CN101783016B - Crown appearance extract method based on shape analysis - Google Patents
Crown appearance extract method based on shape analysis Download PDFInfo
- Publication number
- CN101783016B CN101783016B CN2009102427509A CN200910242750A CN101783016B CN 101783016 B CN101783016 B CN 101783016B CN 2009102427509 A CN2009102427509 A CN 2009102427509A CN 200910242750 A CN200910242750 A CN 200910242750A CN 101783016 B CN101783016 B CN 101783016B
- Authority
- CN
- China
- Prior art keywords
- point
- data
- skeleton
- cluster
- classified
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 80
- 238000006243 chemical reaction Methods 0.000 claims description 13
- 238000007621 cluster analysis Methods 0.000 claims description 9
- 230000005540 biological transmission Effects 0.000 claims description 5
- 241001269238 Data Species 0.000 claims description 4
- 230000015572 biosynthetic process Effects 0.000 claims description 2
- 230000010354 integration Effects 0.000 abstract 1
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000012217 deletion Methods 0.000 description 5
- 230000037430 deletion Effects 0.000 description 5
- 238000000605 extraction Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000004224 protection Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 239000007787 solid Substances 0.000 description 3
- 241000208140 Acer Species 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 2
- 244000055346 Paulownia Species 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000008635 plant growth Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 239000002028 Biomass Substances 0.000 description 1
- 235000018639 Terminalia sericea Nutrition 0.000 description 1
- 241000844454 Terminalia sericea Species 0.000 description 1
- 240000001866 Vernicia fordii Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 230000037237 body shape Effects 0.000 description 1
- 238000011960 computer-aided design Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 244000144985 peep Species 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a crown appearance extract method based on shape analysis. The data treatment of the invention is characterized in that a laser scanner carries out single-sweep on a whole plant to obtain the single-sweep point cloud data of the plant; firstly, a root node is found from the point cloud data, and then the point cloud data is divided into branch point cloud and leaf point cloud by a shortest distance method; a corresponding skeleton is extracted from the branch point cloud, and the skeleton is divided into a plurality of parts according to the branching characteristics of a tree, and classified branch skeletons are utilized to classify the leaf point cloud; Delaunay triangulation is carried out on each classified leaf point cloud to obtain a tetrahedron and a triangular patch; the triangular patch is divided into three classes, such as exterior, interior, boundary and the like, wherein the triangular patch located on the boundary forms a sealed body which is used for representing the shape of the classified leaf point cloud; and the shape integration of each class of the leaf point cloud represents the shape of the whole crown.
Description
Technical field
The invention belongs to the cross discipline technical field that pattern-recognition combines with visualization in scientific computing, relate to the extractive technique of a cloud processing, shape analysis, plant body (abbreviation body) and classification skeleton.
Background technology
Along with the foundation of virtual environment, product design, digital entertainment, the development of historical relics protections and city planning, the reconstruction of 3-D geometric model and being processed into for a research focus.We can say that 3 d geometric modeling can be regarded as except DAB, digital picture, the fourth digit multimedia after the digital video.Three-dimensional geometric model can be represented object surfaces, thereby can differentiate some character of external shape and appearance.
And along with the fast development of data acquisition technology, increasing 3-dimensional digital scanner has obtained popularizing and being applied to growing field.In computer graphics, employed data resolution is more and more higher in the Computer-aided Design Technology, and data volume is also increasing.By object is carried out intensive scanning, can obtain about the more details of object profile, acquisition be point cloud data, and do not have topological link information.Represent so develop new method, handle, rebuild or even draw these complex geometry bodies and just become suitable important.And the reconstruction of geometric model is important in a modern virtual reality research theme.Skeletonizing can be deleted redundant information and keeping characteristics information effectively as a kind of feature extraction and character representation method.In virtual reality, skeletonizing can prevent in roaming lost from the extracting data outbound path to instruct roaming.For the visual of forestry and measurement, skeleton can be represented the topological structure of body accurately so that carry out the identification of trees, and can accurately measure length and the radius of trees.For " thin stem " the shape image etc. in literal, engineering drawing, the fingerprint all is long and narrow band-like image, also usually needs to discern or processing such as reconstruct by the skeleton that calculates them.In Video processing,, can accurately follow the tracks of the motion body by principal character in the skeletal extraction image.
A lot of bodies, as the various trees in the blood vessel in the medical science, tracheae, the forestry, and the skeleton of human body, can represent to become body, therefore the skeletal extraction of body has obtained paying attention to widely, and Ogniewicz utilizes the important tool-Voronoi figure in the computational geometry to extract skeleton.For the point in the plane, can obtain the Voronoi polygon of this point, and Voronoi figure is exactly the polygonal set of these Voronoi.Voronoi limit that the point on body border is produced among the Voronoi figure or Voronoi face are near the center of body, so the skeleton of body can be thought in these limits.But this method generally can only be applicable to triangle grid model, and the skeleton that generates comprises a lot of assorted branches.
Rosenfeld has proposed to open up the skeletonization method of mending refinement.This algorithm at first defines and does not influence that to open up the point of mending character be simple point after those deletions.From this notion, consider the benefit relation of opening up of the 8-neighborhood or the 26-neighborhood of tissue points then, design some deletion templates, make that the tissue points in the template satisfies simple definition of putting.Utilize these deletion templates strip off body from level to level then, till obtaining skeleton.But this method calculated amount is too big, and working time is oversize.Ma has improved the deletion template, and deletion action is carried out parallel processing to accelerate travelling speed.But the result of refinement often comprises a lot of assorted branches.
Borgefors has proposed the skeletonization method based on range conversion.The main basis of range conversion method is its skeletal definition: three-dimensional skeleton is three-dimensional inner set of arriving the point of three-dimensional frontier distance maximum.Voxel code and dendrogram method that Zhou proposes can decompose body effectively, but the topological structure of the topological structure of its skeleton and body are inconsistent at crotch.The Wan method is earlier volume data to be carried out range conversion, then the distance map after the conversion is regarded as an oriented centrality weighted graph, use improves Dijkstra shortest path generating algorithm with the method that the inverse of frontier distance value is set up minimum expansion tree, thereby make the skeleton point not off-center as far as possible that obtains, but center line " disturbance " phenomenon can occur.In virtual reality, trees are more typical things, so rebuild and represent that real trees just become suitable important.And the reconstruction of trees can be used in considerable field and be applied, and comprises the digitizing of vegetation scene, the design of new scene, and in the digital entertainment or the like.The reconstruction of crown appearance is very useful for the modeling of true trees, and obtains paying close attention at some scientific research fields, such as virtual optical is according to the simulation to the plant growth, estimation of trees biomass or the like.
A lot of researchists has obtained certain achievement aspect curve reestablishing at present, but crown appearance is than the general more difficult reconstruction of solid profile, because blocking mutually between leaf of trees and branch, and trees how much and topological aspect high complexity.
It is a challenging job in the computational geometry field that distribution cloud data at random is asked for its external shape always, and in three dimensions, the convex closure profile of cloud data is unique, but this convex closure profile only provides less data appearance information.There are not a large amount of details in the focused data.So the researchist wishes to find the method for better asking for profile.
Jarvis is first researchist who point set convex closure shape on two dimensional surface is generalized to general shape.Edelsbrunner has at first provided the mathematical explication about shape.For three-dimensional point set, Boissonnat proposes come with the Delaunay triangulation the interconnective shape of " carving " point set.The method of Hoppe is to reconstruct curved surface from some cloud at random, and he utilizes a distance field function to reconstruct an implicit surface, and then this implicit surface of triangulation.The Crust method can be rebuild the stream body of smooth closure in the two and three dimensions space, but it need use the important tool Voronoi figure in the computational geometry, and need to calculate Voronoi figure twice, this calculating is quite consuming time, and this method can not handle flex point and angle point, and to need the sampling density of some cloud be very big.Alpha shape method at first is a Delaunay triangulation point set, need the user to go to select the size of parameter then, by the difference of this parameter, the detail of the point set profile of Huo Deing also is different at last, certainly this Delaunay calculated amount is too big, also is suitable consuming time.
Summary of the invention
Existing technology can not well be calculated the external shape of tree-like object cloud data, therefore can not solve the expression of tree, measures and identification.In order to solve the deficiencies in the prior art, the objective of the invention is to ask for the external shape of the scan-data of tree, the method for external shape of a kind of calculating tree of the skeleton sorting technique based on tree is provided.
In order to realize purpose of the present invention, a kind of crown appearance extract method provided by the invention based on shape analysis, the step of this method comprises:
Step S1: with the single face of laser scanner scans plant, the scan-data of acquisition is called cloud data;
Step S2: from cloud data, find root node, method with k nearest neighbor is carried out range conversion to whole cloud data, construct point in these cloud datas to the weight graph of root node, each of calculating cloud data put the minor increment of root node, utilizes this minor increment to isolate the cloud data that is positioned on the major branch from whole cloud data;
Step S3: the cloud data that is positioned on the major branch is carried out cluster analysis, generate tree-shaped dendrogram, be used to seek the bifurcation site and the tip position of body, in each cluster, according to distance cluster once more, obtain the sub-cluster in the cluster, to each sub-cluster, calculate the representative point of each sub-cluster with the method for arithmetic mean, as major branch skeleton point;
Step S4: the representative point of all sub-clusters is formed set, have centrality, bifurcated and topological internuncial major branch skeleton with all representative point set structures again;
Step S5: from the leaf node or the title tip position on major branch skeleton top, each branch of skeleton is classified, make sorted each branch length reach certain standard; On the basis that the major branch skeleton is classified, the leaf cloud data that is positioned at tree crown in the cloud data is classified;
Step S6: each sorted leaf cloud data is calculated their shape respectively, at first one of them sorted some cloud is carried out three-dimensional Delaunay triangulation, set up the topological structure of each data point in the some cloud, obtain tri patch and tetrahedron, these tetrahedrons and tri patch are calculated their circumsphere and external radius of a circle respectively, obtain comprising the smallest interval [A of these radiuses, B], the parameter Alpha who gets in Alpha's shape method is interval [A, B] intermediate point, tetrahedron of trying to achieve and tri patch are classified, tetrahedron is categorized as outside tetrahedron and inner tetrahedron, and tri patch is categorized as inner tri patch, the outside tri patch and the tri patch on border, if what the tri patch on border was formed is a closed body, so parameter Alpha is got [A, B] mid point between lower region continues this process, up to the difference of two adjacent parameters less than given abundant little value, fully little value is one of percentage of getting point-to-point transmission minor increment in the scan-data, at this time in these two parameters, the set of the border tri patch of a bigger correspondence is exactly the external shape of whole tree crown.
Wherein, the described root node that finds from cloud data is that the method according to Y coordinate minimum in the data point finds root node.
Wherein, point in these cloud datas of described structure is the method by k nearest neighbor to the weight graph of root node, finds K the point that range data point is nearest, point that these range data points are nearest and data point couple together the formation weight graph, and the power in the weight graph is exactly the distance of point-to-point transmission.Wherein, each of described calculating cloud data put the minor increment of root node, is on the basis of known weight graph, calculate from root node to the data point path power and minimum path, this power and be exactly the minor increment of this data point to root node.
Wherein, described cluster analysis is to carry out cluster analysis according to each the data point p in the major branch cloud data to the known minor increment of root node, forms big cluster, and the point in each big cluster belongs to certain distance range.
Wherein, described cluster once more is in each the big cluster that obtains, and cluster analysis is carried out in the applications distances conversion once more, obtains sub-cluster.
Wherein, described structural classification skeleton, be with each sub-cluster core out point, then with the point of Y coordinate minimum in these central points as root node, calculate in these central points other and put the minimum distance path of this root node, the point on the access path constructs this major branch skeleton in order.
Wherein, described each branch to skeleton classifies, at first find each tip representative point of major branch skeleton, the tip point of choosing Y coordinate maximum is as starting point, opposite direction along major branch skeleton root node to the shortest path of this starting point moves, when arriving first bifurcation, calculate, calculate the length in present path, if the length in path is greater than specified value 1 at present, then at present be classified as a class through the point on the path, choosing next untreated and Y coordinate then is that maximum tip point carries out the same path of recalling, otherwise calculate with first bifurcation be initial point all son branches length and; As if all length of sub of initial point with greater than specified value 1, be that representative point on all son branches of initial point is classified as a class then with first bifurcation, otherwise from present bifurcation, continue to date back to next bifurcation along minimal path, with length of top all branch and add first bifurcation with the distance between bifurcation at present get up till now path with, as if this present path with greater than specified value, all representative points on the then present path are classified as a class, otherwise calculate all with present bifurcation be initial point all son branches length and, as if length of these all son branches with greater than specified value 1, then the representative point with present path process is classified as a class, otherwise the process above just continuing continues to recall up to the length of son branch with greater than specified value 1 along shortest path.
Wherein, described the leaf cloud data that is positioned at tree crown in the cloud data is classified, on the basis of the major branch skeleton being classified, arrive the nearest principle of major branch skeleton according to each data point in the leaf cloud data, whole leaf cloud data is classified, be classified as a class to the nearest leaf number strong point of the same class of major branch skeleton branch, otherwise just adhere to different classes separately.
Wherein, the tetrahedron of described outside and inner tetrahedron are that tetrahedron has unique circumsphere, and if this circumsphere radius then is outside tetrahedron greater than parameter Alpha given in advance, otherwise are exactly inner tetrahedron.
Wherein, the tri patch of the tri patch of described inside, outside and the tri patch on border, each triangular facet has unique circumscribed circle, and this circumradius is one of feature of this triangular facet, and triangular facet is divided into two classes and is positioned on the convex closure and is positioned at convex closure inside; For the triangular facet that is positioned on the convex closure, if the tetrahedron at its place is outside tetrahedron, then this triangular facet is outside triangular facet, otherwise it is the triangular facet on border; For the triangular facet that is positioned at convex closure inside, it is shared by two tetrahedrons, these two adjacent tetrahedrons are if it all is outside tetrahedron, it then on this triangular facet outside triangular facet, if these two tetrahedrons all are inner tetrahedrons, then this triangular facet is exactly inner triangular facet, otherwise this triangular facet is exactly the triangular facet on border.
Beneficial effect of the present invention: the technical matters that the present invention solves is to ask for the shape of trees single-sweep data.Method of the present invention is to classify by the scan-data to whole trees, it is divided into leaf point cloud and limb point cloud, we extract corresponding classification skeleton at limb point cloud, then this classification skeleton is at first carried out cluster, skeleton applications distances after the cluster comes leaf point cloud is classified, and is at last each sorted leaf point cloud is asked for its corresponding shape.The skeleton master is with will representing and measure in the identification of body and tracking, body shape.The skeleton of 3-d tree-like body be widely used in that trees body in the plant growth modeling represents to peep roaming in virtual in the measurement of (description), visual and trees shape, the medical image and organ shape is represented and computer vision in target following etc.
Description of drawings
Fig. 1 illustrates algorithm flow chart of the present invention
Fig. 2 illustrates the single-sweep data that the present invention adopts
Fig. 3 a, Fig. 3 b, Fig. 3 c illustrate that the present invention separates is positioned at cloud data on the limb
Fig. 4 illustrates the representative point that the present invention is based on each cluster that clustering method obtains
Fig. 5 illustrates the skeleton of the limb that the present invention obtains
Fig. 6 illustrates skeleton node-classification flow process
Fig. 7 the present invention is based on the result of the skeleton classification leaf point cloud of limb
Fig. 8 a, Fig. 8 b and Fig. 8 c illustrate the shape that the present invention asks for sorted leaf point cloud respectively
Fig. 9 a, Fig. 9 b and Fig. 9 c illustrate the shape that the present invention is based on the whole scan-data that automatic alpha-shape method obtains
Figure 10 a, Figure 10 b and Figure 10 c illustrate the present invention's asking for the shape of blue or green paulownia data
Figure 11 a and Figure 10 b illustrate limitrophe cloud data and the corresponding profile that the present invention obtains
Figure 12 illustrates automatic Alpha's shape (alpha-shape) method flow
Figure 13 a to Figure 13 h illustrates the detailed information based on the shape of tree-crown of limb skeleton classification results.The grid representation of Figure 13 a to Figure 13 h shape of tree-crown; Figure 13-1 represents to the some cloud of Figure 13-8 shape of tree-crown.
Embodiment
Below in conjunction with accompanying drawing the present invention is described in detail, be to be noted that described embodiment only is intended to be convenient to the understanding of the present invention, and it is not played any qualification effect.
1 method general introduction (overview of approach)
The core of algorithm of the present invention is based on the limb skeleton of range conversion extraction trees scan-data, based on the limb skeleton leaf point cloud is carried out the solid classification, asks for corresponding shape based on sorted each leaf point cloud with the alpha-shape method.As Fig. 1 algorithm flow chart of the present invention is shown, specific algorithm comprises that 6 steps (1. only have 6 steps, please carry out the step mark, simultaneously every width of cloth figure is clearly described among Fig. 1; 2. must occur in the text the literal that has in every width of cloth process flow diagram, guarantee the texts and pictures unanimity):
1, at first, to the scan-datas (the single-sweep data of employing are shown as Fig. 2) of true trees by interactively observation, the point of choosing Y coordinate minimum is as root node, and other data point except that this root node in the scan-data carried out range conversion, wherein we use is that the method for k nearest neighbor is finished this task, soon each data point is connected with K the point nearest apart from this data point and forms corresponding limit, a weight graph has just been formed with point in these limits like this, and the power on each limit is the Euclidean distance between corresponding two end points.The distance of any point-to-point transmission in this scan-data is exactly the shortest path length that connects in this weight graph in all paths of this two points like this.And be very useful to the bee-line at other number of scans strong point by root node, the quantity of generic point cloud is huge, traditional Di Jiesitela bee-line method needs an enough big two-dimensional matrix, this is that computing machine generally holds and can't stand, so we only calculate the bee-line of root node to other number of scans strong points, we are referred to as the R distance.
2, secondly, by the weight graph of range conversion structure, we separate the point that is positioned in the whole data on the major branch.By step 1, each data point all has corresponding R distance, and we can choose suitable threshold λ like this, if the R distance is then put us accordingly and just defined it on major branch less than threshold value λ.As Fig. 3 a, Fig. 3 b and Fig. 3 c corresponding threshold value is shown and is taken as 1.6,1.4 and 0.5.By result's wherein reasonable major branch part of isolating trees of Fig. 3 c as can be seen.The scan-data of input is a single face, and is just incomplete, thus we when using the k nearest neighbor method, if parameter K is too little, the weight graph that then calculates is unconnected, promptly produces the subgraph of several connections.If parameter K is too big, will produce incorrect topological link information.The point that for example will belong to different limbs couples together, in the experiment we to choose K be 100.
3, more secondly, calculate the skeleton point of major branch.For the data point on the trunk of being positioned at that obtains, by local annexation they are converted into a tree-shaped dendrogram, extract the representative point of each cluster, thereby construct one and take into account bifurcated and topological internuncial classification skeleton (classification skeleton), the value of this skeleton is to classify for the leaf point cloud that is positioned at trees scan-data top.At present whole scan-data has been divided into two parts: be positioned at the point on the major branch and be positioned at the epiphyllous point of trees canopy.We choose can represent trunk point cloud among Fig. 3 c calculate the skeleton point.Because we know the bee-line of putting root node on the major branch, so we use the method for cluster the point on these major branches are classified.We at first calculate the maximal value r of point in the bee-line of root node on these major branches
MaxWith minimum value r
Min, then with interval [r
Min, r
Max] be divided into 50 parts, put the minor increment of root node according to each, these points that will be positioned on the trunk are divided into 50 classes.As our finding, be arranged in the point data of a class on the trunk, distance to each other is very little, but be arranged in the point data of a class on the trunk, owing to may belong to different trunks, distance may be very big to each other, for obtaining real relatively skeleton, we need classify once more with being arranged in the point data of a big class after the classification, are divided into different subclasses.The point that belongs to different major branches that is about in the big class differentiates.At each big class, at first take out a bit, it belongs to a new subclass mark, travel through K neighbour of this point, the point that wherein belongs to this big class, it belongs to this subclass our mark, and the point to new this subclass of adding carries out this top process equally then, and this is the ergodic process of a breadth-first.Finish or do not belong to this big class up to corresponding K neighbour of new adding point is processed.We take out not processed new point from the left point of this big class then, and mark its belong to the new subclass of another one, and the ergodic process that begins a new breadth-first finds other the point data that belongs to this new subclass in this big class.Point in this big class all disposes, and promptly their whole subclass institute marks finish.
When we with different big classes with different subclass mark after the point data of trunk, for getting skeleton to the end, need to calculate the representative point of each subclass, here we calculate each subclass mid point as representative point, the skeleton point that expression is following, we also need refine this skeleton point certainly, if the point in the subclass very little, for example be less than 5, we just abandon the corresponding skeleton point of this subclass so, think that this skeleton point does not have too much influence to whole skeleton.The representative point of each cluster that obtains based on clustering method as shown in Figure 4.
Behind known skeleton point, the quantity of skeleton point greatly reduces, we just can use traditional Di Jiesite pulling method and calculate the minor increment that these skeletons are put root node, simultaneously also just obtain the shortest path that these skeletons are put root node, also just drawn the skeleton (skeleton of the limb that obtains is shown as Fig. 5) of this trunk.
4, then, we just can use this limb skeleton the leaf point cloud that is positioned at whole data top is classified.Promptly according to leaf point cloud to the distance of skeleton limb with its cluster.In botany, different limbs belongs to different grades, even the limb of same levels may length also be different, and these different limbs have influenced the external crown shape of trees exactly, therefore, we at first carry out cluster with the skeleton point, and then cluster leaf point cloud.We obtain previous step is the digraph of skeleton, here we take out a leaf node p1 of the digraph of skeleton earlier, the y coordinate that requires this leaf node is maximum in all skeleton nodes, date back to first bifurcation p2 from this leaf node then, if the distance of the R between these two node p1 and p2 is greater than 1, we are classified as a class with the node on these two node p1 and p2 and their paths so, and stop to seek other node that belongs to this class, otherwise our calculating is the length of all individual paths of starting point with this bifurcation p2, if the length of all individual paths and be not less than 1, our all nodes that will be positioned on these individual paths are classified as a class so, and stop to seek other node that belongs to this class.Otherwise as if this length with less than 1, we continue to recall from present this bifurcation p2 and go for next bifurcation p3 so, and calculating is the total length of the limb of starting point with present this bifurcation p3, up to the length of this limb with greater than 1, we are classified as a class with all nodes on the limb, and mark its for handling.We take out at present untreated and the y coordinate is a maximum leaf node then, and the process above continuing was all handled up to all leaf nodes; The flow process of this algorithm such as Fig. 6.
We with the skeleton node-classification after, each that calculate in the leaf point cloud put the distance of skeleton node, and comes this leaf data point of mark with the classification of the skeleton node of minimum distance, our each point that just leaf can be put in the cloud carries out cluster like this; The present invention is based on result such as Fig. 7 of the skeleton classification leaf cloud of cluster.
5, subsequently, calculate leaf point cloud shape after each cluster with automatic Alpha's shape (alpha-shape) method.At first we without skeleton to whole single-sweep data P={p
iClassify, regard whole scan-data as an integral body, it is carried out Delaunay (Delaunay) triangulation, behind the subdivision, whole data have just obtained a kind of topology information, have obtained tetrahedron T={T
jAnd tri patch F={F
kSet, from the outside, the convex closure of these data has been formed in this tetrahedral set, the convex closure of data is unique, but this convex closure looks there is not more details.So we further handle it, calculate each tetrahedral circumsphere radius R (T
j) and each tetrahedral tri patch F={F
kCircumradius r (F
k), then to tetrahedron { T
jClassify, according to being exactly its corresponding circumsphere radius, our first designated parameter alpha value, all then tetrahedrons can be divided into two classes: as if R (T
j)＞α, T so
jIt is exactly outside tetrahedron; Otherwise it is exactly inner tetrahedron.Tetrahedral all { F
kBe divided into three classes: for the face that is positioned on the convex closure, if it belongs to outside tetrahedron, it is exactly an exterior face so; Otherwise if it belongs to inner tetrahedron, it is exactly a boundary surface so.For the face on the non-convex closure, if it is two tetrahedral common factors in outside, then it is an exterior face; If it is two tetrahedral common factors in inside, then it is an inside face; It otherwise is exactly boundary surface.And all boundary surfaces can be formed a grid, and this grid is exactly being similar to crown appearance.Wherein make r
MaxBe all R (T
j) and r (F
k) maximal value, r
MinBe all R (T
j) and r (F
k) minimum value, make A=λ r
Min, B=μ r
Max, λ=0.9, we obtain an interval [A, B] and μ=1.1., and the parameter alpha in Alpha (alpha) the shape method is in this interval value.Obviously as if Alpha alpha＞B, then grid M is exactly a convex closure, if alpha＜A, then grid M will not be a solid.Finding suitable parameters alpha is the process of an iteration, we first initialization alpha is interval [A, B] mid point, at each iterative step, we verify whether all boundary surfaces have formed a stream shape face, if like this, parameter value just reduces, otherwise parameter value just increases, and concrete flow process is seen Figure 12.The result is Figure 10 a, Figure 10 b and Figure 10 c, Figure 11 a and Figure 11 b, and Figure 10 a is the raw data of tung oil tree.Figure 10 b figure is the result of automatic Alpha's shape method, and Figure 10 c figure is local feature.Figure 11 a is the limitrophe point of maple, and Figure 11 b figure is a result calculated.The experimental detail of these two data is as shown in table 1 below.
Table 1 experimental data
| Trees | The number of point | The Alpha value is calculated | Frontier point calculates | Time (second) |
| Maple | 114997 | 4.2354 | 2810 | 1814.16 |
| Blue or green paulownia | 86675 | 0.41399 | 4291 | 2131.03 |
Our other method is with the skeleton of trees leaf point cloud to be classified, and as shown in Figure 7, the method above using at each sorted set is then calculated its corresponding profile.The result is that the different angles of same experimental result are showed shown in Fig. 8 a, Fig. 8 b and Fig. 8 c.Fig. 9 a, Fig. 9 b and Fig. 9 c are that list contrasts with automatic Alpha's shape (ahpha-shape) method result calculated.Figure 13 is the detail view of crown appearance.
The characteristic of method of the present invention and innovation are the characteristics according to trees, choose root node earlier, utilize the range conversion method to try to achieve the bee-line that other puts root node, use this bee-line then and divide two classes, trunk data and leaf data whole cloud data.The bee-line that uses root node is asked the trunk data is carried out cluster analysis, tells big class earlier, and to the data in each big class, cluster is told subclass once more, and this subclass representative belongs to different branches.Obtain the representative point of each subclass, and form skeleton, estimate classification leaf cloud data, then sorted each cloud data is obtained its corresponding shape, obtain the external shape of whole tree with this.
In the modeling of a lot of trees, the external shape of this tree crown can effectively provide information, express structure and the geometric properties of trees, the accuracy of aided modeling, and the external shape of tree crown is for the resource of research trees, measure the geometric scale of trees, for the trees modeling software provides more data.
The above; only be the embodiment among the present invention; but protection scope of the present invention is not limited thereto; anyly be familiar with the people of this technology in the disclosed technical scope of the present invention; can understand conversion or the replacement expected; all should be encompassed in of the present invention comprising within the scope, therefore, protection scope of the present invention should be as the criterion with the protection domain of claims.
Claims (11)
1. the crown appearance extract method based on shape analysis is characterized in that, the step of this method comprises:
Step S1: with the single face of laser scanner scans plant, the scan-data of acquisition is called cloud data;
Step S2: from cloud data, find root node, method with k nearest neighbor is carried out range conversion to whole cloud data, construct point in these cloud datas to the weight graph of root node, each of calculating cloud data put the minor increment of root node, utilizes this minor increment to isolate the cloud data that is positioned on the major branch from whole cloud data;
Step S3: the cloud data that is positioned on the major branch is carried out cluster analysis, generate tree-shaped dendrogram, be used to seek the bifurcation site and the tip position of body, in each cluster, according to distance cluster once more, obtain the sub-cluster in the cluster, to each sub-cluster, calculate the representative point of each sub-cluster with the method for arithmetic mean, as major branch skeleton point;
Step S4: the representative point of all sub-clusters is formed set, have centrality, bifurcated and topological internuncial major branch skeleton with all representative point set structures again;
Step S5: from the leaf node or the title tip position on major branch skeleton top, each branch of skeleton is classified, make sorted each branch length reach certain standard; On the basis that the major branch skeleton is classified, the leaf cloud data that is positioned at tree crown in the cloud data is classified;
Step S6: each sorted leaf cloud data is calculated their shape respectively, at first one of them sorted some cloud is carried out three-dimensional Delaunay triangulation, set up the topological structure of each data point in the some cloud, obtain tri patch and tetrahedron, these tetrahedrons and tri patch are calculated their circumsphere and external radius of a circle respectively, obtain comprising the smallest interval [A of these radiuses, B], the parameter Alpha who gets in Alpha's shape method is interval [A, B] intermediate point, tetrahedron of trying to achieve and tri patch are classified, tetrahedron is categorized as outside tetrahedron and inner tetrahedron, and tri patch is categorized as inner tri patch, the outside tri patch and the tri patch on border, if what the tri patch on border was formed is a closed body, so parameter Alpha is got [A, B] mid point between lower region continues this process, up to the difference of two adjacent parameters less than given abundant little value, fully little value is one of percentage of getting point-to-point transmission minor increment in the scan-data, at this time in these two parameters, the set of the border tri patch of a bigger correspondence is exactly the external shape of whole tree crown.
2. crown appearance extract method according to claim 1 is characterized in that, the described root node that finds from cloud data is that the method according to Y coordinate minimum in the data point finds root node.
3. crown appearance extract method according to claim 1, it is characterized in that, point in these cloud datas of described structure is to the weight graph of root node, it is method by k nearest neighbor, find K the point that range data point is nearest, point that these range data points are nearest and data point couple together the formation weight graph, and the power in the weight graph is exactly the distance of point-to-point transmission.
4. crown appearance extract method according to claim 1, it is characterized in that, each of described calculating cloud data put the minor increment of root node, be on the basis of known weight graph, calculating from root node to the data point path power and minimum path, this power and be exactly the minor increment of this data point to root node.
5. crown appearance extract method according to claim 1, it is characterized in that, described cluster analysis is to carry out cluster analysis according to each the data point p in the major branch cloud data to the known minor increment of root node, forms big cluster, and the point in each big cluster belongs to certain distance range.
6. crown appearance extract method according to claim 1 is characterized in that described cluster once more is in each the big cluster that obtains, and cluster analysis is carried out in the applications distances conversion once more, obtains sub-cluster.
7. crown appearance extract method according to claim 1, it is characterized in that, structure major branch skeleton, be with each sub-cluster core out point, then with the point of Y coordinate minimum in these central points as root node, calculate in these central points other and put the minimum distance path of this root node, the point on the access path constructs this major branch skeleton in order.
8. crown appearance extract method according to claim 1, it is characterized in that, described each branch to skeleton classifies, at first find each tip representative point of major branch skeleton, the tip point of choosing Y coordinate maximum is as starting point, opposite direction along major branch skeleton root node to the shortest path of this starting point moves, when arriving first bifurcation, calculate, calculate the length in present path, if the length in path is greater than specified value 1 at present, then at present be classified as a class through the point on the path, choosing next untreated and Y coordinate then is that maximum tip point carries out the same path of recalling, otherwise calculate with first bifurcation be initial point all son branches length and; As if all length of sub of initial point with greater than specified value 1, be that representative point on all son branches of initial point is classified as a class then with first bifurcation, otherwise from present bifurcation, continue to date back to next bifurcation along minimal path, with length of top all branch and add first bifurcation with the distance between bifurcation at present get up till now path with, as if this present path with greater than specified value, all representative points on the then present path are classified as a class, otherwise calculate all with present bifurcation be initial point all son branches length and, as if length of these all son branches with greater than specified value 1, then the representative point with present path process is classified as a class, otherwise just continue to recall, up to the length of all son branches with greater than specified value 1 along shortest path.
9. crown appearance extract method according to claim 1, it is characterized in that, described the leaf cloud data that is positioned at tree crown in the cloud data is classified, on the basis of the major branch skeleton being classified, arrive the nearest principle of major branch skeleton according to each data point in the leaf cloud data, whole leaf cloud data is classified, be classified as a class to the nearest leaf number strong point of the same class of major branch skeleton branch, otherwise just adhere to different classes separately.
10. crown appearance extract method according to claim 1, it is characterized in that, the tetrahedron of described outside and inner tetrahedron, be that tetrahedron has unique circumsphere, if this circumsphere radius is greater than parameter Alpha given in advance, then be outside tetrahedron, otherwise be exactly inner tetrahedron.
11. crown appearance extract method according to claim 1, it is characterized in that, the tri patch of the tri patch of described inside, outside and the tri patch on border, each triangular facet has unique circumscribed circle, and this circumradius is one of feature of this triangular facet, triangular facet is divided into two classes, is positioned on the convex closure and is positioned at convex closure inside; For the triangular facet that is positioned on the convex closure, if the tetrahedron at its place is outside tetrahedron, then this triangular facet is outside triangular facet, otherwise it is the triangular facet on border; For the triangular facet that is positioned at convex closure inside, it is shared by two tetrahedrons, these two adjacent tetrahedrons are if it all is outside tetrahedron, it then on this triangular facet outside triangular facet, if these two tetrahedrons all are inner tetrahedrons, then this triangular facet is exactly inner triangular facet, otherwise this triangular facet is exactly the triangular facet on border.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009102427509A CN101783016B (en) | 2009-12-16 | 2009-12-16 | Crown appearance extract method based on shape analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009102427509A CN101783016B (en) | 2009-12-16 | 2009-12-16 | Crown appearance extract method based on shape analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101783016A CN101783016A (en) | 2010-07-21 |
| CN101783016B true CN101783016B (en) | 2011-11-30 |
Family
ID=42523002
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009102427509A Expired - Fee Related CN101783016B (en) | 2009-12-16 | 2009-12-16 | Crown appearance extract method based on shape analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101783016B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103136788B (en) * | 2013-03-04 | 2016-01-20 | 重庆大学 | The visual method for reconstructing of a kind of three-dimensional blood vessel bifurcation |
| MX2016000642A (en) * | 2013-08-16 | 2016-09-22 | Landmark Graphics Corp | Identifying matching properties between a group of bodies representing a geological structure and a table of properties. |
| CN103544700B (en) * | 2013-10-11 | 2016-08-17 | 中国科学院深圳先进技术研究院 | Motion plant point cloud data consistency framework extraction method and system |
| CN103871100B (en) * | 2014-04-02 | 2016-09-21 | 中国科学院自动化研究所 | Tree modelling method for reconstructing based on a cloud Yu data-driven |
| CN109285217B (en) * | 2018-09-10 | 2021-01-29 | 中国科学院自动化研究所 | Multi-view image-based procedural plant model reconstruction method |
| CN109919913B (en) * | 2019-02-01 | 2020-12-08 | 浙江大学 | Coronary artery radius calculation method, terminal and storage medium |
| CN111025323B (en) * | 2020-02-17 | 2021-02-26 | 广西大学 | Centering method of cylindrical hedge trimmer based on multi-line laser radar |
| CN112348829B (en) * | 2020-11-02 | 2022-06-28 | 东华理工大学 | Method for separating branches and leaves of ground LiDAR point cloud based on modal point evolution |
| CN112967333B (en) * | 2021-02-04 | 2024-02-09 | 重庆大学 | Complex point cloud skeleton extraction method and system based on grading |
| CN112819963B (en) * | 2021-02-20 | 2022-04-26 | 华中科技大学鄂州工业技术研究院 | Batch differential modeling method for tree branch model and related equipment |
| CN115422856B (en) * | 2022-09-05 | 2023-08-08 | 青岛埃米博创医疗科技有限公司 | CFD teaching-oriented teaching blood vessel model generation method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101488226A (en) * | 2008-01-16 | 2009-07-22 | 中国科学院自动化研究所 | Tree measurement and reconstruction method based on single three-dimensional laser scanning |
-
2009
- 2009-12-16 CN CN2009102427509A patent/CN101783016B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101488226A (en) * | 2008-01-16 | 2009-07-22 | 中国科学院自动化研究所 | Tree measurement and reconstruction method based on single three-dimensional laser scanning |
Non-Patent Citations (5)
| Title |
|---|
| Bucksch, A, etal.《A skeletonization method for point cloud processing 》.《ISPRS Journal of Photogrammetry & Remote Sensing》.2008,115-127. * |
| M. Rutzinger,etal.《DETECTION AND MODELLING OF 3D TREES FROM MOBILE LASER SCANNING DATA 》.《International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences》.2010,第XXXVIII卷520-525. * |
| Pfeifer.N., etal.《Automatic reconstruction of single trees from terrestrial laser scanner data》.《International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences》.2004,第35卷114−119. * |
| Vosselman.G, etal.《Recognising structure in laser scanner point clouds》.《 International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences》.2004,第46卷33-38. * |
| Yotam Livny1,etal.《Automatic Reconstruction of Tree Skeletal Structures from Point Clouds》.《ACM transactions on Graphics (Proceedings SIGGRAPH ASIA 2010))》.2010,1-8. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101783016A (en) | 2010-07-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101783016B (en) | Crown appearance extract method based on shape analysis | |
| CN101887596B (en) | Three-dimensional model reconstruction method of tree point cloud data based on partition and automatic growth | |
| CN105574929A (en) | Single vegetation three-dimensional modeling method based on ground LiDAR point cloud data | |
| CN103871100B (en) | Tree modelling method for reconstructing based on a cloud Yu data-driven | |
| CN101763652B (en) | Three-dimensional framework fast extraction method based on branch feathers | |
| CN103098100B (en) | Based on the three-dimensional model shape analysis method of perception information | |
| CN104331699B (en) | A kind of method that three-dimensional point cloud planarization fast search compares | |
| CN103310481B (en) | A kind of point cloud compressing method based on fuzzy entropy iteration | |
| CN110136246A (en) | Three-dimension Tree Geometric model reconstruction method based on class trunk point | |
| CN103942838A (en) | Point cloud data based single tree three-dimensional modeling and morphological parameter extracting method | |
| CN101650836B (en) | Self-adaptive gridding method and self-adaptive gridding system of geometric curved surfaces of three-dimensional plant organs | |
| CN107146280A (en) | A kind of point cloud building method for reconstructing based on cutting | |
| CN103870845A (en) | Novel K value optimization method in point cloud clustering denoising process | |
| CN104851126B (en) | Threedimensional model dividing method and device based on generalized cylinder | |
| CN103258345A (en) | Method for extracting parameters of tree branches based on ground laser radar three-dimensional scanning | |
| CN104392486A (en) | Point-cloud scene rebuilding method | |
| CN106651900A (en) | Three-dimensional modeling method of elevated in-situ strawberry based on contour segmentation | |
| CN102254343A (en) | Convex hull and OBB (Oriented Bounding Box)-based three-dimensional grid model framework extracting method | |
| CN104166748A (en) | Forest stand growth modeling method based on relation model | |
| CN103559705A (en) | Computer method for comparing similarity of different plant forms | |
| CN105279794B (en) | Reservoir core multi-tissue model construction method based on Micro-CT technologies | |
| CN115018982A (en) | Digital tree twinning method based on foundation laser radar point cloud | |
| Huang et al. | A 3D individual tree modeling technique based on terrestrial LiDAR point cloud data | |
| CN105139452B (en) | A kind of Geologic Curve method for reconstructing based on image segmentation | |
| CN109241628A (en) | Three-dimensional CAD model dividing method based on Graph Spectral Theory and cluster |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20111130 Termination date: 20211216 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |