CN104077770B - A kind of leaf image local auto-adaptive tree structure feature matching method - Google Patents
A kind of leaf image local auto-adaptive tree structure feature matching method Download PDFInfo
- Publication number
- CN104077770B CN104077770B CN201410271272.5A CN201410271272A CN104077770B CN 104077770 B CN104077770 B CN 104077770B CN 201410271272 A CN201410271272 A CN 201410271272A CN 104077770 B CN104077770 B CN 104077770B
- Authority
- CN
- China
- Prior art keywords
- point
- characteristic
- feature
- characteristic point
- coupling
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of leaf image local auto-adaptive tree structure feature matching method, construct DOG Feature Points Matching substrates, using substrate between characteristic point that is having mated and not mating extraction feature cable architecture, can be constrained with the spatial distribution of adaptive combination partial descriptions sub-information and characteristic point by feature cable architecture, form dynamic referring-to relation tree.Effectively prevent existing method irregular deformation occurs in plant leaf blade, the scaling under strong noise point and rotation combination, the problem that obtains matched well result is locally difficult to when high repeat pattern, achieve the plant leaf portion Image Feature Matching method of high robust, its adaptability and the degree of accuracy have reached the requirement of practical application.Meanwhile, the present invention be easy to be combined based on the Feature Descriptor of picture with existing in a large number, expand all kinds of feature matching methods based on image with different qualities, for computer vision in classical problem have great significance.
Description
Technical field
The invention belongs to area of pattern recognition, is related to computer graphics techniques field, more particularly to a kind of plant leaf blade figure
As local auto-adaptive tree structure feature matching method.
Background technology
Image Feature Matching is the corresponding relation of the set of characteristic points that sets up in two pictures, is image mosaic, scene knowledge
Not, the requisite base support section of the application such as image retrieval, three-dimensional modeling.With the work with relatively regular geometry
Industry part is compared, and under nature, the surface profile of plant and resemblance are all increasingly complex.When processing feature is mated, above-mentioned
The performance of difficult point is also more projected.As the blade of one of important plant organ, which is carried out based on Image Feature Matching, tool
Body may be summarized to be several difficult points as follows：(1) plant leaf blade surface has obvious local similar feature, such as texture, blade
Symmetry of color and spatial distribution etc..Such high local repeat pattern is always the difficulty in existing Feature Correspondence Algorithm
Point；(2), in the case of shooting naturally, plant leaf blade itself has the deformation such as significantly irregular coiling, distortion.Due to deformation not
Systematicness, brings difficulty to the unified constraint and specific feature interpretation of Matching Model；(3) under nature, plant leaf portion
The actual photographed environment of image is easily affected by factors such as focal length, shooting angle, so when characteristic matching is actually carried out,
Often to be processed is the noise spot and rotation, the combinatorial problem of scale transformation for mixing in a large number, and this is equally brought to coupling work
Difficult.As above a few class difficult points cause the feature point selection method that plant leaf portion Image Feature Matching has not only needed, in addition it is also necessary to
The feature matching method of more robustness and the degree of accuracy.
In sum, existing plant leaf portion Image Feature Matching method is all under specific shooting condition due to being directed to mostly
Object, disturbs the features such as conversion of little, locus is simple, blade surface deformation is relatively regular with uniform illumination, noise, therefore
Still have several drawbacks, be mainly manifested in：(1) local irregularities' coiling in the face of plant leaf blade, twist distortion, existing method are difficult
It is described to build unified model, it is impossible to introduce effective space geometry constraint rule to improve characteristic matching precision；(2)
In the case where introducing regular geometric constraint, the similitude of simple use local description processes matching problem, it is difficult to
Obtain satisfied matching result；(3) when process much noise point mixes and rotates the combination of scale transformation, the precision of coupling is difficult
To improve.
Content of the invention
For solving the strong noise point introduced during actual photographed present in prior art interference and image rotation, scaling
The problems such as combined transformation, it is an object of the invention to provide a kind of leaf image local auto-adaptive tree structure characteristic matching
Method, can flexibly carry out characteristic matching, and so as to solve in actual characteristic matching application, existing method is in these problems
The problem for being difficult to obtain matched well.
To achieve these goals, the leaf image local auto-adaptive tree structure characteristic matching side that the present invention is provided
Method, constructs DOG Feature Points Matching substrates, using substrate between characteristic point that is having mated and not mating extraction feature knot
Structure, can be constrained with the spatial distribution of adaptive combination partial descriptions sub-information and characteristic point by feature cable architecture, be formed dynamic
The referring-to relation tree of state, specifically includes following steps：
Step S1：The multi-resolution Gaussian difference pyramid substrate of leaf image is built, by the feature for having extracted
Point is mapped in Gaussian difference substrate；
Step S2：In the case where step S1 completes characteristic point mapping, combining space information is set up in set of characteristic points
Go out one and refer to triangular structure；
Step S3：Using the reference triangular structure of step S2, local scale information and the rotation information of blade is calculated,
According to this information architecture partial transformation model, complete the space to characteristic point and divide；
Step S4：From on the reference triangle of step S3, certain point carries out the growth coupling of characteristics tree, points out from known
Send out constantly adaptive to unknown point extension.By combine local feature description, the characteristic point top-stitching feature built in substrate and
Spatial distribution is screened, until completing whole tree structure coupling；
Step S5：The process of S2-S4 is repeated, until filter out in the characteristic point that never cannot mate that triangle is referred to
Structure, coupling are completed.
Wherein, step S1 is comprised the following steps that：
Step S11：By formula：
Gaussian scale-space image is calculated, whereinRepresent to the convolution operation on x and y directions, G (x, y, δ) is one
The Gaussian function of mutative scale.
Step S12：The Gaussian kernel convolved image of two neighboring different scale is subtracted each other, obtain Gauss error image D (x, y,
δ).
Step S13：Introducing two neighboring metric space multiple constant k carries out approximation sample to adjacent picture, by formula：
Pyramid Gauss error image is obtained, is 1.5 or 2 wherein with image drop sampling multiplying power k.
Wherein, step S2 is comprised the following steps that：
Step S21：The set of characteristic points of artwork and the set of characteristic points of target figure is obtained, from the set of characteristic points of artwork
The continuous one section of characteristic point subset of random extraction, using local feature description's, random with target signature point set successively
The continuous characteristic point subset for extracting is mated.
Step S22：Small-scale candidate matches set is formed using ratio screening method, three pairs of characteristic points are therefrom randomly selected,
The correctness that coupling is further verified by the similitude of the three pairs of characteristic edges for being formed.If there is two characteristic edges all to test on triangle
Failure is demonstrate,proved, then deletes the characteristic point folded by two sides, random choosing is then put to continuation again from candidate matches set of characteristic points
Match somebody with somebody, until three pairs of characteristic edges refer to triangular structure by checking so as to be formed.
Wherein, step S3 is comprised the following steps that：
Step S31：Using the constraint with reference to triangle, characteristic point to be matched is carried out drawing according to space distribution situation
Point.
Step S32：Assume that characteristic point distribution to be matched is relatively uniform in figure, calculates optimal net according to reference to triangle
Lattice partitioning parameters, then complete characteristic point distributing position to the mapping of the mesh space of coarseness.
Wherein, step S4 is comprised the following steps that：
Step S41：If ASFor in artwork set of characteristic points with reference to a bit, making on triangle as entirely mating the tree root of tree.
Step S42：From ASSet out, with radius rbranchIn the range of arbitrarily select a point DS, using artwork in distance
VectorAnd characteristic curveA qualified match point D is found out in target figureT.
Step S43：With D in artworkSCentered on, do not include D at whichSFind out in the eight neighborhood of the grid scope at place next
Point ES, relative obtains Corresponding matching position E in target figurevir, with EvirCentered on extract adjacent mesh, take out grid in
All characteristic points are alternative features point.
Step S44:Go out to select optimal match point using the characteristic curve Policy Filtering of local feature and dynamic construction, such as E1 T.
Then, characteristic point E newly to mate1 TContinue coupling for new originating point, the new characteristic point of expanded search and coupling, until complete whole
Coupling tree coupling work.
Present invention effectively prevents scaling and rotation of the existing method under the irregular deformation of plant leaf blade generation, strong noise point
Turn combination, the problem that obtains matched well result is locally difficult to when high repeat pattern, it is achieved that the plant of high robust
Leaf portion Image Feature Matching method, its flexible adaptability and the higher degree of accuracy have reached the requirement of practical application.Meanwhile,
The present invention is easy to be combined based on the Feature Descriptor of picture with existing in a large number, expand all kinds of with different qualities based on figure
The feature matching method of picture, for the classics in the computer visions such as target following, scene modeling, image mosaic, image retrieval
Problem has great significance.
Description of the drawings
Method flow diagrams of the Fig. 1 for leaf image local auto-adaptive tree structure feature matching method.
Fig. 2 is the phoenix tree leaves with irregular curling and deformation.
1st, 3,5,7 layer feature substrate stratal diagrams of the Fig. 3 for Chinese parasol leaf piece.
Feature point extraction and mark figure of the Fig. 4 for Chinese parasol leaf piece.
Fig. 5 is built for the characteristic point subdivision of Chinese parasol leaf piece and with reference to triangle.
Fig. 6 is the tree-like coupling under the irregular deformation and rotation of Chinese parasol leaf piece.
Feature Points Matching line graphs of the Fig. 7 for Chinese parasol leaf piece.
Fig. 8 is the coupling under the high repeat pattern of maple blade and rotation.
Specific embodiment
Each detailed problem in the technology of the present invention method involved is described in detail below in conjunction with accompanying drawing.It is to be noted that
Described embodiment is intended merely to facilitate the understanding of the present invention, and does not play any restriction effect to which.
A kind of leaf image local auto-adaptive tree structure feature matching method, its flow process are as shown in Figure 1.This feature
Matching process constructs DOG Feature Points Matching substrates, extracts spy using substrate between characteristic point that is having mated and not mating
Cable architecture is levied, can be constrained with the spatial distribution of adaptive combination partial descriptions sub-information and characteristic point by feature cable architecture,
Form dynamic referring-to relation tree.
Specifically include following steps：
Step S1：Build the multi-resolution Gaussian difference pyramid substrate of leaf portion image.
By taking the phoenix tree leaves with irregular curling and deformation as shown in Figure 2 as an example, the characteristic point for having extracted is mapped
To in basis space：Gaussian scale-space image is calculated firstWhereinRepresent to x and
Convolution operation on y directions, G (x, y, δ) are the Gaussian functions of a mutative scale.Gauss by two adjacent different scales
The image subtraction of core convolution, draws Gauss error image D (x, y, δ), is re-introduced into two neighboring metric space multiple constant k to phase
Adjacent picture carries out approximation sample, passes throughMeet according to specific needs
Scaling and deformation requirements, select suitable pyramid level.The pyramid response image for finally giving.Wherein, image
Down-sampled multiplying power constant k values are 1.5 or 2.
This down-sampled pyramid of 7 floor height of selection as shown in Figure 3, during k=1.5, the base that the 1st, 3,5,7 layers of phoenix tree leaves
Bottom.
Step S2：Never in matching characteristic point, screening foundation refers to triangular structure.
In the case of completing characteristic point mapping, the set of characteristic points of artwork and the set of characteristic points of target figure is obtained, from original
The continuous one section of characteristic point subset of random extraction in the set of characteristic points of figure, using local feature description's, successively with target
The continuous characteristic point subset that randomly selects in set of characteristic points is mated.The characteristic point of Chinese parasol leaf piece as shown in Figure 4 is carried
Take and mark figure.
Set of characteristic points to be matched is divided into the less subset of granularity, by the quick of simple ratio method between subclass
Filter out candidate matches point, randomly select 3 spaces be relatively close to and Feature Descriptor successful match candidate point.Finally,
Three sides of a triangle are judged using the route characteristic built in Gaussian difference pyramid substrate and screened, if on triangle
Have two characteristic edges all authentication faileds, then delete two sides folded by characteristic point, from candidate matches set of characteristic points again with
Machine choosing is then put to continuing coupling, until three pairs of characteristic edges refer to triangular structure by checking so as to be formed.Finally give one
Individual stable local-reference triangular structure.
The characteristic point subdivision of Chinese parasol leaf piece as shown in Figure 5 and with reference to triangle build.
Step S3：Local scale and rotation matching model are built using feature triangle, completes to cut open space characteristics point
Point.
Using with reference to the corresponding three opposite vertexes (A of triangleS,BS,CS)→(AT,BT,CT) artwork is calculated to target figure
Local scale ratio, according to optimal coupling basalis pair in local scale gain of parameter characteristic matching, follow-up characteristic curve is carried
Take and all extract from this feature basalis, to overcome the matching problem in the case of scaling.
Calculate with reference to respective centre of form O of triangleSAnd OT, further obtain the characteristic growth reference direction of local
WithIn the point search matching process for combining local space, when direction of search judgement is run into, by relatively each
Growth reference direction judged and screened, with effectively overcome coupling in Rotation.
Initialization pantograph ratio and rotary reference direction, it is assumed that characteristic point distribution is relatively uniform, by regular grid to feature
Point distributed areas carry out uniform subdivision.All of characteristic point numbering is mapped to by one lattice structure of setting corresponding
In grid, convenient follow-up quickly space querying and search are compared.Division principle is carried out to the space of characteristic point is：Reduce as far as possible
Feature points in each cell, improve matching speed；Appropriate retains enough cell sizes to increase characteristic edge
Discrimination, increases the differentiation degree of single match.By mesh mapping, characteristic point more will be held in follow-up space search coupling
Local Search and inquiry are carried out easily, while avoiding substantial amounts of erroneous matching.The irregular shape of Chinese parasol leaf piece as shown in Figure 6
Tree-like coupling under becoming and rotating.
Step S4, certain point on reference to triangle, the growth for carrying out characteristics tree using tree-like growth matching strategy
Match somebody with somebody.
If ASIt is a bit on the reference triangle extracted from artwork, makes the tree root which is whole coupling tree.From ASSet out,
With radius rbranchIn the range of arbitrarily select a point DS, using artwork in characteristic point vectorIn base and between characteristic point
The characteristic curve formed on bottomFilter out qualified match point D in target figureT.
With D in artworkSCentered on, do not include D at whichSSubsequent point E is found out in the eight neighborhood of the grid scope at placeS, relative
Obtain Corresponding matching position E in target figurevir, with EvirCentered on extract adjacent mesh, take out grid in all features
Point is alternative features point.
Half r with the regular grid length of sideleafCircle is drawn for radius, the grid intersected with circle is extracted, the institute in grid is taken out
It is alternative characteristic point E to be matched to have characteristic pointn T, filtered out most preferably using the sub and dynamic Characteristic constraint line of local feature description
Match point E1 T, with characteristic point E that newly mates1 TFor new originating point, continue Downward match.Searched with the dynamic of this increment type expansion
Rope decision search whole region, till whole coupling tree completes all couplings that can be extended.
The Feature Points Matching line graph of Chinese parasol leaf piece as shown in Figure 7.
Step S5, Iterative matching.
Step S2～S4 is repeated, the characteristic point for completing to mate is rejected, in remaining point is continually looked for referring to three
Angular structure, sets up new Local Search Matching Model and scans for, until cannot extract new reference triangular structure being
Only, now coupling is finally completed.Coupling under the high repeat pattern of maple blade as shown in Figure 8 and rotation.
The above, the only specific embodiment in the present invention, but protection scope of the present invention is not limited thereto, and appoints
What be familiar with the people of the technology disclosed herein technical scope in, conceivable conversion and remodeling can be managed, also should be regarded as by
It is included within the scope of the present invention.
Claims (5)
1. a kind of leaf image local auto-adaptive tree structure feature matching method, it is characterised in that construct DOG first
Feature Points Matching substrate, using substrate between characteristic point that is having mated and not mating extraction feature cable architecture, by feature
Cable architecture can be constrained with the spatial distribution of adaptive combination partial descriptions sub-information and characteristic point, form dynamic referring-to relation
Tree, specifically includes following steps：
Step S1：The multi-resolution Gaussian difference pyramid substrate of leaf image is built, the characteristic point for having extracted is reflected
It is mapped in Gaussian difference substrate；
Step S2：In the case where step S1 completes characteristic point mapping, combining space information sets up out one in set of characteristic points
Individual reference triangular structure；
Step S3：Using the reference triangular structure of step S2, local scale information and the rotation information of blade is calculated, according to
This information architecture partial transformation model, completes the space to characteristic point and divides；
Step S4：From on the reference triangle of step S3, certain point carries out the growth coupling of characteristics tree, from known point not
Disconnected adaptive to unknown point extension, by with reference to the characteristic point top-stitching feature and space built in local feature description, substrate
Distribution is screened, until completing whole tree structure coupling；
Step S5：The process of S2-S4 is repeated, until filter out in the characteristic point that never cannot mate tying with reference to triangle
Structure, coupling are completed.
2. leaf image local auto-adaptive tree structure feature matching method according to claim 1, its feature exist
In：Step S1 is comprised the following steps that：
Step S11：By formula：
Gaussian scale-space image is calculated, whereinRepresent to the convolution operation on x and y directions, G (x, y, δ) is a mutative scale
Gaussian function；
Step S12：The Gaussian kernel convolved image of two neighboring different scale is subtracted each other, Gauss error image D (x, y, δ) is obtained；
Step S13：Introducing two neighboring metric space multiple constant k carries out approximation sample to adjacent picture, by formula：
Pyramid Gauss error image is obtained, is 1.5 or 2 wherein with image drop sampling multiplying power k.
3. leaf image local auto-adaptive tree structure feature matching method according to claim 1, its feature exist
In：Step S2 is comprised the following steps that：
Step S21：The set of characteristic points of artwork and the set of characteristic points of target figure is obtained, random from the set of characteristic points of artwork
The continuous one section of characteristic point subset of extraction, using local feature description's, successively with randomly select in target signature point set
Continuous characteristic point subset mated；
Step S22：Small-scale candidate matches set is formed using ratio screening method, three pairs of characteristic points is therefrom randomly selected, is passed through
The similitude of the three pairs of characteristic edges for being formed further verifies the correctness of coupling, if there is two characteristic edges all to verify mistake on triangle
Lose, then delete the characteristic point folded by two sides, random selection point mates to continuation again from candidate matches set of characteristic points, directly
To three pairs of characteristic edges by checking, triangular structure is referred to so as to be formed.
4. leaf image local auto-adaptive tree structure feature matching method according to claim 1, its feature exist
In：Step S3 is comprised the following steps that：
Step S31：Using the constraint with reference to triangle, characteristic point to be matched is divided according to space distribution situation；
Step S32：Assume that characteristic point distribution to be matched is relatively uniform in figure, calculates optimal grid point according to reference to triangle
Parameter is cut, and characteristic point distributing position is then completed to the mapping of the mesh space of coarseness.
5. leaf image local auto-adaptive tree structure feature matching method according to claim 1, its feature exist
In：Step S4 is comprised the following steps that：
Step S41：If ASFor in artwork set of characteristic points with reference to a bit, making on triangle as entirely mating the tree root of tree；
Step S42：From ASSet out, with radius rbranchIn the range of arbitrarily select a point DS, using artwork in distance vectorAnd characteristic curveA qualified match point D is found out in target figureT；
Step S43：With D in artworkSCentered on, do not include D at whichSSubsequent point E is found out in the eight neighborhood of the grid scope at placeS,
Relative obtains Corresponding matching position E in target figurevir, with EvirCentered on extract adjacent mesh, take out all in grid
Characteristic point is alternative features point；
Step S44:Go out optimal match point, such as E using the characteristic curve Policy Filtering of local feature and dynamic construction1 T, then, with new
Characteristic point E of coupling1 TContinue coupling, the new characteristic point of expanded search and coupling for new originating point, until completing whole coupling tree
Coupling work.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410271272.5A CN104077770B (en) | 2014-06-17 | 2014-06-17 | A kind of leaf image local auto-adaptive tree structure feature matching method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410271272.5A CN104077770B (en) | 2014-06-17 | 2014-06-17 | A kind of leaf image local auto-adaptive tree structure feature matching method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104077770A CN104077770A (en) | 2014-10-01 |
| CN104077770B true CN104077770B (en) | 2017-03-15 |
Family
ID=51599012
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410271272.5A Expired - Fee Related CN104077770B (en) | 2014-06-17 | 2014-06-17 | A kind of leaf image local auto-adaptive tree structure feature matching method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104077770B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104657986B (en) * | 2015-02-02 | 2017-09-29 | 华中科技大学 | A kind of quasi- dense matching extended method merged based on subspace with consistency constraint |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5625717A (en) * | 1992-06-24 | 1997-04-29 | Mitsubishi Denki Kabushiki Kaisha | Image processing device for processing grey level images |
| JPH09305766A (en) * | 1996-05-17 | 1997-11-28 | Meidensha Corp | Recognition method for two-dimensional object |
| JP2010086540A (en) * | 2008-10-01 | 2010-04-15 | Fuji Xerox Co Ltd | Method, system and program for generating feature descriptor |
| CN103065135A (en) * | 2013-01-25 | 2013-04-24 | 上海理工大学 | License number matching algorithm based on digital image processing |
| CN103177444A (en) * | 2013-03-08 | 2013-06-26 | 中国电子科技集团公司第十四研究所 | Automatic SAR (synthetic-aperture radar) image rectification method |
| CN103177434A (en) * | 2013-04-10 | 2013-06-26 | 浙江大学 | Method for calculating rotation angle of apple by using stem images of apple |
| CN103456022A (en) * | 2013-09-24 | 2013-12-18 | 中国科学院自动化研究所 | High-resolution remote sensing image feature matching method |
-
2014
- 2014-06-17 CN CN201410271272.5A patent/CN104077770B/en not_active Expired - Fee Related
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5625717A (en) * | 1992-06-24 | 1997-04-29 | Mitsubishi Denki Kabushiki Kaisha | Image processing device for processing grey level images |
| JPH09305766A (en) * | 1996-05-17 | 1997-11-28 | Meidensha Corp | Recognition method for two-dimensional object |
| JP2010086540A (en) * | 2008-10-01 | 2010-04-15 | Fuji Xerox Co Ltd | Method, system and program for generating feature descriptor |
| CN103065135A (en) * | 2013-01-25 | 2013-04-24 | 上海理工大学 | License number matching algorithm based on digital image processing |
| CN103177444A (en) * | 2013-03-08 | 2013-06-26 | 中国电子科技集团公司第十四研究所 | Automatic SAR (synthetic-aperture radar) image rectification method |
| CN103177434A (en) * | 2013-04-10 | 2013-06-26 | 浙江大学 | Method for calculating rotation angle of apple by using stem images of apple |
| CN103456022A (en) * | 2013-09-24 | 2013-12-18 | 中国科学院自动化研究所 | High-resolution remote sensing image feature matching method |
Non-Patent Citations (4)
| Title |
|---|
| A Double-layer Word Segmentation Combined with Local Ambiguity Word Grid and CRF;Lei Chen 等;《Computer Science and Technology》;20130331;第20卷(第8期);第1-8页 * |
| MULTI-SPECTRAL REMOTE SENSING IMAGE REGISTRATION VIA SPATIAL RELATIONSHIP ANALYSIS ON SIFT KEYPOINTS;Mahmudul Hasan 等;《Geoscience and Remote Sensing Symposium》;20100730;第1011-1014页 * |
| 基于分水岭算法的作物病害叶片图像分割方法;任玉刚 等;《计算机应用》;20120301;第32卷(第3期);第752-755页 * |
| 改进仿射尺度不变特征变换算法的图像配准;范雪婷 等;《计算机应用》;20140510;第34卷(第5期);第1449-1452页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104077770A (en) | 2014-10-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Livny et al. | Automatic reconstruction of tree skeletal structures from point clouds | |
| Kokkinos et al. | Intrinsic shape context descriptors for deformable shapes | |
| CN104867126B (en) | Based on point to constraint and the diameter radar image method for registering for changing region of network of triangle | |
| Gain et al. | Terrain sketching. | |
| CN102834845B (en) | The method and apparatus calibrated for many camera heads | |
| CN105139379B (en) | Based on the progressive extracting method of classified and layered airborne Lidar points cloud building top surface | |
| CN109816706B (en) | Smooth constraint and triangulation equal-proportion image pair dense matching method | |
| KR101182683B1 (en) | A Visual Shape Descriptor Generating Method Using Sectors and Shape Context of Contour Lines and the Recording Medium thereof | |
| Ying et al. | An intrinsic algorithm for parallel poisson disk sampling on arbitrary surfaces | |
| CN110168608A (en) | The system that 3 dimension words for obtaining physical object indicate | |
| Chen et al. | Robust affine-invariant line matching for high resolution remote sensing images | |
| CN109285117A (en) | A kind of more maps splicing blending algorithm based on map feature | |
| CN107180436A (en) | A kind of improved KAZE image matching algorithms | |
| CN107895370B (en) | Saddle point extraction method based on contour line | |
| CN116452644A (en) | Three-dimensional point cloud registration method and device based on feature descriptors and storage medium | |
| CN107025449A (en) | A kind of inclination image linear feature matching process of unchanged view angle regional area constraint | |
| CN107945221A (en) | A kind of three-dimensional scenic feature representation based on RGB D images and high-precision matching process | |
| CN105354829A (en) | Self-adaptive point cloud data segmenting method | |
| CN113409332B (en) | Building plane segmentation method based on three-dimensional point cloud | |
| CN103824323B (en) | A kind of three-dimensional rebuilding method based on single width two dimensional image | |
| CN104751479A (en) | Building extraction method and device based on TIN data | |
| CN102446356A (en) | Parallel self-adaptive matching method for obtaining remote sensing images with uniformly distributed matching points | |
| CN111161267A (en) | Segmentation method of three-dimensional point cloud model | |
| CN109410183A (en) | Plane extracting method, system, device and storage medium based on point cloud data | |
| CN106295674A (en) | Image target matching detection method and system based on multi-particle swarm algorithm |
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: 20170315 Termination date: 20200617 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |