CN103236050A - Auxiliary bank note and worn coin reestablishing method based on graph clustering - Google Patents
Auxiliary bank note and worn coin reestablishing method based on graph clustering Download PDFInfo
- Publication number
- CN103236050A CN103236050A CN2013101622784A CN201310162278A CN103236050A CN 103236050 A CN103236050 A CN 103236050A CN 2013101622784 A CN2013101622784 A CN 2013101622784A CN 201310162278 A CN201310162278 A CN 201310162278A CN 103236050 A CN103236050 A CN 103236050A
- Authority
- CN
- China
- Prior art keywords
- point
- fragment
- fragmentation pattern
- worn coin
- relation
- 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.)
- Granted
Links
Images
Landscapes
- Image Processing (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses an auxiliary bank note and worn coin reestablishing method based on graph clustering. The auxiliary bank note and worn coin reestablishing method comprises the following steps of scanning worn coin fragments to a computer for processing; extracting feature points and feature vectors in worn coin fragment images by using an ORB (Oriented FAST and Rotated BRIEF (Binary Robust Independent Elementary Features)) algorithm; matching fragment images by using an MSAC (M-Estimator Sample and Consensus) algorithm; carrying out range conversion on edges of fragments and then judging the relation between every two fragments; clustering the worn coin fragments by using a method based on the graph theory to obtain a clustering result; and indicating related persons how to splice and reestablish actual worn coin fragments according to the clustering result. According to the auxiliary bank note and worn coin reestablishing method disclosed by the invention, the defect that a great number of pattern fragments which are completely from the same complete pattern are processed in related technical fields is made up; and the auxiliary bank note and worn coin reestablishing method is simple in operation and is quick and effective in algorithm and can be used for splicing and restoring a large number of worn coin fragments and other picture fragments under similar conditions.
Description
Technical field
The present invention relates to image processing, matrix theory and graph theory, relate in particular to image splicing and reconstruction.More particularly, the present invention relates to use computing machine to assist splicing to recover the method for complete bank note a large amount of worn coin fragments.
Background technology
Bank note has portable advantage as a kind of currency symbol that does not have actual value, after the careless damaged division of bank note, if can be spliced into complete bank note, just can the bank note that renew not lost value.The law that China exchanges about worn coin and the correlation technique of differentiating worn coin are also comparatively perfect, have tens of relevant patents at least.
Yet the prerequisite of exchanging worn coin is that it is spliced into complete bank note, but because number is excessive, number of tiles is too much, the situation that causes being difficult to manual splicing happens occasionally, and worn coin denomination total value is the unaffordable losses of a lot of common people usually up to tens thousand of units under such situation.
Be similar to the such image fragment splicing problem of picture mosaic, a lot of advanced technologies are arranged, splice as utilize profile and color that people such as M. G. Chung propose, P. people such as Franti proposes the method based on k neighbour figure, A. C. Gallagher proposes to solve not specified fragment joining method, E. people such as Justino proposes to utilize the method for reconstructing of characteristic matching, rebuilds method of multiple image etc. when people such as S. Cao propose.
Restore the problem that this a large amount of fragmentation pattern picture all comes from a same complete pattern for worn coin, but do not have a cover total solution.Discover that the shape of bank note fragment, color and profile are not suitable for carrying out the splicing between fragment and the fragment; So how to find the position of fragment on complete pattern exactly, how to find the fragment that belongs to same bank note, to recover all bank note, be problem demanding prompt solution.
Summary of the invention
This purpose is to disclose a kind ofly to be handled and the auxiliary banknote worn coin splicing of graph theory and the method for rebuilding based on image, by Computer Processing worn coin fragmentation pattern picture and provide the result, so that by manually the worn coin fragment of reality being spliced.The present invention is used for a large amount of worn coin fragments are spliced to recover complete bank note, and the picture fragment splicing that also can be used for other analogue recovers.
The present invention is by the following technical solutions to achieve these goals:
A kind of auxiliary banknote worn coin method for reconstructing based on the figure cluster is characterized in that may further comprise the steps:
Step 1: being the back side of all worn coin fragments of background scans and a complete bank note with black, is whole N fragments numberings automatically;
Step 2: passing threshold is handled and is extracted whole N fragmentation pattern pictures;
Step 3: use the ORB algorithm to extract unique point and the proper vector of all fragmentation pattern pictures and complete banknote image, use the MSAC algorithm to carry out the fragment images match;
Step 4: handle by morphology, extract the fragment edge, and be the foreground point with the point on the edge, the image in the zone of fragment and 10 to 30 pixel wide of periphery thereof is carried out signed Euclidean distance conversion;
Step 5: according to marginal point and the corresponding signed distance map of all fragments in the step 4, judge the relation between all fragments, relation is divided three classes: the relation of repelling each other, neighbouring relations, uncertainty relation;
Step 6: structure adjacency matrix A and the matrix R that repels each other,
Step 7: calculate diagonal matrix
And diagonal matrix
, and
,,
, wherein I is unit matrix, calculates
Generalized eigenvalue and with they descending orderings, make k generalized eigenvalue characteristic of correspondence vector be
(column vector) constitutes matrix with preceding k generalized eigenvector
Step 8: every row of matrix Y is carried out normalization respectively obtain matrix X, the quadratic sum that makes every row element of X all is that 1, X is the N*k matrix, makes its i capable
Be i the position of worn coin fragment in the cluster space;
Step 9: based on graph theory method the worn coin image is carried out cluster, specifically be divided into
Step 9.1: design of graphics: each fragment is as a summit
, each summit
Have 3 parameters: the class center
, fragmentation pattern picture size
, summit internal fragment quantity
And additionally define the constant S of a complete banknote image size of representative; To any two summits
With
, the structure limit
, every limit
1 or 2 parameter is arranged: first parameter is the relation of repelling each other between two fragments
, when satisfying condition
The time, the 2nd parameter is the Euclidean distance at two fragment class centers
Step 9.2: in all limits, find minimum Euclidean distance
, if satisfy the relation of repelling each other
And fragmentation pattern picture size
, then with the summit
With
Merge into a new summit
,
3 parameters be recalculated as respectively: the class center
, summit internal fragment quantity
, fragmentation pattern picture size
, and delete the limit
For whole other summits, with
Be example, structure
With
The limit
,
The repellence calculation of parameter be
, if
Be not equal to 1, calculate
Second parameter
Step 9.3: repetitive cycling step 9.2, up to there not being one group of new summit to merge;
Step 10: the cluster result of computing machine output fragmentation pattern picture.
In the such scheme, in the described step 1, computing machine by fragmentation pattern as the sequencing of input computer automatically for each fragment is numbered, and generate the fragmentation pattern picture of a band numbering.
In the such scheme, in the described step 3:
The proper vector of the unique point on the proper vector of the unique point on the fragmentation pattern picture and the complete banknote image is poor, calculates its Euclidean distance, and topography's information of more little then two the proper vector characteristic of correspondence points of distance is more approaching;
Extract unique point and proper vector thereof in the complete banknote image, and be stored in the computing machine;
At each fragmentation pattern picture, extract wherein unique point and proper vector thereof, at each unique point in the fragmentation pattern picture, the unique point of choosing in the complete banknote image with its proper vector Euclidean distance minimum is right as a candidate matches point, in a secondary fragmentation pattern picture, the candidate matches point of all and complete banknote image calculates the relevant position of fragment in complete bank note to the input as the MSAC algorithm;
Described MSAC algorithm may further comprise the steps into:
Step 3.1: at a fragmentation pattern picture, with each unique point wherein, and in characteristic vector space the unique point in the immediate complete banknote, it is right to be configured to a pair of candidate matches point, the right quantity of match point equals the quantity of the unique point of this fragmentation pattern picture, unique point n in the note fragmentation pattern picture, coordinate is
, it is right that the unique point n ' in itself and the complete banknote constitutes a pair of candidate matches point, and the coordinate of n ' is
Step 3.2: randomly draw two points of all candidate matches point centerings to (1,1 ') and (2,2 '), its mid point 1, point 2 are from the fragmentation pattern picture, and point 1 ', point 2 ' are from complete banknote image;
Step 3.3: calculation level 1 and point between 2 apart from d
1,2And put 1 ' and point between 2 ' apart from d
1 ', 2 ', if
Then carry out step 3.2 again, it is right to randomly draw two new match points;
Step 3.4: point 1 constitutes vector v to point 2
1,2, point 1 ' constitutes vector v to point 2 '
1 ', 2 ', compute vector v
1,2With vector v
1 ', 2 'Between angle
, and calculate the point 1 in the fragmentation pattern picture, the coordinate after 2 minutes the rotational transform of point: namely calculate
With
Step 3.5: displacement calculating
,
, wherein
Be the coordinate of point 1 ',
Be the horizontal ordinate after point 1 rotational transform that obtains in the step 3.4, other symbol by that analogy;
Step 3.6: to each the unique point n on the fragment, coordinate is
, the displacement that the angle that use step 3.4 obtains and step 3.5 obtain uses rigid transformation that they are projected in the complete banknote image:
, and calculate the Euclidean distance between the candidate matches point in its subpoint and its complete banknote image
, in the formula in
With
Be the coordinate of a n ', some n and some n ' are that a pair of candidate matches point of describing in the step 3.1 is right; Calculate the matching error of a single point
, wherein
It is a preset threshold value; Angle
And displacement parameter
,
The error of mating can be designated as the matching error of having a few
Step 3.7: above-mentioned steps 3.2 to step 3.6 circulation repeatedly, can both obtain the matching error e of one group of angle and displacement, with the angle of wherein minimum e correspondence and displacement angle and the displacement as the best between fragmentation pattern picture and the complete banknote image at every turn;
Step 3.8: best angle and displacement according to step 3.7 calculates look like to be converted into its correspondence position in complete banknote image with fragmentation pattern;
Step 3.9: at each fragmentation pattern picture, repeat above-mentioned steps 3.1 to step 3.8, convert it to its correspondence position in complete banknote image.
In the such scheme, described step 4 is characterized in that, in the distance map that signed Euclidean distance conversion obtains, the transformation results in the fragment zone is negative, and extra-regional transformation results is positive number, and the transformation results on the edges of regions is 0.
In the such scheme, described step 5, it is characterized in that, given two positions are close, need to judge the fragment of mutual relationship, hereinafter referred to as A fragment and B fragment, can be by the distance value of retrieval A fragment marginal point correspondence in B fragment distance map, or the distance value of B fragment marginal point correspondence in A fragment distance map, judge the relation between them:
The relation of repelling each other: if any marginal point of A fragment at the distance value of fragment B distance map corresponding point less than-10, or any marginal point of B fragment at the distance value of fragment A distance map corresponding point less than-10, then judge to be the relation of repelling each other between A, the B fragment, and no longer judge other relation between A, the B fragment;
Neighbouring relations: if be no less than the absolute value of 25 somes distance value of corresponding point in B fragment distance map less than 3 in the A fragment marginal point, and be no less than the absolute value of 25 somes distance value of corresponding point in A fragment distance map in the B fragment marginal point less than 3, then judge to be neighbouring relations between A, the B fragment;
Uncertainty relation: do not satisfy above-mentioned arbitrary relation as if A, B fragment, then judge to be uncertainty relation between A, the B fragment.
Because so the present invention adopts above technical scheme to possess following beneficial effect:
This purpose is to disclose a kind of based on image processing and the auxiliary banknote worn coin splicing of graph theory and the method for rebuilding, by Computer Processing worn coin fragmentation pattern picture and provide the result, in order to splice by artificial worn coin fragment with reality, the present invention has remedied correlative technology field to handling the disappearance of " a large amount of pattern fragments all come from a same complete pattern " such problem, simple to operate, algorithm is effective fast, can avoid meaningless trial, saves a large amount of time and manpower.The present invention is used for a large amount of worn coin fragments are spliced to recover complete bank note, and the picture fragment splicing that also can be used for other analogue recovers.
Description of drawings
Fig. 1 is the synoptic diagram of scanning worn coin;
Fig. 2 extracts the synoptic diagram of fragmentation pattern picture for threshold process;
Fig. 3 is that fragmentation pattern is as the synoptic diagram of matching result;
Fig. 4 is fragment edge exemplary plot;
Fig. 5 is for to carry out the distance map that signed Euclidean distance conversion obtains to Fig. 4;
Fig. 6 is the corresponding relation synoptic diagram of A fragment marginal point in B fragment distance map;
Fig. 7 is the process flow diagram of whole auxiliary banknote worn coin reconstruction algorithm.
Embodiment
To be described in detail the present invention below, and be to be noted that described embodiment only is intended to be convenient to the understanding of the present invention, and it is not played any restriction effect.
Auxiliary banknote worn coin is rebuild and is divided into 10 steps:
Step 1: being the back side (101) that background (102) scans all worn coin fragments and a complete bank note with black, is whole N fragments numberings (103) automatically,
Particularly, the method of scanning paper currency has a lot, include but not limited to use scanner, the high instrument etc. of clapping, the description of the relevant number of pixels of the present invention is set up based on the basis of the scanning accuracy of having used 72 dpi, complete 100 yuans resolution is 600*297, in the practical application, need carry out convergent-divergent to these numerical value according to resolution sizes, too small resolution can't guarantee image detail, and excessive resolution can increase to be calculated and the storage burden;
Computing machine by fragmentation pattern as the sequencing of input computer automatically for each fragment is numbered, and generate the fragmentation pattern picture of a band numbering, convenient manually according to the actual fragment of fragment numbering retrieval.
At the Renminbi fragment, the present invention only scans the back side of worn coin, because the white space of back side pattern than the front still less, and does not have the bank note numbering, the white space of banknote pattern and the different couplings that all are unfavorable for fragment in the subsequent step of numbering, synoptic diagram such as Fig. 1 of scanning worn coin;
Step 2: passing threshold is handled and is extracted whole N fragmentation pattern pictures,
Particularly, because the scan image background is elected black as, only need to use black background to be removed near 0 threshold process, extract the higher fragmentation pattern picture of brightness; Threshold process is extracted synoptic diagram such as Fig. 2 of fragmentation pattern picture, and black is removed background among the figure, and middle two zones are the fragmentation pattern picture that extracts;
Step 3: use the ORB algorithm to extract unique point and the proper vector of all fragmentation pattern pictures and complete banknote image, use the MSAC algorithm to carry out the fragment images match,
Particularly, the detection of unique point and the calculating of proper vector are based on ORB(Oriented FAST and Rotated BRIEF) algorithm, be quick key point detection algorithm (FAST keypoint detector) and the BRIEF(Binary Robust Independent Elementary Features that proposes in recent years) the two combination of descriptor, the proper vector of the unique point on the proper vector of the unique point on the fragmentation pattern picture and the complete banknote image is poor, calculate its Euclidean distance, topography's information of these two proper vector characteristic of correspondence points of the more little explanation of distance is more approaching;
At first, extract unique point and proper vector thereof in the complete banknote image, and be stored in the computing machine.
At each fragmentation pattern picture, extract wherein unique point and proper vector thereof.At each unique point in the fragmentation pattern picture, the unique point of choosing in the complete banknote image with its proper vector Euclidean distance minimum is right as a candidate matches point.In a secondary fragmentation pattern picture, the candidate matches point of all and complete banknote image calculates the relevant position of fragment in complete bank note to the input as the MSAC algorithm.MSAC(M-estimator sample and consensus) be based on RANSAC(random sample consensus, random sampling consistency algorithm) one improve algorithm, its concrete steps are:
Step 3.1: at a fragmentation pattern picture, with each unique point wherein, and in characteristic vector space the unique point in the immediate complete banknote, it is right to be configured to a pair of candidate matches point, the right quantity of match point equals the quantity of the unique point of this fragmentation pattern picture, unique point n in the note fragmentation pattern picture, coordinate is
, it is right that the unique point n ' in itself and the complete banknote constitutes a pair of candidate matches point, and the coordinate of n ' is
Step 3.2: two points randomly drawing all candidate matches point centerings are right, and for example (1,1 ') and (2,2 '), its mid point 1, point 2 are from the fragmentation pattern picture, and point 1 ', point 2 ' are from complete banknote image;
Step 3.3: calculation level 1 and point between 2 apart from d
1,2And put 1 ' and point between 2 ' apart from d
1 ', 2 ', if
Then carry out step 3.2 again, it is right to randomly draw two new match points;
Step 3.4: point 1 constitutes vector v to point 2
1,2, point 1 ' constitutes vector v to point 2 '
1 ', 2 ', compute vector v
1,2With vector v
1 ', 2 'Between angle
, and calculate the point 1 in the fragmentation pattern picture, the coordinate after 2 minutes the rotational transform of point: namely calculate
With
Step 3.5: displacement calculating
,
, wherein
Be the coordinate of point 1 ',
Be the horizontal ordinate after point 1 rotational transform that obtains in the step 3.4, other symbol by that analogy;
Step 3.6: to each the unique point n on the fragment, coordinate is
, the displacement that the angle that use step 3.4 obtains and step 3.5 obtain uses rigid transformation that they are projected in the complete banknote image:
, and calculate the Euclidean distance between the candidate matches point in its subpoint and its complete banknote image
, wherein
With
Be the coordinate of a n ', some n and some n ' are that a pair of candidate matches point of describing in the step 3.1 is right; Calculate the matching error of a single point
, wherein
It is a preset threshold value; Angle
And displacement parameter
,
The error of mating can be designated as the matching error of having a few
Step 3.7: above-mentioned steps 3.2 to step 3.6 circulation repeatedly (is generally thousands of to tens thousand of times), can both obtain the matching error e of one group of angle and displacement, with the angle of wherein minimum e correspondence and displacement angle and the displacement as the best between fragmentation pattern picture and the complete banknote image at every turn;
Step 3.8: best angle and displacement according to step 3.7 calculates look like to be converted into its correspondence position in complete banknote image with fragmentation pattern;
Step 3.9: at each fragmentation pattern picture, repeat above-mentioned steps 3.1 to step 3.8, convert it to its correspondence position in complete banknote image.
Fragmentation pattern is as synoptic diagram such as Fig. 3 of matching result, 302 sizes that represent a complete bank note among the figure, fragmentation pattern picture after the 301 representative couplings, be placed on the relevant position of complete bank note, 301 are positioned at a corner position of complete bank note just, be the explanation of drawing for convenience, the fragmentation pattern picture that fragments matching is actually each diverse location all returns to its relevant position on complete bank note;
Step 4: handle by morphology, extract the fragment edge, and be the foreground point with the point on the edge, the image in the zone of fragment and 10 to 30 pixel wide of periphery thereof is carried out signed Euclidean distance conversion,
Particularly, use complete 1 mask of 3*3 to carry out morphological erosion to the fragment foreground area, poor with original fragment foreground area again, can obtain the fragment marginal point, fragment marginal point exemplary plot such as Fig. 4.
Need be to what signed Euclidean distance conversion illustrated, in the distance map that obtains, transformation results in the fragment zone is negative, extra-regional transformation results is positive number, transformation results on the edges of regions is 0, and Fig. 4 is carried out distance map such as Fig. 5 that signed Euclidean distance conversion obtains, among the figure in order clearly to observe the edge, change the edge into white and show, the more approaching white expression of other parts color numerical value is more big;
Step 5: according to marginal point and the corresponding signed distance map of all fragments in the step 4, judge the relation between all fragments, relation is divided three classes: the relation of repelling each other, neighbouring relations, uncertainty relation,
Given two positions are close, need to judge the fragment (hereinafter referred to as A fragment and B fragment) of mutual relationship, can be by the distance value of retrieval A fragment marginal point (303) correspondence in B fragment distance map (304), the corresponding relation synoptic diagram of A fragment marginal point in B fragment distance map seen Fig. 6, or the distance value of B fragment marginal point correspondence in A fragment distance map, judge the relation between them.
The relation of repelling each other: if any marginal point of A fragment at the distance value of fragment B distance map corresponding point less than-10, or any marginal point of B fragment at the distance value of fragment A distance map corresponding point less than-10, then judge to be the relation of repelling each other between A, the B fragment, and no longer judge other relation between A, the B fragment.
Neighbouring relations: if be no less than the absolute value of 25 somes distance value of corresponding point in B fragment distance map less than 3 in the A fragment marginal point, and be no less than the absolute value of 25 somes distance value of corresponding point in A fragment distance map in the B fragment marginal point less than 3, then judge to be neighbouring relations between A, the B fragment.
Uncertainty relation: do not satisfy above-mentioned arbitrary relation as if A, B fragment, then judge to be uncertainty relation between A, the B fragment.
Step 6: structure adjacency matrix A and the matrix R that repels each other,
Step 7: calculate diagonal matrix
And diagonal matrix
, and
,
, wherein I is unit matrix, calculates
Generalized eigenvalue and with they descending orderings, k maximum generalized eigenvalue characteristic of correspondence vector is before the order
(column vector) constitutes matrix with preceding k generalized eigenvector
Step 8: every row of matrix Y is carried out normalization respectively obtain matrix X, making the quadratic sum of every row element of X all is 1, particularly, establishes
Be the element of the capable j row of matrix Y i, the element of the capable j row of matrix X i is so
, X is the N*k matrix, makes its i capable
Be i the position of worn coin fragment in the cluster space;
Step 9: based on graph theory method the worn coin image is carried out cluster, specifically be divided into
Step 9.1: design of graphics: each fragment is as a summit
, each summit
Have 3 parameters: the class center
, fragmentation pattern picture size
, summit internal fragment quantity
And additionally define the constant S of a complete banknote image size of representative; To any two summits
With
, the structure limit
, every limit
1 or 2 parameter is arranged: first parameter is the relation of repelling each other between two fragments
, when satisfying condition
The time, the 2nd parameter is the Euclidean distance at two fragment class centers
Step 9.2: in all limits, find minimum Euclidean distance
, if satisfy the relation of repelling each other
And fragmentation pattern picture size
, then with the summit
With
Merge into a new summit
,
3 parameters be recalculated as respectively: the class center
, summit internal fragment quantity
, fragmentation pattern picture size
, and delete the limit
For whole other summits, with
Be example, structure
With
The limit
,
The repellence calculation of parameter be
, if
Be not equal to 1, calculate
Second parameter
Step 9.3: repetitive cycling step 9.2, up to there not being one group of new summit to merge;
Step 10: the cluster result of computing machine output fragmentation pattern picture.
Process flow diagram such as Fig. 7 of whole auxiliary banknote worn coin reconstruction algorithm.
Can indicate the related personnel how the worn coin fragment splicing of reality to be rebuild according to cluster result.
The above is only for the present invention's preferred embodiment; be not in order to limit claim of the present invention and specific use-pattern; other do not break away from the equivalence change of finishing under the disclosed spirit or modify, and all should be included in protection scope of the present invention.
Claims (5)
1. auxiliary banknote worn coin method for reconstructing based on the figure cluster is characterized in that may further comprise the steps:
Step 1: being the back side of all worn coin fragments of background scans and a complete bank note with black, is whole N fragments numberings automatically;
Step 2: passing threshold is handled and is extracted whole N fragmentation pattern pictures;
Step 3: use the ORB algorithm to extract unique point and the proper vector of all fragmentation pattern pictures and complete banknote image, use the MSAC algorithm to carry out the fragment images match;
Step 4: handle by morphology, extract the fragment edge, and be the foreground point with the point on the edge, the image in the zone of fragment and 10 to 30 pixel wide of periphery thereof is carried out signed Euclidean distance conversion;
Step 5: according to marginal point and the corresponding signed distance map of all fragments in the step 4, judge the relation between all fragments, relation is divided three classes: the relation of repelling each other, neighbouring relations, uncertainty relation;
Step 6: structure adjacency matrix A and the matrix R that repels each other,
Step 7: calculate diagonal matrix
And diagonal matrix
, and
,,
, wherein I is unit matrix, calculates
Generalized eigenvalue and with they descending orderings, make k generalized eigenvalue characteristic of correspondence vector be
(column vector) constitutes matrix with preceding k generalized eigenvector
Step 8: every row of matrix Y is carried out normalization respectively obtain matrix X, the quadratic sum that makes every row element of X all is that 1, X is the N*k matrix, makes its i capable
Be i the position of worn coin fragment in the cluster space;
Step 9: based on graph theory method the worn coin image is carried out cluster, specifically be divided into
Step 9.1: design of graphics: each fragment is as a summit
, each summit
Have 3 parameters: the class center
, fragmentation pattern picture size
, summit internal fragment quantity
And additionally define the constant S of a complete banknote image size of representative; To any two summits
With
, the structure limit
, every limit
1 or 2 parameter is arranged: first parameter is the relation of repelling each other between two fragments
, when satisfying condition
The time, the 2nd parameter is the Euclidean distance at two fragment class centers
Step 9.2: in all limits, find minimum Euclidean distance
, if satisfy the relation of repelling each other
And fragmentation pattern picture size
, then with the summit
With
Merge into a new summit
,
3 parameters be recalculated as respectively: the class center
, summit internal fragment quantity
, fragmentation pattern picture size
, and delete the limit
For whole other summits, with
Be example, structure
With
The limit
,
The repellence calculation of parameter be
, if
Be not equal to 1, calculate
Second parameter
Step 9.3: repetitive cycling step 9.2, up to there not being one group of new summit to merge;
Step 10: the cluster result of computing machine output fragmentation pattern picture.
2. a kind of auxiliary banknote worn coin method for reconstructing based on the figure cluster according to claim 1, it is characterized in that, in the described step 1, computing machine by fragmentation pattern as the sequencing of input computer automatically for each fragment is numbered, and generate the fragmentation pattern picture of a band numbering.
3. a kind of auxiliary banknote worn coin method for reconstructing based on the figure cluster according to claim 1 is characterized in that, in the described step 3:
The proper vector of the unique point on the proper vector of the unique point on the fragmentation pattern picture and the complete banknote image is poor, calculates its Euclidean distance, and topography's information of more little then two the proper vector characteristic of correspondence points of distance is more approaching;
Extract unique point and proper vector thereof in the complete banknote image, and be stored in the computing machine;
At each fragmentation pattern picture, extract wherein unique point and proper vector thereof, at each unique point in the fragmentation pattern picture, the unique point of choosing in the complete banknote image with its proper vector Euclidean distance minimum is right as a candidate matches point, in a secondary fragmentation pattern picture, the candidate matches point of all and complete banknote image calculates the relevant position of fragment in complete bank note to the input as the MSAC algorithm;
Described MSAC algorithm may further comprise the steps into:
Step 3.1: at a fragmentation pattern picture, with each unique point wherein, and in characteristic vector space the unique point in the immediate complete banknote, it is right to be configured to a pair of candidate matches point, the right quantity of match point equals the quantity of the unique point of this fragmentation pattern picture, unique point n in the note fragmentation pattern picture, coordinate is
, it is right that the unique point n ' in itself and the complete banknote constitutes a pair of candidate matches point, and the coordinate of n ' is
Step 3.2: randomly draw two points of all candidate matches point centerings to (1,1 ') and (2,2 '), its mid point 1, point 2 are from the fragmentation pattern picture, and point 1 ', point 2 ' are from complete banknote image;
Step 3.3: calculation level 1 and point between 2 apart from d1,2 and put 1 ' with point 2 ' between apart from d1 ', 2 ', as if
Then carry out step 3.2 again, it is right to randomly draw two new match points;
Step 3.4: point 1 constitutes vector v 1,2 to point 2, and point 1 ' constitutes a vector v 1 ', 2 ' to point 2 ', compute vector v1,2 with vector v 1 ', 2 ' between angle
, and calculate the point 1 in the fragmentation pattern picture, the coordinate after 2 minutes the rotational transform of point: namely calculate
With
Step 3.5: displacement calculating
,
, wherein
Be the coordinate of point 1 ',
Be the horizontal ordinate after point 1 rotational transform that obtains in the step 3.4, other symbol by that analogy;
Step 3.6: to each the unique point n on the fragment, coordinate is
, the displacement that the angle that use step 3.4 obtains and step 3.5 obtain uses rigid transformation that they are projected in the complete banknote image:
, and calculate the Euclidean distance between the candidate matches point in its subpoint and its complete banknote image
, in the formula in
With
Be the coordinate of a n ', some n and some n ' are that a pair of candidate matches point of describing in the step 3.1 is right; Calculate the matching error of a single point
, wherein
It is a preset threshold value; Angle
And displacement parameter
,
The error of mating can be designated as the matching error of having a few
Step 3.7: above-mentioned steps 3.2 to step 3.6 circulation repeatedly, can both obtain the matching error e of one group of angle and displacement, with the angle of wherein minimum e correspondence and displacement angle and the displacement as the best between fragmentation pattern picture and the complete banknote image at every turn;
Step 3.8: best angle and displacement according to step 3.7 calculates look like to be converted into its correspondence position in complete banknote image with fragmentation pattern;
Step 3.9: at each fragmentation pattern picture, repeat above-mentioned steps 3.1 to step 3.8, convert it to its correspondence position in complete banknote image.
4. a kind of auxiliary banknote worn coin method for reconstructing based on the figure cluster according to claim 1, it is characterized in that: described step 4, it is characterized in that, in the distance map that signed Euclidean distance conversion obtains, transformation results in the fragment zone is negative, extra-regional transformation results is positive number, and the transformation results on the edges of regions is 0.
5. a kind of auxiliary banknote worn coin method for reconstructing based on the figure cluster according to claim 1, it is characterized in that: described step 5, it is characterized in that, given two positions are close, need to judge the fragment of mutual relationship, hereinafter referred to as A fragment and B fragment, can be by the distance value of retrieval A fragment marginal point correspondence in B fragment distance map, or the distance value of B fragment marginal point correspondence in A fragment distance map, judge the relation between them:
The relation of repelling each other: if any marginal point of A fragment at the distance value of fragment B distance map corresponding point less than-10, or any marginal point of B fragment at the distance value of fragment A distance map corresponding point less than-10, then judge to be the relation of repelling each other between A, the B fragment, and no longer judge other relation between A, the B fragment;
Neighbouring relations: if be no less than the absolute value of 25 somes distance value of corresponding point in B fragment distance map less than 3 in the A fragment marginal point, and be no less than the absolute value of 25 somes distance value of corresponding point in A fragment distance map in the B fragment marginal point less than 3, then judge to be neighbouring relations between A, the B fragment;
Uncertainty relation: do not satisfy above-mentioned arbitrary relation as if A, B fragment, then judge to be uncertainty relation between A, the B fragment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310162278.4A CN103236050B (en) | 2013-05-06 | 2013-05-06 | A kind of auxiliary bank note worn coin method for reconstructing based on figure cluster |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310162278.4A CN103236050B (en) | 2013-05-06 | 2013-05-06 | A kind of auxiliary bank note worn coin method for reconstructing based on figure cluster |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103236050A true CN103236050A (en) | 2013-08-07 |
| CN103236050B CN103236050B (en) | 2015-08-12 |
Family
ID=48884089
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310162278.4A Expired - Fee Related CN103236050B (en) | 2013-05-06 | 2013-05-06 | A kind of auxiliary bank note worn coin method for reconstructing based on figure cluster |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103236050B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679634A (en) * | 2013-12-10 | 2014-03-26 | 锐达互动科技股份有限公司 | Method for splicing recovery of two-dimensional irregular fragments |
| CN104915949A (en) * | 2015-04-08 | 2015-09-16 | 华中科技大学 | Image matching algorithm of bonding point characteristic and line characteristic |
| CN107180412A (en) * | 2017-06-15 | 2017-09-19 | 北京工业大学 | Transverse and longitudinal based on floor projection and seed point constraint K mean cluster shreds scraps of paper method for reconstructing |
| CN107229953A (en) * | 2017-06-06 | 2017-10-03 | 西南石油大学 | A kind of broken document joining method based on DFS with improvement central cluster method |
| CN108805811A (en) * | 2018-05-30 | 2018-11-13 | 山东师范大学 | A kind of natural image intelligence pattern splicing method and system based on nonconvex quadratic programming |
| CN109978006A (en) * | 2019-02-25 | 2019-07-05 | 北京邮电大学 | Clustering method and device |
| CN111784697A (en) * | 2020-06-30 | 2020-10-16 | 北京百度网讯科技有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100246976A1 (en) * | 2009-02-20 | 2010-09-30 | Samsung Electronics Co., Ltd. | Method of creating a composite image |
| CN102855649A (en) * | 2012-08-23 | 2013-01-02 | 山东电力集团公司电力科学研究院 | Method for splicing high-definition image panorama of high-pressure rod tower on basis of ORB (Object Request Broker) feature point |
-
2013
- 2013-05-06 CN CN201310162278.4A patent/CN103236050B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100246976A1 (en) * | 2009-02-20 | 2010-09-30 | Samsung Electronics Co., Ltd. | Method of creating a composite image |
| CN102855649A (en) * | 2012-08-23 | 2013-01-02 | 山东电力集团公司电力科学研究院 | Method for splicing high-definition image panorama of high-pressure rod tower on basis of ORB (Object Request Broker) feature point |
Non-Patent Citations (4)
| Title |
|---|
| HAIRONG LIU 等: "Automated Assembly of Shredded Pieces From Multiple Photos", 《IEEE TRANSACTIONS ON MULTIMEDIA》 * |
| P. FRANTI 等: "Fast agglomerative clustering using a -nearest neighbor graph", 《IEEE TRANSACTION ON PATTERN ANALYSIS AND MACHING INTELLIGENCE》 * |
| S.X.YU 等: "Understanding popout through repulsion", 《IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| 庄俊东: "基于数字图像处理的人民币碎片拼接方法的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679634A (en) * | 2013-12-10 | 2014-03-26 | 锐达互动科技股份有限公司 | Method for splicing recovery of two-dimensional irregular fragments |
| CN104915949A (en) * | 2015-04-08 | 2015-09-16 | 华中科技大学 | Image matching algorithm of bonding point characteristic and line characteristic |
| CN104915949B (en) * | 2015-04-08 | 2017-09-29 | 华中科技大学 | A kind of image matching method of combination point feature and line feature |
| CN107229953A (en) * | 2017-06-06 | 2017-10-03 | 西南石油大学 | A kind of broken document joining method based on DFS with improvement central cluster method |
| CN107180412A (en) * | 2017-06-15 | 2017-09-19 | 北京工业大学 | Transverse and longitudinal based on floor projection and seed point constraint K mean cluster shreds scraps of paper method for reconstructing |
| CN107180412B (en) * | 2017-06-15 | 2020-10-16 | 北京工业大学 | Horizontal and vertical shredded paper sheet reconstruction method based on horizontal projection and K-means clustering |
| CN108805811A (en) * | 2018-05-30 | 2018-11-13 | 山东师范大学 | A kind of natural image intelligence pattern splicing method and system based on nonconvex quadratic programming |
| CN108805811B (en) * | 2018-05-30 | 2022-06-24 | 山东师范大学 | Natural image intelligent picture splicing method and system based on non-convex quadratic programming |
| CN109978006A (en) * | 2019-02-25 | 2019-07-05 | 北京邮电大学 | Clustering method and device |
| CN109978006B (en) * | 2019-02-25 | 2021-02-19 | 北京邮电大学 | Face image clustering method and device |
| CN111784697A (en) * | 2020-06-30 | 2020-10-16 | 北京百度网讯科技有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
| CN111784697B (en) * | 2020-06-30 | 2024-03-08 | 北京百度网讯科技有限公司 | Image processing method, device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103236050B (en) | 2015-08-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103236050A (en) | Auxiliary bank note and worn coin reestablishing method based on graph clustering | |
| CN108010045A (en) | Visual pattern characteristic point error hiding method of purification based on ORB | |
| Trulls et al. | Dense segmentation-aware descriptors | |
| CN108596867A (en) | A kind of picture bearing calibration and system based on ORB algorithms | |
| CN104036480B (en) | Quick elimination Mismatching point method based on surf algorithm | |
| CN105654421B (en) | Based on the projective transformation image matching method for converting constant low-rank texture | |
| CN105303615A (en) | Combination method of two-dimensional stitching and three-dimensional surface reconstruction of image | |
| CN104616297A (en) | Improved SIFI algorithm for image tampering forensics | |
| CN102982524B (en) | Splicing method for corn ear order images | |
| CN115082617B (en) | Multi-view optimization-based pipeline three-dimensional reconstruction method, device and storage medium | |
| CN104240231A (en) | Multi-source image registration based on local structure binary pattern | |
| CN105913435A (en) | Multidimensional remote sensing image matching method and multidirectional remote sensing image matching system suitable for large area | |
| CN107945221A (en) | A kind of three-dimensional scenic feature representation based on RGB D images and high-precision matching process | |
| CN103226831A (en) | Image matching method utilizing block Boolean operation | |
| Shi et al. | Self-supervised shape alignment for sports field registration | |
| Liu et al. | Improved rectangle template matching based feature point matching algorithm | |
| CN117710711B (en) | Optical and SAR image matching method based on lightweight depth convolution network | |
| Bellavia et al. | Image orientation with a hybrid pipeline robust to rotations and wide-baselines | |
| Li et al. | Solving a special type of jigsaw puzzles: Banknote reconstruction from a large number of fragments | |
| Yuan et al. | Structure flow-guided network for real depth super-resolution | |
| Liang et al. | Robust hashing with local tangent space alignment for image copy detection | |
| Jin et al. | Registration of UAV images using improved structural shape similarity based on mathematical morphology and phase congruency | |
| Schmid et al. | Building descriptors from local feature neighborhoods for applications in semi-dense 3D reconstruction | |
| CN107993230A (en) | Distorted image detection method based on triangle gridding comprehensive characteristics | |
| Barrera et al. | Multimodal template matching based on gradient and mutual information using scale-space |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150812 Termination date: 20160506 |