CN113989352A - Blade point cloud framework extraction method and device - Google Patents
Blade point cloud framework extraction method and device Download PDFInfo
- Publication number
- CN113989352A CN113989352A CN202111088873.9A CN202111088873A CN113989352A CN 113989352 A CN113989352 A CN 113989352A CN 202111088873 A CN202111088873 A CN 202111088873A CN 113989352 A CN113989352 A CN 113989352A
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- blade
- skeleton
- relevant
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000605 extraction Methods 0.000 title claims description 32
- 238000000034 method Methods 0.000 claims abstract description 68
- 238000004422 calculation algorithm Methods 0.000 claims description 33
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 238000003064 k means clustering Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 230000003044 adaptive effect Effects 0.000 claims description 7
- 238000012847 principal component analysis method Methods 0.000 claims description 7
- 239000013598 vector Substances 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 5
- 238000007635 classification algorithm Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000000452 restraining effect Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 6
- 238000004458 analytical method Methods 0.000 abstract description 4
- 230000000875 corresponding effect Effects 0.000 description 24
- 241000196324 Embryophyta Species 0.000 description 15
- 238000000513 principal component analysis Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 11
- 210000003462 vein Anatomy 0.000 description 6
- 241001057636 Dracaena deremensis Species 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 240000008042 Zea mays Species 0.000 description 4
- 235000002017 Zea mays subsp mays Nutrition 0.000 description 4
- 230000007547 defect Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 235000005824 Zea mays ssp. parviglumis Nutrition 0.000 description 3
- 235000005822 corn Nutrition 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 240000006394 Sorghum bicolor Species 0.000 description 1
- 235000011684 Sorghum saccharatum Nutrition 0.000 description 1
- 235000021307 Triticum Nutrition 0.000 description 1
- 244000098338 Triticum aestivum Species 0.000 description 1
- 235000016383 Zea mays subsp huehuetenangensis Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 235000009973 maize Nutrition 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 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
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method and a device for extracting a leaf point cloud framework, which comprise the following steps: acquiring normal information of each point in the initial point cloud of the blade, clustering the initial point cloud of the blade based on the normal information, and acquiring a plurality of point cloud subsets; establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the extension direction of the blade according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located; determining the skeleton constraint points of each relevant point set by using a self-adaptive weighting operator to construct a skeleton constraint point set; and acquiring a main curve of the skeleton constraint point set to be used as a blade point cloud skeleton. The method and the device for extracting the blade point cloud framework can accurately extract the blade framework from the blade point cloud data, can better reflect the three-dimensional shape structure of the blade, have better robustness on the missing point cloud data, do not need manual adjustment in the later period, and can provide technical support for phenotypic big data processing, automatic phenotypic analysis and the like.
Description
Technology neighborhood
The invention relates to the field of three-dimensional image processing technology, in particular to a method and a device for extracting a leaf point cloud framework.
Background
The leaves are important components of plants, and the difference of the shape and the structure directly influences the growth and the final yield of the plants. The skeleton is an abstract expression form of three-dimensional data and a model, and can intuitively express the topological connection relation and the geometric structure of an object. How to extract a skeleton with semantic information from disordered plant point cloud data becomes a difficult problem of plant three-dimensional phenotype and three-dimensional reconstruction research.
The classical three-dimensional point cloud framework extraction method mainly comprises an extraction algorithm based on a Laplace matrix and an L1 median framework. Because plants have specific shape and structure characteristics, the method for extracting the skeleton by directly adopting the skeleton extraction method is difficult to obtain an ideal result, and the method needs to be improved and perfected on the basis of the original method by combining the shape and structure characteristics of the plants.
Due to the fact that a large number of shelters exist among organs of the plant and the material properties of the surfaces of plant leaves are complex, the collected plant three-dimensional point cloud data usually have defects. In addition, plant leaves often have specific shape and structural characteristics, depending on such factors as variety and environment. Therefore, the accurate extraction result of the point cloud framework of the plant leaf is difficult to obtain by directly applying the framework extraction algorithm generalized by the computer graphics neighborhood.
Disclosure of Invention
The invention provides a leaf point cloud framework extraction method and device, which are used for solving the defect that an ideal plant leaf point cloud framework extraction result is difficult to obtain in the prior art and realizing the robust extraction of a corn leaf framework lacking point cloud data.
In a first aspect, the invention provides a method for extracting a leaf point cloud skeleton, which comprises the following steps:
acquiring normal information of each point in the initial point cloud of the blade, clustering the initial point cloud of the blade based on the normal information, and acquiring a plurality of point cloud subsets;
establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located;
determining a skeleton constraint point corresponding to each relevant point set by using a self-adaptive weighting operator to construct a skeleton constraint point set;
and acquiring a main curve of the skeleton constraint point set to serve as a blade point cloud skeleton.
According to the method for extracting the blade point cloud framework, the establishment of the distance field of each point cloud subset comprises the following steps:
determining a directed bounding box for each of the point cloud subsets;
distances between each point in the point cloud subsets and an orthogonal plane perpendicular to the blade elongation direction are calculated and normalized, thereby establishing a distance field for each of the point cloud subsets.
According to the method for extracting the blade point cloud framework, the directional bounding box of each point cloud subset is determined, and the method comprises the following steps:
calculating three characteristic vectors of each point cloud subset based on a principal component analysis method;
performing standard orthogonalization on the three eigenvectors to obtain an orthogonal matrix;
constructing a Cartesian coordinate system by using the orthonormal basis of the orthogonal matrix, and calculating the maximum position and the minimum position in three directions X, Y, Z from the origin in the Cartesian coordinate system to construct 6 planes of the directional bounding box according to all the maximum positions and the minimum positions.
According to the method for extracting the leaf point cloud framework, provided by the invention, the normal information of each point in the leaf initial point cloud is obtained, and the leaf initial point cloud is clustered based on the normal information to obtain a plurality of point cloud subsets, and the method comprises the following steps:
denoising and downsampling the initial point cloud of the blade to remove noise points and obtain a blade point cloud;
determining a neighborhood of each point in the blade point cloud based on a K nearest neighbor classification algorithm;
determining the normal direction of each point according to the neighborhood of each point based on a principal component analysis method;
based on a minimum spanning tree algorithm, adjusting the normal direction of each point in the blade point cloud, and acquiring the normal direction information of each point;
based on a K-means clustering algorithm, clustering the leaf point cloud according to the normal information of each point to obtain a plurality of point cloud subsets.
According to the method for extracting the leaf point cloud framework, the self-adaptive weighting operator comprises a spatial distance weight, a normal difference weight and a point cloud integrity weight;
the determining the skeleton constraint point corresponding to each relevant point set by using the adaptive weighting operator includes:
according to the spatial distribution characteristics of the relevant point set of each section, the central point of the relevant point set of each section is calculated in a self-adaptive mode;
and moving the central point according to the missing state of the point cloud data in the relevant point set of each section to obtain the corresponding skeleton constraint point of each relevant point set.
According to the method for extracting the blade point cloud framework, the expression of the spatial distance weight is as follows:
the expression of the normal difference weight is:
the expression of the point cloud integrity weight is as follows:
wherein p isiIs a set of relevant points S of the cross sectionkPoint i of (1); q. q.sjIs piK neighbor set of points N (p)i) J-th point in (1); alpha is an influence parameter on the spatial distance; w'k,iIs a spatial distance weight;
is the leaf surface orientation at the kth section; n ispiIs a set of relevant points S of the cross sectionkMidpoint piNormal direction of (2); beta is an influencing parameter on the normal difference; w ″)k,iIs the normal difference weight; lmRepresenting the length of the mth side of a polygon formed by a concave shell, wherein the concave shell is generated by projecting a relevant point set of a cross section onto an orthogonal surface which is perpendicular to the extending direction of the blade; r is the ratio of the longest side of the concave shell to the total side length; epsilon is a threshold value for judging point cloud data missing.
According to the blade point cloud framework extraction method provided by the invention, the central point is moved according to the missing state of the point cloud data in the relevant point set of each section so as to obtain the moved central point in the corresponding framework constraint point of each relevant point set, and the determination formula is as follows:
wherein, ckIs a set of relevant points S of the cross sectionkCorresponding skeleton constraint points; e.g. of the typekIs the direction in which the center point is moved according to the point cloud integrity in the set of related points.
According to the method for extracting the blade point cloud framework, which is provided by the invention, the main curve of the framework constraint point set is obtained, and the method comprises the following steps:
step 1, restraining a skeleton to a point set CmK neighbor distance calculation is carried out on all the constraint points in the Voronoi region, outliers are removed by utilizing a preset distance threshold value, and a Voronoi region V is constructed1,…,Vm；
Step 2, based on principal component analysis method, taking the first principal component as direction, and respectively taking length from the centroid to positive and negative directionsIs t times VmA line segment of standard deviation length as the first principal component line segment LmAnd calculating said CmTo said LmThe projected distance of (a);
step 3, obtaining the CmRemote point V infAnd the remote point V is adjustedfTo said CmIn (3), generating a new skeleton constraint point set Cm+1(ii) a The deflection point VfIs a set of distance principal component line segments L1,…,Lm+1-the farthest point;
step 4, in the step Cm+1In calculating a new first principal component line segment Lm+1And combining said Lm+1Add to the set of principal component line segments { L }1,…,Lm+1In the method, a new principal component line segment set { L } is obtained1,…,Lm+1}；
Step 5, enabling m +1, and iteratively executing the step 1 to the step 4 until the number of the line segments in the new principal component line segment set reaches a preset number threshold value or a preset energy function value is minimum;
and 6, constructing a Hamilton path by using the new principal component line segment set obtained in the step 5 based on a greedy algorithm, and optimizing the Hamilton path based on a 2-opt optimization algorithm to obtain the master curve.
In a second aspect, the present invention further provides a device for extracting a leaf point cloud skeleton, including:
the blade point cloud grouping unit is used for acquiring normal information of each point in the blade initial point cloud, clustering the blade initial point cloud based on the normal information and acquiring a plurality of point cloud subsets;
the section point set dividing unit is used for establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located;
the constraint point set generating unit is used for determining the skeleton constraint points corresponding to each relevant point set by using a self-adaptive weighting operator so as to construct a skeleton constraint point set;
and the point cloud framework generation unit is used for acquiring a main curve of the framework constraint point set to serve as a blade point cloud framework.
In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the steps of the blade point cloud skeleton extraction method as described in any one of the above.
In a fourth aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the blade point cloud skeleton extraction method as described in any one of the above.
According to the method and the device for extracting the leaf point cloud framework, provided by the invention, the self-adaptive weighting operator is designed to calculate the framework constraint point set of the leaf point cloud, and the main curve is introduced to optimize the framework constraint point set, so that the three-dimensional shape structure of a leaf can be better reflected, the missing point cloud data has better robustness, manual adjustment in the later period is not needed, and technical support can be provided for phenotype big data processing, automatic phenotype analysis and the like.
Drawings
In order to more clearly illustrate the technical solutions of the present invention or the prior art, the drawings needed for the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for extracting a leaf point cloud skeleton according to the present invention;
FIG. 2 is a schematic diagram of a subset of point clouds generated after clustering initial point clouds of leaves;
FIG. 3 is a schematic diagram of skeleton extraction under ideal conditions;
FIG. 4 is a schematic diagram of the distance field distribution of the multi-point cloud set provided by the present invention;
FIG. 5 is a second schematic flow chart of the method for extracting a leaf point cloud skeleton according to the present invention;
FIG. 6 is a schematic structural diagram of a leaf point cloud skeleton extraction device provided by the present invention;
fig. 7 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present invention without any creative work belong to the protection scope of the present invention.
It should be noted that in the description of the embodiments of the present invention, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element. The terms "upper", "lower", and the like, indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only for convenience in describing the present invention and simplifying the description, but do not indicate or imply that the referred devices or elements must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention. Unless expressly stated or limited otherwise, the terms "mounted," "connected," and "connected" are intended to be inclusive and mean, for example, that they may be fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art as the case may be.
The terms first, second and the like in the description and in the claims of the present application are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that embodiments of the application may be practiced in sequences other than those illustrated or described herein, and that the terms "first," "second," and the like are generally used herein in a generic sense and do not limit the number of terms, e.g., the first term can be one or more than one. In addition, "and/or" in the specification and claims means at least one of connected objects, a character "/" generally means that a preceding and succeeding related objects are in an "or" relationship.
The following describes a method and an apparatus for extracting a blade point cloud skeleton according to an embodiment of the present invention with reference to fig. 1 to 6.
Fig. 1 is a schematic flow diagram of a method for extracting a blade point cloud skeleton according to the present invention, as shown in fig. 1, including but not limited to the following steps:
step 101: acquiring normal information of each point in the initial point cloud of the blade, clustering the initial point cloud of the blade based on the normal information, and acquiring a plurality of point cloud subsets;
step 102: establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located;
step 103: determining a skeleton constraint point corresponding to each relevant point set by using a self-adaptive weighting operator to construct a skeleton constraint point set;
step 104: and acquiring a main curve of the skeleton constraint point set to serve as a blade point cloud skeleton.
As an optional embodiment, the method takes the example of processing the cloud data of the corn plant point in the silking period by adopting a crop single plant high-flux phenotype platform MVS-Pheno platform.
After the corn plant point cloud data is obtained, the corn plant point cloud data is segmented by using point cloud labeling software Label3 Dmean to obtain leaf initial point cloud.
The MVS-Pheno platform is the corn plant point cloud data obtained by utilizing a multi-view imaging principle, and is influenced by factors such as an acquired light environment, blade shielding and the like, so that the obtained corn plant point cloud data inevitably has a small amount of loss.
The existing three-dimensional point cloud framework extraction method is limited by the requirement of point cloud precision, and an ideal blade point cloud framework cannot be obtained. Therefore, the invention provides a novel method for extracting a blade point cloud framework.
In step 101, obtaining normal information of each point in the initial point cloud of the blade, and clustering the initial point cloud of the blade based on the normal information, the following steps may be adopted:
firstly, determining a neighborhood of each point in the initial point cloud of the blade based on a K-Nearest Neighbor classification algorithm (KNN); then, based on Principal Component Analysis (PCA), determining the normal direction of each point according to the neighborhood of each point; further adjusting the normal direction of each point in the initial point cloud of the blade based on a minimum spanning tree algorithm, and acquiring the normal direction information of each point; and finally, clustering the blade point cloud according to the normal information of each point based on a K-means clustering algorithm, so as to obtain a plurality of point cloud subsets.
Fig. 2 is a schematic diagram of clustering the initial point clouds of the leaves to generate point cloud subsets, and as shown in fig. 2, for the initial point cloud of the first leaf (or the initial point cloud of the second leaf), the initial point cloud of the first leaf (or the initial point cloud of the second leaf) can be divided into a first point cloud subset combination (or a second point cloud subset combination) composed of three point cloud subsets in a K-means clustering manner, generally, one initial point cloud of the leaf is divided into 3-4 point cloud subsets, so that the shape change of the region where each point cloud subset is located can be relatively smooth.
Fig. 3 is a schematic diagram of skeleton extraction in an ideal case, and as shown in fig. 3, the skeleton of the three-dimensional object can be determined by calculating the center point of a curve obtained by intersecting a cross section perpendicular to the skeleton direction with the three-dimensional object. However, in the euclidean space, the intersection curve of the cross section perpendicular to the blade extension direction (skeleton direction) and the three-dimensional object is not easy to be explicitly calculated.
Wherein the cross sections are determined according to a preset step length, and each cross section is perpendicular to the extending direction of the blade.
In order to solve the problem, in step 102 of the method for extracting the blade point cloud skeleton provided by the invention, a curve obtained by intersecting a cross section and a three-dimensional object is replaced by acquiring a related point set of an interface perpendicular to the extending direction of the blade, so that the skeleton constraint point is determined according to the determined related point set.
As an alternative embodiment, the computation of the set of relevant points for the cross-section is performed in step 102 primarily by distance field methods.
Fig. 4 is a schematic diagram of the distribution of the distance field of the multi-point cloud subset provided by the present invention, as shown in fig. 4, a distance field is generated for each point cloud subset, and a related point set corresponding to each cross section can be extracted from the initial point cloud of the leaf by the distance field, and each related point set can be regarded as a set of orthogonal surfaces (i.e., cross sections) perpendicular to the elongation direction of the leaf.
It should be noted that, as shown in fig. 4, the entire blade initial point cloud is divided into a plurality of relatively gentle point cloud subsets, so that the orthogonal plane perpendicular to the blade extending direction can be regarded as one end of the directional bounding box perpendicular to the blade extending direction where the orthogonal plane is located.
Further, in order to overcome influences of uneven sampling of the initial point cloud of the leaf, data missing and the like, in step 103, the framework constraint point of the relevant point set of each section is calculated by designing a set of adaptive weighting operators. For example, the center of each section related point set can be calculated according to the spatial distribution characteristics of the blade initial point cloud; and carrying out appropriate movement on the calculated central point according to the missing condition of the point cloud data, so that the finally obtained skeleton constraint point is closer to the vein point in the middle of the leaf.
Finally, in step 104, according to the skeleton constrained point set constructed by all the skeleton constrained points, a main curve of the skeleton constrained point set with the leaf is calculated, and then a skeleton curve closer to the vein position is obtained to be used as a leaf point cloud skeleton.
According to the leaf point cloud framework extraction method provided by the invention, a self-adaptive weighting operator is designed to calculate the framework constraint point set of the leaf point cloud, and a principal curve is introduced to optimize the framework constraint point set, so that the three-dimensional shape structure of the leaf can be better reflected, the missing point cloud data has better robustness, manual adjustment in the later period is not needed, and technical support can be provided for phenotype big data processing, automatic phenotype analysis and the like.
Fig. 5 is a second schematic flowchart of the method for extracting a leaf point cloud skeleton according to the present invention, and as an alternative embodiment, as shown in fig. 5, the establishing a distance field of each point cloud subset includes:
determining a directed bounding box for each of the point cloud subsets;
distances between each point in the point cloud subsets and an orthogonal plane perpendicular to the blade elongation direction are calculated and normalized, thereby establishing a distance field for each of the point cloud subsets.
In general, the invention employs an improved distance field method for computing a set of correlated points for each cross-section. The distance field is constructed by calculating the distance between the point in each cloud subset and the directional bounding box, and the invention adopts the method to determine the relevant point set corresponding to each section, thereby greatly reducing the calculation cost compared with the prior method.
As shown in fig. 5, the above method for determining the directional bounding box of each point cloud subset may include, but is not limited to, the following steps:
calculating three characteristic vectors of each point cloud subset based on a PCA algorithm;
performing standard orthogonalization on the three eigenvectors to obtain an orthogonal matrix;
constructing a Cartesian coordinate system by using the orthonormal basis of the orthogonal matrix, and calculating the maximum position and the minimum position in three directions X, Y, Z from the origin in the Cartesian coordinate system to construct 6 planes of the directional bounding box according to all the maximum positions and the minimum positions.
Specifically, for a certain point cloud subset P ═ P with relatively gentle changes of original blade correlation obtained by K-means clusteringi(ii) a After i is 1,2, …, n, three feature vectors epsilon corresponding to the point cloud data in the point cloud subset are calculated by PCA algorithm1、ε2、ε3。
Wherein, the PCA algorithm transforms point cloud data by converting each point into three eigenvectors epsilon1、ε2、ε3Formed into a new X, Y, Z coordinate system, and the origin of the new coordinate system and the global origin are the same, i.e. three eigenvectors epsilon are returned1、ε2、ε3And vectors of an X axis, a Y axis and a Z axis of the plane self coordinate system in the point cloud global coordinate system are respectively.
Further, three feature vectors ε1、ε2、ε3Orthogonal matrix Q ═ epsilon 'is formed by standard orthogonalization'1，ε′2，ε′3]。
Then, with an orthonormal base of epsilon'1、ε′2、ε′3And constructing a new Cartesian coordinate system, determining 6 planes in the Cartesian coordinate system according to the maximum position and the minimum position respectively determined on each axis, and combining and rescuing the 6 planes to obtain the directional bounding box to be constructed.
After determining the directional bounding box of each point cloud subset, calculating each point P in P for any point cloud subset PiNormalizing the distance to the orthogonal plane perpendicular to the blade extension direction to obtain a distance field D of the partial point cloud dataiThe calculation formula is as formula 1:
pi∈P；i＝1,2,…,n；
wherein D isiRepresents a point piNormalized distance in the distance field; σ represents an orthogonal plane perpendicular to the blade extension direction; d (p)iσ) denotes piAt a distance from the orthogonal plane sigma.
Further, a set of correlated points corresponding to each cross-section can be extracted from each subset of clouds by the determined distance field.
Based on the content of the foregoing embodiment, as an optional embodiment, as shown in fig. 5, the above obtaining normal information of each point in the leaf initial point cloud, and clustering the leaf initial point cloud based on the normal information to obtain a plurality of point cloud subsets mainly includes, but is not limited to, the following steps:
denoising and downsampling the initial point cloud of the blade to remove noise points and obtain a blade point cloud;
determining a neighborhood of each point in the blade point cloud based on a KNN algorithm;
based on a PCA algorithm, determining the normal direction of each point according to the neighborhood of each point;
based on a minimum spanning tree algorithm, adjusting the normal direction of each point in the blade point cloud, and acquiring the normal direction information of each point;
based on a K-means clustering algorithm, clustering the leaf point cloud according to the normal information of each point to obtain a plurality of point cloud subsets.
According to the method for extracting the blade point cloud framework, before the obtained initial point cloud data of the blade is processed, noise points in all the point cloud data are removed in a denoising and downsampling mode, so that the number of the points is reduced under the condition that the quality of the point cloud data is not reduced, and the subsequent calculation efficiency is improved.
Then, a neighborhood of each point is determined by using a KNN algorithm, wherein the KNN algorithm refers to that if a point in the initial point cloud data of the blade belongs to a certain category with most of K most similar (i.e. nearest) points in the feature space, the point is also assigned to the category.
The invention adopts the algorithm, and the category of the point to be classified is determined according to the category of the nearest point or points in the classification decision, so that the neighborhood of each point can be determined.
The normal to each point in the initial point cloud of blades is estimated using the PCA algorithm.
The invention adopts PCA algorithm to carry out normal estimation of each point, and can effectively reduce the dimension of the point cloud.
And finally, adjusting the normal direction of each point in the point cloud through a minimum spanning tree algorithm to meet the consistency of the normal direction in the direction, and finally obtaining the normal direction information of all the points in the point cloud data.
For the preprocessed blade point cloud data, dividing a single blade point cloud into a plurality of point cloud subset parts (generally, the blades are divided into 3-4 parts) in a K-means clustering mode, so that the shape change of the part where each point cloud subset is located is gentle.
According to the leaf point cloud framework extraction method, the sample in the whole leaf initial point cloud is clustered and divided into a plurality of point cloud subsets according to the normal information of the sample in the whole leaf initial point cloud, and then the calculation of the relevant point set of the cross section is carried out on each point cloud subset.
Based on the content of the above embodiments, as an optional embodiment, the adaptive weighting operator includes a spatial distance weight, a normal difference weight, and a point cloud integrity weight.
With reference to fig. 5, the determining, by using the adaptive weighting operator, the skeleton constraint point corresponding to each relevant point set mainly includes:
according to the spatial distribution characteristics of the relevant point set of each section, the central point of the relevant point set of each section is calculated in a self-adaptive mode;
and moving the central point according to the missing state of the point cloud data in the relevant point set of each section to obtain the corresponding skeleton constraint point of each relevant point set.
Wherein the expression of the spatial distance weight is:
the expression of the normal difference weight is:
the expression of the point cloud integrity weight is as follows:
wherein p isiIs a set of relevant points S of the cross sectionkPoint i of (1); q. q.sjIs piK neighbor set of points N (p)i) J-th point in (1); alpha is an influence parameter on the spatial distance; w'k,iIs a spatial distance weight;
is the leaf surface orientation at the kth section; n ispiIs a set of relevant points S of the cross sectionkMidpoint piNormal direction of (2); beta is an influencing parameter on the normal difference; w ″)k,iIs the normal difference weight; lmRepresenting the length of the mth side of a polygon formed by a concave shell, wherein the concave shell is generated by projecting a relevant point set of a cross section onto an orthogonal surface which is perpendicular to the extending direction of the blade; r is the ratio of the longest side of the concave shell to the total side length; epsilon is a threshold value for judging point cloud data missing.
Moving the central point according to the missing state of the point cloud data in the relevant point set of each section to obtain the moved central point in the corresponding skeleton constraint point of each relevant point set, wherein the determination formula is as follows:
wherein, ckIs a set of relevant points S of the cross sectionkCorresponding skeleton constraint points; e.g. of the typekIs the direction in which the center point is moved according to the point cloud integrity in the set of related points.
In the method for extracting the leaf point cloud framework provided by the invention, all distance fields can divide the whole leaf initial point cloud into a plurality of subsets SkEach subset SkThe relevant point set of a section can be approximated to a section point set perpendicular to the extending direction of the blade, and the specific form is as follows:
Sk＝{pi|Di∈[tk,tk+1],k＝1,2,…num，
i＝1,2,…,n}，
where num is the maximum number of sections set (which may be set to 100 in the test); t is tkIndicating the distance corresponding to the kth cross-section.
On the basis, the invention provides a framework constraint point extraction method based on a self-adaptive weighting operator, so as to effectively overcome the influences of uneven sampling of leaf point cloud data, data missing and the like.
The method provided by the invention is based on the self-adaptive weighting operator to calculate the related point set S of each sectionkSkeleton constraint point ckSubstantially, based on the spatial distribution characteristics of the initial point cloud of the blade, a set S of points associated with each section is calculated adaptivelykA center point of (a); and carrying out appropriate movement on the calculated central point according to the missing condition of the point cloud data so as to ensure that the skeleton constraint point ckThe vein point closer to the middle of the corn leaf is calculated by the formula:
wherein the content of the first and second substances,w′k,iis the spatial distance weight; w ″)k,iIs the normal difference weight; w'k,iIs the point cloud integrity weight; e.g. of the typekThe moving direction of the central point is determined according to the integrity of the point cloud, and a point set S can be adoptedkIs calculated for the third principal component direction.
1) Space distance weight w'k,iThe method is mainly used for solving the defect of extracted framework constraint point offset caused by uneven sampling of initial point cloud data of the blade, and the calculation formula is as follows:
wherein p isiIs a set of relevant points S of the cross sectionkPoint i of (1); q. q.sjIs piK neighbor set of points N (p)i) J-th point in (1); alpha is an influence parameter on the spatial distance, and can be set as the average distance of all points in the initial point cloud data of the whole blade.
2) Normal difference weight w ″k,iThe method is mainly obtained by calculating the difference between the normal direction of a current point and the orientation of the blade, and aims to enable an extracted skeleton constraint point to be located inside initial point cloud data of the blade, wherein the calculation formula is as follows:
wherein the content of the first and second substances,
the orientation of the leaf surface at the k section can be determined by the point set SkEstimating the second principal component of (a); n ispiIs a set of relevant points S of the cross sectionkMidpoint piNormal direction of (2); β is an influencing parameter on the normal difference and can be set to 1.
3) Point cloud integrity weight w'k,iThe method is mainly used for solving the problem that the extracted skeleton constraint points form interference due to the fact that the initial point cloud data of the leaves are missingTo a problem of (a).
For a complete initial blade point cloud, the weight w ″'k,iSet to 0; as the missing part increases, the point cloud integrity weight also increases accordingly.
For the missing leaf point cloud, the skeleton constraint points extracted by using the spatial distance weight and the normal difference weight only generate larger deviation at the missing part, so the point cloud integrity weight is used as compensation, and the skeleton constraint points can be moved towards the missing direction through the weight to carry out calibration.
Alternatively, the invention projects the cross-sectional correlation point set onto an orthogonal plane perpendicular to the blade elongation direction and finds its concave shell. The concave shell is used as a polygon, the side length corresponding to a large number of missing parts is longer, and the ratio r of the longest side of the concave shell to the total side length can be calculated to judge the missing condition
Wherein lmThe length of the mth side of the polygon formed by the concave shell is represented, and epsilon is a threshold value for judging the missing of the initial point cloud data, and the setting can be set to 0.1 as in the experiment.
The method for extracting the blade point cloud framework can extract the framework information of the complete blade point cloud, and especially can improve the framework extraction effect of the blade lacking in the point cloud due to the introduction of the point cloud integrity weight.
Based on the content of the foregoing embodiment, as an alternative embodiment, as shown in fig. 5, the above acquiring the main curve of the skeleton constraint point set includes:
step 1, restraining a skeleton to a point set CmK neighbor distance calculation is carried out on all the constraint points in the Voronoi region, outliers are removed by utilizing a preset distance threshold value, and a Voronoi region V is constructed1,…,Vm；
Step 2, based on a principal component analysis method, taking the first principal component as a direction, and respectively taking the length of t times V from the centroid to the positive and negative directions thereofmOf length of standard deviationLine segment as the first principal component line segment LmAnd calculating said CmTo said LmThe projected distance of (a);
step 3, obtaining the CmRemote point V infAnd the remote point V is adjustedfTo said CmIn (3), generating a new skeleton constraint point set Cm+1(ii) a The deflection point VfIs a set of distance principal component line segments L1,…,Lm+1-the farthest point;
step 4, in the step Cm+1In calculating a new first principal component line segment Lm+1And combining said Lm+1Add to the set of principal component line segments { L }1,…,Lm+1In the method, a new principal component line segment set { L } is obtained1,…,Lm+1}；
Step 5, enabling m +1, and iteratively executing the step 1 to the step 4 until the number of the line segments in the new principal component line segment set reaches a preset number threshold value or a preset energy function value is minimum;
and 6, constructing a Hamilton path by using the new principal component line segment set obtained in the step 5 based on a greedy algorithm, and optimizing the Hamilton path based on a 2-opt optimization algorithm to obtain the master curve.
The principal curve is a nonlinear generalization of linear principal component analysis, and the sum of the distances from points in the data set to the curve is minimized mainly by finding a curve, which is essentially a smooth curve passing through the middle position of the data set.
The extracted skeleton constraint point set is calculated for each point cloud subset, and the constraint point sets among different point cloud subsets may have some discontinuous situations, so that a skeleton curve closer to the vein position can be obtained by calculating main curves of the skeleton constraint point sets of all the point cloud subsets.
As an alternative embodiment, the invention provides a way to iteratively insert new line segments to compute the principal curves of the skeleton constraint point sets to optimally adjust the maize leaf skeleton:
firstly, extracting principal component line segments from a sample set through a PCA algorithm; then, the sample set is divided again according to the conditions; adjusting the corresponding Voronoi area to obtain a new principal component line segment set until the specified line segment number is met; and finally, constructing a Hamiltonian path through a greedy algorithm to obtain a final main curve.
The method specifically comprises but is not limited to the following steps:
(1) initialization
Taking the skeleton constraint point set C of all the sub-leaves as a sample set CmAnd using it to form an initial Voronoi region Vm(including all skeletal constraints). Through each region VmPerforming principal component analysis, taking the first principal component as direction, and respectively taking the length of 1.5 times V from the centroid to the positive and negative directionsmThe line segment of standard deviation length is used as the initial principal component line segment LmAnd calculating the projection distance from each skeleton constraint point in the C to the principal component line segment. The projection distance calculation formula is as follows:
d(b,L(t))＝inf‖b-L(t)‖
wherein, b represents one point in the C skeleton constraint point set C; l (t), t epsilon R represents a parameter equation of a straight line where the principal component line segment is located; inf represents rounding down; | represents the norm.
(2) Inserting new principal component line segments
Firstly, finding out an off-principal component line segment set { L ] from a skeleton constraint point set C1,…,LmThe farthest remote point VfWhile the remote point VfThree or more points around the sample set C, and a new sample set C is constructed using the pointsm+1(ii) a Then, in a new sample set Cm+1To calculate its first principal component line segment Lm+1Adding the main component line segment set to obtain { L1,…,Lm+1And recalculate their new Voronoi regions V1,…,Vm+1Until a set maximum number of line segments N is reached or the energy function as follows is minimized:
wherein sigma2Is the variance of the set of skeleton constraint points C, l is the sum of the lengths of all principal component line segments, M represents the number of points in the set of skeleton constraint points C, CkSet of skeleton-bound points, LiA corresponding principal component line segment.
(3) Generating a master curve
Since the skeleton generated by calculating the main curve of the skeleton constraint point set may shrink at the blade base and the blade tip, the skeleton may be processed by following processes after obtaining the skeleton, such as: calibrating the positions of the two ends of the framework, comprising:
and (3) taking the head and the tail of the skeleton constraint point set as the initial point and the end point of the skeleton, and searching the direction from the extension lines of the two ends of the skeleton to the constraint points of the head and the tail skeletons.
After the head and tail end points are calibrated, the framework extraction precision and the precision of subsequent phenotype (such as leaf length) calculation can be improved.
Experiments show that the leaf point cloud framework extraction method provided by the invention can be used for generating the framework according to the leaf initial point cloud of the long and narrow plant, can be used for extracting and obtaining the high-precision framework, is suitable for the leaves of corn, sorghum, rice, wheat and the like, and particularly aims at solving the defect that the prior art cannot effectively solve the problem of high-precision framework modeling of a few missing plant leaf point clouds, and the difference between the obtained extraction result and the extraction result of complete leaf point cloud data is not large.
Fig. 6 is a schematic structural diagram of a blade point cloud skeleton extraction apparatus provided by the present invention, as shown in fig. 6, the apparatus mainly includes a blade point cloud grouping unit 61, a cross-section point set partitioning unit 62, a constraint point set generating unit 63, and a point cloud skeleton generating unit 64, where:
the blade point cloud grouping unit 61 is mainly used for acquiring normal information of each point in the blade initial point cloud, clustering the blade initial point cloud based on the normal information, and acquiring a plurality of point cloud subsets;
the section point set dividing unit 62 is mainly configured to establish a distance field of each point cloud subset, determine a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determine a relevant point set of each section according to the distance field in which each section is located;
the constraint point set generating unit 63 determines the skeleton constraint points corresponding to each relevant point set mainly by using a self-adaptive weighting operator to construct a skeleton constraint point set;
the point cloud skeleton generating unit 64 is mainly used for acquiring a main curve of the skeleton constraint point set as a blade point cloud skeleton.
Specifically, the blade point cloud grouping unit 61 is mainly used to perform the following functions: determining the neighborhood of each point in the initial point cloud of the blade based on a KNN algorithm; then, determining the normal direction of each point according to the neighborhood of each point based on a PCA algorithm; further adjusting the normal direction of each point in the initial point cloud of the blade based on a minimum spanning tree algorithm, and acquiring the normal direction information of each point; and finally, clustering the blade point cloud according to the normal information of each point based on a K-means clustering algorithm, so as to obtain a plurality of point cloud subsets.
The section point set partitioning unit 62 mainly uses a distance field method to calculate a relevant point set corresponding to a section, each point cloud subset correspondingly generates a distance field, and the relevant point set corresponding to each section can be extracted from the leaf initial point cloud through the distance field.
The constraint point set generation unit 63 calculates skeleton constraint points of the relevant point set of each section by means of an adaptive weighting operator, for example: calculating the center of a relevant point set of each section according to the spatial distribution characteristics of the initial point cloud of the blade; and carrying out appropriate movement on the calculated central point according to the missing condition of the point cloud data, so that the finally obtained skeleton constraint point is closer to the vein point in the middle of the leaf.
The point cloud skeleton generating unit 64 is mainly configured to calculate a main curve of the skeleton constraint point set with the leaf according to the skeleton constraint point set, and further obtain a skeleton curve closer to the vein position as a leaf point cloud skeleton.
According to the leaf point cloud framework extraction device provided by the invention, a self-adaptive weighting operator is designed to calculate the framework constraint point set of the leaf point cloud, and a principal curve is introduced to optimize the framework constraint point set, so that the three-dimensional shape structure of the leaf can be better reflected, the missing point cloud data has better robustness, manual adjustment in the later period is not needed, and technical support can be provided for phenotype big data processing, automatic phenotype analysis and the like.
It should be noted that, in the method for extracting a blade point cloud framework according to the embodiment of the present invention, when the method is specifically executed, the method for extracting a blade point cloud framework according to any one of the embodiments described above may be executed, and details of this embodiment are not described herein.
Fig. 7 is a schematic structural diagram of an electronic device provided in the present invention, and as shown in fig. 7, the electronic device may include: a processor (processor)710, a communication Interface (Communications Interface)720, a memory (memory)730, and a communication bus 740, wherein the processor 710, the communication Interface 720, and the memory 730 communicate with each other via the communication bus 740. Processor 710 may invoke logic instructions in memory 730 to perform a method of blade point cloud skeleton extraction, the method comprising: acquiring normal information of each point in the initial point cloud of the blade, clustering the initial point cloud of the blade based on the normal information, and acquiring a plurality of point cloud subsets; establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located; determining a skeleton constraint point corresponding to each relevant point set by using a self-adaptive weighting operator to construct a skeleton constraint point set; and acquiring a main curve of the skeleton constraint point set to serve as a blade point cloud skeleton.
In addition, the logic instructions in the memory 730 can be implemented in the form of software functional units and stored in a computer readable storage medium when the software functional units are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, which when executed by a computer, enable the computer to perform the method for extracting a blade point cloud skeleton provided by the above methods, the method comprising: acquiring normal information of each point in the initial point cloud of the blade, clustering the initial point cloud of the blade based on the normal information, and acquiring a plurality of point cloud subsets; establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located; determining a skeleton constraint point corresponding to each relevant point set by using a self-adaptive weighting operator to construct a skeleton constraint point set; and acquiring a main curve of the skeleton constraint point set to serve as a blade point cloud skeleton.
In yet another aspect, the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, the computer program being implemented by a processor to perform the method for extracting a blade point cloud skeleton provided in the foregoing embodiments, the method including: acquiring normal information of each point in the initial point cloud of the blade, clustering the initial point cloud of the blade based on the normal information, and acquiring a plurality of point cloud subsets; establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located; determining a skeleton constraint point corresponding to each relevant point set by using a self-adaptive weighting operator to construct a skeleton constraint point set; and acquiring a main curve of the skeleton constraint point set to serve as a blade point cloud skeleton.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, it is clear to those skilled in the art that the embodiments may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the invention has been described in detail with reference to the foregoing embodiments, it will be appreciated by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for extracting a blade point cloud skeleton is characterized by comprising the following steps:
acquiring normal information of each point in the initial point cloud of the blade, clustering the initial point cloud of the blade based on the normal information, and acquiring a plurality of point cloud subsets;
establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located;
determining a skeleton constraint point corresponding to each relevant point set by using a self-adaptive weighting operator to construct a skeleton constraint point set;
and acquiring a main curve of the skeleton constraint point set to serve as a blade point cloud skeleton.
2. The method of blade point cloud skeleton extraction of claim 1, wherein the establishing of the distance field for each of the point cloud subsets comprises:
determining a directed bounding box for each of the point cloud subsets;
distances between each point in the point cloud subsets and an orthogonal plane perpendicular to the blade elongation direction are calculated and normalized, thereby establishing a distance field for each of the point cloud subsets.
3. The method of claim 2, wherein the determining a directed bounding box for each of the point cloud subsets comprises:
calculating three characteristic vectors of each point cloud subset based on a principal component analysis method;
performing standard orthogonalization on the three eigenvectors to obtain an orthogonal matrix;
constructing a Cartesian coordinate system by using the orthonormal basis of the orthogonal matrix, and calculating the maximum position and the minimum position in three directions X, Y, Z from the origin in the Cartesian coordinate system to construct 6 planes of the directional bounding box according to all the maximum positions and the minimum positions.
4. The method for extracting the leaf point cloud framework of claim 1, wherein the obtaining normal information of each point in the leaf initial point cloud, and clustering the leaf initial point cloud based on the normal information to obtain a plurality of point cloud subsets comprises:
denoising and downsampling the initial point cloud of the blade to remove noise points and obtain a blade point cloud;
determining a neighborhood of each point in the blade point cloud based on a K nearest neighbor classification algorithm;
determining the normal direction of each point according to the neighborhood of each point based on a principal component analysis method;
based on a minimum spanning tree algorithm, adjusting the normal direction of each point in the blade point cloud, and acquiring the normal direction information of each point;
based on a K-means clustering algorithm, clustering the leaf point cloud according to the normal information of each point to obtain a plurality of point cloud subsets.
5. The method of claim 1, wherein the adaptive weighting operators comprise a spatial distance weight, a normal difference weight, and a point cloud integrity weight;
the determining the skeleton constraint point corresponding to each relevant point set by using the adaptive weighting operator includes:
according to the spatial distribution characteristics of the relevant point set of each section, the central point of the relevant point set of each section is calculated in a self-adaptive mode;
and moving the central point according to the missing state of the point cloud data in the relevant point set of each section to obtain the corresponding skeleton constraint point of each relevant point set.
6. The method of extracting a leaf point cloud skeleton according to claim 5, wherein the expression of the spatial distance weight is:
the expression of the normal difference weight is:
the expression of the point cloud integrity weight is as follows:
wherein p isiIs a set of relevant points S of the cross sectionkPoint i of (1); q. q.sjIs piK neighbor set of points N (p)i) J-th point in (1); alpha is an influence parameter on the spatial distance; w'k,iIs a spatial distance weight;
is the leaf surface orientation at the kth section; n ispiIs a set of relevant points S of the cross sectionkMidpoint piNormal direction of (2); beta is an influencing parameter on the normal difference; w ″)k,iIs the normal difference weight; lmRepresenting the length of the mth side of a polygon formed by a concave shell, wherein the concave shell is generated by projecting a relevant point set of a cross section onto an orthogonal surface which is perpendicular to the extending direction of the blade; r is the ratio of the longest side of the concave shell to the total side length; epsilon is a threshold value for judging point cloud data missing.
7. The blade point cloud skeleton extraction method of claim 6, wherein the central point is moved according to the missing state of the point cloud data in the relevant point set of each section to obtain the moved central point in the corresponding skeleton constraint point of each relevant point set, and the determination formula is as follows:
wherein, ckIs a set of relevant points S of the cross sectionkCorresponding skeleton constraint points; e.g. of the typekAccording to the point cloud integrality of the related point set to the central pointThe direction of movement is made.
8. The method of claim 1, wherein the obtaining of the main curve of the skeleton-constrained point set comprises:
step 1, restraining a skeleton to a point set CmK neighbor distance calculation is carried out on all the constraint points in the Voronoi region, outliers are removed by utilizing a preset distance threshold value, and a Voronoi region V is constructed1,…,Vm；
Step 2, based on a principal component analysis method, taking the first principal component as a direction, and respectively taking the length of t times V from the centroid to the positive and negative directions thereofmA line segment of standard deviation length as the first principal component line segment LmAnd calculating said CmTo said LmThe projected distance of (a);
step 3, obtaining the CmRemote point V infAnd the remote point V is adjustedfTo said CmIn (3), generating a new skeleton constraint point set Cm+1(ii) a The deflection point VfIs a set of distance principal component line segments L1,…,Lm+1-the farthest point;
step 4, in the step Cm+1In calculating a new first principal component line segment Lm+1And combining said Lm+1Add to the set of principal component line segments { L }1,…,Lm+1In the method, a new principal component line segment set { L } is obtained1,…,Lm+1}；
Step 5, enabling m +1, and iteratively executing the step 1 to the step 4 until the number of the line segments in the new principal component line segment set reaches a preset number threshold value or a preset energy function value is minimum;
and 6, constructing a Hamilton path by using the new principal component line segment set acquired in the step 5 based on a greedy algorithm, and optimizing the Hamilton path based on a 2-opt optimization algorithm to acquire the master curve.
9. A blade point cloud skeleton extraction element, its characterized in that includes:
the blade point cloud grouping unit is used for acquiring normal information of each point in the blade initial point cloud, clustering the blade initial point cloud based on the normal information and acquiring a plurality of point cloud subsets;
the section point set dividing unit is used for establishing a distance field of each point cloud subset, determining a plurality of sections perpendicular to the blade extension direction according to a preset step length, and determining a relevant point set of each section according to the distance field where each section is located;
the constraint point set generating unit is used for determining the skeleton constraint points corresponding to each relevant point set by using a self-adaptive weighting operator so as to construct a skeleton constraint point set;
and the point cloud framework generation unit is used for acquiring a main curve of the framework constraint point set to serve as a blade point cloud framework.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the blade point cloud skeleton extraction method steps of any one of claims 1 to 8 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111088873.9A CN113989352A (en) | 2021-09-16 | 2021-09-16 | Blade point cloud framework extraction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111088873.9A CN113989352A (en) | 2021-09-16 | 2021-09-16 | Blade point cloud framework extraction method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113989352A true CN113989352A (en) | 2022-01-28 |
Family
ID=79735972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111088873.9A Pending CN113989352A (en) | 2021-09-16 | 2021-09-16 | Blade point cloud framework extraction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113989352A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101251932A (en) * | 2008-03-18 | 2008-08-27 | 中国科学院软件研究所 | Plant leaf analogy method based on physics |
| CN106446343A (en) * | 2016-07-25 | 2017-02-22 | 北京航空航天大学 | Method for automatically extracting parameterized profile line of straight-line blade of radial-flow impeller |
| US20170193692A1 (en) * | 2015-12-30 | 2017-07-06 | Shenzhen Institutes Of Advanced Technology Chinese Academy Of Sciences | Three-dimensional point cloud model reconstruction method, computer readable storage medium and device |
| US20190274257A1 (en) * | 2018-03-08 | 2019-09-12 | Regents Of The University Of Minnesota | Crop biometrics detection |
| CN112991300A (en) * | 2021-03-04 | 2021-06-18 | 中国林业科学研究院资源信息研究所 | Single-wood skeleton extraction and visualization method based on neighborhood characteristics |
| CN113298833A (en) * | 2021-05-20 | 2021-08-24 | 山东大学 | Target object point cloud characteristic line and surface extraction method and system |
-
2021
- 2021-09-16 CN CN202111088873.9A patent/CN113989352A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101251932A (en) * | 2008-03-18 | 2008-08-27 | 中国科学院软件研究所 | Plant leaf analogy method based on physics |
| US20170193692A1 (en) * | 2015-12-30 | 2017-07-06 | Shenzhen Institutes Of Advanced Technology Chinese Academy Of Sciences | Three-dimensional point cloud model reconstruction method, computer readable storage medium and device |
| CN106446343A (en) * | 2016-07-25 | 2017-02-22 | 北京航空航天大学 | Method for automatically extracting parameterized profile line of straight-line blade of radial-flow impeller |
| US20190274257A1 (en) * | 2018-03-08 | 2019-09-12 | Regents Of The University Of Minnesota | Crop biometrics detection |
| CN112991300A (en) * | 2021-03-04 | 2021-06-18 | 中国林业科学研究院资源信息研究所 | Single-wood skeleton extraction and visualization method based on neighborhood characteristics |
| CN113298833A (en) * | 2021-05-20 | 2021-08-24 | 山东大学 | Target object point cloud characteristic line and surface extraction method and system |
Non-Patent Citations (1)
| Title |
|---|
| 赵元棣;温维亮;郭新宇;肖伯祥;陆声链;孙智慧: "基于参数化的玉米叶片三维模型主脉提取", 农业机械学报, vol. 43, no. 4, 25 April 2012 (2012-04-25) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108830931B (en) | Laser point cloud simplification method based on dynamic grid k neighborhood search | |
| EP2811424B1 (en) | Method and apparatus for training an estimator for estimating a pose of an articulated object | |
| US7995841B2 (en) | Hybrid graph model for unsupervised object segmentation | |
| Huang et al. | Shape decomposition using modal analysis | |
| Kempthorne et al. | Surface reconstruction of wheat leaf morphology from three-dimensional scanned data | |
| CN109493372A (en) | The product point cloud data Fast global optimization method for registering of big data quantity, few feature | |
| Carneiro et al. | Robust left ventricle segmentation from ultrasound data using deep neural networks and efficient search methods | |
| EP3667611A1 (en) | Information processing device, model data creation program, and model data creation method | |
| CN109685841B (en) | Registration method and system of three-dimensional model and point cloud | |
| CN110990775B (en) | Multi-view clustering method based on multi-manifold dual graph regularized non-negative matrix factorization | |
| CN111210466A (en) | Multi-view point cloud registration method and device and computer equipment | |
| CN115239760B (en) | Target tracking method, system, equipment and storage medium | |
| CN104406594A (en) | Measuring algorithm for relative position and posture of rendezvous and docking spacecraft | |
| CN112200915A (en) | Front and back deformation amount detection method based on target three-dimensional model texture image | |
| Hu et al. | Curve skeleton extraction from 3D point clouds through hybrid feature point shifting and clustering | |
| CN116452648A (en) | Point cloud registration method and system based on normal vector constraint correction | |
| CN108229502B (en) | Method for extracting base points of blades in three-dimensional point cloud data of tree canopy layer | |
| CN112991300B (en) | Single wood skeleton extraction and visualization method based on neighborhood characteristics | |
| JP5175754B2 (en) | Linear transformation matrix calculation apparatus, method thereof, and program thereof | |
| CN111428555B (en) | Joint-divided hand posture estimation method | |
| CN113989352A (en) | Blade point cloud framework extraction method and device | |
| CN113609914A (en) | Obstacle identification method and device and vehicle control system | |
| CN112270746A (en) | Aluminum alloy 3D printing point cloud simplification algorithm based on neighborhood covariance characteristic parameter threshold | |
| Khan et al. | Geometric correction of deformed chromosomes for automatic Karyotyping | |
| Muñoz et al. | Manifold learning for coherent design interpolation based on geometrical and topological descriptors |
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 |