CN107610120B - A kind of multiple dimensioned building surface Methodology for Entities Matching and system - Google Patents
A kind of multiple dimensioned building surface Methodology for Entities Matching and system Download PDFInfo
- Publication number
- CN107610120B CN107610120B CN201710891628.9A CN201710891628A CN107610120B CN 107610120 B CN107610120 B CN 107610120B CN 201710891628 A CN201710891628 A CN 201710891628A CN 107610120 B CN107610120 B CN 107610120B
- Authority
- CN
- China
- Prior art keywords
- matching
- point pair
- matched
- entity
- points
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 239000013598 vector Substances 0.000 claims abstract description 20
- 238000012545 processing Methods 0.000 claims abstract description 8
- 239000011159 matrix material Substances 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 239000004576 sand Substances 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 238000013519 translation Methods 0.000 claims description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 4
- 230000009897 systematic effect Effects 0.000 claims description 3
- 238000012986 modification Methods 0.000 claims description 2
- 230000004048 modification Effects 0.000 claims description 2
- 238000006386 neutralization reaction Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 6
- 238000006243 chemical reaction Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 5
- 238000001228 spectrum Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002269 spontaneous effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of multiple dimensioned building surface Methodology for Entities Matching and systems, comprising: S100 obtains feature point set to be matched from matched data source；The possible matched characteristic point pair of S200 building, calculates the main cluster confidence level vector of characteristic point pair；S300 corrects the matching relationship of characteristic point, obtains face entity matching result；This step include: S310 from may matched characteristic point centering obtain optimal matching points collection；S320 carries out robust processing with the minimum target of the population deviation of all optimal matching points, using least square method, obtains the location error of optimal matching points；Face entity matching result is obtained based on location error；S330 judges current best match point to whether the matched face entity in equal source terminates if so, matching is correct；If it is not, repeating sub-step S310.The present invention overcomes the influences of position deviation opposite Entities Matching, and propose the calculation method of multi-to-multi matching similarity.
Description
Technical Field
The invention belongs to the field of geographic information science, and particularly relates to a multi-scale building face entity matching method and system based on a pairwise constrained spectrum matching method and a least square method.
Background
Entity matching is a fundamental technology in the field of spatial data processing and application, and is widely applied to updating, maintaining and fusing spatial data. With the rise of 'spontaneous geographic information' (VGI), entity matching becomes a key technology for improving and evaluating the data quality of the VGI. However, geospatial data from different sources often are inconsistent in terms of situational, expression scale, geometry, and semantics, making it a challenge to obtain accurate matching results. Especially the matching of multi-scale data because complex matching (one-to-many and many-to-many matching) exists widely in multi-scale matching and features between matching data are more blurred.
To date, there has been a great deal of research in the field of entity matching. Obtaining candidate matching by a buffer area growing method and an area overlapping method, and then matching entities by using geometrical information, attribute information and topological information; matching entities using context information of the entities; and introducing an optimization model, a logistic regression model, a BP neural network and an SVM to improve the matching accuracy and the like. However, the buffer growing method can only obtain 1:1 matching candidates, which cannot determine which elements in the buffer need to be aggregated for matching; the area overlap method is completely useless when there is a large displacement deviation in the spatial data. Even through preprocessing such as projection conversion, format conversion and map correction, the matching entities are only roughly aligned, and coordinate deviation still exists, so that a large number of correct matches are missed, and entities with small areas are easily ignored by errors. Therefore, the complex matching in the multi-scale matching processing is difficult to effectively solve in the existing entity matching result, which affects the applicability of the matching technology.
Disclosure of Invention
The invention provides a multi-scale building face entity matching method and system based on a pairwise constrained spectrum matching method and a least square method, aiming at the problems that position deviation of a same-name entity in multi-scale matching is large and candidate matching cannot be obtained directly through an area overlapping method.
The invention provides a multi-scale building surface entity matching method, which comprises the following steps:
s100, acquiring any entity in the matched data source, and recording as SiEstablishing siThe set of all the face entities in the buffer area is marked as Bi(ii) a Extraction of siAnd BiObtaining feature points P and Q to be matched from the feature points on the surface entity outline; the characteristic points are contour points of which the steering angle is not smaller than a preset angle;
s200, constructing possibly matched feature point pairs (p)u,qv) Obtaining a characteristic point pair set L, calculating L maximization confidence coefficient indication vector x of each characteristic point pair*(ii) a Wherein p isuDenotes the u-th feature point in P, qvRepresenting the v-th characteristic point in Q, and taking 1,2, asV. taking 1,2,. and n in sequencet，nsAnd ntThe number of characteristic points in P and Q respectively;
s300, correcting the matching relation of the feature points to obtain a face entity matching result; the method further comprises the following steps:
s310 is according to x*Acquiring a best matching point pair set from the feature point pair sets which are possibly matched currently by using a greedy method, wherein the initial value of the feature point pair set which is possibly matched currently is L;
s320, aiming at the minimum total deviation of all the optimal matching point pairs, performing robust processing by using a least square method to obtain position errors of the optimal matching point pairs; obtaining a face entity matching result based on the position error;
s330, judging whether the current best matching point pair is a face entity with source matching, if so, matching correctly, and ending; if not, the face entity source collision point pairs are removed from the feature point pair set which may be currently matched, and the substep S310 is repeatedly executed.
Preferably, step S1001 is preceded by a step of preprocessing the matching data source to eliminate systematic errors of the matching data source.
Further, the steering angleWherein: (x)k-1,yk-1)、(xk,yk)、(xk+1,yk+1) Coordinates representing three adjacent contour points; θ represents the steering angle of the middle contour point of the three adjacent contour points.
Further, in step S200, the calculating an L-maximized confidence indicator vector of each feature point pair in L specifically includes:
s210, constructing an undirected graph by taking each characteristic point pair in the L as a vertex;
s220, forming a correlation matrix M by the values of all the undirected graph edges, wherein the correlation matrix is a symmetric matrix, and the diagonal element value is 0;
s230, singular value decomposition is carried out on M, and the eigenvector corresponding to the maximum eigenvalue is x*。
Further, the value M (a, b) of the undirected graph edge is w1Sdis(a,b)+w2Slen(a,b)+w3Scos(a, b), wherein M (a, b) represents the value of the undirected graph edge with the feature point pair a and b as vertices; w is a1、w2、w3For the weight coefficients, the values of which are determined by sample data training, w1+w2+w3＝1；Sdis(a,b)、Slen(a,b)、Scos(a, b) represent the similarity of the distance, length and direction of a and b, respectively.
Further, the substep S310 specifically includes:
s311 assigns the feature point pair set which is possible to be matched currently to LkTaking the current x*The median maximum value is LkThe corresponding characteristic point pair (p) in (1)max，qmax) And is stored in C, excluding LkMiddle and (p)max，qmax) Pairs of conflicting feature points, at the same time, at x*Deleting the corresponding elements of the conflicted characteristic point pairs; wherein k represents the number of iterations, and the value is taken from 0;
s312 for current x*Repeating the substep S311 until LkAll the middle characteristic point pairs are excluded, and the final C is the best matching point pair set;
and, step S320 further includes:
s321, constructing an objective function with minimum overall deviation of all best matching point pairsWherein,andrespectively representing the distance of the position error of the jth best matching point pair in the t-1 th iteration and the t-th iteration relative to the average position deviation; the average position deviation at the t-th iteration is recorded as Is composed ofThe weight of (a) is determined,a scheme weight function for IGG 1;
s322, an objective function is solved by iteration through a least square method, and the position error of the optimal matching point pair is obtained;
s323 will SiTranslation according to position error, BiS after neutralization and translationiFace entity B with overlapping rate not less than 0.5i' i.e. siA matching face entity;
in addition, step S330 specifically includes:
judging whether the current best matching point pair is a face entity with source matching, if so, matching correctly, and ending; if not, the face entity source collision point pair is removed from the feature point pair set that may be currently matched, and the sub-step S311 is repeated.
Further, the invention relates to a multi-scale building surface entity matching method, which further comprises the step of judging whether the matching degree is determined according to the total similarity of the matching entities, and the method specifically comprises the following steps:
s410 according to SiThe sequence of the middle outline points is used for sorting the final best matching point pairs in the best matching point pair set to obtain a sorted best matching point pair setWherein,
s420, calculating the total similarity of the matching surface entitiesWherein, M (c)j',cj+1') is denoted by cj' and cj+1' is the value of the undirected graph edge of the vertex; djIs composed ofAndwhen j is equal to ncWhen the temperature of the water is higher than the set temperature,
s430 judging whether the surface entities are matched according to the total similarity, if S (S)i,Bi') is not less than a preset threshold, Bi' and siMatching is successful; otherwise, the matching is unsuccessful.
The invention provides a multi-scale building surface entity matching system, which comprises:
a characteristic point extraction module for obtaining any entity in the matched data source and recording as siEstablishing siThe set of all the face entities in the buffer area is marked as Bi(ii) a Extraction of siAnd BiObtaining feature points P and Q to be matched from the feature points on the surface entity outline; the characteristic points are contour points of which the steering angle is not smaller than a preset angle;
a main clustering confidence coefficient acquisition module for constructing possibly matched feature point pairs (p)u,qv) Obtaining a characteristic point pair set L, and calculating an L maximized confidence coefficient indication vector x of the characteristic point pairs in the L*(ii) a Wherein p isuDenotes the u-th feature point in P, qvRepresenting the v-th characteristic point in Q, and taking 1,2, asV. taking 1,2,. and n in sequencet，nsAnd ntThe number of characteristic points in P and Q respectively;
the correction module is used for correcting the matching relation of the feature points to obtain a face entity matching result;
the modification module further comprises sub-modules:
a best matching point pair obtaining submodule for obtaining the best matching point pair according to x*Acquiring a best matching point pair set from the feature point pair sets which are possibly matched currently by using a greedy method, wherein the initial value of the feature point pair set which is possibly matched currently is L;
the surface entity matching submodule is used for performing robust processing by using a least square method to obtain the position error of the optimal matching point pair by taking the minimum total deviation of all the optimal matching point pairs as a target; obtaining a face entity matching result based on the position error;
and the judging module is used for judging whether the current best matching point pair is from the source-matched face entity.
Compared with the prior art, the invention has the following advantages and beneficial effects:
(1) the influence of the position offset on the matching of the surface entity can be overcome.
(2) The matching accuracy is high, and can generally reach more than 96% as long as the parameters are properly set.
(3) The method carries out characteristic point correction based on the least square method of the weight of the robust estimation IGG1, and has certain robustness on the geometric change of a matched entity.
(4) The parameters set by the invention comprise a buffer query distance, an IGG1 weight threshold and a matching threshold. The buffer query distance represents the position deviation distance of the same-name entity, the IGG1 weight represents the geometric change condition of the same-name entity, the matching threshold represents the similarity threshold of the same-name entity, the physical significance is obvious, and the practicability is good.
Drawings
FIG. 1 is a detailed flow diagram of the method of the present invention;
fig. 2 is a schematic diagram of matching point pair connections.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention and/or the technical solutions in the prior art, the following description will explain specific embodiments of the present invention with reference to the accompanying drawings. It is obvious that the drawings in the following description are only examples of the invention, and that for a person skilled in the art, other drawings and embodiments can be derived from them without inventive effort.
Aiming at the problems of position offset, incapability of accurately acquiring one-to-many and many-to-many matching and the like in multi-scale building surface matching, the invention provides a multi-scale building surface entity matching method, wherein in the specific implementation process of the method, firstly, a surface entity is decomposed into a contour point set for representation; secondly, acquiring an L-maximized confidence coefficient indication vector of the corresponding relation of the feature points according to a pairwise constrained spectrum matching method based on geometric similarity; thirdly, iteratively correcting the corresponding relation of the characteristic points by using a least square method based on IGG1 weight to enhance the detection capability of geometric change of the entities with the same name; finally, the total similarity of the matching entities is calculated, and whether the matching entities are matched or not is judged by comparing with a threshold value. The method can overcome the influence of position offset on matching, and can be used for matching 1:0, 1:1 and 1: the matching of M and M to N achieves higher precision.
Referring to fig. 1, the method comprises the following specific steps:
the first step is as follows: and preprocessing a matching data source, wherein the matching feature source comprises a face entity to be matched.
The step further comprises format conversion, topology inspection and geometric coordinate conversion of the matching data source, and aims to eliminate systematic errors of the matching data source. Format conversion, topology inspection and geometric coordinate conversion are all conventional preprocessing means for matching data sources.
The second step is that: and extracting the characteristic points on the entity outline of the surface to be matched in the matched data source.
The method further comprises the following steps:
2.1 converting the entity of the surface to be matched into a contour point set.
The method comprises the following specific steps:
obtaining any surface entity s in matched data sourcei，siRepresenting the ith surface entity in the matched data source and establishing a surface entity siThe buffer area of (2) for querying all the surface entities in the buffer areaIn the present embodiment, the buffer radius d is set to 30 m. Extraction of siAnd BiObtaining a set of contour points.
2.2 extracting the characteristic points from the contour point set according to the steering angle of the contour point to obtain a characteristic point set to be matched.
The method comprises the following specific steps:
and calculating the steering angle theta of each contour point in the contour point set, and extracting the contour points with the steering angles not smaller than a preset angle, namely the feature points. Will siAnd BiThe characteristic points on the contour are respectively recorded asAnd
the calculation formula of the steering angle θ is as follows:
in formula (1):
(xk-1,yk-1)、(xk,yk)、(xk+1,yk+1) Coordinates representing three adjacent contour points;
θ represents the steering angle of the middle contour point of the three adjacent contour points.
The preset angle is an empirical value and is determined through sample data training. Tests prove that the preset angle is preferably 3-6 degrees.
And thirdly, acquiring possibly matched feature point pairs from the feature points to be matched, and calculating L maximum confidence coefficient indicating vectors of the feature point pairs by using a pairwise constrained spectrum matching method based on geometric similarity.
The method further comprises the following steps:
and 3.1, acquiring possibly matched feature point pairs in the P and the Q, and constructing an undirected graph G by taking the feature point pairs as vertexes.
Constructing pairs of possibly matching characteristic points (P) in P and Qu,qv) Wherein u is 1,2s，v＝1,2,...,ntAll possible matching feature point pair sets are marked as L, and the length of L is marked as nlWherein n isl≤ns×nt，nsDenotes siNumber of feature points on the contour, ntIs represented by BiNumber of feature points on the outline. L is the vertex of the undirected graph G, i.e. a ═ pu,qv) And b ═ pw,qz) Is any two vertices of the undirected graph G, a, b ∈ L, and a ≠ b.
3.2 calculate the value M (a, b) of the undirected graph G edge, M (a, b) constituting nlThe dimensional symmetric matrix M, i.e. the incidence matrix M.
M (a, b) is calculated as follows:
M(a,b)＝w1Sdis(a,b)+w2Slen(a,b)+w3Scos(a,b) (2)
in formula (2):
w1、w2、w3for the weight coefficients, the values of which are determined by sample data training, w1+w2+w31 in the present embodiment, w1＝w2＝w3＝1/3；
Sdis(a,b)、Slen(a,b)、Scos(a, b) represent the similarity of the distance, length and direction of a and b, respectively, as calculated as follows:
in formulae (3) to (5):
represents the length of the vector;
dthis a distance threshold value which is generally taken according to a data scale and data quality, or the position deviation of a matched data sample is taken, or the buffer area half in the substep 2.1 is directly taken;
represents a point puAnd point pwThe length of the line segment to be connected,representing point qvAnd point qzThe length of the line segment to be connected,representing line segmentsAndaverage Hausdorff distance (Hausdorff distance);
m _ HD is transformed from the classical HD distance,is calculated as follows:
in formula (6):
representing line segmentsAndthe distance of the hausdorff of (c),representing line segmentsAndthe Hausdorff distance;
psto representAt an arbitrary point, qtTo representAt any point above.
M (a, b) is ∈ [0,1], where M (a, b) ═ 1 denotes that a and b are absolutely related; m (a, b) ═ 0 means that a and b are absolutely uncorrelated, and the diagonal elements M (a, a) ═ M (b, b) ═ 0 in the matrix M.
In entity matching, the direction is an important similarity measurement index, and the rotation transformation of a larger angle between entities with the same name is a small probability condition, so that partial point pairs can be excluded by setting a distance threshold value during calculation, and the calculation efficiency is improved.
3.3 calculating L-maximized confidence indication vectors for pairs of feature points based on the correlation matrix M.
Correctly matched pairs of characteristic points should be consistent, so that the total association of the correctly matched set of pairs of characteristic points is maximal. Let C be the best matching feature point pair, i.e. total confidence SC＝∑a,b∈CM (a, b) takes the maximum value. For this type of problem, a spectral matrix can be used to solve C.
Will SCConversion:
optimal solution x*Is such that SCObtaining a binary vector, x, which takes the maximum value*Can be expressed as:
x*＝argmax(xTMx) (8)
argmax(xTmx) denotes let xTMx obtains the variable corresponding to the maximum value. The binary vector refers to that the elements of the vector are 0 and 1, wherein 0 represents that the corresponding characteristic point pairs are not matched, and 1 represents that the corresponding characteristic point pairs are matched.
According to Rayleigh-Ritz theorem, when x is the maximum eigenvalue λ of the correlation matrix MmaxCorresponding feature vector, SCTaking the maximum value. Therefore, the correlation matrix M is subjected to SVD decomposition (singular value decomposition) to obtain the maximum eigenvalue λ of the correlation matrix MmaxCorresponding feature vectors, i.e. optimal solution x*I.e., L-maximized confidence indication vectors for pairs of feature points in L.
And fourthly, correcting the corresponding relation of the characteristic point pairs by using a least square method based on IGG1 weight.
The method further comprises the following steps:
4.1 according to the L maximization confidence coefficient indication vector of the feature point pair, obtaining the best matching relation of the feature points from the feature point pair set which is possibly matched currently by utilizing a greedy method, wherein the initial value of the feature point pair set which is possibly matched currently is L.
The method further comprises the following steps:
4.1a assigning the current set of possibly matching pairs of feature points to LkTaking the current x*L corresponding to the medium maximum valuekCharacteristic point pair (p) in (1)max，qmax) Store in C, exclude LkMiddle and (p)max，qmax) The conflicting pairs of characteristic points, correspondingly, x*Also excluding the L-maximized confidence indication vector for the conflicting pair of feature points. The conflicting characteristic point pair fingers comprise a characteristic point pmaxOr qmaxThe characteristic point pairs of (1). Where k represents the number of iterations, starting from 0. L iskAnd current x*Is a symbiotic relationship, i.e. current x*Is LkIndicates the vector.
4.1b vs. Current x*Continue to execute substep 4.1a until LkThe middle characteristic point pairs are all excludedThe final C, i.e. the best matching point pair set with the largest geometric similarity, is denoted as CWherein, c1、c2、……Representing respective pairs of best matching points, ncFor the best match point logarithm, nc＝min(ns,nt)。
And 4.2, acquiring the matching relation of the surface entity according to the optimal matching point pair.
In the entity matching, the geometric characteristics of the same-name entities are changed, and the best matching point pairs in the C possibly have weak corresponding relations, or a few best matching point pairs are in error matching. Therefore, the robust processing is performed by the least square method.
Order toSet of position errors, i.e. s, for best-matching pairs of pointsiAnd BiIs a position shift distance of, wherein (Δ x)1,Δy1)、(Δx2,Δy2)、…、Are respectively No. 1, No. 2, No. … and No. ncThe position error between two matching points in the best matching point pair.
The overall deviation minimum formula is as follows:
in formula (9):
andrespectively representing the distance of the position error of the jth best matching point pair in the t-1 th iteration and the t-th iteration relative to the average position deviation;
the average position deviation at the t-th iteration is recorded asWherein,
is composed ofThe formula of the weight of (2) is:
in formula (10):
p represents an equivalent weight function.
Using IGG1 scheme weight function:
in formula (11):
when in useWhen the best matching point pair is a outlier point pair, the best matching point pair may be a wrong match, and the weight of the best matching point pair is setIs 0;
when in useWhen the point pair is a suspicious point pair, the jth best matching point pair is indicated, the local contour change corresponding to the best matching point pair is large, and the best matching point pair possibly has a weak corresponding relation;
when in useWhen the point pair is a credible point pair, the jth best matching point pair is described as a credible point pair, the local contour change corresponding to the best matching point pair is in a credible range, and the weight of the best matching point pair is setIs 1;
σ1and σ2And taking a value as a threshold value according to experience.
As shown in FIG. 2, the connecting lines indicate the corresponding relationship of the matching point pairs, wherein a1、a2、a3、a4As a pair of trusted points, a7And a8The point pairs are suspicious points; a is5And a6Are outlier pairs.
Solving for position error (Deltax) using an iterative method1,Δy1)、(Δx2,Δy2)、…、
Will siTranslating according to the position error, obtaining and BiSurface entity B with middle area overlapping rate not less than 0.5i', i.e., the face entity match result.
4.3 judging whether the best matching point pair is all from Bi' if yes, the face entity is correctly matched; if not, eliminating the face entity source conflict point pairs from the feature point pair set which can be matched currently, adding 1 to the iteration number k, and repeatedly executing the substep 4.1 until the face entity matching is correct. The plane entity source conflict point pair finger comprises Bi' neutral and siPairs of points on the non-matching surface entity. And fifthly, calculating the total similarity of the matching entities and judging whether the matching is successful.
The method further comprises the following steps:
5.1 calculation of Bi' and siTotal similarity of (c).
According to siSequencing the contour points, sequencing the best matching point pair set in the C to obtain the sequenced best matching point pair setWherein,and if the matching is multi-to-multi-surface entity matching, the contour points are discontinuous, and then the elements in the C are sequenced through Graham scanning.
Bi' and siTotal similarity of S (S)i,Bi') is:
in formula (11):
when j is equal to ncWhen the temperature of the water is higher than the set temperature,
M(cj',cj+1') is obtained by calculation of formula (1);
djis the normalized distance of the distance between the two,when h is ncWhen the temperature of the water is higher than the set temperature,
and 5.2 judging whether the surface entities are matched according to the total similarity. If S (S)i,Bi') is not less than a preset threshold, Bi' and siMatching is successful; otherwise, the matching is unsuccessful.
The following is a specific embodiment of the present invention. The following examples are only for explaining the present invention, the scope of the present invention shall include the full contents of the claims, and the full contents of the claims of the present invention can be realized by those skilled in the art through the following examples.
Claims (8)
1. A multi-scale building surface entity matching method is characterized by comprising the following steps:
s100, acquiring any entity in the matched data source, and recording as SiEstablishing siThe set of all the face entities in the buffer area is marked as Bi(ii) a Extraction of siAnd BiObtaining feature points P and Q to be matched from the feature points on the surface entity outline; the characteristic points are contour points of which the steering angle is not smaller than a preset angle;
s200, constructing possibly matched feature point pairs (p)u,qv) ObtainingA characteristic point pair set L, and an L maximum confidence coefficient indication vector x of each characteristic point pair is calculated*(ii) a Wherein p isuDenotes the u-th feature point in P, qvRepresenting the v-th characteristic point in Q, and taking 1,2, asV. taking 1,2,. and n in sequencet，nsAnd ntThe number of characteristic points in P and Q respectively;
s300, correcting the matching relation of the feature points to obtain a face entity matching result; the method further comprises the following steps:
s310 is according to x*Acquiring a best matching point pair set from the feature point pair sets which are possibly matched currently by using a greedy method, wherein the initial value of the feature point pair set which is possibly matched currently is L;
s320, aiming at the minimum total deviation of all the optimal matching point pairs, performing robust processing by using a least square method to obtain position errors of the optimal matching point pairs; obtaining a face entity matching result based on the position error;
s330, judging whether the current best matching point pair is a face entity with source matching, if so, matching correctly, and ending; if not, the face entity source collision point pairs are removed from the feature point pair set which may be matched currently, and S310 is executed repeatedly.
2. The multi-scale building face entity matching method of claim 1, wherein:
the method comprises a step of preprocessing the matching data source before the step S100, so as to eliminate systematic errors of the matching data source.
3. The multi-scale building face entity matching method of claim 1, wherein:
the steering angleWherein:
(xk-1,yk-1)、(xk,yk)、(xk+1,yk+1) Coordinates representing three adjacent contour points; theta denotes three adjacent profilesThe steering angle of the mid-profile point of the point.
4. The multi-scale building face entity matching method of claim 1, wherein:
in S200, the calculating an L-maximized confidence indicator vector for each feature point pair in L specifically includes:
s210, constructing an undirected graph by taking each characteristic point pair in the L as a vertex;
s220, forming a correlation matrix M by the values of all the undirected graph edges, wherein the correlation matrix is a symmetric matrix, and the diagonal element value is 0;
s230, singular value decomposition is carried out on M, and the eigenvector corresponding to the maximum eigenvalue is x*。
5. The multi-scale building face entity matching method of claim 4, wherein:
the value of the undirected graph edge, M (a, b) ═ w1Sdis(a,b)+w2Slen(a,b)+w3Scos(a, b), wherein M (a, b) represents the value of the undirected graph edge with the feature point pair a and b as vertices; w is a1、w2、w3For the weight coefficients, the values of which are determined by sample data training, w1+w2+w3＝1；Sdis(a,b)、Slen(a,b)、Scos(a, b) represent the similarity of the distance, length and direction of a and b, respectively.
6. The multi-scale building face entity matching method of claim 1, wherein:
s310 specifically comprises the following steps:
s311 assigns the feature point pair set which is possible to be matched currently to Lk，LkRepresenting all possible matched feature point pair sets of the k iteration, and taking the current x*The median maximum value is LkThe corresponding characteristic point pair (p) in (1)max，qmax) And storing the characteristic point pairs in C, wherein C is the best matching characteristic point pair set; excluding LkMiddle and (p)max，qmax) Phase impulsePairs of salient feature points, at the same time, at x*Deleting the corresponding elements of the conflicted characteristic point pairs; wherein k represents the number of iterations, and the value is taken from 0;
s312 for current x*Repeatedly executing S311 until LkAll the middle characteristic point pairs are excluded, and the final C is the best matching point pair set;
and, S320 further includes:
s321, constructing an objective function with minimum overall deviation of all best matching point pairsWherein,andrespectively representing the distance of the position error of the jth best matching point pair in the t-1 th iteration and the t-th iteration relative to the average position deviation; the average position deviation at the t-th iteration is recorded as Is composed ofThe weight of (a) is determined, project weight functions for the geodetic and geophysical institute;
s322, an objective function is solved by iteration through a least square method, and the position error of the optimal matching point pair is obtained;
s323 will SiTranslation according to position error, BiS after neutralization and translationiFace entity B with overlapping rate not less than 0.5i' i.e. siA matching face entity;
and, S330 specifically is:
judging whether the current best matching point pair is a face entity with source matching, if so, matching correctly, and ending; if not, the face entity source collision point pair is removed from the feature point pair set which may be currently matched, and S311 is repeated.
7. The multi-scale building face entity matching method of claim 1, wherein:
after S300 is completed, the method further includes a step of determining whether to match according to the total similarity of the matching entities, specifically:
s410 according to SiThe sequence of the middle outline points is used for sorting the final best matching point pairs in the best matching point pair set to obtain a sorted best matching point pair setWherein,
s420, calculating the total similarity of the matching surface entitiesWherein, M (c)j',cj+1') is denoted by cj' and cj+1' is the value of the undirected graph edge of the vertex; djIs composed ofAndwhen j is equal to ncWhen the temperature of the water is higher than the set temperature,
s430 judging whether the surface entities are matched according to the total similarity, if S (S)i,Bi') is not less than a preset threshold, Bi' and siMatching is successful; otherwise, the matching is unsuccessful.
8. A multi-scale building face entity matching system is characterized by comprising:
a characteristic point extraction module for obtaining any entity in the matched data source and recording as siEstablishing siThe set of all the face entities in the buffer area is marked as Bi(ii) a Extraction of siAnd BiObtaining feature points P and Q to be matched from the feature points on the surface entity outline; the characteristic points are contour points of which the steering angle is not smaller than a preset angle;
a main clustering confidence coefficient acquisition module for constructing possibly matched feature point pairs (p)u,qv) Obtaining a characteristic point pair set L, and calculating an L maximized confidence coefficient indication vector x of the characteristic point pairs in the L*(ii) a Wherein p isuDenotes the u-th feature point in P, qvRepresenting the v-th characteristic point in Q, and taking 1,2, asV. taking 1,2,. and n in sequencet，nsAnd ntThe number of characteristic points in P and Q respectively;
the correction module is used for correcting the matching relation of the feature points to obtain a face entity matching result;
the modification module further comprises sub-modules:
a best matching point pair obtaining submodule for obtaining the best matching point pair according to x*Acquiring a best matching point pair set from the feature point pair sets which are possibly matched currently by using a greedy method, wherein the initial value of the feature point pair set which is possibly matched currently is L;
the surface entity matching submodule is used for performing robust processing by using a least square method to obtain the position error of the optimal matching point pair by taking the minimum total deviation of all the optimal matching point pairs as a target; obtaining a face entity matching result based on the position error;
and the judging module is used for judging whether the current best matching point pair is from the source-matched face entity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710891628.9A CN107610120B (en) | 2017-09-27 | 2017-09-27 | A kind of multiple dimensioned building surface Methodology for Entities Matching and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710891628.9A CN107610120B (en) | 2017-09-27 | 2017-09-27 | A kind of multiple dimensioned building surface Methodology for Entities Matching and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107610120A CN107610120A (en) | 2018-01-19 |
| CN107610120B true CN107610120B (en) | 2019-08-20 |
Family
ID=61058980
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710891628.9A Active CN107610120B (en) | 2017-09-27 | 2017-09-27 | A kind of multiple dimensioned building surface Methodology for Entities Matching and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107610120B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111375B (en) * | 2019-03-29 | 2021-05-04 | 中国地质大学（武汉） | Image matching gross error elimination method and device under Delaunay triangulation network constraint |
| CN110807797B (en) * | 2019-10-22 | 2022-03-22 | 中国测绘科学研究院 | Multi-source heterogeneous surface entity and point entity matching method considering global optimization and storage medium thereof |
| CN112861341A (en) * | 2021-02-04 | 2021-05-28 | 江西师范大学 | Multi-scale surface element matching method and system combining shape and environmental characteristics |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020966A (en) * | 2012-12-04 | 2013-04-03 | 南京大学 | Automatic registration method of aviation and ground LiDAR data based on building contour constraint |
| CN103279955A (en) * | 2013-05-23 | 2013-09-04 | 中国科学院深圳先进技术研究院 | Image matching method and system |
| CN104751475A (en) * | 2015-04-16 | 2015-07-01 | 中国科学院软件研究所 | Feature point optimization matching method for static image object recognition |
-
2017
- 2017-09-27 CN CN201710891628.9A patent/CN107610120B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020966A (en) * | 2012-12-04 | 2013-04-03 | 南京大学 | Automatic registration method of aviation and ground LiDAR data based on building contour constraint |
| CN103279955A (en) * | 2013-05-23 | 2013-09-04 | 中国科学院深圳先进技术研究院 | Image matching method and system |
| CN104751475A (en) * | 2015-04-16 | 2015-07-01 | 中国科学院软件研究所 | Feature point optimization matching method for static image object recognition |
Non-Patent Citations (1)
| Title |
|---|
| 《中误差和邻近关系的多尺度面实体匹配算法研究》;刘坡等;《测绘学报》;20140430;第43卷(第4期);419-425 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107610120A (en) | 2018-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111028277B (en) | SAR and optical remote sensing image registration method based on pseudo-twin convolution neural network | |
| Ma et al. | Non-rigid point set registration with robust transformation estimation under manifold regularization | |
| Williams | An efficient, variational approximation of the best fitting multi-Bernoulli filter | |
| CN111795704A (en) | Method and device for constructing visual point cloud map | |
| CN107633226B (en) | Human body motion tracking feature processing method | |
| CN102169545B (en) | Detection method for changes of high-resolution remote sensing images | |
| Chen et al. | Non-rigid point set registration via coherent spatial mapping | |
| CN109544603B (en) | Target tracking method based on deep migration learning | |
| CN107610120B (en) | A kind of multiple dimensioned building surface Methodology for Entities Matching and system | |
| CN112070832B (en) | Non-cooperative target pose tracking method and device based on point cloud DCA (distributed computing architecture) features | |
| CN109544610A (en) | A kind of method for registering images based on convolutional neural networks | |
| Zhou et al. | Nonrigid feature matching for remote sensing images via probabilistic inference with global and local regularizations | |
| CN106709943A (en) | Point cloud registration method based on optimal transmission | |
| CN115527050A (en) | Image feature matching method, computer device and readable storage medium | |
| CN103700107A (en) | Image sharp degree distribution-based characteristic point matching method | |
| Shen et al. | A novel affine covariant feature mismatch removal for feature matching | |
| Nieto et al. | Simultaneous estimation of vanishing points and their converging lines using the EM algorithm | |
| Yang et al. | Non-rigid point set registration via global and local constraints | |
| CN111104922A (en) | Feature matching algorithm based on ordered sampling | |
| CN116883463A (en) | Three-dimensional registration reconstruction method based on multi-domain multi-dimensional feature map | |
| Zhao et al. | Visual homing by robust interpolation for sparse motion flow | |
| Wang | LPOT: Locality-Preserving Gromov–Wasserstein Discrepancy for Nonrigid Point Set Registration | |
| Chen et al. | An application of improved RANSAC algorithm in visual positioning | |
| CN114723791A (en) | Feature point matching method and device, electronic equipment and storage medium | |
| Chen et al. | Mismatch removal for remote sensing images based on non-rigid transformation and local geometrical constraint |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |