-tree point cloud data are obtained by calculating formulas 3 and 4.
Wherein, formula 2, formula 3, formula 4 are as follows respectively:
in a specific embodiment, the calculating module 23 specifically performs reverse order arrangement on all the single-tree point cloud data according to the height values thereof to obtain a reverse order data set; acquiring the single-tree point cloud data of the maximum elevation value in the reverse data set; transferring the single-tree point cloud data with the maximum elevation in the inverted data set and all the single-tree point cloud data in the maximum crown radius of the single-tree point cloud data from the inverted data set to the target tree data set according to the maximum crown radius parameter of the single-tree point cloud data; and circularly executing until all the single-tree point cloud data in the reverse data set are traversed.
In a specific embodiment, the extracting module 24 specifically calculates distances between all the single-tree point cloud data and the single-tree point cloud data with the maximum elevation difference; taking all the single-wood point cloud data in the 2-time maximum peripheral range as a data set to be processed; traversing the single wood point cloud data in the data set to be processed according to the elevation values in a reverse order; the single-tree point cloud data with the maximum elevation value in the data set to be processed are divided into a target tree data set, and the single-tree point cloud data at infinity are divided into a non-target tree data set.
It can be understood that the range selection is performed on all the single-tree point cloud data, that is, all the point cloud data are not traversed, so that the processing efficiency of the algorithm is effectively improved. Meanwhile, traversing the selected single tree point cloud data in a reverse order according to the elevation values, and dividing the single tree point cloud data at infinity corresponding to the currently traversed target tree point cloud data into non-target tree data sets, so that two data are processed at one time, and the processing efficiency of the algorithm is further improved.
In a specific embodiment, the single-tree splitting apparatus further includes a detection module 25, which detects the target tree data set, specifically, determines whether target tree point cloud data exists in the target tree data set, and whether a difference between a tree top elevation value of any single-tree point cloud data in the target tree data set and a tree top elevation value of the target tree point cloud data is less than 1.5 m; if the point cloud data exists, the single-tree point cloud data is classified into the target tree point cloud data set, and the detection operation is finished; if not, continuing to execute the detection operation; judging whether the number of elements of the target tree point cloud data set is less than 5; if yes, marking as an error tree and deleting; if not, marking as the target tree.
It can be understood that each time the division is completed, there is a case that there is some error, and therefore, it is necessary to detect each divided tree and determine whether it is a correct tree.
In a specific embodiment, the single tree splitting apparatus further includes a fitting module 26, which performs single tree modeling fitting on the target tree data sets, specifically, performs least square linear fitting and standard gaussian function fitting on all the target tree data sets to obtain fitting values of each target tree data set; and marking the target tree data set with the fitting value less than 0.05 as an abnormal tree and deleting the abnormal tree.
It can be understood that after the above steps are completed, some trees with a part of wrong segmentation due to too long outer branches of the trees or noisy points existing in the data center exist, but the point cloud structure of the trees deviates from the normal tree characteristics, so that single-tree modeling fitting needs to be performed on all the trees.
In a specific embodiment, the single-tree segmentation apparatus further includes a modification module 27, which performs tree-form modification on the target tree, specifically, by traversing the target tree dataset, the single-tree point cloud data in the non-target tree dataset, whose characteristic line distance from the target tree is less than 0.1m, is assigned to the target tree dataset.
It can be understood that, in the segmentation result, since the algorithm performs segmentation from high to low, some point cloud data that originally belongs to the tree is segmented first by the tree higher next to the tree, so that the shape of the tree is missing, and therefore, the tree shape correction is required.
Traversing each tree, calculating the distance of non-tree point clouds, and classifying the point clouds with the distance less than 0.1m from the characteristic line of the tree as the tree point clouds, wherein the formula is shown as formula 5:
the embodiment of the invention has the following beneficial effects:
by utilizing the single-wood model constructed by the embodiment of the invention, all real point clouds of the single-wood sample are input into the single-wood model to obtain the reference form of the tree species, and the parameter threshold value required by the algorithm is extracted from the form to achieve de-empirical thresholding.
The invention also provides a computer readable storage medium, which includes a stored computer program, wherein when the computer program runs, the apparatus on which the computer readable storage medium is located is controlled to execute the method for modeling single trees as described above.
While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention.
It will be understood by those skilled in the art that all or part of the processes of the above embodiments may be implemented by hardware related to instructions of a computer program, and the computer program may be stored in a computer readable storage medium, and when executed, may include the processes of the above embodiments. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
Claims (9)
1. A method of modeling a single tree, comprising:
performing single tree clipping on forest point cloud data of a target forest area, and extracting partial single tree point cloud data as a sample data set;
training the sample data set by utilizing a pre-designed 2.5D Manhattan distance model to obtain a target forest region model; training the sample data set by utilizing a pre-designed 2.5D-based Manhattan distance model to obtain a target forest region model, specifically, calculating the Manhattan distance between the single-tree point cloud data with the maximum elevation value in the sample data set and the rest single-tree point cloud data to obtain the Manhattan distance model; performing convex hull calculation by using the Manhattan distance model to obtain a convex hull point set; wherein the convex hull point set is a point set sorted from the single-wood point cloud data of the maximum elevation value in a reverse-time-oriented manner; sequentially traversing the convex hull points until the Manhattan distance of the traversed convex hull points is the maximum value; if the Manhattan distance of the traversed convex hull point is not the maximum value, adding the convex hull point to an upper convex hull point set; performing least square fitting on the upper convex hull point set to obtain k and b characteristic values of a linear model of the target tree; obtaining the outmost Manhattan distance, the average outmost point distance and the maximum outmost distance of the single-wood point cloud data by utilizing the linear model;
respectively inputting all the single-tree point cloud data into the target forest region model to obtain a segmentation threshold value of the target forest region;
and extracting the target tree data sets one by one according to the segmentation threshold.
2. The single-tree modeling method according to claim 1, wherein said single-tree clipping forest point cloud data of the target forest area and extracting part of the single-tree point cloud data as a sample data set, further comprises:
dividing the forest point cloud data into ground points and vegetation point clouds;
interpolating the ground points and generating a digital elevation model;
performing normalization processing on the vegetation point cloud by using the digital elevation model;
wherein the normalized value of the vegetation point cloud is an elevation value; and when the point corresponding to the vegetation point cloud is a tree top point, the tree top elevation value is obtained.
3. The singletree modeling method of claim 2, wherein said inputting all of said singletree point cloud data into said target forest region model, respectively,
arranging all the single-wood point cloud data in a reverse order according to the height values of the single-wood point cloud data to obtain a reverse order data set;
acquiring the single-tree point cloud data of the maximum elevation value in the reverse-order data set;
transferring the single-tree point cloud data with the maximum elevation in the inverted data set and all the single-tree point cloud data in the maximum crown radius of the single-tree point cloud data from the inverted data set to the target tree data set according to the maximum crown radius parameter of the single-tree point cloud data;
and circularly executing until all the single-tree point cloud data in the reverse data set are traversed.
4. The singletree modeling method of claim 2, wherein said extracting said target tree datasets one by one, in particular,
calculating the distance between all the single-tree point cloud data and the single-tree point cloud data with the maximum elevation difference;
taking all the single-wood point cloud data in the 2-time maximum peripheral range as a data set to be processed;
traversing the single-tree point cloud data in the data set to be processed in a reverse order according to the elevation values; the single-tree point cloud data with the maximum elevation value in the data set to be processed are divided into a target tree data set, and the single-tree point cloud data at infinity are divided into a non-target tree data set.
5. The method of singletree modeling according to claim 1, further comprising detecting said target tree dataset, in particular,
judging whether target tree point cloud data exist in the target tree data set or not, and whether the difference value of the tree top elevation value of any single tree point cloud data in the target tree data set and the target tree point cloud data is smaller than 1.5m or not;
if the point cloud data exists, the single-tree point cloud data is classified into the target tree point cloud data set, and the detection operation is finished; if not, continuing to execute the detection operation;
judging whether the number of elements of the target tree point cloud data set is less than 5, if so, marking as an error tree and deleting; if not, marking as the target tree.
6. The method of singletree modeling according to claim 5, further comprising performing a singletree modeling fit to said target tree dataset, in particular,
performing least square linear fitting and standard Gaussian function fitting on all the target tree data sets to obtain a fitting value of each target tree data set;
and marking the target tree data set with the fitting value less than 0.05 as an abnormal tree and deleting the abnormal tree.
7. The method of singletree modeling according to claim 5, further comprising tree-state modification of said target tree, in particular,
and through traversing the target tree data set, attributing the single-tree point cloud data with the distance less than 0.1m from the characteristic line of the target tree in the non-target tree data set to the target tree data set.
8. A single-wood modeling apparatus, comprising:
the processing module is used for performing single-tree clipping on the forest point cloud data of the target forest area and extracting part of the single-tree point cloud data as a sample data set;
the modeling module is used for training the sample data set by utilizing a pre-designed 2.5D Manhattan distance model to obtain a target forest region model; training the sample data set by utilizing a pre-designed 2.5D-based Manhattan distance model to obtain a target forest region model, specifically, calculating the Manhattan distance between the single-tree point cloud data with the maximum elevation value in the sample data set and the rest single-tree point cloud data to obtain the Manhattan distance model; performing convex hull calculation by using the Manhattan distance model to obtain a convex hull point set; wherein the convex hull point set is a point set sorted from the single-wood point cloud data of the maximum elevation value in a reverse-time-oriented manner; sequentially traversing the convex hull points until the Manhattan distance of the traversed convex hull points is the maximum value; if the Manhattan distance of the traversed convex hull point is not the maximum value, adding the convex hull point to an upper convex hull point set; performing least square fitting on the upper convex hull point set to obtain k and b characteristic values of a linear model of the target tree; obtaining the Manhattan distance of the outermost layer, the average outermost layer point distance and the maximum outermost layer distance of the single-wood point cloud data by utilizing the linear model;
the calculation module is used for respectively inputting all the single-tree point cloud data into the target forest region model to obtain a segmentation threshold value of the target forest region;
and the extraction module is used for extracting the target tree data sets one by one according to the segmentation threshold.
9. A computer-readable storage medium, comprising a stored computer program, wherein the computer program, when executed, controls an apparatus in which the computer-readable storage medium is located to perform a method of modeling singletrees according to any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910625385.3A CN110428438B (en) | 2019-07-11 | 2019-07-11 | Single-tree modeling method and device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910625385.3A CN110428438B (en) | 2019-07-11 | 2019-07-11 | Single-tree modeling method and device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110428438A CN110428438A (en) | 2019-11-08 |
| CN110428438B true CN110428438B (en) | 2022-09-02 |
Family
ID=68409232
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910625385.3A Active CN110428438B (en) | 2019-07-11 | 2019-07-11 | Single-tree modeling method and device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110428438B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111553987B (en) * | 2020-05-19 | 2023-11-21 | 北京数字绿土科技股份有限公司 | Shan Mudian cloud editing method and system based on mass data |
| CN111898688B (en) * | 2020-08-04 | 2023-12-05 | 沈阳建筑大学 | Airborne LiDAR data tree classification method based on three-dimensional deep learning |
| CN117131061B (en) * | 2023-08-17 | 2024-05-17 | 广州市城市规划勘测设计研究院有限公司 | Tree warehouse-in information processing method, device, terminal and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107274417A (en) * | 2017-07-05 | 2017-10-20 | 电子科技大学 | A kind of single wooden dividing method based on airborne laser point cloud aggregation |
| CN107390681A (en) * | 2017-06-21 | 2017-11-24 | 华南理工大学 | A kind of mobile robot real-time location method based on laser radar and map match |
| CN108198190A (en) * | 2017-12-28 | 2018-06-22 | 北京数字绿土科技有限公司 | A kind of single wooden dividing method and device based on point cloud data |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8577611B2 (en) * | 2010-03-30 | 2013-11-05 | Weyerhaeuser Nr Company | System and method for analyzing trees in LiDAR data using views |
-
2019
- 2019-07-11 CN CN201910625385.3A patent/CN110428438B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107390681A (en) * | 2017-06-21 | 2017-11-24 | 华南理工大学 | A kind of mobile robot real-time location method based on laser radar and map match |
| CN107274417A (en) * | 2017-07-05 | 2017-10-20 | 电子科技大学 | A kind of single wooden dividing method based on airborne laser point cloud aggregation |
| CN108198190A (en) * | 2017-12-28 | 2018-06-22 | 北京数字绿土科技有限公司 | A kind of single wooden dividing method and device based on point cloud data |
Non-Patent Citations (1)
| Title |
|---|
| 机载激光雷达人工林单木分割方法比较和精度分析;李平昊等;《林业科学》;20181231;第54卷(第12期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110428438A (en) | 2019-11-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10896349B2 (en) | Text detection method and apparatus, and storage medium | |
| CN110533084B (en) | Multi-scale target detection method based on self-attention mechanism | |
| US10679146B2 (en) | Touch classification | |
| CN110428438B (en) | Single-tree modeling method and device and storage medium | |
| CN102982534B (en) | Canny edge detection dual threshold acquiring method based on chord line tangent method | |
| JP6487576B2 (en) | Fingerprint ridge point recognition method and apparatus | |
| CN111091095B (en) | Method for detecting ship target in remote sensing image | |
| CN110619333B (en) | Text line segmentation method, text line segmentation device and electronic equipment | |
| JPWO2019026104A1 (en) | Information processing apparatus, information processing program, and information processing method | |
| CN110570442A (en) | Contour detection method under complex background, terminal device and storage medium | |
| CN111768392A (en) | Target detection method and device, electronic equipment and storage medium | |
| CN113269791A (en) | Point cloud segmentation method based on edge judgment and region growth | |
| CN113240623A (en) | Pavement disease detection method and device | |
| CN114155200A (en) | Remote sensing image change detection method based on convolutional neural network | |
| CN110598581B (en) | Optical music score recognition method based on convolutional neural network | |
| CN114898118A (en) | Automatic statistical method and system for power transmission line house removal amount based on multi-source point cloud | |
| CN114092814A (en) | Unmanned plane navel orange tree image target identification and statistics method based on deep learning | |
| CN112580624B (en) | Multi-directional text region detection method and device based on boundary prediction | |
| CN108428234B (en) | Interactive segmentation performance optimization method based on image segmentation result evaluation | |
| CN113159103A (en) | Image matching method, image matching device, electronic equipment and storage medium | |
| CN113989823B (en) | Image table restoration method and system based on OCR coordinates | |
| CN114399780A (en) | Table detection method, table detection model training method and device | |
| CN112418313B (en) | Big data online noise filtering system and method | |
| CN111860625B (en) | Ancient wood building bucket arch automatic category identification method and system | |
| CN111698700B (en) | Method and device for judging working state of cell |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |