CN104574519B - Multi-source resident's terrain feature exempts from the automatic sane matching process of threshold value - Google Patents
Multi-source resident's terrain feature exempts from the automatic sane matching process of threshold value Download PDFInfo
- Publication number
- CN104574519B CN104574519B CN201510053743.XA CN201510053743A CN104574519B CN 104574519 B CN104574519 B CN 104574519B CN 201510053743 A CN201510053743 A CN 201510053743A CN 104574519 B CN104574519 B CN 104574519B
- Authority
- CN
- China
- Prior art keywords
- key element
- matching
- face
- plane
- factor combination
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/30—Polynomial surface description
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Physics (AREA)
- Mathematical Analysis (AREA)
- Algebra (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
Exempt from the automatic sane matching process of threshold value the present invention relates to multi-source resident's terrain feature, effectively solve accurately to obtain 1 existed between face key element of the same name:1 matching relationship, 1:N matching relationships and M:N matching relationships, do not require that two groups of face key elements to be matched are accurately close in position again, allow the presence of larger position deviation, the problem of even nonconforming position deviation, method is to read two groups of settlement place data to be matched respectively with computer, one group of settlement place is plane of reference factor combination, R is designated as, another group of settlement place is target face factor combination, is designated as T;Generation candidate matches pair;Judge key element matching relationship in face of the same name;The conflict probe of matching result and elimination, the inventive method is reliable and stable, easy to operate, do not require that two groups of face key elements to be matched are accurately close in position, allow the presence of larger position deviation, threshold value need not be set, the workload of pretreatment is reduced, effective for map vector merging, map rejuvenation and change detection.
Description
Technical field
The present invention relates to map, the data integration and production of map are referred mainly to, particularly relating to a kind of multi-source resident ground will
Plain exempts from the automatic sane matching process of threshold value.
Background technology
Along with the fast development of space data collection technology, people have accumulated abundant spatial data, even
Covering same area, different departments also acquire the data for differing from one another for different needs repeatedly.For example, house property department and survey
Paint department and all acquire settlement place data, but attribute information (such as ownership, the area of settlement place data are more valued by house property department
Deng), and the precision of data is more valued by Mapping departments, if it is possible to the data of different departments are carried out into effective integration, it becomes possible to collect
The advantage of various version space data, repeats to pay wages such that it is able to save substantial amounts of field operation.In order to empty to multi-source
Between data carry out accurately integrated, multiplexing and shared, its core and prerequisite steps seek to that separate sources space number can be realized
Automatic identification and matching according to key element of the same name.Therefore, present invention is particularly directed to the face factor data of separate sources in map vector
The invention work of (mainly by taking resident's terrain feature as an example) Develop Data matching.
In various face key element matching algorithms, it can be common that simplify the complexity of face key element matching problem by dimension-reduction treatment
Property, determine face by comparing the position relationship of the distance between face key element character pair point or barycenter, barycenter and face key element
The matching relationship of key element;Or extract the profile of face key element, determined apart from upper proximity by calculating contour line
With relation, frequently with distance measure have Fr é chet distances, Hausdorff distances etc., or calculate contour line in shape
Similarity degree determines matching relationship, its similarity measure such as shape letter of transfer number, and Fourier descriptor etc., a small number of algorithms extract faces
Skeleton line inside key element, by the matching relationship for contrasting the shape similarity of skeleton line to determine face key element.Also some calculate
Rule is matched using the interior zone of face key element, and common characteristic index is the overlapping area between the key element of calculating face,
The ratio that two face key elements to be matched its overlapping areas accounts for respective area is bigger, then they be key element of the same name possibility also
It is bigger.Certainly, some matching algorithms are not used alone the above some characteristic index, but can be combined various indexs
Use, but its difficult point is in the accurate determination of weights.Also some algorithms can not only consider over there in key element matching process
The matching similarity degree of face key element to be matched, further accounts for the match condition of adjacent other face key elements, further increases matching
Algorithm adapts to the ability of complex situations.
Although having been achieved with plentiful and substantial progress in the key element matching problem of face, existing algorithm is to some difficulties
Still without preferably solution, one is, it is difficult to accurately obtain the M existed between face key element of the same name:N matching relationships;Such as those drops
Dimension algorithm and the method based on contour line, due to lacking the utilization of opposite key element interior zone, when there is M:During N matching relationships,
Obtaining combinatorial surface key element contour line on the whole will become time-consuming and difficulty, and these algorithms all cannot be to complicated M:N matchings are closed
System is automatically processed, and can only process 1:N, even 1:1 match condition.The second is, it is desirable to face key element of the same name is in position
As close possible to many algorithms often use distance as matching threshold;Although some algorithms are obtained in that between face key element of the same name
M:N matching relationships, but such method has an important prerequisite, i.e., between two groups of face key elements to be matched in position compared with
It is close, or clearly meets map making specification, but in practice, due to the popularity and complexity of data source, treats
The position deviation of uniformity is there may be between the face key element of matching, i.e. the universal deviation in some places is big, and some places
Deviation is small.Even if being corrected by the coordinate system of globality, accuracy registration is still unable between two groups of face key elements, in these cases
These algorithms also will be hard to work, therefore, its improve and innovate it is imperative.
The content of the invention
For above-mentioned situation, for the purpose for overcoming the defect of prior art, the present invention is just to provide a kind of multi-source settlement place
Face key element exempts from the automatic sane matching process of threshold value, can effectively solve accurately to obtain 1 existed between face key element of the same name:
1 matching relationship, 1:N matching relationships and M:N matching relationships, and do not require that two groups of face key elements to be matched are accurately close in position,
Allow the presence of larger position deviation, the problem of even nonconforming position deviation.
The technical scheme that the present invention is solved is to comprise the following steps:
Step one, data acquisition:
Read two groups of settlement place data to be matched respectively with computer, one group of settlement place is plane of reference factor combination, is designated as
R, another group of settlement place is target face factor combination, is designated as T;
Step 2, generation candidate matches pair:
For plane of reference factor combination R to be matched and target face factor combination T, by the friendship of locus overlapping relation
Repeatedly judge, select candidate matches face factor combination from plane of reference factor combination and target face factor combination successively, and generate time
Choosing matching is right, and candidate matches are to comprising 1:1、1:N、M:1 and M:Various matchings that may be present are closed between the face key element of the same name such as N
System;
The judgement of step 3, key element matching relationship in face of the same name:
The matching relationship of face key element of the same name is determined using the matching similarity of two-way calculating face key element, i.e., for the plane of reference
For factor combination R and target face factor combination T, when can be optimal for other side mutually according to the matching similarity for being calculated
Matching face factor combination, and the two matching corresponding relation is set up;
In order to effectively process various complex match corresponding relations, described matching similarity its to calculate thinking be by weighing apparatus
Matching degree of the overlapped degree to judge between the key element of face between the key element of amount separate sources face;Even if in order to ensure same
When name face key element has larger position deviation, still it is effectively matched, then is extracted the envelope square of plane of reference factor combination first
The global shape feature of shape and individual features point, obtains the matching corresponding relation of characteristic point of the same name in face key element of the same name automatically, will
Plane of reference factor combination R carries out position skew so that two groups of face factor combinations to be matched in position can be further registering, after
And the matching degree of plane of reference factor combination R and target face factor combination T is judged by weighing overlapping degree again;
The calculating of face key element matching similarity is carried out from both direction respectively, right to obtain matching, method is, first referring to
On the basis of face factor combination R, according to the matching similarity for being calculated, best match target face key element is found from target face key element
Combination T, then again on the basis of target face factor combination T, according to the matching similarity for being calculated, from plane of reference factor combination
In find best match plane of reference factor combination, when the best match face factor combination be the plane of reference factor combination R when, then refer to
Face factor combination R and target face factor combination T-shaped are into one group of best match face factor combination each other;
The computational methods of described matching similarity are:By using degree overlapped between the key element of separate sources face
To weigh the matching degree of face key element, if the area of plane of reference factor combination R is Ar, the area of target face factor combination T is At,
The overlapping area of plane of reference factor combination R and target face factor combination T is Ao, then overlapping similarity Sim is:
Judge the matching degree between candidate matches pair by calculating overlap similarity, extract plane of reference factor combination
The global shape feature of enclosure rectangle and individual features point, using the matching corresponding relation of characteristic point of the same name in face key element of the same name,
Plane of reference factor combination R is carried out into position skew so that two groups of face factor combinations to be matched in position can be further registering,
The matching degree of plane of reference factor combination R and target face factor combination T is judged by weighing overlapping degree, to target face key element
Each face key element in combination T calculates the weight after the preceding movement with plane of reference factor combination R of plane of reference factor combination R movements respectively
Folded similarity, averages, and is designated as matching similarity ρ, and formula is as follows:
In formula (2), ArIt is the area of plane of reference factor combination, AtIt is the area of target face factor combination, ApFor R is translated
The lap area of the former two, AqThe lap area of the latter two is translated for R;
Step 4, the conflict probe of matching result and elimination:
According to matching similarity value size, the matching result to clashing is given up or is retained, i.e. reservation
It is big with Similarity value, and give up that matching similarity is less, so as to the contradiction for realizing matching result is eliminated.
The inventive method is reliable and stable, easy to operate, can be applied to face key element of the same name under the multiple dimensioned situation of multi-source from
Dynamic matching algorithm, can either accurately obtain the matching relationships of 1 ︰ 1,1 ︰ N matching relationships and the M ︰ N matchings existed between face key element of the same name
Relation, and do not require that two groups of face key elements to be matched are accurately close in position, it is allowed to there are larger position deviation, or even right and wrong
The position deviation of uniformity, threshold value is set without user, can adapt to increasingly complex situation, therefore can be referred to as a kind of steady
Strong matching process (algorithm).The method has expanded the applicability of map conflation field face key element matching technique, makes it to data strip
The requirement and limitation of part are further reduced, and reduce the workload of pretreatment, promote the integrated, shared of spatial data and multiplexing, are had
, in map vector merging, map rejuvenation and change detection, with very strong practicality, economic and social benefit is huge for effectiveness.
Brief description of the drawings
Fig. 1 is generation schematic diagram of the candidate matches of the present invention to combination.
Fig. 2 is present invention determine that the schematic diagram of best match face factor combination.
Fig. 3 is the contradiction elimination figure of matching result of the present invention.
Fig. 4 is two groups of the invention face factor data collection schematic diagrames to be matched.
Fig. 5 is that matching relationship of the present invention is 1:1 matching result schematic diagram.
Fig. 6 is that matching relationship of the present invention is 1:The matching result schematic diagram of N.
Fig. 7 is that matching relationship of the present invention is M:1 matching result schematic diagram.
Fig. 8 is that matching relationship of the present invention is M:The matching result schematic diagram of N.
Fig. 9 is trying hard to for inventive algorithm resistance face elements position deviation.
Specific embodiment
This bright specific embodiment is described further below in conjunction with concrete condition.
The present invention can be realized in specific implementation by following steps:
Step one, data acquisition:
Read two groups of settlement place data to be matched respectively with computer, one group of settlement place is plane of reference factor combination, is designated as
R, another group of settlement place is target face factor combination, is designated as T;
Step 2, generation candidate matches pair:
Various matching relationships, including 1 are there may be between face key element of the same name:1、1:N、M:1 and M:N, in order to correct
Various matching relationships between face key element of the same name are obtained, firstly generate potential candidate matches to the greatest extent should be able to may be used to set, and the set
By all correct matchings to being included, for two groups of face factor data set to be matched, one group is referred to as reference on energy ground
Face elements combination, another group is referred to as target face elements combination, and algorithm is alternately successively from plane of reference elements combination and target face
Candidate matches face factor combination is selected in elements combination, and forms candidate matches to (as shown in Figure 1), plane of reference key element R has three
Individual face key element, i.e. the first plane of reference key element R1, the second plane of reference key element R2With the 3rd plane of reference key element R3, target face factor combination T
In have three face key elements, i.e. first object face key element T1, the second target face key element T2With the 3rd target face key element T3;Order is chosen
Each face key element in plane of reference elements combination, when in target face elements combination with the first plane of reference key element R1In the presence of intersecting
Relation is first object face key element T1, first group of candidate matches can be obtained to (R1):(T1), then again in plane of reference key element collection
Found in conjunction and first object face key element T1The plane of reference key element that there is overlapping relation is the first plane of reference key element R1, second reference
Face key element R2, then can obtain the second group is paired into (R again1,R2):(T1), further obtain and the first plane of reference key element R1, second
Plane of reference key element R2Intersecting target face key element, can obtain the 3rd group of candidate matches to being (R1,R2):(T1,T2,T3), until most
The 4th candidate matches are obtained afterwards to being (R1,R2,R3):(T1,T2,T3);When two conditions are met, the generation of candidate matches pair
Terminate, the two conditions are:One is that cannot further obtain candidate matches to (as shown in Figure 2), is obtaining the 4th group of candidate
After pairing, each face key element has participated in computing, and cannot reentry new candidate matches pair;Two is the candidate for preparing for
Pairing had been generated before this;
But it is final matching result that the generation of candidate matches pair terminates not representing, such as the 4th group of candidate matches to for
(R1,R2,R3):(T1,T2,T3), in addition it is also necessary to further extracted and (R from target face factor combination1,R2,R3) combination phase the most
As (T1,T2) combination, it is only final correct matching result;
The judgement of step 3, key element matching relationship in face of the same name:
The calculating of face key element matching similarity is carried out from both direction respectively, right to obtain matching, method is, first referring to
On the basis of face factor combination R, according to the matching similarity for being calculated, best match target face key element is found from target face key element
Combination T, then again on the basis of target face factor combination T, according to the matching similarity for being calculated, from plane of reference factor combination
In find best match plane of reference factor combination, when the best match face factor combination be the plane of reference factor combination R when, then refer to
Face factor combination R and target face factor combination T-shaped are into one group of best match face factor combination each other;
Obtain candidate matches to after, it is necessary to weigh between plane of reference factor combination and target face factor combination
With degree, and from candidate matches centering optimal matching result is determined accordingly, in order to make full use of the border and inside of face key element
Information, the matching degree of face key element is weighed using degree overlapped between the key element of face, as above sets plane of reference factor combination
It is R, target face factor combination is T, if the area of plane of reference factor combination R is Ar, the area of target face factor combination T is At, ginseng
The overlapping area for examining face factor combination R and target face factor combination T is Ao, then overlapping similarity Sim is:
Judge the matching degree between candidate matches pair by calculating overlap similarity, its precondition is that the two is in place
Putting can coincide enough to good, but in a practical situation, and the two may have larger position deviation, as shown in Figure 2
A kind of extreme case, it is assumed that have a candidate matches pair, its plane of reference factor combination is (R1,R2), target face factor combination is
(T1-T6), the two in position deviation it is very big, matching degree therebetween cannot be now reflected by weighing overlapping degree,
For this reason, it may be necessary to solve in the case of position deviation is larger, how best match face factor combination is determined using overlap similarity
Problem;
As shown in Figure 2, when by T1To T3When combining, its shape and (R1,R2) rather similar, they are matched each other
Possibility it is maximum;But question now is how to could correctly by T1To T3Combine, for (R1,R2) for, T1-
T6In certain or several face key elements combination with (R1,R2) there may be M:N matching relationships, list when according to the method for exhaustion
The combination of its all candidate matches faces key element, its combined number hasIt is individual, it is clear that according to exhaustion
Method is therefrom found out and (R1,R2) matching face factor combination the most similar will be very time-consuming, therefore, face key element to be carried out
Auto-matching, it is necessary to find a kind of method that can quickly and accurately determine best match face factor combination first;
Notice for face key element of the same name, their global shape feature will be rather similar, thus can be by following
Four steps determine candidate matches face factor combination quickly:
(1) the global shape feature of plane of reference factor combination R is extracted:Global shape feature refers to the enclosure rectangle and phase of R
Characteristic point is answered, as shown in Fig. 2 finding first in plane of reference factor combination R, between each summit two farthest summits of distance, point
Wei not r1、r3, then respectively in line segment r1、r3Left side and right side find apart from line segment r1、r3Point with maximum distance, is r2
And r4, then by r1、r2、r3And r4Four feature point groups into rectangle be clearly plane of reference factor combination R enclosure rectangle, envelope square
The a line and r of shape1r3Line it is parallel, another a line then with r1r3Line is vertical, and this enclosure rectangle has and similarity transformation
Unrelated the characteristics of, when plane of reference factor combination R is integrally shifted, rotates and scaled, the enclosure rectangle for being extracted is still by r1、
r2、r3And r4Four feature point groups into;
(2) for target face factor combination T, find and r1、r2、r3And r4Best match corresponding vertex:Method is, suitable
Sequence chooses each the face key element in target face factor combination T, then sequentially chooses each summit of the face key element, if select
One summit is t1(as shown in Figure 2), finds a summit p so that by point t in T1The vectorial U constituted to p with by point r1
To r2The vector similarity value of the vectorial V for being constituted is maximum, i.e., the most similar between vector, herein the meter of vector similarity α
Calculate formula as follows:
The summit can be found in fig. 2 for t2, summit t can also be respectively found according to same steps3And t4, because to
Amount t1t3With vectorial r1r3It is the most similar, vectorial t1t4With vectorial r1r4It is the most similar.
According to the method described above, according to each summit of face key element in T, the other three summit can be found, and in succession
Three maximum vector similarities are calculated, three average values of maximum vector similarity is taken as resultant vector similarity, works as synthesis
When vector similarity value is maximum, that is, obtain plane of reference factor combination and target face factor combination global shape feature
Optimal matching points relation, as shown in Fig. 2 r1、r2、r3And r4Four characteristic points respectively with t1、t2、t3And t4Four characteristic points are deposited
In matching corresponding relation;
(3) matching similarity is calculated:After optimal matching points are obtained by global shape feature, r is calculated1、r2、r3With
r4Four mean places of characteristic point, are designated as r, similarly also calculate t1、t2、t3And t4Four mean places of characteristic point, are designated as
T, R is moved integrally at the t of position from position r points, now will preferably be realized between R and T it is registering, in can be by
Weigh overlapping degree to judge the matching degree of R and T, each the face key element in T is calculated respectively before R is moved and after R movements
Overlap similarity, average, be designated as matching similarity ρ, formula is as follows:
In above formula, ArIt is the area of plane of reference factor combination, AtIt is the area of target face factor combination, ApFor R is translated
The lap area of the former two, AqThe lap area of the latter two is translated for R;
(4) calculate each the face key element in target face factor combination T matching similarity, and with size be ordered pair
It is ranked up with similarity:Each the face key element in target face factor combination T is chosen successively according to descending, when plane of reference key element
Its matching similarity value diminishes with the target face factor combination chosen for combination, then stop adding the face key element in T, otherwise then
The face key element in T is continuously added, then the face factor combination for being constituted as best match face factor combination;As shown in Fig. 2 work as will
Plane of reference factor combination R is moved to when at dashed box position, and single face key element T is calculated respectively1To T6With the matching similarity of R,
Assuming that matching similarity is according to T1To T6Order reduce successively, then choose T first1Face key element, matching similarity now is false
ρ 1 is set to, then T is added in target face factor combination2Face key element, as a result finds to work as T1With T2Combine, then calculate
During with the matching similarity of R, the value is increased than ρ 1, then receive T1With T2Combination and continuously add T3Face key element, as a result finds
Matching similarity further increases, then continuously add T4Face key element, as a result finds that matching similarity is reduced, then stop adding and appoint
What face key element, finally can successively select T1、T2And T3Three face key elements, and they are grouped together into R1And R2Optimal
With face factor combination;
The conflict probe of step 4 matching result and elimination:
After candidate matches according to formula (3) to carrying out the calculating of matching similarity, matching result appearance is may result in
Conflicting phenomenon, is respectively (R as shown in figure 3, there may be six groups of matching relationships1):(T1) matching similarity 0.82,
(R2):(T2) matching similarity 0.82, (R3):(T3) matching similarity 0.85, (R1,R2):(T1,T2) matching similarity 0.9,
(R1,R3):(T1,T3) matching similarity 0.82, (R1,R2,R3):(T1,T2,T3) matching similarity 0.85, it is clear that matching result it
Between exist it is conflicting and conflict phenomenon, therefore, can be given up to matching result according to matching similarity value size
Or retain, the matching result that should be retained first in Fig. 3 is (R1,R2):(T1,T2), because its matching similarity is most
Greatly, matching result (R is then retained again3):(T3), contradiction and the conflict of matching result can be preferably solved using the method, from
And realize the sane matching of the multiple dimensioned face key element of multi-source.
The inventive method achieves very satisfied Advantageous Effects, the embodiment provided with Fig. 4 by practical application
As a example by, concrete condition is as follows:
One group of separate sources being given by Fig. 4 a, 4b, resident's terrain feature of different periods verify inventive algorithm,
Wherein, with reference to residential feature class See Figure 4a, a 703 face key element is had, target residential feature class is shown in Fig. 4 b, has 820
Individual face key element.Look that two groups of data are more similar although overall, in fact have many face key elements not only have in shape compared with
Big difference, there is also nonconforming range deviation in position, and some local deviations are larger, and some places are then smaller.
(1) ability of algorithm process complex match relation
Algorithm can be to 1:1,1:N, M:N, 1:0,0:The various matching relationships such as 1 carry out the Auto-matching of face key element, and table 1 is united
The match condition and matching similarity of various matching relationships in the example are counted, to the example, reference data is concentrated and has 7 faces
Key element there occurs error hiding, and 10 face key elements there occurs and Lou match, 12 face key element its match-type identification mistakes.Target data
Concentration has 10 face key elements and there occurs error hiding, and 16 face key elements there occurs and Lou match, 13 face key elements its final matchings
Type error.Total matching accuracy rate is 1- (7+10+12+10+16+13)/(703+820)=94.9%.
It is 1 that Fig. 5 illustrates matching relationship:Plane of reference factor data collection and target face factor data collection when 1, Fig. 6 displayings
Matching relationship is 1:Plane of reference factor data collection and target face factor data collection during N, Fig. 7 illustrate matching relationship for M:1
When plane of reference factor data collection and target face factor data collection, Fig. 8 illustrates matching relationship for M:Plane of reference during N wants prime number
According to collection and target face factor data collection.
The statistics and matching similarity of the various match-type match conditions of table 1
(2) algorithm resists the ability of face elements position deviation
The inventive method does not require two groups of face factor datas in position very close to even if two groups of face key elements to be matched
There is nonuniformity position deviation phenomenon, i.e., it is very big in some local position deviations in some local position deviation very littles, still
Correct matching result can be obtained.Fig. 9 is a partial enlarged drawing in Fig. 4, and the matching result at this is shown in Table 2, matching knot
Fruit has 1:1、1:N and M:Tri- kinds of match-types of N.Fig. 9 is it can be found that plane of reference factor data collection for observation (see solid line polygon)
No. 13 key elements (see black matrix mark) with target face factor data collection (see dashed polygon) in No. 58 key elements (see italic mark
Note) range deviation is very big, and equally, No. 197 that No. 338 key elements of plane of reference factor data collection are concentrated with target face factor data will
Plain range deviation is also quite big, but checks the matching result of table 2, it can be seen that algorithm still can carry out correct to this two
Match somebody with somebody.As shown in Table 2, for the example, no matter less than normal in face elements position of the same name algorithm is, or elements position deviation in face of the same name compared with
Greatly, can obtain correct matching result, this indicate that algorithm can preferably resist face key element of the same name between nonuniformity
Position deviation, reduces the requirement and limitation to data so that algorithm has wider application scenario.
The matching result list of table 2
Source key element | Target component | Match-type | Matching similarity |
29 | 191 | 1:1 | 0.58 |
27 | 207 | 1:1 | 0.72 |
338 | 197 | 1:1 | 0.41 |
117 | 183 | 1:1 | 0.66 |
26 | 214 | 1:1 | 0.65 |
119 | 165 | 1:1 | 0.56 |
13 | 58,59,60,61,62 | 1:N | 0.54 |
52 | 82,85 | 1:N | 0.37 |
1,2,32 | 8,12 | M:N | 0.57 |
25,27,28 | 13,14,15,16,18,20 | M:N | 0.52 |
By it is above-mentioned it should be apparent that the method have the benefit that:
1st, 1 between face key element of the same name can be obtained:1、1:N and M:The various matching relationships such as N;
Even if the 2, still can when face key element of the same name has larger position deviation even nonconforming position deviation
Face key element of the same name is effectively matched;
3rd, unlike other method, it is necessary to set threshold value, and the present invention is without the setting of threshold value, and the present invention is by carrying out
Bi-directional matching, when plane of reference key element and target face key element can be relatively the best match object of other side, determines that them
Matching relationship, avoiding problems the setting of threshold value.Whole process is automatically performed.
In sum, the applicability of map conflation field face key element matching technique has been expanded in the invention, makes it to data strip
The requirement and limitation of part are further reduced, and reduce the workload of pretreatment, promote the integrated, shared of spatial data and multiplexing, are had
Effectiveness is detected in map vector merging, map rejuvenation and change, with very strong practicality.
Claims (1)
1. a kind of multi-source resident terrain feature exempts from the automatic sane matching process of threshold value, it is characterised in that by step in detail below
Realize:
Step one, data acquisition:
Read two groups of settlement place data to be matched respectively with computer, one group of settlement place parameter is plane of reference factor combination, is designated as
R, another group of settlement place parameter is target face factor combination, is designated as T;
Step 2, generation candidate matches pair:
Can correctly to obtain various matching relationships between face key element of the same name, potential candidate matches are firstly generated to set, should
By all correct matchings to being included, for two groups of face factor data set to be matched, one group is referred to as reference for set
Face elements combination, another group is referred to as target face elements combination, and crossover is successively from plane of reference elements combination and target face key element collection
Candidate matches face factor combination is selected in conjunction, candidate matches pair are formed, plane of reference key element R has three face key elements, i.e., the first reference
Face key element R1, the second plane of reference key element R2With the 3rd plane of reference key element R3, there are three face key elements in target face factor combination T, i.e.,
One target face key element T1, the second target face key element T2With the 3rd target face key element T3;It is every in order selection plane of reference elements combination
One face key element, when in target face elements combination with the first plane of reference key element R1Exist overlapping relation for first object face will
Plain T1, first group of candidate matches is obtained to (R1):(T1), then being found in plane of reference elements combination again will with first object face
Plain T1The plane of reference key element that there is overlapping relation is the first plane of reference key element R1, the second plane of reference key element R2, obtain second group of matching
To being (R1,R2) ︰ (T1), then obtain and the first plane of reference key element R1, the second plane of reference key element R2Intersecting target face key element, obtains
3rd group of candidate matches are to being (R1,R2):(T1,T2,T3), until finally obtaining the 4th candidate matches to being (R1,R2,R3) ︰
(T1,T2,T3);When two conditions are met, the generation of candidate matches pair terminates, and the two conditions are:One is further to obtain
Candidate matches pair are taken, the 4th group of candidate matches is being obtained to rear, each face key element has participated in computing, cannot reentry new
Candidate matches pair;Two is the candidate matches for preparing for having been generated before this;
When the 4th group of candidate matches are to being (R1,R2,R3) ︰ (T1,T2,T3), further to be extracted from target face factor combination
With (R1,R2,R3) combination (T the most similar1,T2) combination, it is finally correct matching result;
The judgement of step 3, key element matching relationship in face of the same name:
The calculating of face key element matching similarity is carried out from both direction respectively, right to obtain matching, method is first to be wanted with the plane of reference
On the basis of element combination R, according to the matching similarity for being calculated, best match target face factor combination is found from target face key element
T, then again on the basis of target face factor combination T, according to the matching similarity for being calculated, looks for from plane of reference factor combination
To best match plane of reference factor combination, when the best match face factor combination is plane of reference factor combination R, then the plane of reference will
Element combination R and target face factor combination T-shaped are into one group of best match face factor combination each other;
Match journey between plane of reference factor combination and target face factor combination, it is necessary to weigh to after candidate matches are obtained
Degree, and from candidate matches centering determine optimal matching result accordingly, weighed using degree overlapped between the key element of face
The matching degree of face key element, is R according to the plane of reference factor combination for having set above, and target face factor combination is T, if the plane of reference will
The area of element combination R is Ar, the area of target face factor combination T is At, plane of reference factor combination R and target face factor combination T's
Overlapping area is Ao, then overlapping similarity Sim is:
Overlap similarity and judge the matching degree between candidate matches pair by calculating, its precondition be the two in position
Can coincide, when there are a candidate matches pair, its plane of reference factor combination is (R1,R2), target face factor combination is (T1-T6),
The two in position deviation it is very big, matching degree therebetween cannot be now reflected by weighing overlapping degree, in position
In the case of deviation is larger, candidate matches face factor combination is determined by following steps:
(1) the global shape feature of plane of reference factor combination R is extracted:Global shape feature refers to the enclosure rectangle of R and corresponding spy
Levy a little, two farthest summits of distance, respectively r between each summit are found in plane of reference factor combination R first1、r3, then
Respectively in line segment r1、r3Left side and right side find apart from line segment r1、r3Point with maximum distance, is r2And r4, then by r1、r2、
r3And r4Four feature point groups into plane of reference factor combination R enclosure rectangle, a line and r of enclosure rectangle1r3Line put down
OK, another a line then with r1r3Line is vertical, the characteristics of this enclosure rectangle has unrelated with similarity transformation, when plane of reference key element
Combination R is integrally shifted, rotates and scaled, and the enclosure rectangle for being extracted is still by r1、r2、r3And r4Four feature point groups into;
(2) for target face factor combination T, find and r1、r2、r3And r4Best match corresponding vertex:Method is that order is chosen
Each face key element in target face factor combination T, then each summit of the face key element is sequentially chosen, if a top of selection
Point is t1, a summit p is found in T so that by point t1The vectorial U constituted to p with by point r1To r2The vectorial V for being constituted
Vector similarity value it is maximum, i.e., the most similar between vector, the computing formula of vector similarity α is as follows:
The summit is found for t2, find summit t respectively according to same steps3And t4;
According to the method described above, according to each summit of face key element in T, the other three summit, and successive computations can be found
Three maximum vector similarities, take three average values of maximum vector similarity as resultant vector similarity, work as resultant vector
When similarity value is maximum, that is, obtain plane of reference factor combination and target face factor combination optimal of global shape feature
With point to relation, r1、r2、r3And r4Four characteristic points respectively with t1、t2、t3And t4There is matching corresponding relation in four characteristic points;
(3) matching similarity is calculated:After optimal matching points are obtained by global shape feature, r is calculated1、r2、r3And r4Four
The mean place of characteristic point, is designated as r, equally calculates t1、t2、t3And t4Four mean places of characteristic point, are designated as t, by R from
Moved integrally at the t of position at position r points, now will preferably realize registering between R and T, commented by weighing overlapping degree
Sentence the matching degree of R and T, the overlap similarity before R is moved and after R movements is calculated each the face key element in T respectively, make even
Average, is designated as matching similarity ρ, and formula is as follows:
Wherein, ArIt is the area of plane of reference factor combination, AtIt is the area of target face factor combination, ApThe weight of the former two is translated for R
Folded area, AqThe lap area of the latter two is translated for R;
(4) matching similarity is calculated each the face key element in target face factor combination T, and phase is matched by ordered pair of size
It is ranked up like degree:Each the face key element in target face factor combination T is chosen successively according to descending, when plane of reference factor combination
Its matching similarity value diminishes with the target face factor combination chosen, then stop adding the face key element in T, otherwise then continue
Add the face key element in T, the face factor combination for being constituted as best match face factor combination;When by plane of reference factor combination R
Move to when at the t of position, single face key element T is calculated respectively1To T6With the matching similarity of R, if matching similarity be according to
T1To T6Order reduce successively, then choose T first1Face key element, matching similarity now is ρ 1, then target face key element group
T is added in conjunction2Face key element, by T1With T2Combine, when calculating the matching similarity with R, the value increases than ρ 1, then connect
By T1With T2Combination and continuously add T3Face key element, when matching similarity further increases, then continuously adds T4Face key element,
Reduced with similarity, then stop adding any face key element, T is finally selected successively1、T2And T3Three face key elements, combine
As R1And R2Best match face factor combination;
The conflict probe of step 4 matching result and elimination:
After candidate matches according to formula (3) to carrying out the calculating of matching similarity, matching result can be caused to there are six groups of matchings
Relation, is respectively (R1):(T1) matching similarity 0.82, (R2):(T2) matching similarity 0.82, (R3):(T3) matching similarity
0.85、(R1,R2):(T1,T2) matching similarity 0.9, (R1,R3):(T1,T3) matching similarity 0.82, (R1,R2,R3):(T1,
T2,T3) matching similarity 0.85, it is clear that exist between matching result it is conflicting and conflict phenomenon, therefore, can according to
With similarity value size, matching result is given up or is retained, the matching result for being retained first is (R1,R2):
(T1,T2), because its matching similarity is maximum, matching result (R is then retained again3):(T3), can be preferably using the method
Contradiction and the conflict of matching result are solved, so as to realize the sane matching of the multiple dimensioned face key element of multi-source.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510053743.XA CN104574519B (en) | 2015-01-31 | 2015-01-31 | Multi-source resident's terrain feature exempts from the automatic sane matching process of threshold value |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510053743.XA CN104574519B (en) | 2015-01-31 | 2015-01-31 | Multi-source resident's terrain feature exempts from the automatic sane matching process of threshold value |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104574519A CN104574519A (en) | 2015-04-29 |
CN104574519B true CN104574519B (en) | 2017-05-31 |
Family
ID=53090480
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510053743.XA Expired - Fee Related CN104574519B (en) | 2015-01-31 | 2015-01-31 | Multi-source resident's terrain feature exempts from the automatic sane matching process of threshold value |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104574519B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106886984B (en) * | 2017-03-07 | 2019-10-08 | 国家基础地理信息中心 | A kind of vector element bearing calibration and device |
CN107292919B (en) * | 2017-06-01 | 2019-10-18 | 中国地质大学(武汉) | A kind of multiple dimensioned settlement place matching process of space-oriented subdivision |
CN109615646A (en) * | 2018-12-17 | 2019-04-12 | 石家庄爱赛科技有限公司 | Stereo matching system and solid matching method |
CN111782739A (en) * | 2019-04-04 | 2020-10-16 | 西安四维图新信息技术有限公司 | Map updating method and device |
CN115238800B (en) * | 2022-07-27 | 2023-04-28 | 湖北省地质环境总站 | Vector planar element matching method based on OCSVM |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101833765A (en) * | 2010-04-30 | 2010-09-15 | 天津大学 | Characteristic matching method based on bilateral matching and trilateral restraining |
CN103927785A (en) * | 2014-04-22 | 2014-07-16 | 同济大学 | Feature point matching method for close-range shot stereoscopic image |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2857771B1 (en) * | 2003-07-17 | 2005-09-23 | Thales Sa | AUTOMATIC IDENTIFICATION SYSTEM OF TARGETS |
US20070183688A1 (en) * | 2006-02-03 | 2007-08-09 | Gary Hollfelder | Data management system and method |
-
2015
- 2015-01-31 CN CN201510053743.XA patent/CN104574519B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101833765A (en) * | 2010-04-30 | 2010-09-15 | 天津大学 | Characteristic matching method based on bilateral matching and trilateral restraining |
CN103927785A (en) * | 2014-04-22 | 2014-07-16 | 同济大学 | Feature point matching method for close-range shot stereoscopic image |
Non-Patent Citations (2)
Title |
---|
利用集合矩和叠置分析进行多尺度面要素自动匹配;赵东保 等;《武汉大学学报 信息科学版》;20111130;第36卷(第11期);第1371-1375页 * |
多源点要素的全局一致性自动匹配;刘继宝 等;《测绘与空间地理信息》;20110630;第34卷(第3期);第27-29,32页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104574519A (en) | 2015-04-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104574519B (en) | Multi-source resident's terrain feature exempts from the automatic sane matching process of threshold value | |
CN106156082B (en) | A kind of ontology alignment schemes and device | |
Huh et al. | Detecting conjugate-point pairs for map alignment between two polygon datasets | |
CN103826298B (en) | Wireless sensor network positioning and computing method for collaborative iterative optimization | |
CN100485662C (en) | Characteristic analytical method for product point clouds surface based on dynamic data access model | |
Kim et al. | A new method for matching objects in two different geospatial datasets based on the geographic context | |
CN103337068B (en) | The multiple subarea matching process of spatial relation constraint | |
CN107679566A (en) | A kind of Bayesian network parameters learning method for merging expert's priori | |
CN108168564A (en) | A kind of Data Association based on LHD grey relational grades | |
CN107492120A (en) | Point cloud registration method | |
Zhang et al. | The absolute degree of grey incidence for grey sequence base on standard grey interval number operation | |
CN106600537B (en) | A kind of anisotropy three-dimensional space interpolation method of anti-distance weighting | |
CN103617163B (en) | Quick target association method based on cluster analysis | |
CN104573333A (en) | Method for optimizing of model selection based on clustering analysis | |
CN107909194A (en) | System level testing designs Multipurpose Optimal Method | |
CN103489193A (en) | High-resolution remote-sensing image change detection method facing targets and based on integrating strategy | |
CN105021199B (en) | Multi-model self-adapting method for estimating state and system based on LS | |
CN104700368A (en) | Self-adaptive sliding method of displacement field of digital image relevant method based on kernel function | |
CN104361600B (en) | motion recognition method and system | |
Yi-Bo | Based on DS evidence theory of information fusion improved method | |
CN102706348A (en) | Gravimetric map fast matching method based on triangle | |
Rosas | Multi-map comparison for group concept mapping: An approach for examining conceptual congruence through spatial correspondence | |
CN102930531B (en) | Detection method for repetition structure of building surface image | |
CN104899371B (en) | Survey of Transmission Line method and device | |
CN106846160A (en) | One kind arable land accounts for complementary space checking method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170531 Termination date: 20180131 |
|
CF01 | Termination of patent right due to non-payment of annual fee |