CN105844290A - Method of matching same objects in image and apparatus thereof - Google Patents
Method of matching same objects in image and apparatus thereof Download PDFInfo
- Publication number
- CN105844290A CN105844290A CN201610153710.7A CN201610153710A CN105844290A CN 105844290 A CN105844290 A CN 105844290A CN 201610153710 A CN201610153710 A CN 201610153710A CN 105844290 A CN105844290 A CN 105844290A
- Authority
- CN
- China
- Prior art keywords
- image
- match point
- match
- point
- point set
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The invention relates to a method of matching a plurality of same objects in an image and an apparatus thereof. The method of matching the plurality of same objects in the image comprises the following steps of extracting a characteristic point set of a given object and taking as a first characteristic point set, and extracting a characteristic point set of an image to be matched and taking as a second characteristic point set; matching characteristic points in the first characteristic point set and the second characteristic point set so as to acquire a matching point set on the image to be matched; and using a clustering algorithm to classify the matching point set according to a distribution density so as to form a plurality of matching point subsets. By using the technical scheme in the invention, classification is accurate.
Description
Technical field
It relates to field of image recognition, particularly relate to mate the method for multiple same object in image and
Device.
Background technology
The object etc. searched in the picture in image is applied widely.Such as when graphic user interface has
When having multiple the most identical very much like given object, generally require and graphic user interface is comprised
All coupling target queries are out.
Correlation technique provides a kind of multi-targets recognition algorithm, and it is special that this algorithm forms SIFT with SUSAN angle point
Levy a little, use staircase chart to realize Scale invariant as pyramid structure, set up unified overdetermination for all match points
System of linear equations and equation system matrix number is carried out simplification make its dimension reduce half, obtain augmentation square
Battle array.Augmented matrix is carried out rank transformation, can therefrom extract multiobject stable according to the characteristic of Coordinate Conversion
Normal point, it is achieved that the multiobject match point of sharp separation.But said method robustness is not enough.Separately
Outward, the match point after separating is processed without reference to relevant scope acquisition.
Summary of the invention
The disclosure provides a kind of mates the method and device of multiple same object in image, it is possible to obtain with
The match point of tag along sort, classifies more accurate.
First aspect according to disclosure embodiment, it is provided that a kind of mate the side of multiple same object in image
Method, including: the feature point set of the given object of extraction, as fisrt feature point set, extracts image to be matched
Feature point set is as second feature point set；
The characteristic point that described fisrt feature point set is concentrated with described second feature point is mated, obtains institute
State the coupling point set on image to be matched；
Utilize clustering algorithm that described coupling point set is classified according to distribution density, form multiple match point
Subset.
In an embodiment, the method also includes: travel through the plurality of match point subset, respectively according to every
The size of individual match point subset and described given object obtain in described image to be matched with described given object
The scope of the destination object matched.
In an embodiment, travel through the plurality of match point subset, respectively according to each match point subset and
The size of described given object obtains the target pair in described image to be matched with described given match objects
The scope of elephant includes:
From match point subset Q, arbitrarily choose two match point A and B, calculate described match point A and B
Between distance as the first distance；
Obtain respectively in described given object with described match point C and D corresponding for match point A and B, meter
Calculate the distance between described match point C and D as second distance；
Calculate the ratio between described first distance and described second distance as the first matching ratio；
According in described first matching ratio, described match point A and described given object with described match point
The coordinate figure of the described match point C that A is corresponding and the size of described given object, obtain with described to
Determine the scope of the destination object of match objects.
In an embodiment, the scope of described destination object is determined by equation below:
RectA=(xa-xc*p,ya-yc*p,w*p,h*p)；
The wherein corresponding rectangular extent of RectA, in RectA, four elements are followed successively by: a left side for rectangular extent
Inferior horn abscissa, the lower left corner vertical coordinate of rectangular extent, the width of rectangular extent and rectangular extent
Highly；
xaFor described match point A coordinate in x-axis；
xcFor described match point C coordinate in x-axis；
yaFor described match point A coordinate on the y axis；
ycFor described match point C coordinate on the y axis；
P is described matching ratio；
W is the width pixel value of described given object；
H is the height pixel value of described given object.
In an embodiment, described given object includes icon and button.
In an embodiment, described clustering algorithm includes DBSCAN algorithm, OPTICS algorithm or DENCLUE
Algorithm.
In an embodiment, SIFT algorithm is used to extract the feature point set of given object as fisrt feature
Point set, and extract the feature point set of image to be matched as second feature point set.
In an embodiment, SIFT algorithm includes: utilizes Gaussian convolution to check described given image and carries out
Process, obtain multiscale space image；Described multiscale space image is carried out difference of Gaussian process, structure
Build Gaussian difference scale space image；Detect the Local Extremum of described Gaussian difference scale space image,
Utilize matching three-dimensional quadratic function that described Local Extremum is accurate to sub-pixel, use threshold method and
Hessian matrix method carries out screening to described Local Extremum and obtains feature point set.
Second aspect according to disclosure embodiment, it is provided that a kind of mate the dress of multiple same object in image
Put, including:
Feature point set extraction unit, for extracting the feature point set of given object as fisrt feature point set,
Extract the feature point set of image to be matched as second feature point set；
Coupling point set acquiring unit, for concentrate described fisrt feature point set and described second feature point
Characteristic point is mated, and obtains the coupling point set on described image to be matched；
Taxon, is used for utilizing clustering algorithm to classify described coupling point set according to distribution density,
Form multiple match point subset.
In an embodiment, this device also includes: scope acquiring unit, is used for traveling through the plurality of coupling
Point subset, obtains described figure to be matched according to the size of each match point subset and described given object respectively
With the scope of the destination object of described given match objects in Xiang.
In an embodiment, described scope acquiring unit is configured that and arbitrarily chooses from match point subset Q
Two match point A and B, calculate the distance between described match point A and B as the first distance；
Obtain respectively in described given object with described match point C and D corresponding for match point A and B, meter
Calculate the distance between described match point C and D as second distance；
Calculate the ratio between described first distance and described second distance as the first matching ratio；
According in described first matching ratio, described match point A and described given object with described match point
The coordinate figure of the described match point C that A is corresponding and the size of described given object, obtain with described to
Determine the scope of the destination object of match objects.
In an embodiment, described clustering algorithm includes DBSCAN algorithm, OPTICS algorithm or DENCLUE
Algorithm.
In an embodiment, described feature point set extraction unit is configured to perform SIFT algorithm.
In an embodiment, described feature point set extraction unit is configured that and utilizes Gaussian convolution verification described
Given image processes, and obtains multiscale space image；Described multiscale space image is carried out Gauss
Difference processing, builds Gaussian difference scale space image；Detect described Gaussian difference scale space image
Local Extremum, utilizes matching three-dimensional quadratic function that described Local Extremum is accurate to sub-pixel, adopts
By threshold method and Hessian matrix method, described Local Extremum is carried out screening and obtain feature point set.
According to the technical scheme of disclosure embodiment, can effectively reduce the interference of noise, classification can be made more
Add accurately.It addition, according to the technical scheme of other embodiments, it is provided that obtain mesh according to the match point separated
The method of mark object range, more meets the purpose of Object identifying.
It should be appreciated that it is only exemplary and explanatory that above general description and details hereinafter describe
, the disclosure can not be limited.
Accompanying drawing explanation
Accompanying drawing herein is merged in description and constitutes the part of this specification, it is shown that meet this
Bright embodiment, and for explaining the principle of the present invention together with description.
Fig. 1 is according to the stream of the method for multiple same object in the coupling image shown in an exemplary embodiment
Cheng Tu；
Fig. 2 is according to the schematic diagram of multiple same object in the coupling image shown in an exemplary embodiment；
Fig. 3 is according to the stream of the method for multiple same object in the coupling image shown in an exemplary embodiment
Cheng Tu；
Fig. 4 is the flow chart according to the method calculating destination object scope shown in an exemplary embodiment；
Fig. 5 is according to the frame of the device of multiple same object in the coupling image shown in an exemplary embodiment
Figure.
Detailed description of the invention
Here will illustrate exemplary embodiment in detail, its example represents in the accompanying drawings.Following
When description relates to accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represents same or analogous
Key element.Embodiment described in following exemplary embodiment does not represent the institute consistent with the present invention
There is embodiment.On the contrary, they only with as appended claims describes in detail, the one of the present invention
The example of the apparatus and method that a little aspects are consistent.
Fig. 1 is according to the stream of the method for multiple same object in the coupling image shown in an exemplary embodiment
Cheng Tu.The present embodiment is applicable to match multiple and given given object shapes from image to be matched
The situation of the object that content is consistent.As it is shown in figure 1, it is multiple identical in coupling image described in the present embodiment
The method of object includes:
In step s 110, extract the feature point set of given object as fisrt feature point set, extract and treat
The feature point set of figure picture is as second feature point set.
It should be noted that described given object can be the pictorial element arbitrarily with clear and definite profile, such as
Icon, button etc..Described image to be matched is the image containing multiple described given object, Ke Yiwei
Graphic user interfaces etc., such as, have image or the button of word or icon.
Those of ordinary skill in the art it should be explicitly made clear at this point, extracts the feature point set of given object as first
Feature point set, and the feature point set extracting image to be matched can use multiple side as second feature point set
Method performs.Such as, can based on SIFT algorithm, SURF algorithm, the marginal point algorithm of wavelet transformation, with
And Harris Corner Detection Algorithm etc., this is not restricted by the disclosure.
Illustrate as a example by using SIFT algorithm below.
The feature point set using SIFT algorithm to extract given image comprises the steps that and utilizes Gaussian convolution verification described
Given image processes, and obtains multiscale space image；Described multiscale space image is carried out Gauss
Difference processing, builds Gaussian difference scale space image；Detect described Gaussian difference scale space image
Local Extremum, utilizes matching three-dimensional quadratic function that described Local Extremum is accurate to sub-pixel, adopts
By threshold method and Hessian matrix method, described Local Extremum is carried out screening and obtain feature point set.
In the step s 120, the feature that described fisrt feature point set and described second feature point are concentrated is clicked on
Row coupling, obtains the coupling point set on described image to be matched.
The present embodiment can use any Feature Points Matching, and the disclosure is without limitation.Such as, can calculate
The Euclidean distance of the characteristic vector of two characteristic points is as two width images (such as given object and image to be matched)
The similarity determination tolerance of middle characteristic point.Such as, draw characteristic point A determining object, treated by traversal
Figure picture calculates the Euclidean distance that all validity feature points are corresponding, if distance is less than certain threshold value, then sentences
This point fixed is the match point of an A.
For SIFT algorithm, due to SIFT feature point translation, scaling invariance can solution never
With the compatibility issue of resolution display content identical image, for different resolution but display content is identical
Or very much like image to be matched and given object can obtain corresponding match point such that it is able to realize not
Compatibility with image in different resolution.
In step s 130, described coupling point set is carried out classification and can form multiple match point subset.Have many
Plant clustering algorithm to can be used for coupling point set is classified, such as partitioning, stratification, density algorithm etc..
Partition clustering method major part is based on distance, and needs given number of partitions K, for dividing
Number uncertain situation in district's then needs to be iterated analyzing, thus obtains optimal solution.For given K,
First algorithm provides an initial group technology, and the method by iterating changes packet later, makes
Obtain the packet scheme after improving each time the best.And the good standard of what is called is exactly: same point
Record in group is the nearest more good, and the record in different grouping is the most remote more good.The related algorithm bag of partitioning
Include K-MEANS algorithm, K-MEDOIDS algorithm, CLARANS algorithm etc..
Hierarchy clustering method (hierarchical methods) can be based on distance or based on density or
Connective.This method is decomposed as given data set carries out level, until certain condition meets
Till.Some extensions of hierarchy clustering method have also contemplated that subspace clustering.The defect of hierarchical method is,
Once a step (merge or divide) completes, and it cannot be revoked, and therefore it can not be righted the wrong
Decision.The related algorithm of stratification includes BIRCH algorithm, CURE algorithm, CHAMELEON algorithm etc..
The guiding theory of density algorithm is exactly, as long as the density of the point in a region is greater than certain threshold value,
Just it is added in the most close cluster.Density algorithm include DBSCAN algorithm, OPTICS algorithm,
DENCLUE algorithm etc..Method based on density with a fundamental difference of other method is: it is not base
In various distances, but based on density.Density algorithm can overcome algorithm based on distance
The shortcoming that can only find the cluster of " similar round ", has well eliminating for some away from the noise clustered
Effect, can find the cluster of arbitrary shape in noisy spatial database.Below with DBSCAN algorithm
As a example by categorizing process according to an embodiment of the invention is described.
DBSCAN(Density-Based Spatial Clustering of Applications with
Noise) it is a representational density-based algorithms of comparison.With division and hierarchy clustering method
Difference, it bunch will be defined as the maximum set of the point that density is connected, it is possible to having the most highdensity district
Territory is divided into bunch, and can find the cluster of arbitrary shape in the spatial database of noise.
According to an embodiment, utilize DBSCAN algorithm that match point carries out classification and obtain multiple match point subset
Process as follows.
Match point from image to be matched is concentrated to appoint and is taken a untreated point.
Centered by the point that this is chosen, scan in the range of radius E.Point set will be mated at radius E
In the range of the quantity of point compare with threshold value MinPts (MinPts > 1).If the point searched
Quantity more than this threshold value, then this point chosen is core point, searches in this core point and radius E
All formation one bunch.To bunch in other repeat the above steps in addition to this core point, carry out expanding bunch,
By that analogy, until complete expansion bunch a little.Then, choose next untreated point, repeat on
State step.Radius E and threshold value MinPts can be chosen according to concrete application, such as, can be empirical value.
If the quantity of the point searched is less than this threshold value, this point chosen is marginal point, then choose next
Individual untreated point, repeat the above steps.
After being processed a little, match point is classified according to their distribution density on image
Divide out, obtain multiple match point subset.
According to embodiments of the invention, feature extraction algorithm is combined with clustering algorithm, can be effectively improved algorithm
Robustness, reduces the interference that noise is brought, classifies more accurate.
Such as, Fig. 2 is according to the showing of multiple same object in the coupling image shown in an exemplary embodiment
Being intended to, the part 210 in the left side dashed box of Fig. 2 is image to be matched, and little in the dashed box of the upper right corner " is robbed
Take by force " button 220 is given object, after image 210 to be matched mates with " plunder " button 220, root
Three classes can be divided into according to its match point of distribution density, lay respectively at three " plunder " in described image to be matched
Button position.
According to an embodiment, disclosed method may also include step S140.In step 140, traversal institute
State multiple match point subset, obtain institute according to the size of each match point subset and described given object respectively
State the scope with the destination object of described given match objects in image to be matched.The method of this step can
Having various implementation, the disclosure is not restricted.
Such as, illustrate as a example by match point subset Q in the plurality of match point subset below.
From match point subset Q, arbitrarily choose two match point A and B, calculate described match point A and B
Between distance as the first distance.
Obtain respectively in described given object with described match point C and D corresponding for match point A and B, meter
Calculate the distance between described match point C and D as second distance.It can be readily appreciated that to given object (example
" plunder " button that right side as shown in, Fig. 2 is little) when carrying out arithmetic operation, choose this given object
Bottom left vertex is as zero.For image to be matched, zero is chosen and there is no spy
Do not limit.
Calculate the ratio between described first distance and described second distance as the first matching ratio.
According in described first matching ratio, described match point A and described given object with described match point
The coordinate figure of the described match point C that A is corresponding and the size of described given object, obtain with described to
Determine the scope of the destination object of match objects.
As such, it is possible to obtain in described image to be matched with the destination object of described given match objects
Scope.
Travel through the plurality of match point subset, perform to operate as above to each match point subset, can obtain respectively
Take the scope with the destination object of described given match objects in described image to be matched.
Fig. 3 is to mate the method for multiple same object in image according to a kind of shown in an exemplary embodiment
Flow chart.
As it is shown on figure 3, the method for multiple same object in image of mating described in the present embodiment includes:
In step S301, input image to be matched.
In step S303, extract SIFT feature point.
The present embodiment, as a example by using SIFT algorithm, can specifically use following operation: utilize Gaussian convolution core
Described given image is processed, obtains multiscale space image；Described multiscale space image is entered
Row difference of Gaussian processes, and builds Gaussian difference scale space image；Detect described Gaussian difference scale space
The Local Extremum of image, utilizes matching three-dimensional quadratic function that described Local Extremum is accurate to sub-pix
Level, uses threshold method and Hessian matrix method described Local Extremum to be carried out screening and obtains feature point set.
In step S305, the given object of input.
In step S307, extract SIFT feature point.This step uses SIFT algorithm to extract characteristic point can
Identical with step S303, therefore not to repeat here.
It should be noted that in practical operation, may wait for step S303 and be performed both by knot with step S307
Step S309 could be performed after bundle.That is, after step S303 performs to terminate, need to wait for step S307 and perform
After end, then carry out step S309.Equally, after step S307 performs to terminate, step S303 is needed to wait for
After execution terminates, then carry out step S309.
It addition, it should be noted that, step S301-step S303, and step S305-step S307,
Between two groups of steps, do not limit its sequencing, it is possible to executed in parallel.
In step S309, mate image to be matched and the characteristic point of given object.
In step S311, obtain the match point on image to be matched.
In step S313, match point is classified, form multiple match point subset.
In step S315, travel through the plurality of match point subset.
In step S317, the characteristic point to each match point subset, calculate matching ratio.
In step S319, calculate the scope of destination object corresponding with aforementioned match point.
The method of the scope calculating destination object can include multiple, and this is not restricted by the disclosure.
Such as, algorithm as shown in Figure 4 can be used to calculate.Fig. 4 is according to an exemplary embodiment
The flow chart calculating destination object scope illustrated.As shown in Figure 4, according to the calculating target of the present embodiment
The method of object range includes:
In step S410, from match point subset Q, arbitrarily choose two match point A and B, calculate institute
State the distance between match point A and B as the first distance；
In the step s 420, corresponding with described match point A and B is obtained in described given object respectively
Join C and D, calculate the distance between described match point C and D as second distance；
In step S430, calculate the ratio between described first distance and described second distance as the
One matching ratio.
Matching ratio calculates and can calculate in the following way, such as:
P=Euclidean (a (x, y), b (x, y))/Euclidean (c (x, y), d (x, y))
A (x, y) be in image to be matched from match point subset Q any one match point；
B (x, y) be image to be matched removes from match point subset Q a (x, y) beyond any one
Join a little；
C (x, y) in given object with a (x, Corresponding matching y) mated point；
D (x, y) in given object with b (x, Corresponding matching y) mated point；
((x, y), b (x, y)) is that (x, y) with b (x, Euclidean distance y) for a to a to Euclidean；
((x, y), d (x, y)) is that (x, y) with d (x, Euclidean distance y) for c to c to Euclidean；
P is matching ratio.
In step S440, according to described first matching ratio, described match point A and described given object
In with the coordinate figure of described match point C corresponding for match point A and the size of described given object,
With the scope of the destination object of described given match objects in described image to be matched.
Specifically, this step can be adopted with the following method:
RectA=(xa-xc*p,ya-yc*p,w*p,h*p)；
The wherein corresponding rectangular extent of RectA, in RectA, four elements are followed successively by: a left side for rectangular extent
Inferior horn abscissa, the lower left corner vertical coordinate of rectangular extent, the width of rectangular extent and rectangular extent
Highly；
xaFor described match point A coordinate in x-axis；
xcFor described match point C coordinate in x-axis；
yaFor described match point A coordinate on the y axis；
ycFor described match point C coordinate on the y axis；
P is described matching ratio；
W is the width pixel value of described given object；
H is the height pixel value of described given object.
So, can get in described image to be matched with described given object mutually according to this match point subset
The scope of the destination object joined.Match point subset similarly, for other also can perform in this way,
The most also can perform to operate as above, to obtain respectively in described image to be matched with described to each match point subset
The scope of the destination object of given match objects.
Technical scheme according to the disclosure can effectively remove the noise in match point.It addition, based on match point
Density classify to being in diverse location match point in image, can orient that position is different but display
The scope of the icon that content is identical.
Fig. 5 is according to the frame of the device of multiple same object in the coupling image shown in an exemplary embodiment
Figure.As it is shown in figure 5, the device of multiple same object in image that mates described in the present embodiment includes feature
Point set extraction unit 510, coupling point set acquiring unit 520, taxon 530.
Feature point set extraction unit 510 is for extracting the feature point set of given object as fisrt feature point
Collection, extracts the feature point set of image to be matched as second feature point set.
Coupling point set acquiring unit 520 is for concentrating described fisrt feature point set with described second feature point
Characteristic point mate, obtain the coupling point set on described image to be matched.
Taxon 530, for classifying described coupling point set, forms multiple match point subset.
An embodiment according to the disclosure, in coupling image, the device of multiple same object may also include scope
Acquiring unit 540.Scope acquiring unit 540 is used for traveling through the plurality of match point subset, basis respectively
It is given right with described that the size of each match point subset and described given object obtains in described image to be matched
The scope of the destination object as matching.
According to an embodiment, scope acquiring unit 540 is configured that and arbitrarily chooses from match point subset Q
Two match point A and B, calculate the distance between described match point A and B as the first distance；Respectively
Obtain in described given object with described match point C and D corresponding for match point A and B, calculate described
Join the distance between C and D as second distance；Calculate described first distance and described second distance it
Between ratio as the first matching ratio；According to described first matching ratio, described match point A and described
The coordinate figure of described match point C corresponding with described match point A in given object and described given right
The size of elephant, obtains the scope with the destination object of described given match objects in described image to be matched.
According to an embodiment, the scope of destination object can be determined by equation below:
RectA=(xa-xc*p,ya-yc*p,w*p,h*p)；
The wherein corresponding rectangular extent of RectA, in RectA, four elements are followed successively by: a left side for rectangular extent
Inferior horn abscissa, the lower left corner vertical coordinate of rectangular extent, the width of rectangular extent and rectangular extent
Highly；
xaFor described match point A coordinate in x-axis；
xcFor described match point C coordinate in x-axis；
yaFor described match point A coordinate on the y axis；
ycFor described match point C coordinate on the y axis；
P is described matching ratio；
W is the width pixel value of described given object；
H is the height pixel value of described given object.
As it was previously stated, described given object can be the pictorial element arbitrarily with clear and definite profile, such as icon,
Button etc..Described image to be matched is the image containing multiple described given object, can be that figure is used
Interface, family etc..
According to example embodiment, taxon 530 be configured to utilize clustering algorithm to described coupling point set by
Classify according to distribution density, form multiple match point subset, described clustering algorithm include DBSCAN algorithm,
OPTICS algorithm or DENCLUE algorithm.
According to example embodiment, described feature point set extraction unit 510 is configured to perform SIFT algorithm.
According to example embodiment, described feature point set extraction unit 510 is configured that and utilizes Gaussian convolution core
Described given image is processed, obtains multiscale space image；Described multiscale space image is entered
Row difference of Gaussian processes, and builds Gaussian difference scale space image；Detect described Gaussian difference scale space
The Local Extremum of image, utilizes matching three-dimensional quadratic function that described Local Extremum is accurate to sub-pix
Level, uses threshold method and Hessian matrix method described Local Extremum to be carried out screening and obtains feature point set.
About the device in above-described embodiment, the concrete mode that wherein unit performs to operate is having
Close in the embodiment of the method and be described in detail, explanation will be not set forth in detail herein.
In the coupling image that the present embodiment provides, the device of multiple same object can perform each method of the present invention in fact
Execute the method for multiple same object in the coupling image that example is provided, possess execution method corresponding function mould
Block and beneficial effect.
Those skilled in the art, after considering description and putting into practice invention disclosed herein, will readily occur to this
Other embodiment of invention.The application is intended to any modification, purposes or the adaptability of the present invention
Change, these modification, purposes or adaptations are followed the general principle of the present invention and include these public affairs
Open undocumented common knowledge in the art or conventional techniques means.Description and embodiments only by
Being considered as exemplary, true scope and spirit of the invention are pointed out by claim below.
It should be appreciated that the invention is not limited in described above and illustrated in the accompanying drawings accurately
Structure, and various modifications and changes can carried out without departing from the scope.The scope of the present invention is only by institute
Attached claim limits.
Claims (16)
1. one kind mates the method for multiple same object in image, it is characterised in that including:
The feature point set of the given object of extraction, as fisrt feature point set, extracts the characteristic point of image to be matched
Collection is as second feature point set；
The characteristic point that described fisrt feature point set is concentrated with described second feature point is mated, obtains institute
State the coupling point set on image to be matched；
Utilize clustering algorithm that described coupling point set is classified according to distribution density, form multiple match point
Subset.
Method the most according to claim 1, it is characterised in that the method also includes: traversal institute
State multiple match point subset, obtain institute according to the size of each match point subset and described given object respectively
State the scope with the destination object of described given match objects in image to be matched.
Method the most according to claim 2, it is characterised in that travel through the plurality of coupling idea
Collection, obtains in described image to be matched according to the size of each match point subset and described given object respectively
Scope with the destination object of described given match objects:
From match point subset Q, arbitrarily choose two match point A and B, calculate described match point A and B
Between distance as the first distance；
Obtain respectively in described given object with described match point C and D corresponding for match point A and B, meter
Calculate the distance between described match point C and D as second distance；
Calculate the ratio between described first distance and described second distance as the first matching ratio；
According in described first matching ratio, described match point A and described given object with described match point
The coordinate figure of the described match point C that A is corresponding and the size of described given object, obtain with described to
Determine the scope of the destination object of match objects.
Method the most according to claim 3, it is characterised in that the scope of described destination object is led to
Cross equation below to determine:
RectA=(xa-xc*p,ya-yc*p,w*p,h*p)；
The wherein corresponding rectangular extent of RectA, in RectA, four elements are followed successively by: a left side for rectangular extent
Inferior horn abscissa, the lower left corner vertical coordinate of rectangular extent, the width of rectangular extent and rectangular extent
Highly；
xaFor described match point A coordinate in x-axis；
xcFor described match point C coordinate in x-axis；
yaFor described match point A coordinate on the y axis；
ycFor described match point C coordinate on the y axis；
P is described matching ratio；
W is the width pixel value of described given object；
H is the height pixel value of described given object.
Method the most according to claim 1, it is characterised in that described given object includes icon
And/or button.
Method the most according to claim 1, it is characterised in that described clustering algorithm includes
DBSCAN algorithm, OPTICS algorithm or DENCLUE algorithm.
Method the most according to claim 1, it is characterised in that use SIFT algorithm to extract given
The feature point set of object is as fisrt feature point set and extracts the feature point set of image to be matched as second
Feature point set.
Method the most according to claim 7, it is characterised in that described SIFT algorithm includes:
Utilize Gaussian convolution to check described given image to process, obtain multiscale space image；
Described multiscale space image is carried out difference of Gaussian process, builds Gaussian difference scale space figure
Picture；
Detect the Local Extremum of described Gaussian difference scale space image, utilize matching three-dimensional quadratic function
Described Local Extremum is accurate to sub-pixel, uses threshold method and Hessian matrix method to described office
Portion's extreme point carries out screening and obtains feature point set.
9. one kind mates the device of multiple same object in image, it is characterised in that including:
Feature point set extraction unit, for extracting the feature point set of given object as fisrt feature point set,
Extract the feature point set of image to be matched as second feature point set；
Coupling point set acquiring unit, for concentrate described fisrt feature point set and described second feature point
Characteristic point is mated, and obtains the coupling point set on described image to be matched；
Taxon, is used for utilizing clustering algorithm to classify described coupling point set according to distribution density,
Form multiple match point subset.
Device the most according to claim 9, it is characterised in that this device also includes: scope obtains
Take unit, be used for traveling through the plurality of match point subset, respectively according to each match point subset and described give
Determine the size of object and obtain in described image to be matched the model with the destination object of described given match objects
Enclose.
11. devices according to claim 10, it is characterised in that described scope acquiring unit configures
For:
From match point subset Q, arbitrarily choose two match point A and B, calculate described match point A and B
Between distance as the first distance；
Obtain respectively in described given object with described match point C and D corresponding for match point A and B, meter
Calculate the distance between described match point C and D as second distance；
Calculate the ratio between described first distance and described second distance as the first matching ratio；
According in described first matching ratio, described match point A and described given object with described match point
The coordinate figure of the described match point C that A is corresponding and the size of described given object, obtain with described to
Determine the scope of the destination object of match objects.
12. devices according to claim 11, it is characterised in that it is characterized in that, described target
The scope of object is determined by equation below: RectA=(xa-xc*p,ya-yc*p,w*p,h*p)；
The wherein corresponding rectangular extent of RectA, in RectA, four elements are followed successively by: a left side for rectangular extent
Inferior horn abscissa, the lower left corner vertical coordinate of rectangular extent, the width of rectangular extent and rectangular extent
Highly；
xaFor described match point A coordinate in x-axis；
xcFor described match point C coordinate in x-axis；
yaFor described match point A coordinate on the y axis；
ycFor described match point C coordinate on the y axis；
P is described matching ratio；
W is the width pixel value of described given object；
H is the height pixel value of described given object.
13. devices according to claim 9, it is characterised in that described given object includes icon
And/or button.
14. devices according to claim 9, it is characterised in that described clustering algorithm includes DBSCAN
Algorithm, OPTICS algorithm or DENCLUE algorithm.
15. devices according to claim 9, it is characterised in that described feature point set extraction unit
It is configured to perform SIFT algorithm.
16. devices according to claim 15, it is characterised in that described feature point set extraction unit
It is configured that
Utilize Gaussian convolution to check described given image to process, obtain multiscale space image；
Described multiscale space image is carried out difference of Gaussian process, builds Gaussian difference scale space image；
Detect the Local Extremum of described Gaussian difference scale space image, utilize matching three-dimensional quadratic function
Described Local Extremum is accurate to sub-pixel, uses threshold method and Hessian matrix method to described office
Portion's extreme point carries out screening and obtains feature point set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610153710.7A CN105844290B (en) | 2016-03-16 | 2016-03-16 | The method and device of multiple same objects in matching image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610153710.7A CN105844290B (en) | 2016-03-16 | 2016-03-16 | The method and device of multiple same objects in matching image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105844290A true CN105844290A (en) | 2016-08-10 |
| CN105844290B CN105844290B (en) | 2019-10-25 |
Family
ID=56587272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610153710.7A Active CN105844290B (en) | 2016-03-16 | 2016-03-16 | The method and device of multiple same objects in matching image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105844290B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778781A (en) * | 2016-12-14 | 2017-05-31 | 维沃移动通信有限公司 | A kind of image processing method and device |
| WO2018121414A1 (en) * | 2016-12-30 | 2018-07-05 | 中兴通讯股份有限公司 | Electronic device, and target image recognition method and apparatus |
| CN109543536A (en) * | 2018-10-23 | 2019-03-29 | 北京市商汤科技开发有限公司 | Image identification method and device, electronic equipment and storage medium |
| CN113743424A (en) * | 2021-09-15 | 2021-12-03 | 昆明理工大学 | DBSCAN cluster analysis method for extracting characteristic points of full tailings dynamic flocculation sedimentation image |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100074528A1 (en) * | 2008-09-23 | 2010-03-25 | Microsoft Corporation | Coherent phrase model for efficient image near-duplicate retrieval |
| CN102855473A (en) * | 2012-08-21 | 2013-01-02 | 中国科学院信息工程研究所 | Image multi-target detecting method based on similarity measurement |
| CN102982528A (en) * | 2008-09-25 | 2013-03-20 | 佳能株式会社 | Image processing apparatus and image processing method |
| CN104766084A (en) * | 2015-04-10 | 2015-07-08 | 南京大学 | Nearly copied image detection method based on multi-target matching |
-
2016
- 2016-03-16 CN CN201610153710.7A patent/CN105844290B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100074528A1 (en) * | 2008-09-23 | 2010-03-25 | Microsoft Corporation | Coherent phrase model for efficient image near-duplicate retrieval |
| CN102982528A (en) * | 2008-09-25 | 2013-03-20 | 佳能株式会社 | Image processing apparatus and image processing method |
| CN102855473A (en) * | 2012-08-21 | 2013-01-02 | 中国科学院信息工程研究所 | Image multi-target detecting method based on similarity measurement |
| CN104766084A (en) * | 2015-04-10 | 2015-07-08 | 南京大学 | Nearly copied image detection method based on multi-target matching |
Non-Patent Citations (2)
| Title |
|---|
| 杨小影等: "一种基于特征点集的图像匹配算法", 《计算机与网络》 * |
| 甘玲等: "《一种基于点匹配的图像重复区域检测方法》", 《微电子学与计算机》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778781A (en) * | 2016-12-14 | 2017-05-31 | 维沃移动通信有限公司 | A kind of image processing method and device |
| CN106778781B (en) * | 2016-12-14 | 2018-08-17 | 维沃移动通信有限公司 | A kind of image processing method and device |
| WO2018121414A1 (en) * | 2016-12-30 | 2018-07-05 | 中兴通讯股份有限公司 | Electronic device, and target image recognition method and apparatus |
| CN108268881A (en) * | 2016-12-30 | 2018-07-10 | 中兴通讯股份有限公司 | Electronic equipment, target image identification method and device |
| CN109543536A (en) * | 2018-10-23 | 2019-03-29 | 北京市商汤科技开发有限公司 | Image identification method and device, electronic equipment and storage medium |
| CN113743424A (en) * | 2021-09-15 | 2021-12-03 | 昆明理工大学 | DBSCAN cluster analysis method for extracting characteristic points of full tailings dynamic flocculation sedimentation image |
| CN113743424B (en) * | 2021-09-15 | 2023-06-27 | 昆明理工大学 | Quantitative characterization method for dynamic flocculation sedimentation dispersion degree of flocs in dynamic flocculation sedimentation of full tailings |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105844290B (en) | 2019-10-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gilani et al. | Learning from millions of 3D scans for large-scale 3D face recognition | |
| CN109753885B (en) | Target detection method and device and pedestrian detection method and system | |
| CN105224942B (en) | RGB-D image classification method and system | |
| CN105894047B (en) | A kind of face classification system based on three-dimensional data | |
| Feng et al. | Robust and efficient algorithms for separating latent overlapped fingerprints | |
| CN105956560A (en) | Vehicle model identification method based on pooling multi-scale depth convolution characteristics | |
| CN102074015A (en) | Two-dimensional image sequence based three-dimensional reconstruction method of target | |
| Sudderth et al. | Depth from familiar objects: A hierarchical model for 3D scenes | |
| Li et al. | JointsGait: A model-based gait recognition method based on gait graph convolutional networks and joints relationship pyramid mapping | |
| CN105844290A (en) | Method of matching same objects in image and apparatus thereof | |
| Alidoost et al. | Knowledge based 3D building model recognition using convolutional neural networks from LiDAR and aerial imageries | |
| CN106845375A (en) | A kind of action identification method based on hierarchical feature learning | |
| CN110147841A (en) | The fine grit classification method for being detected and being divided based on Weakly supervised and unsupervised component | |
| Liu et al. | 3D Point cloud analysis | |
| Ng et al. | Matching of interest point groups with pairwise spatial constraints | |
| Bartolo et al. | Scribbles to vectors: preparation of scribble drawings for CAD interpretation | |
| Wang et al. | Hierarchical image segmentation ensemble for objectness in RGB-D images | |
| Trottier et al. | Sparse dictionary learning for identifying grasp locations | |
| Tang et al. | Exploiting distinctive topological constraint of local feature matching for logo image recognition | |
| JP2018136946A (en) | Three-dimensional cad model part search method and three-dimensional cad model search method | |
| Díaz et al. | Automatic clump splitting for cell quantification in microscopical images | |
| CN109598262A (en) | A kind of children's facial expression recognizing method | |
| JP5986681B2 (en) | Method and apparatus for generating model shape descriptors | |
| Liu et al. | An improved local descriptor based object recognition in cluttered 3D point clouds | |
| Yoon et al. | Recognizing 3D objects with 3D information from stereo vision |
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 |