CN113269825A - Forest breast diameter value extraction method based on foundation laser radar technology - Google Patents
Forest breast diameter value extraction method based on foundation laser radar technology Download PDFInfo
- Publication number
- CN113269825A CN113269825A CN202110368040.1A CN202110368040A CN113269825A CN 113269825 A CN113269825 A CN 113269825A CN 202110368040 A CN202110368040 A CN 202110368040A CN 113269825 A CN113269825 A CN 113269825A
- Authority
- CN
- China
- Prior art keywords
- breast diameter
- points
- forest
- point
- ground
- 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.)
- Granted
Links
- 210000000481 breast Anatomy 0.000 title claims abstract description 269
- 238000000605 extraction Methods 0.000 title claims abstract description 85
- 238000005516 engineering process Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 claims abstract description 85
- 238000003064 k means clustering Methods 0.000 claims abstract description 50
- 238000012545 processing Methods 0.000 claims abstract description 30
- 239000002023 wood Substances 0.000 claims abstract description 29
- 238000005520 cutting process Methods 0.000 claims abstract description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 35
- 230000006870 function Effects 0.000 claims description 26
- 230000008859 change Effects 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 7
- 230000005484 gravity Effects 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims 1
- 238000011835 investigation Methods 0.000 abstract description 10
- 238000007726 management method Methods 0.000 abstract description 7
- 238000004519 manufacturing process Methods 0.000 abstract description 6
- 210000000038 chest Anatomy 0.000 description 21
- 238000005259 measurement Methods 0.000 description 14
- 230000011218 segmentation Effects 0.000 description 14
- 238000009826 distribution Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 7
- 238000011160 research Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 238000010606 normalization Methods 0.000 description 5
- 238000004088 simulation Methods 0.000 description 5
- 230000007547 defect Effects 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 3
- 230000001172 regenerating effect Effects 0.000 description 3
- 241000218657 Picea Species 0.000 description 2
- 235000011611 Pinus yunnanensis Nutrition 0.000 description 2
- 241000018652 Pinus yunnanensis Species 0.000 description 2
- 241001138418 Sequoia sempervirens Species 0.000 description 2
- 241001116500 Taxus Species 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 210000003127 knee Anatomy 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 241001290610 Abildgaardia Species 0.000 description 1
- 235000009051 Ambrosia paniculata var. peruviana Nutrition 0.000 description 1
- 235000003097 Artemisia absinthium Nutrition 0.000 description 1
- 240000001851 Artemisia dracunculus Species 0.000 description 1
- 235000017731 Artemisia dracunculus ssp. dracunculus Nutrition 0.000 description 1
- 235000003261 Artemisia vulgaris Nutrition 0.000 description 1
- 235000004758 Bergkiefer Nutrition 0.000 description 1
- 239000002028 Biomass Substances 0.000 description 1
- 241000722731 Carex Species 0.000 description 1
- 244000025670 Eleusine indica Species 0.000 description 1
- 235000014716 Eleusine indica Nutrition 0.000 description 1
- 241000744304 Elymus Species 0.000 description 1
- 241000451538 Elymus kamoji Species 0.000 description 1
- 235000014820 Galium aparine Nutrition 0.000 description 1
- 241000234435 Lilium Species 0.000 description 1
- 241001656752 Lilium philadelphicum Species 0.000 description 1
- 241001668545 Pascopyrum Species 0.000 description 1
- 235000010450 Pino mugo Nutrition 0.000 description 1
- 235000005205 Pinus Nutrition 0.000 description 1
- 241000218602 Pinus <genus> Species 0.000 description 1
- 241001136577 Pinus mugo Species 0.000 description 1
- 235000002914 Pinus uncinata Nutrition 0.000 description 1
- 229910000831 Steel Inorganic materials 0.000 description 1
- 244000152045 Themeda triandra Species 0.000 description 1
- 241001585668 Trisetum bifidum Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 239000001138 artemisia absinthium Substances 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 235000003499 redwood Nutrition 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- KDYFGRWQOYBRFD-UHFFFAOYSA-N succinic acid Chemical compound OC(=O)CCC(O)=O KDYFGRWQOYBRFD-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
- G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention is suitable for the technical field of forestry engineering, and provides a method for extracting a breast diameter value of a forest based on a ground-based laser radar technology, wherein the method for extracting the breast diameter value of the forest based on the ground-based laser radar technology comprises the following steps: acquiring ground radar forest ground point cloud data; performing point cloud elevation processing on the ground radar forest land point cloud data, and extracting breast diameter slice points; cutting the breast diameter slicing points according to improved k-means clustering, and extracting single-tree breast diameter slicing points; and carrying out ellipse fitting processing on the single-wood breast diameter slicing points to obtain a breast diameter value extraction result. The method can rapidly realize batch extraction of the breast diameter points of the trees through the improved k-means clustering, does not need prior knowledge of the number of the trees, the sample plot size and the like, has the advantages of high automation degree and strong noise resistance, and has certain practical application reference value for researching the foundation laser radar in forestry resource investigation and production management application.
Description
Technical Field
The invention belongs to the technical field of forestry engineering, and particularly relates to a method for extracting a forest breast height diameter value based on a foundation laser radar technology.
Background
The diameter at breast height is the diameter of a trunk section at a distance of 1.3m from a rhizome, is one of important parameters for evaluating the growth state of the forest, and has important significance for researching forest biomass estimation, forest resource monitoring, dynamic management and the like by accurate extraction. The traditional breast diameter measuring method mainly adopts a breast diameter ruler or a diameter checking ruler to manually measure in a field contact manner, the method is simple to operate, but has the defects of long operation time, high labor intensity and the like, and can not meet the requirements of information and intelligent development of modern forestry. Under the dual drive of sensors and national requirements, a high-resolution fine remote sensing technology is rapidly developed, and forest parameter inversion methods based on multiple remote sensing technologies such as unmanned aerial vehicle photogrammetry, airborne laser radar and ground-based laser radar are developed, wherein the ground laser radar technology adopts a non-contact high-speed laser measurement method, three-dimensional space and reflection intensity information-point cloud data with high precision and high density under the forest are directly obtained on the premise of not damaging trees, the method has the advantages that other remote sensing technologies cannot be compared, a brand new data acquisition means is provided for forest resource investigation and forest research, and the method becomes a hotspot for rapidly extracting and researching forest parameters.
At present, a large amount of research is carried out by scholars at home and abroad aiming at the extraction of the breast height diameter of a forest by using point cloud data of a foundation laser radar, and the method can be summarized into 3 steps: extracting breast diameter slices. Namely, point clouds with a certain thickness at 1.3m of the forest are separated from a huge sample plot point cloud data set. And secondly, segmenting the point cloud of the slices. The single-tree breast diameter slice point cloud is the minimum unit for calculating the breast diameter value, the point cloud extracted in the step I contains the coexistence phenomenon of a plurality of single-tree breast diameter slices, and the slice point cloud segmentation is to gather the same tree breast diameter points into a set, separate the different tree breast diameter points and prepare for the subsequent breast diameter value calculation. Estimating the chest diameter value. And (3) taking a circle or an ellipse as a breast diameter section model, and solving model parameters (radius R/long and short half shafts/a and b) to obtain a breast diameter value estimation result. Compared with the step I, the research on the segmentation of the point cloud of the slice is relatively weak, and the method mainly has the following two defects: firstly, human intervention exists, and the automation degree needs to be improved. The prior knowledge of the forest tree number and other sample plot investigation is usually needed, and the automation degree of the chest diameter extraction process is reduced. And high extraction precision of the breast diameter point of the single wood. Noise points also exist in the single-tree breast diameter slice point cloud after the segmentation processing, the noise points are very close to the breast diameter point, the noise points are difficult to remove by a conventional method, and the noise points are used as effective sample points to participate in breast diameter value estimation, so that the breast diameter extraction result is seriously influenced.
Therefore, the problems of low breast diameter slice segmentation precision and low automation degree exist in the conventional forest breast diameter extraction method.
Disclosure of Invention
The embodiment of the invention aims to provide a method for extracting a breast diameter value of a forest based on a foundation laser radar technology, and aims to solve the problems of low breast diameter slice segmentation precision and low automation degree of the conventional forest breast diameter extraction method.
The embodiment of the invention is realized in such a way that a method for extracting the breast diameter value of a forest based on the ground-based laser radar technology comprises the following steps:
acquiring ground radar forest ground point cloud data;
performing point cloud elevation processing on the ground radar forest land point cloud data, and extracting breast diameter slice points;
cutting the breast diameter slicing points according to improved k-means clustering, and extracting single-tree breast diameter slicing points;
and carrying out ellipse fitting processing on the single-wood breast diameter slicing points to obtain a breast diameter value extraction result.
Another objective of embodiments of the present invention is to provide a device for extracting a forest breast height value based on a ground-based laser radar technology, including:
the forest land point cloud data acquisition unit is used for acquiring forest land point cloud data of the ground radar;
the breast diameter slice point extraction unit is used for performing point cloud elevation processing on the foundation radar forest land point cloud data and extracting breast diameter slice points;
the single-tree breast diameter slice point extraction unit is used for cutting the breast diameter slice points according to the improved k-means clustering and extracting single-tree breast diameter slice points; and
and the chest diameter value extraction result obtaining unit is used for carrying out ellipse fitting processing on the single-wood chest diameter slicing points to obtain a chest diameter value extraction result.
It is a further object of an embodiment of the present invention to provide a computer device, comprising a memory and a processor, wherein the memory has stored therein a computer program, which, when executed by the processor, causes the processor to perform the steps of the method for ground-based lidar technology forest breast diameter value extraction.
Another object of an embodiment of the present invention is a computer-readable storage medium having stored thereon a computer program, which, when executed by a processor, causes the processor to perform the steps of the method for ground-based lidar technology-based forest breast diameter value extraction.
According to the method for extracting the breast diameter value of the forest based on the ground-based laser radar technology, point cloud elevation processing is carried out on ground-based radar forest land point cloud data, breast diameter slicing points are extracted, the breast diameter slicing points are cut according to improved k-means clustering, single-tree breast diameter slicing points are extracted, ellipse fitting processing is carried out on the single-tree breast diameter slicing points, and a breast diameter value extraction result is obtained. The method can rapidly realize batch extraction of the breast diameter points of the trees through the improved k-means clustering, does not need prior knowledge such as the number of the trees, the sample plot size and the like, has the advantages of high automation degree and strong noise resistance, and has certain practical application reference value for researching the foundation laser radar in forestry resource investigation and production management application.
Drawings
Fig. 1 is a flowchart of a method for extracting a breast diameter value of a forest based on a ground-based laser radar technology according to an embodiment of the present invention;
fig. 2 is a flowchart of another method for extracting a breast diameter value of a forest based on ground-based laser radar technology according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a k value determination simulation experiment by an "inflection point" method according to an embodiment of the present invention: (a) simulating experimental data; (b) different k values and target function relation curves;
fig. 4 is a flowchart of another method for extracting a breast diameter value of a forest based on ground-based laser radar technology according to an embodiment of the present invention;
fig. 5 is a flowchart of another method for extracting a breast diameter value of a forest based on the ground-based laser radar technology according to an embodiment of the present invention;
fig. 6 is a flowchart of a method for extracting a breast diameter value of a forest based on a ground-based laser radar technology according to an embodiment of the present invention;
fig. 7 is a schematic diagram of circle fitting results of different parameter solving methods provided by the embodiment of the present invention: (a) a least squares method; (b) a consistency algorithm is randomly adopted; (c) an LS-RAN method;
fig. 8 is a block diagram of a device for extracting a breast height value of a forest based on the ground-based laser radar technology according to an embodiment of the present invention;
fig. 9 is a schematic diagram of a distribution of measurement sites and a sample point cloud data according to an embodiment of the present invention: (a) sample station distribution (triangle identification); (b) a sample point cloud side view;
fig. 10 is a schematic diagram of sample-scene and actual measurement data acquisition according to an embodiment of the present invention: (a) measuring the breast diameter; (b) numbering sample plots and forest trees;
fig. 11 is a schematic diagram of segmentation of an improved mean clustering single-tree breast diameter slice according to an embodiment of the present invention: (a) extracting a result of the breast diameter section points; (b) slicing point segmentation results;
FIG. 12 is a schematic diagram of an error source of the number 9 single-wood extraction error measurement provided by the embodiment of the present invention;
fig. 13 is a schematic diagram of linear regression analysis of the extracted diameter at breast height and measured values according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the examples of the invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
It should be understood that although the terms first, second, etc. may be used to describe various information in embodiments of the present invention, the information should not be limited by these terms. These terms are only used to distinguish one type of information from another.
The trees are influenced by competition relations such as competition for space resources, and the growth state of the trees has a self-dredging phenomenon. The breast diameter section of the forest tree shows typical spatial characteristics in a point cloud space: the single-tree breast diameter slice point clouds are closely connected into a similar circle and are gathered into a cluster, and different single-tree breast diameter slice point clouds are separated from each other and keep a certain distance. Therefore, a k-means clustering algorithm with high efficiency and high convergence speed can be introduced to realize the point cloud segmentation of the single-tree breast diameter section. Firstly, intercepting breast diameter slicing points according to normalized elevation point cloud data, then segmenting the breast diameter slicing points through improved k-means clustering, extracting single-tree breast diameter slicing points, and finally fitting circles to the discrete single-tree breast diameter slicing points obtained through segmentation and calculating single-tree breast diameter values.
The breast diameter is one of important parameters for representing the quality of a standing place and the growth state of forest trees. The embodiment of the invention provides a method for extracting a forest breast diameter value based on a ground-based laser radar technology, which improves k-means clustering and aims to solve the problems of insufficient breast diameter extraction precision, low automation degree and the like in the prior art. Firstly, optimizing a selection principle of an initial seed point by using a constraint condition, and avoiding that a clustering result falls into local optimum due to random seed point selection; secondly, adaptively determining the number of the clustering target categories by adopting an inflection point method, and improving the automation degree of single-tree breast diameter point segmentation; and finally, calculating a clustering center according to the inverse distance weighting, eliminating a non-target single wood point set, and solving through circular model parameters to realize chest diameter value calculation. The method can rapidly realize batch extraction of the breast diameter points of the trees through the improved k-means clustering, does not need prior knowledge of the number of the trees, the sample plot size and the like, has the advantages of high automation degree and strong noise resistance, and has certain practical application reference value for researching the foundation laser radar in forestry resource investigation and production management application.
As shown in fig. 1, in an embodiment, a method for extracting a breast diameter value of a forest based on ground-based laser radar technology is provided, which may specifically include the following steps:
and S101, acquiring ground radar forest land point cloud data.
And S102, performing point cloud elevation processing on the foundation radar forest land point cloud data, and extracting breast diameter slice points.
In the embodiment of the invention, the trees are communicated into a whole through the ground, and are influenced by the fluctuation of the ground surface, the elevation points at the roots and stems of different trees are different, namely the elevation calculation surfaces of the breast diameter section point cloud are different, so that for the convenience of automatic batch extraction of the breast diameter section point cloud, the elevation of the roots and stems is converted into the uniform elevation surface by adopting grid elevation normalization processing.
In the embodiment of the present invention, as shown in fig. 2, the step S102 includes:
step S201, dividing the forest land into square grids with the side length being a fixed value, taking the lowest point in the index grid as a ground point, and determining the difference value between the elevation of all points in the square grids and the elevation of the ground point.
In the embodiment of the invention, the sample plot is divided into a square grid with the side length being a fixed value, the lowest point in the index grid is taken as a ground point, and the elevations of all points in the grid are differentiated from the elevation of the ground point.
And S202, extracting breast diameter slice points according to any point in the ground radar forest land point cloud data, the difference value of all point elevations in the square grid and the ground point elevation and the thickness of the point cloud slice at the breast diameter.
In the embodiment of the invention, the diameter of breast section point set S1.3Can be described as a point cloud in a certain thickness range at 1.3m in elevation as the formula (1) As shown, Pi is any point in the sample point cloud data M, epsilon is the slice thickness of the point cloud at the breast diameter, and the width of the breast diameter ruler is usually 3.0-5.0 cm.
S1.3＝{pi＝(xi,yi,zi)∈M|zi-1.3|≤ε} (1)
And S103, cutting the breast diameter slice points according to the improved k-means clustering, and extracting single-tree breast diameter slice points.
In the embodiment of the invention, k-means clustering (k-means clustering algorithm) is a clustering analysis method for iterative solution, has the advantages of high efficiency, convenient operation and the like, and is widely applied to the fields of market investigation, data mining and the like. The basic idea is as follows: for an unmarked data set, on the premise of giving a category number k, an initial seed point, namely an initial target cluster center, is randomly selected, the object category attribution is determined by taking the Euclidean distance of the object-cluster center as a measurement index, the average value of all objects in the cluster is taken as the center of a new target cluster, iteration is continuously carried out until the variation range of the cluster center tends to be stable or reaches the expected iteration times, and finally k target cluster subsets are generated. The goal of K-means clustering is to make the inter-cluster sample distance as small as possible and the inter-sample distance as large as possible, i.e., the objective function of K-means clustering algorithm can be expressed as the sum of inter-cluster sample distances as shown in equation (2). Wherein k is the number of target cluster, n is the total number of sample points, xiFor samples belonging to the jth target class cluster, ujRepresenting the center of the jth target class cluster. When the target function J takes the minimum value, the algorithm converges at this time, and the best clustering result exists.
The original k-means clustering algorithm needs the target category number k as prior input, reduces the automation degree of the algorithm to a certain extent, is influenced by randomly selecting initial seed points, and is easy to fall into local optimal clustering results which generate errors. By combining the spatial distribution characteristics of the point clouds of the tree breast diameter slices, the invention eliminates the defect of k-means clustering in single tree breast diameter extraction through the following three improvements, so as to improve the precision and the automation degree of the tree breast diameter extraction.
(1) And optimizing the initial seed point selection principle. According to the characteristic that single tree breast diameter slice point clouds are mutually isolated and keep a certain distance and the distance is far larger than a breast diameter value, selecting initial seed points by taking the minimum value of the forest spacing as a constraint condition, namely if the minimum value of the pairwise distance of the initial seed points is smaller than the minimum value of the forest spacing, continuing to the next step under the condition, and if not, regenerating the initial seed points. By optimizing the selection principle of the seed points, the initial seed points are ensured to be derived from different single-tree breast diameter slicing point cloud sets, and the algorithm clustering precision and the convergence speed are improved.
And (4) self-adapting k value determination by using an inflection point method. The inflection point method is to calculate the J value of the objective function under different target cluster numbers, and then determine the 'inflection point' by detecting the sudden change of the change rate, so that the k value corresponding to the sudden change point is the optimal target cluster number. The principle can be explained as that the number of samples in a target cluster set is gradually reduced along with the increase of the number of target clusters, the distance between the samples and the center of the cluster is closer, and the J value of the target function is reduced through the calculation of a formula (2). When the target cluster number is larger than the optimal cluster number k, the cluster subset is formed by converting the combination of a plurality of clusters into the splitting of a single cluster, and the target function J is suddenly reduced and then kept to be slow, so that the inflection point of the target function is the optimal cluster number. A group of test data is generated through simulation to verify the feasibility and the effectiveness of the 'inflection point' method, the test data is formed by overlapping three two-dimensional normal distribution discrete points and is shown in figure 3a, each normal distribution point set represents one target class cluster, and a corresponding relation curve of a target function J and different target class numbers K is shown in figure 3 b. The optimal k value was determined to be 3 by the knee point method, which is consistent with the preset classification of the simulation data.
(3) And automatically identifying the non-target. Because the growth rates of the forest vegetation are different, the breast diameter slice point cloud obtained by elevation normalization cutting extraction also contains shrubs and non-target single trees with the breast diameter less than 5.0cm, and the target geometry after K-means clustering can be improvedAnd (3) identifying the distance distribution between the sample and the class center: and determining the cluster center position by adopting an inverse distance weighting method to weaken the influence of noise points, simultaneously calculating the average distance Vd and the variance Sd between the sample and the center point in the cluster, and if Vd is less than 2.5cm or Sd is greater than a variance threshold value T, marking the cluster as a non-target cluster. Equation (3) is an inverse distance weighting equation, where CiAs cluster center of class i, pjFor the j-th object point in the i-th class, Sj，Smax，SminRespectively representing the distance from the jth point to the last iteration clustering center, the maximum distance and the minimum distance.
The improved k-means clustering algorithm can adaptively determine a target clustering number k according to the characteristics of a sample data set, and simultaneously, the selection of an initial seed point is more reasonable by taking the forest spacing as a constraint condition, so that the iterative computation is prevented from getting into local optimality, and the improved k-means clustering breast diameter slice point cloud segmentation step is shown in fig. 4, namely the step S103 comprises:
inputting: point cloud of breast diameter slice, variance threshold T
Step S401, calculating objective function values when different k values are taken according to objective functions of a k-means clustering algorithm, and determining inflection points according to the change rate of the objective functions to determine the k value of the optimal objective category number.
In the embodiment of the invention, the objective function value is calculated according to the formula (2) when different k values are taken, the inflection point is searched according to the change rate of the objective function, and the k value of the optimal objective category number is determined in a self-adaptive manner.
Step S402, a group of initial seed points is randomly selected and the minimum value of the distance between the seed points is determined.
Step S403, judging whether the minimum value of the seed point spacing is smaller than the minimum value of the forest spacing; if not, returning to the step S402; if yes, the process proceeds to step S404.
In the embodiment of the invention, a group of initial seed points are randomly selected and the minimum value d of the distance between the seed points is determined, if the d is smaller than the minimum value of the distance between the forest trees, the step S404 is entered, otherwise, the seed points are regenerated again, namely, the step S returns to the step of randomly selecting a group of initial seed points and determining the minimum value of the distance between the seed points.
And S404, classifying all sample points according to the initial seed points by using a distance minimization classification principle.
Step S405, calculating the gravity center of the sample points in the cluster according to an inverse distance weighting formula, updating the gravity center into a new seed point set, and judging whether the seed point set changes or not; if yes, go back to step S404; if not, the process proceeds to step S406.
In the embodiment of the invention, the gravity centers of the sample points in the class cluster are calculated according to the formula (3) and are updated to be new seed points, if the set of the seed points is not changed any more, the step S406 is executed, otherwise, the step S is executed by returning to the step S for classifying all the sample points according to the initial seed points and the distance minimization as the classification principle.
Step S406, respectively calculating the average distance and variance of the distance between the sample point and the class center in the cluster-like geometry, so as to determine the single breast diameter slicing point according to the average distance and variance.
In the embodiment of the invention, the average distance Vd and the variance Sd of the distance between the sample point and the class center in the class cluster geometry are respectively calculated, if Vd is less than 2.5cm or Sd is more than T, the target class cluster is marked as a non-target set, otherwise, the target class cluster is marked as a single-tree breast diameter slice point set.
And (3) outputting: single-wood breast diameter section point set and non-target cluster point set
In the embodiment of the invention, the automation degree of k-means clustering is improved by determining the number of target clustering clusters through an inflection point method, the algorithm is prevented from falling into local optimization by adopting an optimized initial seed point selection principle, and a non-target single wood point set is removed by utilizing the estimated breast diameter value, so that the single wood breast diameter extraction precision is improved.
And S104, carrying out ellipse fitting processing on the single-wood breast diameter slicing points to obtain a breast diameter value extraction result.
In this embodiment of the present invention, as shown in fig. 5, the step S104 includes:
and S501, constructing a breast diameter section form model according to the breast diameter central point coordinates of the single-wood breast diameter slicing points and the breast diameter slicing radius.
In the embodiment of the invention, the shape of the cross section of the breast diameter can be approximately expressed by a two-dimensional plane circular equation, and the mathematical expression of the shape is shown in formula (4). Wherein (a, b) represents the breast diameter center point coordinates and r represents the breast diameter slice radius. The chest diameter extraction value D is 2 r.
(x-a)2+(y-b)2＝r2 (4)
And S502, solving the model parameters of the breast diameter section form model by combining a random sampling consistency algorithm and a least square method to obtain a breast diameter value extraction result.
In the embodiment of the invention, in general, the model parameter solving method includes a Least Square method (LS) and a Random sample consensus (RANSAC), the former determines the optimal model parameters by using a model fitting error Square minimization principle of an integral sample as a constraint condition, and the method has high stability, but the model parameter solving precision is easily influenced by noise points. The method can well weaken the influence of noise points, but is influenced by a sample point selection mode, and the model fitting stability is poor. In order to improve the solving precision of the model parameters, the LS-RAN method which is the best model parameters is solved by combining a random sampling consistency algorithm and a least square principle: firstly, identifying outlier noise points by adopting a random sampling consistency algorithm, and finally solving model parameters of an effective sample set with the noise points removed by utilizing a least square method.
In this embodiment of the present invention, as shown in fig. 6, the step S502 includes:
step S601, identifying outlier noise points according to a random sampling consistency algorithm, and obtaining an effective sample set after the noise points are removed.
And step S602, solving the model parameters of the breast diameter section form model of the effective sample set with the noise points removed according to a least square method to obtain a breast diameter value extraction result.
In the embodiment of the present invention, fig. 7 is a test result, and the LS-RAN method adopted in the present invention well identifies that the noise points are as large as possible to retain the effective points, and the modeling points are well attached to the model curve and uniformly distributed around the model curve.
According to the method for extracting the breast diameter value of the forest based on the ground-based laser radar technology, point cloud elevation processing is carried out on ground-based radar forest land point cloud data, breast diameter slicing points are extracted, the breast diameter slicing points are cut according to improved k-means clustering, single-tree breast diameter slicing points are extracted, ellipse fitting processing is carried out on the single-tree breast diameter slicing points, and a breast diameter value extraction result is obtained. The method can rapidly realize batch extraction of the breast diameter points of the trees through the improved k-means clustering, does not need prior knowledge such as the number of the trees, the sample plot size and the like, has the advantages of high automation degree and strong noise resistance, and has certain practical application reference value for researching the foundation laser radar in forestry resource investigation and production management application.
As shown in fig. 8, in an embodiment, an apparatus for extracting a breast diameter value of a forest based on ground-based laser radar technology is provided, and may specifically include a forest land point cloud data obtaining unit 810, a breast diameter slicing point extracting unit 820, a single-tree breast diameter slicing point extracting unit 830, and a breast diameter value extraction result obtaining unit 840.
And a forest land point cloud data acquisition unit 810, configured to acquire the forest land point cloud data of the ground-based radar.
And a breast diameter slice point extraction unit 820, configured to perform point cloud elevation processing on the ground radar forest land point cloud data, and extract breast diameter slice points.
In the embodiment of the invention, the trees are communicated into a whole through the ground, and are influenced by the fluctuation of the ground surface, the elevation points at the roots and stems of different trees are different, namely the elevation calculation surfaces of the breast diameter section point cloud are different, for the convenience of batch automatic extraction of the breast diameter section point cloud, the elevation of the roots and stems is converted into the uniform elevation surface by adopting grid elevation normalization treatment, and the specific operation is as follows: dividing the sample plot into square grids with fixed side length, and indexing the lowest gridThe points are regarded as ground points, and the elevations of all the points in the grid are differentiated from the elevations of the ground points. Then the point set S of the breast diameter section1.3The point cloud which can be described as a certain thickness range at the elevation of 1.3M is shown in a formula (1), wherein Pi is any point in sample plot point cloud data M, epsilon is the slice thickness of the point cloud at the breast diameter, and the width of a breast diameter ruler is usually taken to be 3.0-5.0 cm.
S1.3＝{pi＝(xi,yi,zi)∈M|zi-1.3|≤ε} (1)
And the single-tree breast diameter slice point extracting unit 830 is configured to cut the breast diameter slice points according to the improved k-means clustering, and extract single-tree breast diameter slice points.
In the embodiment of the invention, k-means clustering (k-means clustering algorithm) is a clustering analysis method for iterative solution, has the advantages of high efficiency, convenient operation and the like, and is widely applied to the fields of market investigation, data mining and the like. The basic idea is as follows: for an unmarked data set, on the premise of giving a category number k, an initial seed point, namely an initial target cluster center, is randomly selected, the object category attribution is determined by taking the Euclidean distance of the object-cluster center as a measurement index, the average value of all objects in the cluster is taken as the center of a new target cluster, iteration is continuously carried out until the variation range of the cluster center tends to be stable or reaches the expected iteration times, and finally k target cluster subsets are generated. The goal of K-means clustering is to make the inter-cluster sample distance as small as possible and the inter-sample distance as large as possible, i.e., the objective function of K-means clustering algorithm can be expressed as the sum of inter-cluster sample distances as shown in equation (2). Wherein k is the number of target cluster, n is the total number of sample points, xiFor samples belonging to the jth target class cluster, uj represents the center of the jth target class cluster. When the target function J takes the minimum value, the algorithm converges at this time, and the best clustering result exists.
The original k-means clustering algorithm needs the target category number k as prior input, reduces the automation degree of the algorithm to a certain extent, is influenced by randomly selecting initial seed points, and is easy to fall into local optimal clustering results which generate errors. By combining the spatial distribution characteristics of the point clouds of the tree breast diameter slices, the invention eliminates the defect of k-means clustering in single tree breast diameter extraction through the following three improvements, so as to improve the precision and the automation degree of the tree breast diameter extraction.
(1) And optimizing the initial seed point selection principle. According to the characteristic that single tree breast diameter slice point clouds are mutually isolated and keep a certain distance and the distance is far larger than a breast diameter value, selecting initial seed points by taking the minimum value of the forest spacing as a constraint condition, namely if the minimum value of the pairwise distance of the initial seed points is smaller than the minimum value of the forest spacing, continuing to the next step under the condition, and if not, regenerating the initial seed points. By optimizing the selection principle of the seed points, the initial seed points are ensured to be derived from different single-tree breast diameter slicing point cloud sets, and the algorithm clustering precision and the convergence speed are improved.
And (4) self-adapting k value determination by using an inflection point method. The inflection point method is to calculate the J value of the objective function under different target cluster numbers, and then determine the 'inflection point' by detecting the sudden change of the change rate, so that the k value corresponding to the sudden change point is the optimal target cluster number. The principle can be explained as that the number of samples in a target cluster set is gradually reduced along with the increase of the number of target clusters, the distance between the samples and the center of the cluster is closer, and the J value of the target function is reduced through the calculation of a formula (2). When the target cluster number is larger than the optimal cluster number k, the cluster subset is formed by converting the combination of a plurality of clusters into the splitting of a single cluster, and the target function J is suddenly reduced and then kept to be slow, so that the inflection point of the target function is the optimal cluster number. A group of test data is generated through simulation to verify the feasibility and the effectiveness of the 'inflection point' method, the test data is formed by overlapping three two-dimensional normal distribution discrete points and is shown in figure 3a, each normal distribution point set represents one target class cluster, and a corresponding relation curve of a target function J and different target class numbers K is shown in figure 3 b. The optimal k value was determined to be 3 by the knee point method, which is consistent with the preset classification of the simulation data.
(3) Non-target automatic identificationOtherwise. Because the growth rates of the forest vegetation are different, the breast diameter slice point cloud obtained by elevation normalization cutting extraction also contains shrubs and non-target single trees with the breast diameter smaller than 5.0cm, and the distance distribution between the sample and the class center in the target geometry after the K-means clustering can be improved for identification: and determining the cluster center position by adopting an inverse distance weighting method to weaken the influence of noise points, simultaneously calculating the average distance Vd and the variance Sd between the sample and the center point in the cluster, and if Vd is less than 2.5cm or Sd is greater than a variance threshold value T, marking the cluster as a non-target cluster. Equation (3) is an inverse distance weighting equation, where CiAs cluster center of class i, pjIs the jth object point, Sj,Smax,SminRespectively representing the distance between the jth point and the last iteration clustering center, the maximum value and the minimum value of the distance.
The improved k-means clustering algorithm can adaptively determine a target clustering number k according to the characteristics of a sample data set, and simultaneously, the selection of an initial seed point is more reasonable by taking the forest spacing as a constraint condition, so that the iterative computation is prevented from falling into local optimization, and the improved k-means clustering breast diameter slice point cloud segmentation steps are as follows:
____________________________________________
inputting: point cloud of breast diameter slice, variance threshold T
Step 1: calculating an objective function value J when different k values are taken according to a formula (2), searching an inflection point according to the change rate of the objective function, and adaptively determining the k value of the optimal objective category number;
step 2: randomly selecting a group of initial seed points and determining a minimum value d of the distance between the seed points, if d is smaller than the minimum value of the distance between the forest trees, entering the next step, otherwise, regenerating the seed points;
step 3: classifying all sample points according to the initial seed points by using a distance minimization classification principle;
step 4: calculating the center of gravity C of the sample point in the cluster according to the formula (3)iAnd updating the seed points to be new seed points, if the seed point set does not change any more, entering the next Step, otherwise, returning to Step 3;
step 5: respectively calculating the average distance Vd and the variance Sd of the distance between the sample point and the class center in the class cluster geometry, if Vd is less than 2.5cm or Sd is more than T, marking the target class cluster as a non-target set, otherwise, marking the target class cluster as a single-tree breast diameter slice point set
And (3) outputting: single-wood breast diameter section point set and non-target cluster point set
___________________________________________
In the embodiment of the invention, the automation degree of k-means clustering is improved by determining the number of target clustering clusters through an inflection point method, the algorithm is prevented from falling into local optimization by adopting an optimized initial seed point selection principle, and a non-target single wood point set is removed by utilizing the estimated breast diameter value, so that the single wood breast diameter extraction precision is improved.
A chest diameter value extraction result obtaining unit 840, configured to perform ellipse fitting processing on the single-wood chest diameter slice points to obtain a chest diameter value extraction result.
In the embodiment of the invention, the shape of the cross section of the breast diameter can be approximately expressed by a two-dimensional plane circular equation, and the mathematical expression of the shape is shown in formula (4). Wherein (a, b) represents the breast diameter center point coordinates and r represents the breast diameter slice radius. The chest diameter extraction value D is 2 r.
(x-a)2+(y-b)2＝r2 (4)
Generally, a model parameter solving method includes a Least Square method (LS) and a Random sample consensus (RANSAC), the former determines optimal model parameters by using a model fitting error Square minimization principle of an integral sample as a constraint condition, and the method has high stability, but the model parameter solving accuracy is easily influenced by noise points. The method can well weaken the influence of noise points, but is influenced by a sample point selection mode, so that the fitting stability of the model is poor. In order to improve the solving precision of the model parameters, the LS-RAN method which is the best model parameters is solved by combining a random sampling consistency algorithm and a least square principle: firstly, identifying outlier noise points by adopting a random sampling consistency algorithm, and finally solving model parameters of an effective sample set with the noise points removed by utilizing a least square method. The LS-RAN method adopted by the invention well identifies that the noise points are as large as possible and the effective points can be reserved, and the modeling points are well attached to the model curve and are uniformly distributed around the model curve.
According to the forest breast diameter value extraction device provided by the embodiment of the invention, point cloud elevation processing is carried out on ground radar forest land point cloud data, breast diameter slicing points are extracted, the breast diameter slicing points are cut according to improved k-means clustering, single-tree breast diameter slicing points are extracted, ellipse fitting processing is carried out on the single-tree breast diameter slicing points, and a breast diameter value extraction result is obtained. The method can rapidly realize the batch extraction of the forest breast diameter points through the improved k-means clustering, does not need prior knowledge of forest number, sample plot size and the like, has the advantages of high automation degree and strong noise resistance, and has certain practical application reference value for researching application of the foundation laser radar in forestry resource investigation and production management.
Test examples:
1. study area overview and data acquisition
The research area is located in the northern grid and county of Shangri-La county of the prefecture of the Discelebration, Yunnan province, and the geographic range is 99 degrees 29 to 100 degrees 11 'of east longitude and 27 degrees 53 to 28 degrees 37' of northern latitude. The Hoxan belongs to a high and cold mountain area, the average altitude is 3200 meters, the maximum altitude is 5090 meters, the minimum altitude is 2520 meters, the terrain is mainly mountainous, and the gradient change range is large. The vertical change of the climate is particularly obvious, the average annual temperature is 6.6 ℃, and the annual precipitation is 600 mm. The main arbors in the research area are spruce (Picea asperta), redwood (sequoia), fir (fir), mountain pine (Pinus desata), yew (Taxus), Yunnan pine (Pinus yunnanensis) and the like, and the shrubs and vines are wheat grass (Elymus tandtorium), goose grass (Roegneria kamoji), common bluegrass (Trisetum bifidum), sedge (Carex), wormwood (armoise), wild lily (lily root) and the like. A Leica Scanstation P40 foundation laser radar is adopted to obtain the point cloud data of the natural near-maturity forest land in the research area, and the main performance parameters of the instrument are as follows: the angle measurement precision is 8'; the ranging precision is 1.2mm +10ppm, and the scanning speed is 100 ten thousand points/second. During scanning, the height of the instrument is about 1.60m, the scanning range is set to be 360 degrees in the horizontal direction and 270 degrees in the vertical direction, and the scanning mode is fine scanning. In order to obtain complete forest land point cloud data, 5 scanning measuring stations are uniformly distributed according to tree distribution and sample plot environment as shown in fig. 9a, meanwhile, 3 black-and-white targets are distributed on different elevation surfaces within the mutual visibility range of each measuring station, Leica scanning station P40 with Cyclone point cloud data processing commercial software is used as a data preprocessing operation platform, the method mainly comprises measuring station point cloud data registration, point cloud denoising, data cutting and the like, and forest land point cloud with the volume of 30m is selected as test data as shown in fig. 9 b. The point cloud data acquisition time is in 7 th of month in 2017, and the weather is clear and windless.
The measured data is mainly the breast diameter value of the forest in the same sample plot, the breast diameters of the forest which are larger than 5.0cm in the same sample plot are measured one by adopting a breast diameter ruler according to the provisions of GB/T26424 and 2010 forest resource planning design survey technical specification, the measurement mode is as shown in figure 10a, the average value of three times of measurement is taken as the final breast diameter value for improving the actual measurement precision of the breast diameter, meanwhile, the measured forest in the sample plot is numbered and the position of the measured forest is measured, and the number is as shown in figure 10b, so that the subsequent breast diameter extraction precision evaluation is facilitated. And the time synchronization of the actual measurement data acquisition and the point cloud data acquisition is kept.
2. Analysis of results
Taking measured data of the breast diameter value, the number of single trees and the like of the forest trees as standard reference, and adopting a single tree recognition rate P, a root mean square error MSE, an error maximum value M and a decision coefficient R2And comprehensively evaluating the improved k-means clustering single tree breast diameter extraction precision by using the indexes, wherein the evaluation indexes are defined as follows:
(1) the single wood recognition rate P (precision of single tree detection). The single wood number identification result comprises 2 types of errors: under-identified errors are false detection errors which are actually measured as single trees but are not identified correctly; if the over-recognition error is an over-detection error in which the actual measurement is not a single wood but is erroneously recognized as a single wood, the single wood recognition rate can be described by equation (5), where n1To under-identify the error, n2For over-recognition of errors, n total forest trees。
P＝1-(n1+n2)/n*100％ (5)
(2) Root mean square error rmse (root mean square error). The root mean square error describes the discrete degree of the error between the chest diameter value extraction result and the measured value, reflects the stability and the reliability of the chest diameter extraction method, and the smaller the value is, the better the applicability and the reliability of the algorithm is. The root mean square error is defined by the formula (6) wherein diRepresenting the ith single tree breast diameter extraction result; diThe measured value of the ith single tree breast diameter is shown, and n represents the total number of the trees.
(3) Relative error M (relative error). The error describes the deviation between the chest diameter value extraction result and the measured value, the relative error eliminates the influence of the dimension of the measured value on the result, the measurement precision is evaluated more objectively, the error is also a main index for evaluating the accuracy of the chest diameter in forestry investigation, the smaller the relative error is, the higher the precision is, the definition is shown in the formula (7), and the parameter meanings are the same as the above.
M＝|di-Di|/Di*100％ (7)
(4) Determining the coefficient R2(coeffient of determination). The decision coefficient represents the digital characteristic of the relation between one random variable and a plurality of random variables and is used for reflecting the statistical index of the correlation degree between the dependent variable and the independent variable, and the value of the statistical index is closer to 1, which indicates that the correlation is stronger. Can be defined as the ratio of the interpretable variation to the total variation, and is defined as shown in formula (8), wherein
The average value of the single-tree breast diameter extraction,
the measured mean value of the breast diameter of the single tree is obtained.
Fig. 11 shows the result of the extraction and segmentation of the breast diameter, wherein fig. 11a shows the result of the extraction of the point cloud of the breast diameter slice, and the green points represent the extracted breast diameter slice points. And (3) projecting the point cloud of the breast diameter slices obtained by elevation normalization interception in a two-dimensional plane, wherein all single-tree breast diameter slice points are gathered into a cluster, different single-tree breast diameter points are separated from each other and keep a certain distance, and the distance is far greater than the breast diameter value of the forest. Fig. 11b shows the result of point cloud segmentation of the breast diameter slices by the improved k-means clustering algorithm, wherein the black dots are non-target clusters obtained by identification, that is, the single-wood breast diameter slices with the breast diameter smaller than 5.0cm, and the other color dots are correctly extracted single-wood slices. Determining the optimal clustering number k value to be 33 by using an inflection point method, calculating by using an inverse distance weighted clustering center, and identifying 1 point of the non-target single-tree breast diameter type cluster, which is shown as a red circle in the figure, by combining a threshold, wherein the single-tree breast diameter is 4.2 cm. 32 effective target clusters are correctly extracted and obtained, the effective target clusters are consistent with measured data of the same sample, over-recognition and under-recognition phenomena are avoided, and the single wood recognition rate P is 100%. The analysis shows that the improved k-means clustering method can well identify non-target clusters and realize the automatic extraction of the breast diameter of the single tree.
Table 1 shows the breast diameter extraction and accuracy statistics. The maximum value of the absolute error of breast diameter extraction is 0.5cm, the maximum value of the relative error M is 2.00 percent, and the maximum value of the absolute error of single wood extraction with the breast diameter less than 20cm is 0.2 cm. According to the stipulation of national forest resource continuous checking technical regulation: the diameter of the breast height is equal to or more than 20cm, and the error of breast diameter measurement is less than 1.5%; the breast diameter of the tree with the breast diameter less than 20cm has the breast diameter measurement error less than 0.3cm, and the breast diameter extraction except the number 9 is within the specified range. In the aspect of the root mean square error precision evaluation index, the root mean square error RMSE is 0.23cm, which shows that the error fluctuates in a small range, and the breast diameter extraction method has good stability. The diameter at breast height of tree No. 9 was an abnormal value, and the absolute error M was 0.5 cm. The analysis reason is as follows: the damage of the tree body at the breast diameter of the No. 9 tree causes the tree body at the breast diameter to be convex, when the TS-RAN breast diameter section circular equation model is adopted for reconstruction, the convex part of the damage is taken as a noise point to be removed, so that the breast diameter extraction value is smaller than the measured value as shown in figure 12,
TABLE 1 Breast diameter extraction results and accuracy assessment
Note: in table 1, the measured chest diameter results are calculated from the measured circumferential length.
Fig. 13 shows the regression results of the extracted value of the chest diameter and the measured value. Taking the measured value of the chest diameter as an independent variable x, taking the extracted value of the chest diameter of the point cloud data as a dependent variable y, drawing a scatter diagram, carrying out linear regression analysis on the measured value and the extracted value, and calculating to obtain a unitary linear regression model equation as follows: y 0.97769x +0.37498, determining the coefficient R20.97355. The decision coefficient is very close to 1, which shows that the extracted result of the breast diameter based on the point cloud data has obvious linear correlation with the measured value, and shows that the improved k-means clustering point cloud forest breast diameter extracting method can replace a steel tape measuring method to extract forest breast diameter parameters.
In conclusion, the invention uses Leica ScanStation P40 ground laser radar to obtain the natural near-mature forest sample plot point cloud data in Shangri-Lagazan countryside, through the steps of extracting the slice point cloud of breast diameter, segmenting the slice point cloud, calculating the breast diameter value, etc., the high-precision extraction of the breast diameter of the forest is realized, the single-tree breast diameter slice point cloud cluster extraction method for improving k-means cluster is mainly discussed, and the main conclusion is as follows:
(1) the initial seed point selection principle of the k-means clustering algorithm is optimized by taking the forest spacing as a constraint condition, the phenomenon that the algorithm is trapped in local optimization due to improper selection of seed points is avoided, non-target breast diameter extraction of the forest with the breast diameter smaller than 5cm is automatically eliminated through an additional sample point and a cluster center average distance threshold, the single tree recognition rate is 100%, and the single tree extraction precision is improved to a certain extent.
(2) The inflection point method is adopted to adaptively determine the number of the forest trees in the sample plot, the prior input of k values in the traditional k-means clustering is overcome, the automation degree of single tree breast diameter extraction is improved, the influence of forest land canopy density is avoided, and a certain practical application reference value is provided for the rapid batch extraction of the breast diameters of the forest trees in a large range.
(3) By comparing with the measured value of chest diameter, the root mean square error RMSE of the chest diameter extraction result is 0.23cm, the maximum value of the relative error M is 2.00%, and the coefficient R is determined20.97355, indicating that: the method for solving the parameters of the elliptic model by combining the least square method and the random sampling consistency algorithm has higher precision of inverting the breast diameter of the forest, and can replace the traditional manual measurement method to extract the breast diameter of the forest in a certain range.
The breast diameter extraction method provided by the invention has the advantages of high automation degree, convenience in operation and the like, provides a new research idea and method for forest parameter inversion, and has important significance for the application of the foundation laser radar technology in forestry production, operation and management.
In one embodiment, a computer device is proposed, the computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
acquiring ground radar forest ground point cloud data;
performing point cloud elevation processing on the ground radar forest land point cloud data, and extracting breast diameter slice points;
cutting the breast diameter slicing points according to improved k-means clustering, and extracting single-tree breast diameter slicing points;
and carrying out ellipse fitting processing on the single-wood breast diameter slicing points to obtain a breast diameter value extraction result.
In one embodiment, a computer readable storage medium is provided, having a computer program stored thereon, which, when executed by a processor, causes the processor to perform the steps of:
acquiring ground radar forest ground point cloud data;
performing point cloud elevation processing on the ground radar forest land point cloud data, and extracting breast diameter slice points;
cutting the breast diameter slicing points according to improved k-means clustering, and extracting single-tree breast diameter slicing points;
and carrying out ellipse fitting processing on the single-wood breast diameter slicing points to obtain a breast diameter value extraction result.
It should be understood that, although the steps in the flowcharts of the embodiments of the present invention are shown in sequence as indicated by the arrows, the steps are not necessarily performed in sequence as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in various embodiments may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. A method for extracting a breast diameter value of a forest based on a foundation laser radar technology is characterized by comprising the following steps:
acquiring ground radar forest ground point cloud data;
performing point cloud elevation processing on the ground radar forest land point cloud data, and extracting breast diameter slice points;
cutting the breast diameter slicing points according to improved k-means clustering, and extracting single-tree breast diameter slicing points;
and carrying out ellipse fitting processing on the single-wood breast diameter slicing points to obtain a breast diameter value extraction result.
2. The method for extracting the breast diameter value of the forest based on the ground-based laser radar technology as claimed in claim 1, wherein the step of performing point cloud elevation processing on the ground-based radar forest ground point cloud data to extract breast diameter slicing points comprises the following steps of:
dividing a forest land into square grids with side lengths of fixed values, taking the lowest point in the index grid as a ground point, and determining the difference value between the elevation of all points in the square grids and the elevation of the ground point;
and extracting breast diameter section points according to any point in the ground radar forest land point cloud data, the difference value of all point elevations in the square grid and the ground point elevation and the thickness of the point cloud section at the breast diameter.
3. The method for extracting tree breast diameter values based on ground-based laser radar technology according to claim 1, wherein the step of performing cutting processing on the breast diameter slicing points according to improved k-means clustering to extract single tree breast diameter slicing points comprises:
calculating objective function values when different k values are taken according to an objective function of a k-means clustering algorithm, and determining inflection points according to the change rate of the objective function so as to determine the k value of the optimal objective category number;
randomly selecting a group of initial seed points and determining the minimum value of the distance between the seed points;
judging whether the minimum value of the seed point spacing is smaller than the minimum value of the forest spacing;
when the minimum value of the seed point spacing is not smaller than the minimum value of the forest spacing, returning to the step of randomly selecting a group of initial seed points and determining the minimum value of the seed point spacing;
when the minimum value of the seed point intervals is smaller than the minimum value of the forest tree intervals, classifying all sample points according to the initial seed points and by using the distance minimization as a classification principle;
calculating the gravity center of the sample point in the cluster according to an inverse distance weighting formula, updating the gravity center into a new seed point set, and judging whether the seed point set changes or not;
when the seed point set changes, returning to the step of classifying all sample points according to the initial seed points and by using distance minimization as a classification principle when the minimum value of the seed point intervals is smaller than the minimum value of the forest tree intervals;
and when the seed point set does not change, respectively calculating the average distance and the variance of the distances between the sample points and the class centers in the cluster-like geometry, and determining the single breast diameter slicing points according to the average distance and the variance.
4. The method for extracting tree breast diameter value based on ground-based laser radar technology as claimed in claim 3, wherein the objective function of the k-means clustering algorithm is expressed as the sum of the inter-cluster sample distances as shown in formula (2):
where k is the number of target class clusters, n is the total number of sample points, xi is the sample belonging to the jth target class cluster, ujRepresenting the center of the jth target class cluster.
5. The method for extracting tree breast diameter value based on ground-based laser radar technology according to claim 3, wherein the inverse distance weighting formula is shown as formula (3):
wherein, CiAs cluster center of class i, pjFor the j-th object point in the i-th class, Sj,Smax,SminRespectively representing the distance between the jth point and the last iteration clustering center, the maximum value and the minimum value of the distance.
6. The method for extracting tree breast diameter values based on ground-based laser radar technology according to claim 1, wherein the step of performing ellipse fitting processing on the single tree breast diameter slicing points to obtain the breast diameter value extraction result comprises:
constructing a breast diameter section shape model according to the breast diameter central point coordinate of the single-wood breast diameter slicing point and the breast diameter slicing radius;
and solving the model parameters of the breast diameter section form model by combining a random sampling consistency algorithm and a least square method to obtain a breast diameter value extraction result.
7. The method for extracting the breast diameter value of the forest based on the ground-based laser radar technology as claimed in claim 6, wherein the step of solving the model parameters of the breast diameter section morphological model by the combination of the stochastic sampling consistency algorithm and the least square method to obtain the breast diameter value extraction result comprises:
identifying outlier noise points according to a random sampling consistency algorithm, and obtaining an effective sample set after the noise points are removed;
and solving the model parameters of the breast diameter section form model of the effective sample set after the noise points are removed according to a least square method so as to obtain a breast diameter value extraction result.
8. The utility model provides a device based on base laser radar technique forest diameter at breast height value draws which characterized in that includes:
the forest land point cloud data acquisition unit is used for acquiring forest land point cloud data of the ground radar;
the breast diameter slice point extraction unit is used for performing point cloud elevation processing on the foundation radar forest land point cloud data and extracting breast diameter slice points;
the single-tree breast diameter slice point extraction unit is used for cutting the breast diameter slice points according to the improved k-means clustering and extracting single-tree breast diameter slice points; and
and the breast diameter value extraction result obtaining unit is used for carrying out ellipse fitting processing on the single-wood breast diameter slicing points to obtain a breast diameter value extraction result.
9. A computer arrangement, comprising a memory and a processor, the memory having stored thereon a computer program that, when executed by the processor, causes the processor to carry out the steps of the method of ground based lidar technology based forest breast diameter value extraction of any of claims 1 to 7.
10. A computer-readable storage medium, having stored thereon a computer program which, when being executed by a processor, causes the processor to carry out the steps of the method of ground based lidar technology based forest breast diameter value extraction according to any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110368040.1A CN113269825B (en) | 2021-04-06 | 2021-04-06 | Forest breast diameter value extraction method based on foundation laser radar technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110368040.1A CN113269825B (en) | 2021-04-06 | 2021-04-06 | Forest breast diameter value extraction method based on foundation laser radar technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113269825A true CN113269825A (en) | 2021-08-17 |
| CN113269825B CN113269825B (en) | 2022-07-12 |
Family
ID=77228497
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110368040.1A Expired - Fee Related CN113269825B (en) | 2021-04-06 | 2021-04-06 | Forest breast diameter value extraction method based on foundation laser radar technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113269825B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114387288A (en) * | 2022-01-17 | 2022-04-22 | 滁州学院 | Single standing tree three-dimensional information extraction method based on vehicle-mounted laser radar point cloud data |
| CN115222629A (en) * | 2022-08-08 | 2022-10-21 | 西南交通大学 | Single remote sensing image cloud removing method based on cloud thickness estimation and deep learning |
| CN115953607A (en) * | 2023-01-04 | 2023-04-11 | 北京数字绿土科技股份有限公司 | Trunk diameter at breast height extraction method and system using point cloud data |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107705309A (en) * | 2017-10-15 | 2018-02-16 | 南京林业大学 | Forest parameter evaluation method in laser point cloud |
| 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 |
| CN108764531A (en) * | 2018-04-28 | 2018-11-06 | 南京秀强信息科技有限公司 | The cluster that single wooden diameter of a cross-section of a tree trunk 1.3 meters above the ground is extracted based on TLS justifies Optimizing Algorithm for Fitting |
| CN109155151A (en) * | 2016-05-19 | 2019-01-04 | 皇家飞利浦有限公司 | For the mthods, systems and devices based on inconsistency measurement according to the subsets counts of biological data |
| CN109816779A (en) * | 2019-01-30 | 2019-05-28 | 北京林业大学 | A method of artificial forest forest model, which is rebuild, using smart phone obtains single wooden parameter |
| CN110070571A (en) * | 2019-04-28 | 2019-07-30 | 安徽农业大学 | A kind of moso bamboo morphological parameters detection method based on depth camera |
| US20200132651A1 (en) * | 2012-06-01 | 2020-04-30 | Agerpoint, Inc. | Systems and methods for monitoring agricultural products |
| WO2020109666A1 (en) * | 2018-11-30 | 2020-06-04 | Prefor Oy | Method and system for generating forestry data |
| CN112200083A (en) * | 2020-10-10 | 2021-01-08 | 辽宁工程技术大学 | Airborne multi-spectral LiDAR data segmentation method based on multivariate Gaussian mixture model |
| CN112381861A (en) * | 2020-09-18 | 2021-02-19 | 南京航空航天大学 | Forest land point cloud data registration and segmentation method based on foundation laser radar |
| CN112598263A (en) * | 2020-12-19 | 2021-04-02 | 武汉大学 | Centrality and aggregativity measuring method based on city expansion global structure |
-
2021
- 2021-04-06 CN CN202110368040.1A patent/CN113269825B/en not_active Expired - Fee Related
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200132651A1 (en) * | 2012-06-01 | 2020-04-30 | Agerpoint, Inc. | Systems and methods for monitoring agricultural products |
| CN109155151A (en) * | 2016-05-19 | 2019-01-04 | 皇家飞利浦有限公司 | For the mthods, systems and devices based on inconsistency measurement according to the subsets counts of biological data |
| CN107705309A (en) * | 2017-10-15 | 2018-02-16 | 南京林业大学 | Forest parameter evaluation method in laser point cloud |
| 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 |
| CN108764531A (en) * | 2018-04-28 | 2018-11-06 | 南京秀强信息科技有限公司 | The cluster that single wooden diameter of a cross-section of a tree trunk 1.3 meters above the ground is extracted based on TLS justifies Optimizing Algorithm for Fitting |
| WO2020109666A1 (en) * | 2018-11-30 | 2020-06-04 | Prefor Oy | Method and system for generating forestry data |
| CN109816779A (en) * | 2019-01-30 | 2019-05-28 | 北京林业大学 | A method of artificial forest forest model, which is rebuild, using smart phone obtains single wooden parameter |
| CN110070571A (en) * | 2019-04-28 | 2019-07-30 | 安徽农业大学 | A kind of moso bamboo morphological parameters detection method based on depth camera |
| CN112381861A (en) * | 2020-09-18 | 2021-02-19 | 南京航空航天大学 | Forest land point cloud data registration and segmentation method based on foundation laser radar |
| CN112200083A (en) * | 2020-10-10 | 2021-01-08 | 辽宁工程技术大学 | Airborne multi-spectral LiDAR data segmentation method based on multivariate Gaussian mixture model |
| CN112598263A (en) * | 2020-12-19 | 2021-04-02 | 武汉大学 | Centrality and aggregativity measuring method based on city expansion global structure |
Non-Patent Citations (7)
| Title |
|---|
| GÁBOR BROLLY等: "Algorithms for Stem Mapping by Means of Terrestrial Laser Scanning", 《ACTA SILV. LIGN. HUNG》, vol. 5, 31 January 2009 (2009-01-31), pages 122 - 123 * |
| GEZA KIRALY等: "Modelling Single Trees from Terrestrial Laser Scanning Data in A Forest Reserve", 《THE PHOTOGRAMMETRIC JOURNAL OF FINLAND》, vol. 21, no. 1, 31 December 2008 (2008-12-31), pages 37 - 50 * |
| 刘丹等: "K-均值算法聚类数的确定", 《硅谷》, 31 December 2011 (2011-12-31), pages 38 - 39 * |
| 刘鲁霞等: "基于地基激光雷达的亚热带森林单木胸径与树高提取", 《林业科学》, vol. 52, no. 2, 29 February 2016 (2016-02-29), pages 26 - 37 * |
| 吴荣仁: "基于深度学习的点云树木分割及其胸径估计", 《中国优秀博硕士学位论文全文数据库(硕士) 农业科技辑》, no. 07, 15 July 2020 (2020-07-15), pages 43 - 44 * |
| 李真等: "树干与地面点云分类K-means方法的改进", 《东北林业大学学报》, vol. 47, no. 1, 31 January 2019 (2019-01-31), pages 41 - 46 * |
| 步国超等: "基于单站地面激光雷达数据的自适应胸径估计方法", 《激光与光电子学进展》, vol. 53, 31 December 2016 (2016-12-31), pages 082803 - 1 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114387288A (en) * | 2022-01-17 | 2022-04-22 | 滁州学院 | Single standing tree three-dimensional information extraction method based on vehicle-mounted laser radar point cloud data |
| CN115222629A (en) * | 2022-08-08 | 2022-10-21 | 西南交通大学 | Single remote sensing image cloud removing method based on cloud thickness estimation and deep learning |
| CN115222629B (en) * | 2022-08-08 | 2023-05-05 | 西南交通大学 | Cloud thickness estimation and deep learning-based single remote sensing image cloud removal method |
| CN115953607A (en) * | 2023-01-04 | 2023-04-11 | 北京数字绿土科技股份有限公司 | Trunk diameter at breast height extraction method and system using point cloud data |
| CN115953607B (en) * | 2023-01-04 | 2024-02-13 | 北京数字绿土科技股份有限公司 | Trunk breast diameter extraction method and system using point cloud data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113269825B (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108921885B (en) | Method for jointly inverting forest aboveground biomass by integrating three types of data sources | |
| CN113269825B (en) | Forest breast diameter value extraction method based on foundation laser radar technology | |
| CN107274417B (en) | Single tree segmentation method based on airborne laser point cloud aggregation relation | |
| CN106529469B (en) | Unmanned aerial vehicle-mounted LiDAR point cloud filtering method based on self-adaptive gradient | |
| US8537337B2 (en) | Method and apparatus for analyzing tree canopies with LiDAR data | |
| CN112381861A (en) | Forest land point cloud data registration and segmentation method based on foundation laser radar | |
| CN103324945A (en) | Forest point cloud sorting method based on mode recognition | |
| CN107085710A (en) | A kind of single wooden extraction method based on multispectral LiDAR data | |
| CN111091079B (en) | TLS-based method for measuring vegetation advantage single plant structural parameters in friable region | |
| CN108981616B (en) | Method for inverting effective leaf area index of artificial forest by unmanned aerial vehicle laser radar | |
| CN105910556A (en) | Leaf area vertical distribution information extraction method | |
| Pirotti et al. | A comparison of tree segmentation methods using very high density airborne laser scanner data | |
| CN108896021B (en) | Method for extracting artificial forest stand structure parameters based on aerial photogrammetry point cloud | |
| CN110765962A (en) | Plant identification and classification method based on three-dimensional point cloud contour dimension values | |
| CN109766824B (en) | Active and passive remote sensing data fusion classification method based on fuzzy evidence theory | |
| CN112347894A (en) | Single-plant vegetation extraction method based on transfer learning and Gaussian mixture model separation | |
| CN111428784A (en) | Robust segmentation method for deciduous forest tree-level parameter measurement by adopting airborne laser radar | |
| CN109325433A (en) | Introduce the black soil region soybean biomass multi-temporal remote sensing inversion method of terrain factor | |
| CN114863369B (en) | Method, device, equipment and medium for monitoring corn lodging by laser radar | |
| CN115294147A (en) | Method for estimating aboveground biomass of single trees and forests based on unmanned aerial vehicle laser radar | |
| Grishin et al. | Procedure for locating trees and estimating diameters using LiDAR data | |
| CN115880487A (en) | Forest laser point cloud branch and leaf separation method based on deep learning method | |
| CN110569805A (en) | Unmanned aerial vehicle image point cloud-based method for extracting individual forest trees and evaluating quality of forest trees | |
| CN117114147A (en) | Method and device for estimating forest vegetation carbon reserves based on radar and satellite remote sensing | |
| CN115994939A (en) | Tree leaf area estimation method based on ground laser point cloud |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220712 |