CN110765962A - Plant identification and classification method based on three-dimensional point cloud contour dimension values - Google Patents
Plant identification and classification method based on three-dimensional point cloud contour dimension values Download PDFInfo
- Publication number
- CN110765962A CN110765962A CN201911040565.1A CN201911040565A CN110765962A CN 110765962 A CN110765962 A CN 110765962A CN 201911040565 A CN201911040565 A CN 201911040565A CN 110765962 A CN110765962 A CN 110765962A
- Authority
- CN
- China
- Prior art keywords
- plant
- point cloud
- points
- dimensional
- log
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000000034 method Methods 0.000 title claims abstract description 136
- 241000196324 Embryophyta Species 0.000 claims abstract description 256
- 238000005070 sampling Methods 0.000 claims description 22
- 230000008569 process Effects 0.000 claims description 13
- 238000002474 experimental method Methods 0.000 claims description 8
- 230000001419 dependent effect Effects 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 description 17
- 241000723346 Cinnamomum camphora Species 0.000 description 7
- 230000002159 abnormal effect Effects 0.000 description 7
- 238000011160 research Methods 0.000 description 7
- 235000011201 Ginkgo Nutrition 0.000 description 5
- 241000218628 Ginkgo Species 0.000 description 5
- 235000008100 Ginkgo biloba Nutrition 0.000 description 5
- 238000013178 mathematical model Methods 0.000 description 5
- DSSYKIVIOFKYAU-XCBNKYQSSA-N (R)-camphor Chemical compound C1C[C@@]2(C)C(=O)C[C@@H]1C2(C)C DSSYKIVIOFKYAU-XCBNKYQSSA-N 0.000 description 4
- 240000002924 Platycladus orientalis Species 0.000 description 4
- 230000003247 decreasing effect Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 240000005308 Juniperus chinensis Species 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 241000894007 species Species 0.000 description 3
- 241000282994 Cervidae Species 0.000 description 2
- 230000019771 cognition Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 241001665752 Gypsophila repens Species 0.000 description 1
- 244000227633 Ocotea pretiosa Species 0.000 description 1
- 235000004263 Ocotea pretiosa Nutrition 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
The plant identification and classification method based on the three-dimensional point cloud contour dimension value provided by the invention uses the three-dimensional laser scanner to quickly and efficiently acquire the three-dimensional information of the surface of the measured object, the fractal dimension method is used for calculating the fractal dimension value of the plant contour point cloud data, the fractal dimension value of the plant three-dimensional point cloud contour is used for distinguishing the plant species, the intelligent degree is high, the plant identification is efficient and accurate, the method is a brand new technological means for acquiring the three-dimensional coordinate information of an object space, the three-dimensional information of the plant is acquired through a three-dimensional laser scanner, the plant identification work is turned from outdoor to indoor, meanwhile, the plant characteristics are extracted by utilizing the three-dimensional information of the plants, the plant classification is realized by utilizing the plant characteristics, the plant identification and classification method based on the three-dimensional point cloud contour dimension values is small in workload, high in classification speed, high in efficiency, low in cost and less in error, and the accuracy and reliability of the plant identification and classification are greatly improved.
Description
Technical Field
The invention relates to a plant identification and classification method based on point cloud, in particular to a plant identification and classification method based on three-dimensional point cloud contour dimension values, and belongs to the technical field of plant identification and classification methods.
Background
Plants on earth have a very important influence on the ecological environment and are in a very important position in the whole earth ecosystem, so that the improvement of the identification and the recognition of the plants are particularly important. The plant identification and classification is a precondition and basic work for correctly recognizing and researching plants, provides basic research results for the research of botany and ecology, and has wide application prospect for researching the plant identification and classification method which is rapid, accurate and convenient. In addition, urban and rural ecological construction and greening are important contents of urban and rural development, the planting quantity of greening plants is considered in the process, the configuration of plant species is also required to be fully considered, advanced plant identification and classification methods can be fully utilized, meanwhile, in digital urban three-dimensional construction, plant classification has important significance for automatic plant modeling, and the method is an important link in the refined digital urban three-dimensional construction.
The three-dimensional laser radar scanner is a new means and technology for rapidly acquiring spatial information, can directly acquire the three-dimensional space coordinates of the point cloud on the surface of the measured target ground object by utilizing the three-dimensional laser scanner, and has the characteristics of dense point cloud distribution, high sampling density, high acquisition speed and the like. The method is characterized in that only a series of scanning point information related to the shape is reserved in the point cloud data of the plant obtained through the three-dimensional laser point cloud, and the point cloud data is a three-dimensional discrete expression with fractal characteristics, but the points do not contain any topological structure, so that the method for converting the cognition of modern science on plant classification into the representation of the corresponding point cloud data characteristics is a premise for carrying out three-dimensional point cloud plant classification. The recognition of the modern science on plant classification can be combined with the scanned three-dimensional point cloud data, namely, the perceptual cognition of the modern science for classifying and recognizing according to the plant morphology is converted into the feature calculation of the corresponding three-dimensional point cloud data, and the features are closely related to the basis of the modern science for recognizing and classifying different plants, so that the idea of calculating the fractal features of the plants by using the point cloud of the plants is a brand-new idea.
The prior art research on methods for identifying and classifying plant species mainly focuses on plant appearance features, particularly the shape of plant leaves, and comprises the steps of extracting features for describing the shape of the leaves to classify the plants, and describing the boundary and the shape of the plant leaves and other features by using a multi-scale curvature space to classify the plants. In recent years, preliminary discussion of automatic plant classification by using computer technology has gradually appeared, which includes establishing a plant classification recognition model to research plant classification by extracting leaf appearance shape characteristics based on plant leaf shape, size, circularity parameters, leaf margin and the like, but identification based on plant leaves has great difficulty and limitation, describing leaves is difficult to establish an accurate mathematical model, a method for describing plant leaf shapes by using computer graphics and the like can only classify specific types of plants and is not suitable for rapidly recognizing a large number of plants in the nature, and some plants and varieties have similar leaf shapes, and the method based on plant leaf shapes is not suitable for the situation.
In summary, the plant identification and classification method in the prior art mainly has the following defects: firstly, the plant identification and classification method in the prior art mainly depends on manual identification and classification, the classification method needs to consume a large amount of manpower, material resources and financial resources, can complete work only by needing a large amount of manual participation, has large workload of plant identification and classification and low identification efficiency, needs a large amount of expert scholars with rich plant classification experience to participate, and has great influence on classification results by the professional level of plant classification personnel; secondly, the prior art mainly focuses on the plant appearance characteristics, especially the leaf shape, and in recent years, the initial research of using computer technology to automatically classify plants gradually appears, including establishing a plant classification identification model to research the plant classification by extracting leaf appearance shape characteristics based on the leaf shape, size, circularity parameters, leaf margin and the like of plant leaves, but the identification based on the plant leaves has great difficulty and limitation, the leaf description is difficult to establish an accurate mathematical model, the method for describing the shape of the plant leaf by computer graphics and other technologies can only classify specific types of plants, the method is not suitable for quickly identifying a large number of plants in the nature, and some plants and varieties have similar leaf shapes, so that the method based on the leaf shapes of the plants is not suitable; the three-dimensional laser scanner has strong vegetation space detection capability, the structure of the vegetation can be described by utilizing the three-dimensional information of the vegetation acquired by the three-dimensional laser scanner, but most of the prior art is based on optical images or large-area forests, the real three-dimensional information of a single plant outline is not acquired, the fractal dimension value has partial application in ecology, but the research on plant classification based on the fractal feature of three-dimensional point cloud is few, and a reliable method is not available; fourthly, in the prior art, no matter which method is adopted, the problems of large workload of plant identification and classification, low classification speed, low efficiency, high cost, more errors and insufficient accuracy and reliability of plant identification and classification exist.
Disclosure of Invention
Aiming at the defects of the prior art, the plant identification and classification method based on the three-dimensional point cloud profile dimension value provided by the invention uses the three-dimensional laser scanner to quickly and efficiently acquire the three-dimensional information of the surface of the measured object, is a brand-new technological means for acquiring the three-dimensional coordinate information of the object space, acquires the three-dimensional information of the plant through the three-dimensional laser scanner, extracts the plant characteristics by using the three-dimensional information of the plant while turning the plant identification work from the outdoor to the indoor, realizes plant classification by using the plant characteristics, has the advantages of small workload, high classification speed, high efficiency, low cost, less error and greatly improved accuracy and reliability of the plant identification and classification. The method calculates the fractal dimension of the plant contour point cloud data by a Minkowski dimension method, firstly carries out multi-scale cubic box division on the spatial range of the preprocessed point cloud data, counts the number of non-empty cubic boxes corresponding to the division scale, then uses the logarithm value of the statistical result to make a log-log coordinate graph, and finally obtains the slope of a fitting straight line through the points in the log-log coordinate graph, namely the fractal dimension, and uses the fractal dimension of the plant three-dimensional point cloud contour to distinguish the plant species, so that the method has high intelligent degree and high plant identification efficiency and accuracy, overcomes the problems that the prior art mainly depends on manual identification and classification, needs a large amount of manual participation to complete the work, and the professional level of plant classifiers can cause huge influence on the classification result, and is a brand-new plant identification and classification means with high value.
In order to achieve the technical effects, the technical scheme adopted by the invention is as follows:
a plant identification and classification method based on three-dimensional point cloud contour dimension values comprises calculation of plant contour three-dimensional point cloud dimension values and plant classification based on point cloud fractal features,
calculating the fractal dimension value of the plant contour three-dimensional point cloud, calculating the fractal dimension value of the plant contour point cloud data by a Minkowski dimension method, firstly carrying out multi-scale cubic box division on the spatial range of the preprocessed point cloud data, counting the number of non-empty cubic boxes corresponding to the division scale, then using the logarithm value of the statistical result to make a dual-logarithmic coordinate graph, finally obtaining the slope of a fitting straight line through points in the dual-logarithmic coordinate graph, namely the fractal dimension value, calculating the fractal dimension value of the plant contour three-dimensional point cloud, obtaining a plant contour point cloud A by preprocessing the three-dimensional point cloud, calculating the fractal dimension value by the Minkowski dimension method to obtain a dual-logarithmic coordinate graph point set B, and solving the fractal dimension value by fitting the straight line by a least square method;
plant classification based on point cloud fractal features utilizes a Minkowski dimensionality method to calculate fractal values of plant contour point clouds, finds out the fractal value corresponding to each tree through experiments or experiences, distinguishes plant species by utilizing the fractal values of the plant three-dimensional point cloud contours, and identifies and classifies the plants.
A plant identification and classification method based on three-dimensional point cloud contour dimension values is characterized in that plant contour point cloud A is obtained through three-dimensional point cloud pretreatment, plant contour data are collected through a three-dimensional laser scanner, collected multi-station data are spliced, cut and denoised,
the method for acquiring the complete spatial three-dimensional coordinate data of the plant by using the ground three-dimensional laser scanner comprises the following steps: carrying out multi-station laser point cloud measurement on the plant, carrying out data splicing on the three-dimensional point cloud after error correction, cutting the spliced point cloud data according to the plant range, only keeping the three-dimensional point cloud of the main outline of the plant, removing the three-dimensional point cloud of other irrelevant objects, removing repeated points in the data after obtaining the plant three-dimensional point cloud, and finally obtaining the three-dimensional point cloud A of the single plant.
A plant identification and classification method based on three-dimensional point cloud contour fractal dimension value is further characterized in that a Minkowski dimensionality method is used for calculating and obtaining a double-logarithmic coordinate point set B,
cubic box division is carried out on the plant three-dimensional point cloud when the plant fractal dimension value is calculated through a Minkowski dimension method, firstly, the external minimum rectangular parameter of the plant three-dimensional point cloud is obtained, and the spatial range and the cubic box division of the plant three-dimensional point cloud are determinedStarting point coordinate (x) of the pointstart，ystart，zstart) Then setting the initial side length c of the cubic box, carrying out multi-scale cubic box division on the space of the plant, dividing the space range of the plant point cloud by the cubic boxes with different side lengths, carrying out iteration in the whole space division process, determining the iteration number D of the space division, wherein in order to enable enough points in the dual-logarithm coordinate graph point set B, the iteration number D is determined by the following formula,
cube box side length E at each iterationiC × i (i ═ 1, 2, 3.., D), where E isminThe side length of the shortest side of the minimum rectangle externally connected with the plant point cloud, c is the initial side length of the cubic box, i is the current cycle number, EiDetermining the iteration number by taking an integer not greater than the quotient of half of the shortest side length and the initial side length when the side length of the cubic box in the ith cycle is equal to the length of the original side,
at each iteration, first by cube box dimension EiDividing the space where the plant point cloud is located into cubic boxes, judging the cubic boxes containing the plant three-dimensional point cloud in the cubic boxes, and counting the number of the non-empty cubic boxes to be recorded as FEi(A) Simultaneous recording of cube size EiReciprocal 1/E ofiThe logarithmic value of the two forms a point in the point set B of the log-log coordinate graph
And D times of cube box division are completed, and a final double-logarithm coordinate graph point set B is obtained.
A plant identification and classification method based on three-dimensional point cloud contour fractal dimension values, further, a least squares method is used for fitting straight lines to solve the fractal dimension values,
performing least squares method fitting straight line on the point data in the point set B of the obtained final double logarithmic coordinate graph, wherein
Is a dependent variable, ln (1/E)i) Is independent variable, the slope corresponding to the straight line obtained by fitting is the solved fractal dimension value,
the equation of a straight line, y, kx + b, represents all straight lines except the parallel y-axis, and the equation of y, kx + b can be used as an equation for fitting a straight line,
wherein y denotes the ordinate, i.e. the dependent variable
X represents the abscissa, i.e. the argument ln (1/E)i) K represents the slope of the straight line, B represents the intercept of the straight line, the straight line is fitted by using a least squares method, wherein k and B are parameters to be estimated, y and x are observed values, namely point coordinates in a point set B of a log-log coordinate graph, and k and B are taken as approximate values plus errors, so that
For any point in the point set B of the log-log coordinate graph
The error equation is:
b has D points in total, so D error equations are in total, the overall error equation is as follows,
V＝GΔ-e
wherein the parameter meanings are as follows:
V＝[v1v2v3L vD]T
Δ＝[δkδb]T
for equal weight observation, the formula Δ ═ δ according to the least squares principlekδb]TIn (1) should satisfy VTThe minimum V is required, the points in the point set of the double logarithmic coordinates are independently observed, and are obtained according to a method of a function natural extreme value, as shown in the following formula,
finally, the obtained delta is shown as the following formula,
Δ＝(GTG)-1GTe
according to
And solving the slope k of the straight line and the intercept b of the straight line, wherein the slope k corresponding to the straight line obtained by fitting is the solved fractal dimension value.
A plant identification and classification method based on three-dimensional point cloud contour dimension values further comprises the steps of setting cubic box scale division parameters, the cube box side length affects the fit result of the final straight line, as the initial cube box side length increases, when the side length of the cubic box is small, the distance between the corresponding point in the log-log graph and the straight line finally fitted by the least square method is gradually reduced, when the side length of the cubic box is small, the influence of points in the corresponding log-log coordinate graph on the least square fit straight line is reduced, the influence of deviation points in the set of points of the log-log coordinate graph on the least square fit straight line is reduced, the trend of clockwise rotation of the fit straight line is reduced, therefore, the slope of the finally fitted straight line is increased along with the proper increase of the initial side length of the cubic box, namely the calculated plant dimension value is increased.
A plant identification and classification method based on three-dimensional point cloud contour dimension values is characterized by further comprising a screening step of adding points when a plant contour point cloud double-logarithmic coordinate graph point set is obtained by a Minkowski dimension method when points with the same longitudinal coordinate in the double-logarithmic coordinate graph point set are removed, judging whether the points to be added into the double-logarithmic coordinate graph point set exist points with the number of non-empty cubic boxes identical to that of points to be added or not in the point set, judging whether the points with the longitudinal coordinate being the number of the non-empty cubic boxes of the points to be added exist or not in the double-logarithmic coordinate graph point set or not if the points do not exist, adding the points to be added into the double-logarithmic coordinate point set only if the points do not exist in the point set, and otherwise not adding the points into the point set.
A plant identification and classification method based on three-dimensional point cloud contour fractal dimension values is characterized in that a random sampling consistency algorithm is used for removing deviation points, an input item of the random sampling consistency algorithm comprises a group of observation data sets, namely a sample data set A, and a parameterized model B suitable for observation data, the number of necessary parameters for constructing the model B is n, and the number of samples in the sample set A is larger than n. The algorithm randomly selects a group of subsets A with the number of samples n from a sample data set A through cyclic iterationsubCalculating n parameters of model B as correct sample data to obtain model BiRemoving the subset A from the sample set AsubThe other points of (2) judge whether they conform to the model B or not according to the allowable deviation threshold value HiIf the sample number in the internal point set is larger than the sample number threshold h of the internal point set, the model is one of robust models, the sample data in all the internal point sets are used for estimating model parameters, the model is compared with the current optimal model, the optimal model is reserved, the steps are repeated until the iteration times specified by the algorithm are reached, and the parameters corresponding to the optimal model and the corresponding internal point set are obtained.
A plant identification and classification method based on three-dimensional point cloud contour fractal dimension values is further characterized in that all possible linear models in a log-log coordinate graph point set are calculated according to iteration times when a random sampling consistency algorithm is executed, the cycle iteration times N are shown as the following formula, wherein M represents the number of points in the log-log graph point set,
the points in the point set of the log-log coordinate graph are sequentially obtained along with the length of the side of the cubic box in the process of dividing the cubic box space of the plant contour point cloud, when two points are randomly selected from the point set of the log-log coordinate graph, the first point added into the set of the log-log coordinate graph is taken as the first point for obtaining the linear parameter, and then the points added into the set of the log-log coordinate graph are sequentially taken as the points for obtaining the linear parameter.
In the plant classification based on the point cloud fractal characteristics, the point cloud contour fractal dimension values of different plants are obviously different, the fractal dimension value corresponding to each tree can be found out through experiments or experiences, the plant types are distinguished by utilizing the plant three-dimensional point cloud contour fractal dimension values, and the single plant is efficiently identified and classified.
Compared with the prior art, the invention has the advantages that:
1. the plant identification and classification method based on the three-dimensional point cloud contour dimension values, provided by the invention, has the advantages that the three-dimensional laser scanner is used for quickly and efficiently acquiring the three-dimensional information of the surface of a measured object, the method is a brand-new technological means for acquiring the three-dimensional coordinate information of the object space, the three-dimensional laser scanner is used for acquiring the three-dimensional information of plants, the plant characteristics are extracted by using the three-dimensional information of the plants while the plant identification work is turned from the outdoor to the indoor, the plant classification is realized by using the plant characteristics, the workload of the plant identification and classification method based on the three-dimensional point cloud contour dimension values is small, the classification speed is high, the efficiency is high, the cost is low, the errors are few.
2. The invention provides a plant identification and classification method based on three-dimensional point cloud contour dimension values, calculating the fractal dimension of the plant contour point cloud data by Minkowski dimension method, firstly carrying out multi-scale cubic box division on the spatial range of the preprocessed point cloud data, counting the number of non-empty cubic boxes corresponding to the division scale, then using the logarithm value of the statistical result to make a dual-logarithmic coordinate graph, finally obtaining the slope of the fitting straight line through the points in the dual-logarithmic coordinate graph, namely the fractal dimension value, the fractal dimension value of the plant three-dimensional point cloud contour is utilized to distinguish the plant species, the intelligent degree is high, the plant identification is efficient and accurate, the defects that the prior art mainly depends on manual identification and classification, a large amount of labor is needed to complete the work, and the professional level of plant classification personnel can cause huge influence on the classification result, and the method is a brand-new plant identification and classification technological means with high value.
3. The plant identification and classification method based on the three-dimensional point cloud contour fractal dimension solves the problem that when a Minkowski dimension method is used for calculating the plant point cloud fractal dimension, the influence of points with the same initial side length of a cubic box and the same vertical coordinate in a double-logarithmic coordinate graph point set on the fractal dimension value is solved, meanwhile, the deviated points in the double-logarithmic coordinate graph point set are removed through a random sampling consistency algorithm, the accuracy of calculating the plant fractal dimension value by the Minkowski dimension method is improved, and the precision of the plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value is higher.
4. According to the plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value, a series of embodiments and classification experiments are carried out on collected plant point cloud data by using the three-dimensional point cloud contour fractal dimension value characteristics, the feasibility and the reliability of the method are verified, the method can meet the requirement of large-scale plant rapid identification, the obtained plant contour fractal dimension value is high in differentiation degree, and the plant identification and classification are precise and accurate.
Drawings
FIG. 1 is a step diagram of a plant identification and classification method based on three-dimensional point cloud contour dimension values according to the present invention.
FIG. 2 is a flowchart illustrating the determination of the same point on the ordinate in the point set of the log-log coordinates according to the present invention.
FIG. 3 is a flow chart of the random sample consensus algorithm of the present invention to remove outliers.
Detailed Description
The technical scheme of the plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value provided by the invention is further described below with reference to the accompanying drawings, so that the technical scheme can be better understood and implemented by those skilled in the art.
Referring to fig. 1 to 3, the plant identification and classification method based on three-dimensional point cloud contour fractal value provided by the invention comprises plant contour three-dimensional point cloud fractal value calculation and plant classification based on point cloud fractal characteristics, the plant contour three-dimensional point cloud fractal value calculation calculates the fractal value of plant contour point cloud data through a minkowski dimension method, firstly, multi-scale cubic box division is carried out on the spatial range where the preprocessed point cloud data is located, the number of non-empty cubic boxes corresponding to the division scale is counted, then, a logarithm value of a statistical result is used for making a dual-logarithm coordinate graph, finally, the slope of a fitting straight line is obtained through points in the dual-logarithm coordinate graph, namely, the fractal value is obtained, a cubic box division scale parameter setting is added in the process of calculating the fractal dimension value through the minkowski dimension method, error points existing in the dual-logarithm coordinate graph are removed, obtaining the optimal calculation result, wherein the plant contour three-dimensional point cloud fractal dimension value calculation comprises three steps of obtaining a plant contour point cloud A through three-dimensional point cloud pretreatment, obtaining a double-logarithmic coordinate graph point set B through Minkowski dimension method calculation, and solving the fractal dimension value through a least-squares method fitting straight line; plant classification based on point cloud fractal features utilizes a Minkowski dimensionality method to calculate fractal values of plant contour point clouds, finds out the fractal value corresponding to each tree through experiments or experiences, distinguishes plant species by utilizing the fractal values of the plant three-dimensional point cloud contours, and identifies and classifies the plants.
Calculating the fractal dimension value of three-dimensional point cloud of plant contour
The Minkowski dimension is used for calculating the fractal dimension of the outline of a single plant, the Minkowski dimension method of a multi-scale fractal dimension cube covering mode according to the plant space range is used for calculating the fractal dimension of the single plant, and a flow chart is shown in figure 1.
The plant identification and classification method based on the three-dimensional point cloud contour dimension values is mainly divided into three parts, specifically:
(I) preprocessing three-dimensional point cloud to obtain plant outline point cloud A
Gather plant profile data through three-dimensional laser scanner, splice, tailor, the operation of making an uproar of the many stations data of gathering, utilize ground three-dimensional laser scanner to acquire the complete space three-dimensional coordinate data of plant and include: the method comprises the steps of carrying out multi-station laser point cloud measurement on plants, carrying out data splicing on three-dimensional point clouds after error correction, cutting the spliced point cloud data according to the range of the plants, only keeping the three-dimensional point clouds of the main outlines of the plants, and removing the three-dimensional point clouds of other irrelevant objects.
(di) Minkowski dimensionality calculation to obtain dual logarithmic coordinate point set B
When the plant fractal dimension value is calculated by the Minkowski dimension method, cubic box division needs to be carried out on the plant three-dimensional point cloud, so that firstly, the external minimum rectangular parameter of the plant three-dimensional point cloud is obtained, and the space range of the plant three-dimensional point cloud and the starting point coordinate (x) of the cubic box division are determinedstart，ystart，zstart) And then setting the initial side length c of the cubic box, wherein the initial side length c of the cubic box is not suitable to be set too large due to the characteristics of high scanning accuracy, high data density and the like of the ground three-dimensional laser scanner, and the initial side length c of the cubic box is set to 0.1 in the embodiment, and meanwhile, due to the large space range of the plant, when the range of the cubic box is too small, the number of the cubic boxes divided in the space is too large, so that the calculation time is increased, the memory range of a computer is exceeded, and the number of the non-empty cubic boxes in the statistical result when the division size is too small is extremely close to the point cloud number in the plant three-dimensional point cloud, and calculation errors are caused.
The Minkowski dimension calculation method comprises the steps of carrying out multi-scale cubic box division on the space where a plant is located, dividing the space range where the plant point cloud is located through cubic boxes with different side lengths, carrying out iteration in the whole space division process, determining the iteration number D of the space division, determining the iteration number D through the following formula in order to enable enough points to exist in a double-logarithmic coordinate point set B,
cube box side length E at each iterationiC × i (i ═ 1, 2, 3.., D), where E isminThe side length of the shortest side of the minimum rectangle externally connected with the plant point cloud, c is the initial side length of the cubic box, i is the current cycle number, EiDetermining the iteration times by taking an integer not greater than the quotient of half of the shortest side length and the initial side length when the side length of the cubic box in the ith cycle is in the ith cycle, effectively avoiding the problem of too small number of space division cubic boxes caused by too many iteration times, and combining with the step EiThe side length calculation method of the multi-scale cube box in the c × i (i is 1, 2, 3., D) prevents that point data which can sufficiently reflect the structure of the log-log coordinate graph cannot be obtained due to too few iterations.
At each iteration, first by cube box dimension EiDividing the space where the plant point cloud is located into cubic boxes, judging the cubic boxes containing the plant three-dimensional point cloud in the cubic boxes, and counting the number of the non-empty cubic boxes to be recorded as FEi(A) Simultaneous recording of cube size EiReciprocal 1/E ofiThe logarithmic value of the two forms a point in the point set B of the log-log coordinate graph
And D times of cube box division are completed, and a final double-logarithm coordinate graph point set B is obtained.
(III) solving fractal dimension value by fitting straight line through least squares method
After a final point set B of the double logarithmic coordinates is obtained, a least squares method is carried out on the point data to fit a straight line, wherein
Is a dependent variable, ln (1/E)i) And the slope corresponding to the straight line obtained by fitting is the solved fractal dimension value.
And the linear equation y is kx + b and represents all straight lines except the parallel y axis, and the equation y is kx + b and can be used as an equation of the fitting straight line in consideration of the situation that the slope of the fitting straight line of the point set of the log-log coordinates is the fractal dimension of the plant and the dimension of the plant is not infinite.
Wherein y denotes the ordinate, i.e. the dependent variable
X denotes the abscissa, i.e. the argument ln (l/E)i) The method comprises the following steps of (1) fitting a straight line by using a least squares method, wherein k represents the slope of the straight line, B represents the intercept of the straight line, and k and B are parameters to be estimated, y and x are observed values, namely point coordinates in a point set B of a log-log coordinate graph, and k and B can be regarded as approximate values and errors, so that the order of
For any point in the point set B of the log-log coordinate graph
The error equation is:
b has D points in total, so D error equations are in total, the overall error equation is as follows,
V＝GΔ-e
wherein the parameter meanings are as follows:
V＝[v1v2v3L vD]T
Δ＝[δkδb]T
for equal weight observation, the formula Δ ═ δ according to the least squares principlekδb]TIn (1) should satisfy VTThe requirement of V minimum, the points in the point set of the log-log coordinate graph are independently observed, so the points are obtained according to the method of the function natural extremum, as shown in the following formula,
finally, the obtained delta is shown as the following formula,
Δ＝(GTG)-1GTe
thereby according to
And solving the slope k of the straight line and the intercept b of the straight line, wherein the slope k corresponding to the straight line obtained by fitting is the solved fractal dimension value.
The error in the unit weight can be obtained as follows according to the required number of observations being 2,
the accuracy of the slope k of the fitted line and the intercept b of the fitted line is therefore determined by the co-factor matrix (V) of ΔTV)-1Is multiplied by the error in the unit weight of the arithmetic square root of the diagonal element
And (4) calculating.
In the first embodiment, the fractal dimension values of four three-dimensional crown point cloud data of a cinnamomum camphora tree with leaves are calculated according to the above process, the fractal dimension values fluctuate up and down 2.65, wherein the initial side length of a cube box is set to 0.01, when the side length of the cube box is small, corresponding points deviate from the final fitting straight line result, and the distances of the cinnamomum camphora points with less point cloud number from the fitting straight line are far away, the points deviating from the fitting straight line deflect the slope of the fitting straight line in the least squares method by using the points in the point set of the log-log coordinate graph clockwise, so that the slope of the fitting straight line is smaller than that when no such deviation points exist in the point set of the log-log coordinate graph, the dividing scale of the cube box affects the fitting straight line result by using the log-log coordinate graph, and the number of the point cloud affects the sensitivity of the point.
Two, cubic box partition dimension parameter set
When the cubic box is divided, the side length of the cubic box influences the fitting result of the final straight line, and in order to avoid the reduction of the slope of the fitting straight line caused by the undersize of the cubic box, the undersize of the cubic box needs to be prevented. Thus, example two was performed for four sassafras with leaves, and the fractal values were calculated using different initial cube-box side lengths.
Firstly, setting the length of the initial cubic box edge to be 0.02, 0.03, 0.04 and 0.05 in sequence, and calculating the fractal dimension values of the four cinnamomum camphora trees with leaves by using a Minkowski dimension method, and obtaining the fractal dimension values from the results, as the side length of the initial cubic box increases, when the side length of the cubic box is small, the distance between the corresponding point in the log-log graph and the straight line finally fitted by the least square method is gradually reduced, when the side length of the cubic box is small, the influence of points in the corresponding log-log coordinate graph on the least square fit straight line is reduced, the influence of deviation points in the set of points of the log-log coordinate graph on the least square fit straight line is reduced, the trend of clockwise rotation of the fit straight line is reduced, therefore, the slope of the finally fitted straight line is increased along with the proper increase of the initial side length of the cubic box, namely the calculated plant dimension value is increased.
In the second embodiment, the fractal dimension values of the four cinnamomum camphora trees with leaves obtained by the least squares method are increased along with the increment of the initial side length of the cube box, but the number of points in the point set of the corresponding log-log coordinate graph is decreased, so that the number of samples of the fitting straight line by the least squares method is decreased, and the corresponding fractal dimension values are sensitive to the initial side length of the cube box due to the small number of point clouds.
Removing points with the same vertical coordinate in the point set of the double logarithmic coordinates
When the plant outline point cloud is subjected to multi-scale space division through a Minkowski dimension method, the point distribution of the dual-logarithmic coordinate graph changes from sparse to dense along with the increase of the side length of a cubic box during iterative division, the ordinate of the point in the corresponding log-log graph becomes a monotonous decreasing trend along with the increase of the side length of the cube box when the side length of the cube box is smaller, the ordinate of the point in the log-log plot corresponding to a larger cube box edge length is no longer completely monotonically decreasing as the cube box edge length increases further, the points with the same longitudinal coordinate value continuously appear near the adjacent points, the result of fitting straight lines to all the points in the double logarithmic coordinate graph point set by using the least squares method is influenced, because the fitting process does not remove these points, the final fitted straight line will pass through the middle of more points with the same ordinate as much as possible to keep the points as evenly distributed on both sides of the straight line as possible.
Therefore, in the screening step of adding points when the plant contour point cloud dual-logarithmic coordinate graph point set is obtained by the Minkowski dimension method, for the points to be added into the dual-logarithmic coordinate graph point set, whether the points with the number of non-empty cubic boxes identical to that of the points to be added exist in the point set is judged, if the points do not exist, whether the points with the longitudinal coordinates of the points to be added to the dual-logarithmic coordinate graph point set exist is judged, the points to be added are added into the dual-logarithmic coordinate point set only when the points do not exist, and otherwise, the points are not added into the point set. The points with the same ordinate in the log-log graph mostly appear at the neighboring points and the neighboring points with the increased ordinate, so to improve the calculation efficiency, the judgment flow shown in fig. 2 is added at each cubic-box division when the minkowski dimension method calculates the plant point cloud fractal dimension.
The removal of the points with the same ordinate in the point set of the log-log coordinate graph causes the reduction of the precision of the fitted straight line of the least squares method, but compared with the situation that the points with the same ordinate in the point set of the log-log coordinate graph are not removed, the trend that the precision of the fitted straight line of the least squares method is reduced along with the increase of the initial side length of the cube box is obviously reduced, particularly the sample number of the fitted straight line of the least squares is always smaller than the number of the points in the point set of the log-log coordinate graph under the same initial side length parameter of the cube box, the trend of the reduction of the precision is reduced, so the removal of the points with the same ordinate in the point set of the log-log coordinate graph plays a great role in the accurate calculation of the fractal value.
Fourthly, removing deviation points by using random sampling consistency algorithm
The accuracy of calculating the fractal dimension value of the plant contour point cloud based on the Minkowski dimension method cannot be absolutely improved by only properly increasing the initial side length of the cubic box and removing the points with the same vertical coordinate in the double-logarithmic coordinate graph point set, and in addition, the number of the middle points in the double-logarithmic graph point set is reduced, the number of available samples for fitting a straight line by the least-squares method is reduced, the number of redundant observation equations is reduced, and the accuracy is not improved. How to improve the iteration times through the initial side length of the relatively small cubic box so that more points exist in the point set of the dual-logarithm coordinate graph, and meanwhile, the method can remove the deviated points caused by the condition that the side length of the cubic box is too small or too large, and is the key for accurately calculating the point cloud fractal dimension value of the plant.
The random sampling consistency algorithm assumes that a sample contains both correct sample data conforming to the mathematical model and abnormal sample data which cannot be applied to the mathematical model, wherein the abnormal sample data is regarded as noise data, and the algorithm assumes that a set of limited number of correct sample data is given to estimate model parameters which are most suitable for the sample data set. The random sampling consistency algorithm selects correct sample data and parameters of an estimation mathematical model from a sample data set containing abnormal data in an iterative calculation mode, is used as a method for detecting the abnormal data, belongs to a non-deterministic algorithm, obtains a reasonable result under a certain probability, and increases the probability of obtaining the reasonable result by increasing the iteration times.
The input items of the random sampling consistency algorithm comprise a group of observation data sets, namely a sample data set A, and a parameterized model B suitable for observation data, the number of necessary parameters for constructing the model B is n, and the number of samples in the sample set A is more than n. The algorithm randomly selects a group of subsets A with the number of samples n from a sample data set A through cyclic iterationsubCalculating n parameters of model B as correct sample data to obtain model BiRemoving the subset A from the sample set AsubThe other points of (2) judge whether they conform to the model B or not according to the allowable deviation threshold value HiIf the sample data is matched with the correct sample data, adding the point into the correct sample dataIf the internal point set does not meet the requirement, the internal point set and the external point set are added into the external point set where the abnormal data are located, the internal point set and the external point set are finally obtained, when the number of samples in the internal point set is larger than the threshold value h of the number of samples in the internal point set, the model is one of robust models, the model parameters are estimated by using the sample data in all the internal point sets, the model is compared with the current optimal model, the optimal model is reserved, the steps are repeated until the iteration times specified by the algorithm are reached, and the parameters corresponding to the optimal model and the corresponding internal point set are obtained.
The random sampling consistency algorithm has the advantages that the model parameters can be stably estimated, abnormal data are reasonably removed through the random sampling consistency algorithm when the point in the point set of the dual-logarithm coordinate graph is subjected to straight line fitting by using a least squares method, and a straight line equation is accurately fitted under the condition that the initial side length of a cubic box and the same point of a longitudinal coordinate exists in the dual-logarithm point set, so that the fractal dimension value of the plant contour point cloud is obtained.
When abnormal data in the dual-logarithm coordinate graph point set are removed by utilizing a random sampling consistency algorithm, the characteristics of the dual-logarithm coordinate graph point set are fully considered to set reasonable iteration times and other parameters, and the characteristics of the dual-logarithm coordinate graph point set comprise:
(1) the data in the point set of the log-log coordinate graph only conform to the linear model, and the number of necessary parameters of the linear model is 2;
(2) the same point does not exist in the point set of the dual-logarithm coordinate graph, because each point in the set corresponds to one cubic box division, the side length of a cubic box in each cubic box division is different, the abscissa of each point in the point set of the dual-logarithm coordinate graph is different, and the linear model parameter estimation of any two points in the dual-logarithm coordinate graph set is ensured;
(3) the spatial range of the plant contour three-dimensional point cloud is limited and not very large, and the number of points in the point set of the log-log coordinate graph is not too large.
Therefore, all possible linear models in the point set of the log-log coordinate graph are calculated according to the iteration number when the random sampling consistency algorithm is executed, the loop iteration number N is shown as the following formula, wherein M represents the number of points in the point set of the log-log graph,
the points in the point set of the log-log coordinate graph are sequentially obtained along with the length of the side of the cubic box in the process of dividing the cubic box space of the plant contour point cloud, so when two points are randomly selected from the point set of the log-log coordinate graph, the first point added into the set of the log-log coordinate graph is taken as the first point for obtaining the linear parameter, and then the points added into the set of the log-log coordinate graph behind the first point are sequentially taken as the points for obtaining the linear parameter.
A flow chart of removing deviation points and fitting straight lines by random sampling consistency after a point set of a log-log coordinate graph is obtained by dividing a plant contour point cloud by using a multi-scale cube box is shown in fig. 3.
The random sampling consistency algorithm is used, a fixed threshold is set to remove the deviation points, the precision of calculating the fractal dimension value of the plant contour point cloud by the Minkowski dimension method is obviously improved, and the precision of the fitted straight line in unit weight error, slope error, intercept error and the like is greatly improved.
The random sampling consistency algorithm is used, the fixed threshold is set to remove the points subjected to the deviating point treatment, so that the influence of too many points with the same vertical coordinate on the fitting straight line is reduced, meanwhile, the continuity of the points in the point set of the log-log coordinate graph is not damaged, the precision of the final fitting straight line is improved, the random sampling consistency algorithm effectively removes the deviating points in the point set of the log-log coordinate graph, and the calculation precision of the fractal dimension value is improved.
Fifthly, plant classification based on point cloud fractal features
The embodiment distinguishes the classes of the ginkgo tree with the leaf, the camphor tree and the Chinese juniper by utilizing a Minkowski dimension method based on a random sampling consistency iteration threshold, the specific embodiment shows that the fractal dimension values of the point clouds of the ginkgo tree with the leaf, the Chinese juniper with the deer and the camphor tree are obviously distributed in different range intervals, the fractal dimension value of the camphor tree is near 2.6, the fractal dimension value of the Chinese juniper with the deer is near 2.4, and the fractal dimension value of the ginkgo tree is near 2.1, and the fact that the tree species can be distinguished by utilizing the fractal dimension values is proved, so that the fractal dimension value corresponding to each tree species can be found out through experiments or experiences, the plant species can be distinguished by utilizing the three-dimensional contour point cloud fractal dimension values of the plant, and the plant species can be identified and classified.
The invention firstly provides a process for calculating the fractal dimension value of a plant point cloud by using a Minkowski dimension method, and carries out relevant experimental verification, secondly compares the influence of points with the same vertical coordinates in a set of cubic box initial side length and a double-logarithmic coordinate graph points on the fractal dimension value, provides a process for removing the deviated points in the set of the double-logarithmic coordinate graph points based on a random sampling consistency algorithm, lays a foundation for the accurate calculation of the fractal dimension value of the three-dimensional point cloud of the plant outline, and distinguishes the categories of the ginkgo tree, the Chinese arborvitae and the Chinese arborvitae by using the Minkowski dimension method, the specific embodiments can show that the fractal dimension values of the point cloud of the ginkgo tree with leaves, the Chinese arborvitae and the Chinese arborvitae are obviously distributed in different range intervals, prove that the tree species can be distinguished by using the fractal dimension value, so that the corresponding fractal dimension value of each tree can be found out through experiments or experiences, the plant species are distinguished by the aid of the plant three-dimensional point cloud contour dimension dividing values, the plants are identified and classified, the dimension dividing values among different plants are distinguished obviously, the plant contour dimension dividing values are effectively and feasibly obtained and calculated by the aid of the three-dimensional laser point cloud, and the plant identification and classification method based on the three-dimensional point cloud contour dimension dividing values is practical, accurate and efficient.
Claims (9)
1. A plant identification and classification method based on three-dimensional point cloud contour dimension values is characterized by comprising the following steps: comprises the steps of calculating the fractal dimension value of a three-dimensional point cloud of a plant outline and classifying the plant based on the fractal feature of the point cloud,
calculating the fractal dimension value of the plant contour three-dimensional point cloud, calculating the fractal dimension value of the plant contour point cloud data by a Minkowski dimension method, firstly carrying out multi-scale cubic box division on the spatial range of the preprocessed point cloud data, counting the number of non-empty cubic boxes corresponding to the division scale, then using the logarithm value of the statistical result to make a dual-logarithmic coordinate graph, finally obtaining the slope of a fitting straight line through points in the dual-logarithmic coordinate graph, namely the fractal dimension value, calculating the fractal dimension value of the plant contour three-dimensional point cloud, obtaining a plant contour point cloud A by preprocessing the three-dimensional point cloud, calculating the fractal dimension value by the Minkowski dimension method to obtain a dual-logarithmic coordinate graph point set B, and solving the fractal dimension value by fitting the straight line by a least square method;
plant classification based on point cloud fractal features utilizes a Minkowski dimensionality method to calculate fractal values of plant contour point clouds, finds out the fractal value corresponding to each tree through experiments or experiences, distinguishes plant species by utilizing the fractal values of the plant three-dimensional point cloud contours, and identifies and classifies the plants.
2. The plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value as claimed in claim 1, wherein: in the plant contour point cloud A obtained by three-dimensional point cloud pretreatment, plant contour data is collected by a three-dimensional laser scanner, and the collected multi-station data is spliced, cut and denoised,
the method for acquiring the complete spatial three-dimensional coordinate data of the plant by using the ground three-dimensional laser scanner comprises the following steps: carrying out multi-station laser point cloud measurement on the plant, carrying out data splicing on the three-dimensional point cloud after error correction, cutting the spliced point cloud data according to the plant range, only keeping the three-dimensional point cloud of the main outline of the plant, removing the three-dimensional point cloud of other irrelevant objects, removing repeated points in the data after obtaining the plant three-dimensional point cloud, and finally obtaining the three-dimensional point cloud A of the single plant.
3. The plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value as claimed in claim 1, wherein: in the set of bi-logarithmic coordinate points B computed by the minkowski dimensionality method,
cubic box division is carried out on the plant three-dimensional point cloud when the plant fractal dimension value is calculated through a Minkowski dimension method, firstly, the external minimum rectangular parameter of the plant three-dimensional point cloud is obtained, and the space range of the plant three-dimensional point cloud and the starting point coordinate (x) of the cubic box division are determinedstart，ystart，zstart) Then setting the initial side length c of the cubic box, carrying out multi-scale cubic box division on the space of the plant, dividing the space range of the plant point cloud by the cubic boxes with different side lengths, carrying out iteration in the whole space division process, determining the iteration number D of the space division, wherein in order to enable enough points in the dual-logarithm coordinate graph point set B, the iteration number D is determined by the following formula,
cube box side length E at each iterationiC × i (i ═ 1, 2, 3.., D), where E isminThe side length of the shortest side of the minimum rectangle externally connected with the plant point cloud, c is the initial side length of the cubic box, i is the current cycle number, EiDetermining the iteration number by taking an integer not greater than the quotient of half of the shortest side length and the initial side length when the side length of the cubic box in the ith cycle is equal to the length of the original side,
at each iteration, first by cube box dimension EiDividing the space where the plant point cloud is located into cubic boxes, judging the cubic boxes containing the plant three-dimensional point cloud in the cubic boxes, and counting the number of the non-empty cubic boxes to be recorded as FEi(A) Simultaneous recording of cube size EiReciprocal 1/E ofiThe logarithmic value of the two forms a point in the point set B of the log-log coordinate graph
And D times of cube box division are completed, and a final double-logarithm coordinate graph point set B is obtained.
4. The plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value as claimed in claim 1, wherein: in the least squares method to fit a straight line to solve the fractal dimension values,
performing least squares method fitting straight line on the point data in the point set B of the obtained final double logarithmic coordinate graph, wherein
Is a dependent variable, ln (1/E)i) Is independent variable, the slope corresponding to the straight line obtained by fitting is the solved fractal dimension value,
the equation of a straight line, y, kx + b, represents all straight lines except the parallel y-axis, and the equation of y, kx + b can be used as an equation for fitting a straight line,
wherein y denotes the ordinate, i.e. the dependent variable
X represents the abscissa, i.e. the argument ln (1/E)i) K represents the slope of the straight line, B represents the intercept of the straight line, the straight line is fitted by using a least squares method, wherein k and B are parameters to be estimated, y and x are observed values, namely point coordinates in a point set B of a log-log coordinate graph, and k and B are taken as approximate values plus errors, so that
For any point in the point set B of the log-log coordinate graph
The error equation is:
b has D points in total, so D error equations are in total, the overall error equation is as follows,
V＝GΔ-e
wherein the parameter meanings are as follows:
V＝[v1v2v3L vD]T
Δ＝[δkδb]T
for equal weight observation, the formula Δ ═ δ according to the least squares principlekδb]TIn (1) should satisfy VTThe minimum V is required, the points in the point set of the double logarithmic coordinates are independently observed, and are obtained according to a method of a function natural extreme value, as shown in the following formula,
finally, the obtained delta is shown as the following formula,
Δ＝(GTG)-1GTe
according to
And solving the slope k of the straight line and the intercept b of the straight line, wherein the slope k corresponding to the straight line obtained by fitting is the solved fractal dimension value.
5. The plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value as claimed in claim 1, wherein: in the cubic box scale division parameter setting, the side length of a cubic box influences the fitting result of a final straight line, along with the increase of the side length of an initial cubic box, when the side length of the cubic box is small, the distance between a corresponding point in a dual-logarithmic coordinate graph and the straight line finally fitted by a least square method is gradually reduced, the appropriate increase of the initial side length of the cubic box can reduce the influence of a point in the corresponding dual-logarithmic coordinate graph on a least square fitting straight line when the side length of the cubic box is small, the influence of a deviation point in a dual-logarithmic coordinate graph point set on the straight line fitted by the least square is reduced, the clockwise rotation trend of the fitted straight line can be reduced, therefore, the slope of the finally fitted straight line is increased along with the appropriate increase of the initial side length of the cubic box, namely, the calculated plant score value is increased.
6. The plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value as claimed in claim 1, wherein: the point with the same longitudinal coordinate in the point set of the double logarithmic coordinates is removed by adding a screening step of points when the point set of the plant contour point cloud double logarithmic coordinates is obtained by a Minkowski dimension method, for the points to be added into the point set of the double logarithmic coordinates, judging whether the points with the same number of non-empty cubic boxes as the points to be added exist in the point set or not, judging whether the points with the non-empty cubic boxes with the longitudinal coordinate as the points to be added exist in the point set of the double logarithmic coordinates or not exists in the point set of the double logarithmic coordinates or not if the points do not exist in the point set of the double logarithmic coordinates, the points to be added are added into the point set of the double logarithms, otherwise, the points are not added into the point set.
7. The method for identifying and classifying plants based on the three-dimensional point cloud contour fractal dimension value as claimed in claim 6, wherein: in the process of removing the deviation points by the random sampling consistency algorithm, an input item of the random sampling consistency algorithm comprises a group of observation data sets, namely a sample data set A, and a parameterized model B suitable for observation data, the number of necessary parameters for constructing the model B is n, and the number of samples in the sample set A is more than n. The algorithm randomly selects a group of subsets A with the number of samples n from a sample data set A through cyclic iterationsubCalculating n parameters of model B as correct sample data to obtain model BiRemoving the subset A from the sample set AsubThe other points of (2) judge whether they conform to the model B or not according to the allowable deviation threshold value HiIf the sample number in the internal point set is larger than the sample number threshold h of the internal point set, the model is one of robust models, the sample data in all the internal point sets are used for estimating model parameters, the model is compared with the current optimal model, the optimal model is reserved, the steps are repeated until the iteration times specified by the algorithm are reached, and the parameters corresponding to the optimal model and the corresponding internal point set are obtained.
8. The method for identifying and classifying plants based on the three-dimensional point cloud contour fractal dimension value as claimed in claim 7, wherein: calculating all possible linear models in the point set of the log-log coordinate graph aiming at the iteration times when the random sampling consistency algorithm is executed, wherein the loop iteration time N is shown as the following formula, wherein M represents the number of points in the point set of the log-log graph,
the points in the point set of the log-log coordinate graph are sequentially obtained along with the length of the side of the cubic box in the process of dividing the cubic box space of the plant contour point cloud, when two points are randomly selected from the point set of the log-log coordinate graph, the first point added into the set of the log-log coordinate graph is taken as the first point for obtaining the linear parameter, and then the points added into the set of the log-log coordinate graph are sequentially taken as the points for obtaining the linear parameter.
9. The plant identification and classification method based on the three-dimensional point cloud contour fractal dimension value as claimed in claim 1, wherein: in the plant classification based on the point cloud fractal characteristics, point cloud contour fractal dimension values of different plants are obviously different, the corresponding fractal dimension value of each tree can be found out through experiments or experiences, plant types are distinguished by utilizing the plant three-dimensional point cloud contour fractal dimension values, and single plants are efficiently identified and classified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911040565.1A CN110765962A (en) | 2019-10-29 | 2019-10-29 | Plant identification and classification method based on three-dimensional point cloud contour dimension values |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911040565.1A CN110765962A (en) | 2019-10-29 | 2019-10-29 | Plant identification and classification method based on three-dimensional point cloud contour dimension values |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110765962A true CN110765962A (en) | 2020-02-07 |
Family
ID=69334528
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911040565.1A Withdrawn CN110765962A (en) | 2019-10-29 | 2019-10-29 | Plant identification and classification method based on three-dimensional point cloud contour dimension values |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110765962A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111444927A (en) * | 2020-03-26 | 2020-07-24 | 广州市炜城智能科技有限公司 | Method for identifying animal and plant pests by using laser technology in multidimensional manner |
| CN111898684A (en) * | 2020-07-31 | 2020-11-06 | 陈艳 | Biological species identification method based on multi-dimensional point cloud data |
| CN112991424A (en) * | 2021-04-07 | 2021-06-18 | 重庆大学 | Fractal dimension calculation method and device based on octree algorithm |
| CN113096027A (en) * | 2021-03-02 | 2021-07-09 | 吉林农业大学 | Farmland soil layer horizontal correction and removal method based on point cloud |
| CN113344956A (en) * | 2021-06-21 | 2021-09-03 | 深圳市武测空间信息有限公司 | Ground feature contour extraction and classification method based on unmanned aerial vehicle aerial photography three-dimensional modeling |
| CN113610017A (en) * | 2021-08-11 | 2021-11-05 | 吉林农业大学 | Antler cap type identification method based on mid-infrared spectrum and SVM |
| CN115423835A (en) * | 2022-11-02 | 2022-12-02 | 中汽创智科技有限公司 | Rod-shaped object point cloud data processing method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100322477A1 (en) * | 2009-06-04 | 2010-12-23 | Peter Schmitt | Device and method for detecting a plant |
| CN105574929A (en) * | 2015-12-15 | 2016-05-11 | 电子科技大学 | Single vegetation three-dimensional modeling method based on ground LiDAR point cloud data |
| CN106651900A (en) * | 2016-12-27 | 2017-05-10 | 中国农业大学 | Three-dimensional modeling method of elevated in-situ strawberry based on contour segmentation |
| CN109409429A (en) * | 2018-10-28 | 2019-03-01 | 南京林业大学 | A kind of tree species classification method based on laser radar point cloud data |
| CN109801332A (en) * | 2018-12-29 | 2019-05-24 | 湖北大学 | Determine the method and apparatus of Minkowski dimension |
-
2019
- 2019-10-29 CN CN201911040565.1A patent/CN110765962A/en not_active Withdrawn
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100322477A1 (en) * | 2009-06-04 | 2010-12-23 | Peter Schmitt | Device and method for detecting a plant |
| CN105574929A (en) * | 2015-12-15 | 2016-05-11 | 电子科技大学 | Single vegetation three-dimensional modeling method based on ground LiDAR point cloud data |
| CN106651900A (en) * | 2016-12-27 | 2017-05-10 | 中国农业大学 | Three-dimensional modeling method of elevated in-situ strawberry based on contour segmentation |
| CN109409429A (en) * | 2018-10-28 | 2019-03-01 | 南京林业大学 | A kind of tree species classification method based on laser radar point cloud data |
| CN109801332A (en) * | 2018-12-29 | 2019-05-24 | 湖北大学 | Determine the method and apparatus of Minkowski dimension |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111444927A (en) * | 2020-03-26 | 2020-07-24 | 广州市炜城智能科技有限公司 | Method for identifying animal and plant pests by using laser technology in multidimensional manner |
| CN111444927B (en) * | 2020-03-26 | 2023-05-26 | 广州市炜城智能科技有限公司 | Method for identifying animal and plant pests by using multi-dimensional morphology of laser technology |
| CN111898684A (en) * | 2020-07-31 | 2020-11-06 | 陈艳 | Biological species identification method based on multi-dimensional point cloud data |
| CN113096027A (en) * | 2021-03-02 | 2021-07-09 | 吉林农业大学 | Farmland soil layer horizontal correction and removal method based on point cloud |
| CN113096027B (en) * | 2021-03-02 | 2024-01-23 | 吉林农业大学 | Point cloud-based farmland soil layer horizontal correction and removal method |
| CN112991424A (en) * | 2021-04-07 | 2021-06-18 | 重庆大学 | Fractal dimension calculation method and device based on octree algorithm |
| CN112991424B (en) * | 2021-04-07 | 2024-02-13 | 重庆大学 | Fractal dimension calculation method and device based on octree algorithm |
| CN113344956A (en) * | 2021-06-21 | 2021-09-03 | 深圳市武测空间信息有限公司 | Ground feature contour extraction and classification method based on unmanned aerial vehicle aerial photography three-dimensional modeling |
| CN113344956B (en) * | 2021-06-21 | 2022-02-01 | 深圳市武测空间信息有限公司 | Ground feature contour extraction and classification method based on unmanned aerial vehicle aerial photography three-dimensional modeling |
| CN113610017A (en) * | 2021-08-11 | 2021-11-05 | 吉林农业大学 | Antler cap type identification method based on mid-infrared spectrum and SVM |
| CN113610017B (en) * | 2021-08-11 | 2024-02-02 | 吉林农业大学 | Deer horn cap type identification method based on mid-infrared spectrum and SVM |
| CN115423835A (en) * | 2022-11-02 | 2022-12-02 | 中汽创智科技有限公司 | Rod-shaped object point cloud data processing method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110765962A (en) | Plant identification and classification method based on three-dimensional point cloud contour dimension values | |
| CN106815847B (en) | Trees dividing method and single tree extracting method based on laser radar point cloud | |
| CN112381861B (en) | Forest land point cloud data registration and segmentation method based on foundation laser radar | |
| Lefsky et al. | Volume estimates of trees with complex architecture from terrestrial laser scanning | |
| CN111340826B (en) | Aerial image single tree crown segmentation algorithm based on super pixels and topological features | |
| CN111832506B (en) | Remote sensing discrimination method for rebuilding vegetation based on long time sequence vegetation index | |
| WO2014054042A1 (en) | Device and method for detecting plantation rows | |
| CN113269825B (en) | Forest breast diameter value extraction method based on foundation laser radar technology | |
| CN111950589B (en) | Point cloud region growing optimization segmentation method combined with K-means clustering | |
| CN115690081A (en) | Tree counting method, system, storage medium, computer equipment and terminal | |
| CN116704333B (en) | Single tree detection method based on laser point cloud data | |
| Liu et al. | Development of a machine vision algorithm for recognition of peach fruit in a natural scene | |
| 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 | |
| CN111060922B (en) | Tree point cloud extraction method based on airborne laser radar point cloud spatial distribution characteristics | |
| CN116433846A (en) | Forest land cloud branch and leaf separation method based on size and local chaotic distance | |
| CN111696122A (en) | Crop phenotype parameter extraction method and device | |
| CN116563714A (en) | Method and system for automatically distinguishing growth stage of rice | |
| CN116805413A (en) | Automatic calculation method for oil tea seedling stage phenotype characteristics based on three-dimensional point cloud | |
| CN115187979A (en) | Single-wood point cloud branch and leaf separation method based on graph theory | |
| CN114612787B (en) | Urban green space deep learning extraction method supported by scale change strategy | |
| CN112381029B (en) | Method for extracting airborne LiDAR data building based on Euclidean distance | |
| CN116416523A (en) | Machine learning-based rice growth stage identification system and method | |
| CN111768101B (en) | Remote sensing cultivated land change detection method and system taking account of physical characteristics | |
| CN115410036A (en) | Automatic classification method for key element laser point clouds of high-voltage overhead transmission line |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200207 |
|
| WW01 | Invention patent application withdrawn after publication |