CN108763263A - A kind of trade-mark searching method - Google Patents
A kind of trade-mark searching method Download PDFInfo
- Publication number
- CN108763263A CN108763263A CN201810292297.1A CN201810292297A CN108763263A CN 108763263 A CN108763263 A CN 108763263A CN 201810292297 A CN201810292297 A CN 201810292297A CN 108763263 A CN108763263 A CN 108763263A
- Authority
- CN
- China
- Prior art keywords
- window
- trade
- image
- mark
- trade mark
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V10/507—Summing image-intensity values; Histogram projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of trade-mark searching method, including:Obtain all sample trade marks;Multi resolution feature extraction is carried out to the image of sample trade mark;According to the feature of extraction, image feature base is established for the image section of sample trade mark;Obtain trade mark to be measured;According to the step identical as sample trade mark, Multi resolution feature extraction is carried out to trade mark to be measured;According to the characteristics of image of trade mark to be measured, similar area detection segmentation is carried out in image feature base respectively, obtains similarity retrieval result.The present invention is more detailed to the extraction of graphic feature, and fast and effective and discrimination higher overcomes the larger problem of recognition result error of the existing technology;Rate matched is improved, acquisition similarity retrieval result is more excellent, reduces the workload of auditor, improves work efficiency, and improves the accuracy of query result.
Description
Technical field
The present invention relates to a kind of trade-mark searching methods, belong to technical field of information retrieval.
Background technology
Trade mark is the mark of company, product or service, and it is one to melt with the commercial quality of enterprise, service quality, management
Body plays very important effect in industry and commerce society, is an important attribute of company and products thereof, has uniqueness.
To make trade mark obtain legal protection, it is necessary to trademark office's official register.With China's expanding economy and globalization process
Accelerate, trade mark quantity cumulative year after year.It is the key problem of trade-mark administration to prevent repeated registration or similar brand registration.In order to protect
The legitimate rights and interests of registered trademark hit the counterfeit illegal activities for usurping registered trademark, need to retrieve trade mark to be registered,
It is compared with registered trade mark, determines that the two is differed or is not similar, just there is registration and qualification.
Due to the sustainable growth of trade mark registration amount and examination amount, trademark database also increases substantially therewith, and traditional people
Work search method needs a large amount of manpower and materials in judgement, this leads to the increasing for examining difficulty, examines the lengthening of time, and
Examine that accuracy rate and quality can all decreased significantly, therefore traditional manual method has been difficult to cope with increasing trade mark Shen
Please with infringement case.
Invention content
The object of the present invention is to provide a kind of trade-mark searching method, the image characteristics extraction the step of in, use is improved
Gradient orientation histogram feature extraction, the specification of multiple dimensioned sliding window and sliding step reasonable set;In similar area detection point
In the step of cutting, global Similarity matching is carried out first, is improved the strategy design of rate matched, is divided into inquiry and similarity distance measurement
Two parts of method;Candidate similar area segmentation is carried out again, setting adaptivity threshold value so that similar area is divided as far as possible
Accurately.
A kind of trade-mark searching method, includes the following steps:
One, all sample trade marks are obtained;
Two, Multi resolution feature extraction is carried out to the image of sample trade mark;
Three, according to the feature of extraction, image feature base is established for the image section of sample trade mark;
Four, trade mark to be measured is obtained;
Five, according to the step identical as sample trade mark, Multi resolution feature extraction is carried out to trade mark to be measured;
Six, according to the characteristics of image of trade mark to be measured, the detection point of phase similar area is carried out in image feature base respectively
It cuts, obtains similarity retrieval result.
The Multi resolution feature extraction, including:Gradient orientation histogram feature extraction based on Fuzzy Quantifying;It is more
The specification of scale sliding window and the reasonable set of sliding step.
The Multi resolution feature extraction, includes the following steps:
(a) specification and sliding step of self-defined multiple dimensioned sliding window;
(b) each sliding window is by the size of the multiple dimensioned sliding window defined according to step (a) with the image upper left corner
Point from left to right slides according to sliding step, obtains a series of local window images from top to bottom successively;
(c) each local window image zooming-out area image feature for being obtained in step (b).
The gradient orientation histogram feature extraction based on Fuzzy Quantifying, includes the following steps:
1. for any image window, the gradient of calculated level and vertical direction;
2. quantifying gradient direction, gradient orientation histogram is obtained;
3. calculating normalized gradient direction histogram.
The calculating normalized gradient direction histogram, one of specially following three kinds of methods:
Method one:Method for normalizing based on target pixel points sum.
Method two:Method for normalizing based on region area parameter.
Method three:The method for normalizing combined based on both target pixel points sum and region area parameter.
The similar area detection segmentation, includes the following steps:
(1) characteristic window matches between global scale;
(2) candidate similar area segmentation.
The candidate similar area segmentation, includes the following steps:
1. using the RANSAC algorithm debugs matching based on scale-space consistency model;
2. going out similar area according to adaptivity Threshold segmentation.
The scale-space consistency model is specially:
If a pair of of match window { (x1,y1),(x1′,y1′)}:{(x2,y2),(x2′,y2') (wherein (x1,y1)、 (x1′,
y1') window A is indicated respectivelyiThe upper left corner and bottom right angular coordinate, (x2,y2)、(x2′,y2') indicate window BjThe upper left corner and bottom right
Angular coordinate), then there are space transform modelsSo thatIt can solve
L。
The adaptivity threshold value obtains the big of similar window to the initial threshold matrix of each position and above-mentioned screening
It is small related with number.
The definition mode of the adaptivity threshold value is as follows:
If T0For initial threshold matrix, size 10*10, general broad in the middle, surrounding is small, and the specification of specific sliding window has
It closes, if the gross area of all similar windows is s, then adaptive threshold matrix is T=κ T0(s/100)α, κ=0.2, α here
=0.7 is empirical value, and the adjustment of adaptability should be carried out with the running parameter of sliding window specification.
The present invention is more detailed to the extraction of graphic feature by providing a kind of trade-mark searching method, quickly and effectively and knows
Not rate higher overcomes the larger problem of recognition result error of the existing technology;Rate matched is improved, similarity is obtained
Retrieval result is more excellent, reduces the workload of auditor, improves work efficiency, and improves the accuracy of query result.
Description of the drawings
Fig. 1:The gradient direction schematic diagram of embodiment 1-3 quantizations.
Fig. 2:Multiple dimensioned similar window weight is superimposed schematic diagram in embodiment 1,4,5.
Specific embodiment:
Embodiment 1:
A kind of trade-mark searching method, includes the following steps:
One, all sample trade marks are obtained;
Two, Multi resolution feature extraction is carried out to the image of sample trade mark;
1. the specification and sliding step of self-defined multiple dimensioned sliding window:If input picture Iw×h, a variety of scales of sliding window are fixed
Justice such as table 1.1 (in experiment, σ1=0.8, σ2=0.6, σ3=0.4), sliding step parameter μ (μ takes 0.1 or 0.2 in experiment), it is sliding
Window horizontal direction step-length stepx=w μ, vertical direction step-length stepy=h μ.
1.1 multi-scale sliding window mouth size tables
2. according to the size of multiple dimensioned sliding window defined above, by each sliding window with image Iw×hThe upper left corner is
Point, according to sliding step stepx、stepyIt from left to right slides from top to bottom successively, it is (total to obtain a series of local window images
T) set R={ Ri, i=0,1 ..., t.
3. for each local window image R obtained in 2iExtract area image feature fi。
(1) feature one:Gradient orientation histogram feature based on Fuzzy Quantifying.
1. for any image window Ri, the gradient of calculated level and vertical direction.
Computational methods:[Gh,Gv]=gradient (Ri), using direction template [- 1,0,1], calculate RiMiddle any pixel point
The horizontal gradient G of (x, y)h(x, y) and vertical gradient Gv(x,y)。
Then orientation angle θ=arctan (G of (x, y) pointv/Gh), 0~360 degree of value.
2. quantifying gradient direction, gradient orientation histogram is obtained.By the gradient direction obtained in 1. according to Figure of description 1
8 directions of signal are quantified, and the gradient direction of statistics all pixels point obtains gradient orientation histogram.
It is proposed that a kind of Fuzzy Quantifying, a gradient direction is quantized in its two adjacent bin, i.e., by one
A direction is projected to the representation in components in two neighboring direction, adjacent if the gradient direction of certain pixel (x, y) is θ (x, y)
Two Bin be respectively θk、θk+1, then the gradient direction point be quantized to θkComponent beIt is quantized to θk+1Component be1. the gradient direction obtained in is quantified according to above-mentioned Fuzzy Quantifying, statistics all pixels point
Blur gradients direction obtains gradient orientation histogram.
Finally, RiGradient orientation histogram be
3. calculating normalized gradient direction histogram.
Method one:Method for normalizing based on target pixel points sum.
RiGradient orientation histogramNormalization histogram isThis is straight
Square figure method for normalizing makes feature have good consistency of scale, while embodying each gradient direction relative statistic distribution letter
Breath.
4. histogram feature encodes.R is 3. obtained by stepiNormalization histogram
Wherein 0 < huj< 1, j=0,1 ..., 7.Money is calculated in order to save computer
Source encodes above-mentioned floating data.
Histogram normalization after according to each section gradient point non-uniform probability distribution principle calculate quantized interval (0,
0.098), (0.098,0.134), (0.134,0.18), (0.18,0.24), (0.24,1), the calculating of the quantized interval is by current
Sample set carries out statistics experiment with computing and obtains.The data for falling in this 5 sections are encoded respectively:0000,0001,
0011,0111,1111.
It is 4 × 8=to cascade to obtain as a segment length by the code word of each bin after coding
32 binary stringsThat is fi。
Three, according to the feature of extraction, image feature base is established for the image section of sample trade mark;
Four, trade mark to be measured is obtained;
Five, according to the step identical as sample trade mark, Multi resolution feature extraction is carried out to trade mark to be measured;
Six, according to the characteristics of image of trade mark to be measured, the detection point of phase similar area is carried out in image feature base respectively
It cuts, obtains similarity retrieval result.
1, characteristic window matches between global scale
To retrieve imageWith arbitrary image in databaseFor:To retrieving imageIn
Arbitrary sliding window Ai, image in ergodic data libraryIn all window B for meeting similar possible conditionj, j=k1,
k2..., the similarity distance being calculated isFind out most like windowSuch as
Fruit similarity distance then marks this pair of similar window, i.e. d within the scope of similar threshold valuemin-i< Tsim, TsimFor empirical value, in this example
Middle value is 0.4-0.6.
Here similarity distance calculates as follows:If sliding window AiBinary feature string of the feature vector after coding is fi, sliding window Bj
Binary feature string of the feature vector after coding is gj, then AiAnd Bi-jBetween similarity distance dijBy Hamming distance into
Row calculates:Wherein fi kIndicate binary string fiKth position,Indicate binary string gjKth position,
Indicate that xor operation, the value of α are equal to fiAnd gjThe inverse of length.
Here similar possible condition is as follows:
(1) window BjCenter in AiIn a certain range of near center location, permission transformation range is u=0.5
(deviation range, window center position are calculated according to the ratio of graphic aspect, and offset is also calculated according to the ratio of length and width, here, are permitted
Perhaps deviation range is long or wide half, it is proposed that value range 0.4~0.6), i.e.,And
SimilarlyAnd
(2) A is setiLength-width ratioBjLength-width ratioThen haveAndI.e. similar window must
There must be similar length-width ratio.
Matching set { the A of A windows similar with B is obtained by aforesaid operationsi:Bj, due to being the lookup mould between global scale
Formula, wherein there may be the matchings pair for not meeting Space Consistency.Correctly matching will be screened from all these results below
As a result.
2, candidate similar area segmentation
2.1 using the RANSAC algorithm debugs matching based on scale-space consistency model.
By searching for matching between the scale in global scope, some correct match windows can be found, also contain one
The matching of a little mistakes, one is scale matching error, another kind is location matches mistake, using the side of scale-space consistency
Method eliminates erroneous matching.
Using the matching pair of improved RANSAC (random sampling consistency) algorithm debug, it is retained on scale and empty
Between all consistent matching pair on position, steps are as follows:
(1) is to matched data set { Ai:BjIn any pair of match window, calculate transformation matrix L, be denoted as model
M, model are defined as follows:
Transformation model:If a pair of of match window { (x1,y1),(x1′,y1′)}:{(x2,y2),(x2′,y2') (wherein (x1,
y1)、(x1′,y1') window A is indicated respectivelyiThe upper left corner and bottom right angular coordinate, (x2,y2)、(x2′,y2') indicate window BjUpper left
Angle and bottom right angular coordinate), then there are space transform modelsSo that
L can be solved;
Wherein, ɑ1、ɑ2For the relevant zooming parameter of specific matching window, tx、tyIt is and the relevant translation of specific matching window
Parameter.
Wherein, a in transformation matrix L1And a2The length scale ratio and width pantograph ratio of two comparison windows, t are indicated respectivelyx
And tyThe transverse translation distance and longitudinal translation distance of two window centers are indicated respectively.
(2) calculates the projection error of all data and model M in data set, if error is less than threshold value, interior point set I is added;
(3) if element number is more than optimal interior point set I_best in point set I in is current, then I_best=I is updated;
(4) all data in ergodic datas set, repeat the above steps.
(5) sample in the optimal interior point set I_best of is correct matched sample, finally obtains correct matched sample collection
Close I_best={ Ai:Bj}。
2.2 go out similar area according to adaptivity Threshold segmentation.
ForMatrix is defined respectively
(1) for I_best={ Ai:BjAny pair of match window { (x1,y1),(x1′,y1′)}:{(x2,y2),
(x2′,y2') (wherein (x1,y1)、(x1′,y1') window A is indicated respectivelyiThe upper left corner and bottom right angular coordinate, (x2,y2)、(x2′,
y2') indicate window BjThe upper left corner and bottom right angular coordinate), similarity distance dij, define weighting coefficient ωij=min (2,
2.67-3.33dij), then have
(2) traversal I_best={ Ai:BjIn all matched samples repeat (1), updateWith
Multiple dimensioned similar window weight is superimposed schematic diagram as shown in Figure of description 2.(the deeper mark superposition value of color is more
It is small)
(3) willWithIt is reduced into CA by sampling10×10And CB10×10.
(4) initial threshold matrix is defined
T0Setting it is related with the specification of specific sliding window.It is located at set I_best={ Ai:BjAll belong to's
The gross area of all windows is sA, then adaptive threshold matrix is TA=κ T0(sA/(100w1h1))α, in set I_best=
{Ai:BjAll belong toAll windows the gross area be sB, then adaptive threshold matrix is TB=κ T0(sB/
(100w2h2))α, κ=0.2, α=0.7 are empirical value here, and adaptability should be carried out with the running parameter of sliding window specification
Adjustment.
Then there is similar area subdivision matrix
It is not 0 part expression in matrix
Candidate similar area in image.
Embodiment 2:
A kind of trade-mark searching method, includes the following steps:
One, all sample trade marks are obtained;
Two, Multi resolution feature extraction is carried out to the image of sample trade mark;
1. the specification and sliding step of self-defined multiple dimensioned sliding window:If input picture Iw×h, a variety of scales of sliding window are fixed
Justice such as table 1.1 (in experiment, σ1=0.8, σ2=0.6, σ3=0.4), sliding step parameter μ (μ takes 0.1 or 0.2 in experiment), it is sliding
Window horizontal direction step-length stepx=w μ, vertical direction step-length stepy=h μ.
1.1 multi-scale sliding window mouth size tables
2. according to the size of multiple dimensioned sliding window defined above, by each sliding window with image Iw×hThe upper left corner is
Point, according to sliding step stepx、stepyIt from left to right slides from top to bottom successively, it is (total to obtain a series of local window images
T) set R={ Ri, i=0,1 ..., t.
3. for each local window image R obtained in 2iExtract area image feature fi。
(1) feature one:Gradient orientation histogram feature based on Fuzzy Quantifying.
1. for any image window Ri, the gradient of calculated level and vertical direction.
Computational methods:[Gh,Gv]=gradient (Ri), using direction template [- 1,0,1], calculate RiMiddle any pixel point
The horizontal gradient G of (x, y)h(x, y) and vertical gradient Gv(x,y)。
Then orientation angle θ=arctan (G of (x, y) pointv/Gh), 0~360 degree of value.
2. quantifying gradient direction, gradient orientation histogram is obtained.By the gradient direction obtained in 1. according to Figure of description 1
8 directions of signal are quantified, and the gradient direction of statistics all pixels point obtains gradient orientation histogram.
It is proposed that a kind of Fuzzy Quantifying, a gradient direction is quantized in its two adjacent bin, i.e., by one
A direction is projected to the representation in components in two neighboring direction, adjacent if the gradient direction of certain pixel (x, y) is θ (x, y)
Two Bin be respectively θk、θk+1, then the gradient direction point be quantized to θkComponent beIt is quantized to θk+1Component be1. the gradient direction obtained in is quantified according to above-mentioned Fuzzy Quantifying, statistics all pixels point
Blur gradients direction obtains gradient orientation histogram.
Finally, RiGradient orientation histogram be
3. calculating normalized gradient direction histogram.
Method two:Method for normalizing based on region area parameter.
RiSize be wi×hi, gradient orientation histogramArea parametersNormalization histogram based on area parameters is
Area parameters will make feature have relatively good consistency of scale by area evolution to calculate.Joined based on area
Several histogram method for normalizing, had not only contained the abundant degree of marginal information in characteristic window, but also can reflect each gradient side
To statistical distribution information, the variation of single bin does not interfere with the value of other bin.The disadvantage is that the otherness between each bin may
It reduces, for the window that edge is abundant, the value of each bin is relatively large, and there are multiple higher values;And it is diluter for edge
The value of thin window, each bin is smaller, and there are multiple smaller values.
4. histogram feature encodes.R is 3. obtained by stepiNormalization histogram
Wherein 0 < huj< 1, j=0,1 ..., 7.Money is calculated in order to save computer
Source encodes above-mentioned floating data.
Histogram normalization after according to each section gradient point non-uniform probability distribution principle calculate quantized interval (0,
0.098), (0.098,0.134), (0.134,0.18), (0.18,0.24), (0.24,1), the calculating of the quantized interval is by current
Sample set carries out statistics experiment with computing and obtains.The data for falling in this 5 sections are encoded respectively:0000,0001,
0011,0111,1111.
It is 4 × 8=to cascade to obtain as a segment length by the code word of each bin after coding
32 binary stringsThat is fi。
Three, according to the feature of extraction, image feature base is established for the image section of sample trade mark;
Four, trade mark to be measured is obtained;
Five, according to the step identical as sample trade mark, Multi resolution feature extraction is carried out to trade mark to be measured;
Six, according to the characteristics of image of trade mark to be measured, the detection point of phase similar area is carried out in image feature base respectively
It cuts, obtains similarity retrieval result.
Remaining is the same as embodiment 1.
Embodiment 3:
A kind of trade-mark searching method, includes the following steps:
One, all sample trade marks are obtained;
Two, Multi resolution feature extraction is carried out to the image of sample trade mark;
1. the specification and sliding step of self-defined multiple dimensioned sliding window:If input picture Iw×h, a variety of scales of sliding window are fixed
Justice such as table 1.1 (in experiment, σ1=0.8, σ2=0.6, σ3=0.4), sliding step parameter μ (μ takes 0.1 or 0.2 in experiment), it is sliding
Window horizontal direction step-length stepx=w μ, vertical direction step-length stepy=h μ.
1.1 multi-scale sliding window mouth size tables
2. according to the size of multiple dimensioned sliding window defined above, by each sliding window with image Iw×hThe upper left corner is
Point, according to sliding step stepx、stepyIt from left to right slides from top to bottom successively, it is (total to obtain a series of local window images
T) set R={ Ri, i=0,1 ..., t.
3. for each local window image R obtained in 2iExtract area image feature fi。
(1) feature one:Gradient orientation histogram feature based on Fuzzy Quantifying.
1. for any image window Ri, the gradient of calculated level and vertical direction.
Computational methods:[Gh,Gv]=gradient (Ri), using direction template [- 1,0,1], calculate RiMiddle any pixel point
The horizontal gradient G of (x, y)h(x, y) and vertical gradient Gv(x,y)。
Then orientation angle θ=arctan (G of (x, y) pointv/Gh), 0~360 degree of value.
2. quantifying gradient direction, gradient orientation histogram is obtained.By the gradient direction obtained in 1. according to Figure of description 1
8 directions of signal are quantified, and the gradient direction of statistics all pixels point obtains gradient orientation histogram.
It is proposed that a kind of Fuzzy Quantifying, a gradient direction is quantized in its two adjacent bin, i.e., by one
A direction is projected to the representation in components in two neighboring direction, adjacent if the gradient direction of certain pixel (x, y) is θ (x, y)
Two Bin be respectively θk、θk+1, then the gradient direction point be quantized to θkComponent beIt is quantized to θk+1Component be1. the gradient direction obtained in is quantified according to above-mentioned Fuzzy Quantifying, statistics all pixels point
Blur gradients direction obtains gradient orientation histogram.
Finally, RiGradient orientation histogram be
3. calculating normalized gradient direction histogram.
Method three:The method for normalizing combined based on both target pixel points sum and region area parameter.
Based on the above analysis, two kinds of method for normalizing are combined, have not only ensured the relative independentability between each bin, but also
Take into account the otherness of each bin statistical distributions.
RiSize be wi×hi, gradient orientation histogramBased on returning for object pixel sum
One, which changes histogram, isBased on area parametersNormalization histogram be
Then it is defined as in conjunction with the normalization histogram of the two:
0 < w1,w2< 1;w1+w2=1
Wherein α=0.125 is the mean value of 8 direction normalization histograms.
4. histogram feature encodes.R is 3. obtained by stepiNormalization histogram
Wherein 0 < huj< 1, j=0,1 ..., 7.In order to save computer computing resource, above-mentioned floating data is encoded.
Histogram normalization after according to each section gradient point non-uniform probability distribution principle calculate quantized interval (0,
0.098), (0.098,0.134), (0.134,0.18), (0.18,0.24), (0.24,1), the calculating of the quantized interval is by current
Sample set carries out statistics experiment with computing and obtains.The data for falling in this 5 sections are encoded respectively:0000,0001,
0011,0111,1111.
It is 4 × 8 to cascade to obtain as a segment length by the code word of each bin after coding
=32 binary stringsThat is fi。
Three, according to the feature of extraction, image feature base is established for the image section of sample trade mark;
Four, trade mark to be measured is obtained;
Five, according to the step identical as sample trade mark, Multi resolution feature extraction is carried out to trade mark to be measured;
Six, according to the characteristics of image of trade mark to be measured, the detection point of phase similar area is carried out in image feature base respectively
It cuts, obtains similarity retrieval result.
Remaining is the same as embodiment 1.
Embodiment 4:
A kind of trade-mark searching method, includes the following steps:
One, all sample trade marks are obtained;
Two, Multi resolution feature extraction is carried out to the image of sample trade mark;
Three, according to the feature of extraction, image feature base is established for the image section of sample trade mark;
Four, trade mark to be measured is obtained;
Five, according to the step identical as sample trade mark, Multi resolution feature extraction is carried out to trade mark to be measured;
Six, according to the characteristics of image of trade mark to be measured, the detection point of phase similar area is carried out in image feature base respectively
It cuts, obtains similarity retrieval result.
1, characteristic window matches between global scale
To retrieve imageWith arbitrary image in databaseFor:To retrieving imageIn
Arbitrary sliding window Ai, image in ergodic data libraryIn all window B for meeting similar possible conditionj, j=k1,
k2..., the similarity distance being calculated isFind out most like windowSuch as
Fruit similarity distance then marks this pair of similar window, i.e. d within the scope of similar threshold valuemin-i< Tsim, TsimFor empirical value, in this example
Middle value is 0.45-0.55.
Here similarity distance calculates as follows:If sliding window AiBinary feature string of the feature vector after coding is fi, sliding window Bj
Binary feature string of the feature vector after coding is gj, then AiAnd Bi-jBetween similarity distance dijBy Hamming distance into
Row calculates:Wherein fi kIndicate binary string fiKth position,Indicate binary string gjKth position,Indicate that xor operation, the value of α are equal to fiAnd gjThe inverse of length.
Here similar possible condition is as follows:
(1) window BjCenter in AiIn a certain range of near center location, permission transformation range is u=0.5
(deviation range, window center position are calculated according to the ratio of graphic aspect, and offset is also calculated according to the ratio of length and width, here, are permitted
Perhaps deviation range is long or wide half, it is proposed that value range 0.45-0.55), i.e.,And
SimilarlyAnd
(2) A is setiLength-width ratioBjLength-width ratioThen haveAndI.e. similar window must
There must be similar length-width ratio.
Matching set { the A of A windows similar with B is obtained by aforesaid operationsi:Bj, due to being the lookup mould between global scale
Formula, wherein there may be the matchings pair for not meeting Space Consistency.Correctly matching will be screened from all these results below
As a result.
2, candidate similar area segmentation
2.1 using the RANSAC algorithm debugs matching based on scale-space consistency model.
By searching for matching between the scale in global scope, some correct match windows can be found, also contain one
The matching of a little mistakes, one is scale matching error, another kind is location matches mistake, using the side of scale-space consistency
Method eliminates erroneous matching.
Using the matching pair of improved RANSAC (random sampling consistency) algorithm debug, it is retained on scale and empty
Between all consistent matching pair on position, steps are as follows:
(1) is to matched data set { Ai:BjIn any pair of match window, calculate transformation matrix L, be denoted as model
M, model are defined as follows:
Transformation model:If a pair of of match window { (x1,y1),(x1′,y1′)}:{(x2,y2),(x2′,y2') (wherein (x1,
y1)、(x1′,y1') window A is indicated respectivelyiThe upper left corner and bottom right angular coordinate, (x2,y2)、(x2′,y2') indicate window BjUpper left
Angle and bottom right angular coordinate), then there are space transform modelsSo that
L can be solved.
Wherein, ɑ1、ɑ2For the relevant zooming parameter of specific matching window, tx、tyIt is and the relevant translation of specific matching window
Parameter.
(2) calculates the projection error of all data and model M in data set, if error is less than threshold value, interior point set I is added;
(3) if element number is more than optimal interior point set I_best in point set I in is current, then I_best=I is updated;
(4) all data in ergodic datas set, repeat the above steps.
(5) sample in the optimal interior point set I_best of is correct matched sample, finally obtains correct matched sample collection
Close I_best={ Ai:Bj}。
2.2 go out similar area according to adaptivity Threshold segmentation.
ForMatrix is defined respectively
(1) for I_best={ Ai:BjAny pair of match window { (x1,y1),(x1′,y1′)}:{(x2,y2),
(x2′,y2') (wherein (x1,y1)、(x1′,y1') window A is indicated respectivelyiThe upper left corner and bottom right angular coordinate, (x2,y2)、(x2′,
y2') indicate window BjThe upper left corner and bottom right angular coordinate), similarity distance dij, define weighting coefficient ωij=min (2,
2.67- 3.33dij), then have
(2) traversal I_best={ Ai:BjIn all matched samples repeat (1), updateWith
Multiple dimensioned similar window weight is superimposed schematic diagram as shown in Figure of description 2.(the deeper mark superposition value of color is more
It is small)
(3) willWithIt is reduced into CA by sampling10×10And CB10×10.
(4) initial threshold matrix is defined
T0Setting it is related with the specification of specific sliding window.It is located at set I_best={ Ai:BjAll belong to's
The gross area of all windows is sA, then adaptive threshold matrix is TA=κ T0(sA/(100w1h1))α, in set I_best=
{Ai:BjAll belong toAll windows the gross area be sB, then adaptive threshold matrix is TB=κ T0(sB/
(100w2h2))α, κ=0.2, α=0.7 are empirical value here, and adaptability should be carried out with the running parameter of sliding window specification
Adjustment.
Then there is similar area subdivision matrix
It is not 0 part expression in matrix
Candidate similar area in image.
Remaining is the same as embodiment 1.
Embodiment 5:
A kind of trade-mark searching method, includes the following steps:
One, all sample trade marks are obtained;
Two, Multi resolution feature extraction is carried out to the image of sample trade mark;
Three, according to the feature of extraction, image feature base is established for the image section of sample trade mark;
Four, trade mark to be measured is obtained;
Five, according to the step identical as sample trade mark, Multi resolution feature extraction is carried out to trade mark to be measured;
Six, according to the characteristics of image of trade mark to be measured, the detection point of phase similar area is carried out in image feature base respectively
It cuts, obtains similarity retrieval result.
1, characteristic window matches between global scale
To retrieve imageWith arbitrary image in databaseFor:To retrieving imageIn
Arbitrary sliding window Ai, image in ergodic data libraryIn all window B for meeting similar possible conditionj, j=k1,
k2..., the similarity distance being calculated isFind out most like windowSuch as
Fruit similarity distance then marks this pair of similar window, i.e. d within the scope of similar threshold valuemin-i< Tsim, TsimFor empirical value, in this example
Middle value is 0.45-0.55.
Here similarity distance calculates as follows:If sliding window AiBinary feature string of the feature vector after coding is fi, sliding window Bj
Binary feature string of the feature vector after coding is gj, then AiAnd Bi-jBetween similarity distance dijBy Hamming distance into
Row calculates:Wherein fi kIndicate binary string fiKth position,Indicate binary string gjKth position,Indicate that xor operation, the value of α are equal to fiAnd gjThe inverse of length.
Here similar possible condition is as follows:
(1) window BjCenter in AiIn a certain range of near center location, permission transformation range is u=0.5
(deviation range, window center position are calculated according to the ratio of graphic aspect, and offset is also calculated according to the ratio of length and width, here, are permitted
Perhaps deviation range is long or wide half, it is proposed that value range 0.45-0.55), i.e.,And
SimilarlyAnd
(2) A is setiLength-width ratioBjLength-width ratioThen haveAndI.e. similar window is necessary
There is similar length-width ratio.
Matching set { the A of A windows similar with B is obtained by aforesaid operationsi:Bj, due to being the lookup mould between global scale
Formula, wherein there may be the matchings pair for not meeting Space Consistency.Correctly matching will be screened from all these results below
As a result.
2, candidate similar area segmentation
2.1 using the RANSAC algorithm debugs matching based on scale-space consistency model.
By searching for matching between the scale in global scope, some correct match windows can be found, also contain one
The matching of a little mistakes, one is scale matching error, another kind is location matches mistake, using the side of scale-space consistency
Method eliminates erroneous matching.
Using the matching pair of improved RANSAC (random sampling consistency) algorithm debug, it is retained on scale and empty
Between all consistent matching pair on position, steps are as follows:
(1) is to matched data set { Ai:BjIn any pair of match window, calculate transformation matrix L, be denoted as model
M, model are defined as follows:
Transformation model:If a pair of of match window { (x1,y1),(x1′,y1′)}:{(x2,y2),(x2′,y2') (wherein (x1,
y1)、(x1′,y1') window A is indicated respectivelyiThe upper left corner and bottom right angular coordinate, (x2,y2)、(x2′,y2') indicate window BjUpper left
Angle and bottom right angular coordinate), then there are space transform modelsSo that
L can be solved.
Wherein, ɑ1、ɑ2For the relevant zooming parameter of specific matching window, tx、tyIt is and the relevant translation of specific matching window
Parameter.
(2) calculates the projection error of all data and model M in data set, if error is less than threshold value, interior point set I is added;
(3) if element number is more than optimal interior point set I_best in point set I in is current, then I_best=I is updated;
(4) all data in ergodic datas set, repeat the above steps.
(5) sample in the optimal interior point set I_best of is correct matched sample, finally obtains correct matched sample collection
Close I_best={ Ai:Bj}。
2.2 go out similar area according to adaptivity Threshold segmentation.
ForMatrix is defined respectively
(1) for I_best={ Ai:BjAny pair of match window { (x1,y1),(x1′,y1′)}:{(x2,y2),
(x2′,y2') (wherein (x1,y1)、(x1′,y1') window A is indicated respectivelyiThe upper left corner and bottom right angular coordinate, (x2,y2)、(x2′,
y2') indicate window BjThe upper left corner and bottom right angular coordinate), similarity distance dij, define weighting coefficient ωij=min (2,
2.67- 3.33dij), then have
(2) traversal I_best={ Ai:BjIn all matched samples repeat (1), updateWith
Multiple dimensioned similar window weight is superimposed schematic diagram as shown in Figure of description 2.(the deeper mark superposition value of color is more
It is small)
(3) willWithIt is reduced into CA by sampling10×10And CB10×10.
(4) initial threshold matrix is defined
T0Setting it is related with the specification of specific sliding window.It is located at set I_best={ Ai:BjAll belong to's
The gross area of all windows is sA, then adaptive threshold matrix is TA=κ T0(sA/(100w1h1))α, in set I_best=
{Ai:BjAll belong toAll windows the gross area be sB, then adaptive threshold matrix is TB=κ T0(sB/
(100w2h2))α, κ=0.2, α=0.7 are empirical value here, and adaptability should be carried out with the running parameter of sliding window specification
Adjustment.
Then there is similar area subdivision matrix
It is not 0 part expression in matrix
Candidate similar area in image.Remaining is the same as embodiment 2.
Claims (10)
1. a kind of trade-mark searching method, it is characterized in that:Specially following steps:
One, all sample trade marks are obtained;
Two, Multi resolution feature extraction is carried out to the image of sample trade mark;
Three, according to the feature of extraction, image feature base is established for the image section of sample trade mark;
Four, trade mark to be measured is obtained;
Five, according to the step identical as sample trade mark, Multi resolution feature extraction is carried out to trade mark to be measured;
Six, according to the characteristics of image of trade mark to be measured, the detection segmentation of phase similar area is carried out in image feature base respectively, is obtained
Obtain similarity retrieval result.
2. a kind of trade-mark searching method as described in claim 1, it is characterized in that:The Multi resolution feature extraction, including:Base
In the gradient orientation histogram feature extraction of Fuzzy Quantifying;The specification of multiple dimensioned sliding window and the reasonable set of sliding step.
3. a kind of trade-mark searching method as described in claim 1, it is characterized in that:The Multi resolution feature extraction, including such as
Lower step:
(a) specification and sliding step of self-defined multiple dimensioned sliding window;
(b) size of the multiple dimensioned sliding window defined according to step (a) is pressed by each sliding window using the image upper left corner as starting point
It is from left to right slided from top to bottom successively according to sliding step, obtains a series of local window images;
(c) each local window image zooming-out area image feature for being obtained in step (b).
4. a kind of trade-mark searching method as claimed in claim 2, it is characterized in that:The gradient based on Fuzzy Quantifying
Direction histogram feature extraction, includes the following steps:
1. for any image window, the gradient of calculated level and vertical direction;
2. quantifying gradient direction, gradient orientation histogram is obtained;
3. calculating normalized gradient direction histogram.
5. a kind of trade-mark searching method as claimed in claim 4, it is characterized in that:The calculating normalized gradient direction Histogram
Figure, one of specially following three kinds of methods:
Method one:Method for normalizing based on target pixel points sum.
Method two:Method for normalizing based on region area parameter.
Method three:The method for normalizing combined based on both target pixel points sum and region area parameter.
6. a kind of trade-mark searching method as described in claim 1, it is characterized in that:The similar area detection segmentation, including
Following steps:
(1) characteristic window matches between global scale;
(2) candidate similar area segmentation.
7. a kind of trade-mark searching method as claimed in claim 6, it is characterized in that:The candidate similar area segmentation, including
Following steps:
1. using the RANSAC algorithm debugs matching based on scale-space consistency model;
2. going out similar area according to adaptivity Threshold segmentation.
8. a kind of trade-mark searching method as claimed in claim 7, it is characterized in that:The scale-space consistency model tool
Body is:
If a pair of of match window { (x1,y1),(x1′,y1′)}:{(x2,y2),(x2′,y2') (wherein (x1,y1)、(x1′,y1') point
It Biao Shi not window AiThe upper left corner and bottom right angular coordinate, (x2,y2)、(x2′,y2') indicate window BjThe upper left corner and the lower right corner are sat
Mark), then there are space transform modelsSo thatL can be solved;
Wherein, a in transformation matrix L1And a2The length scale ratio and width pantograph ratio of two comparison windows, t are indicated respectivelyxAnd tyPoint
The transverse translation distance and longitudinal translation distance of two window centers are not indicated.
9. a kind of trade-mark searching method as claimed in claim 7, it is characterized in that:The adaptivity threshold value and each position
The size and number that initial threshold matrix and screening obtain similar window are related.
10. a kind of trade-mark searching method as claimed in claim 9, it is characterized in that:The definition mode of the adaptivity threshold value
It is as follows:
If T0For initial threshold matrix, size 10*10, general broad in the middle, surrounding is small, related with the specification of specific sliding window, if institute
It is s to have the gross area of similar window, then adaptive threshold matrix is T=κ T0(s/100)α, κ=0.2 here, α=0.7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810292297.1A CN108763263A (en) | 2018-04-03 | 2018-04-03 | A kind of trade-mark searching method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810292297.1A CN108763263A (en) | 2018-04-03 | 2018-04-03 | A kind of trade-mark searching method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108763263A true CN108763263A (en) | 2018-11-06 |
Family
ID=63981143
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810292297.1A Pending CN108763263A (en) | 2018-04-03 | 2018-04-03 | A kind of trade-mark searching method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108763263A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113624231A (en) * | 2021-07-12 | 2021-11-09 | 北京自动化控制设备研究所 | Inertial vision integrated navigation positioning method based on heterogeneous image matching and aircraft |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104462380A (en) * | 2014-12-11 | 2015-03-25 | 北京中细软移动互联科技有限公司 | Trademark retrieval method |
CN105139023A (en) * | 2015-07-24 | 2015-12-09 | 福州大学 | Seed identification method based on multi-scale feature fusion and extreme learning machine |
CN105160654A (en) * | 2015-07-09 | 2015-12-16 | 浙江工商大学 | Towel label defect detecting method based on feature point extraction |
US20170299393A1 (en) * | 2016-04-19 | 2017-10-19 | Honda Research Institute Europe Gmbh | Navigation system and method for error correction |
CN107622270A (en) * | 2016-07-13 | 2018-01-23 | 中国电信股份有限公司 | Image similarity calculation method and device, method for retrieving similar images and system |
-
2018
- 2018-04-03 CN CN201810292297.1A patent/CN108763263A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104462380A (en) * | 2014-12-11 | 2015-03-25 | 北京中细软移动互联科技有限公司 | Trademark retrieval method |
CN105160654A (en) * | 2015-07-09 | 2015-12-16 | 浙江工商大学 | Towel label defect detecting method based on feature point extraction |
CN105139023A (en) * | 2015-07-24 | 2015-12-09 | 福州大学 | Seed identification method based on multi-scale feature fusion and extreme learning machine |
US20170299393A1 (en) * | 2016-04-19 | 2017-10-19 | Honda Research Institute Europe Gmbh | Navigation system and method for error correction |
CN107622270A (en) * | 2016-07-13 | 2018-01-23 | 中国电信股份有限公司 | Image similarity calculation method and device, method for retrieving similar images and system |
Non-Patent Citations (2)
Title |
---|
厉丹等: "基于改进相位相关与特征点配准的多图拼接算法", 《计算机科学》 * |
阳树洪: "灰度图像阈值分割的自适应和快速算法研究", 《中国优秀博硕士学位论文全文数据库(博士)信息科技辑》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113624231A (en) * | 2021-07-12 | 2021-11-09 | 北京自动化控制设备研究所 | Inertial vision integrated navigation positioning method based on heterogeneous image matching and aircraft |
CN113624231B (en) * | 2021-07-12 | 2023-09-12 | 北京自动化控制设备研究所 | Inertial vision integrated navigation positioning method based on heterogeneous image matching and aircraft |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108830279A (en) | A kind of image characteristics extraction and matching process | |
CN107515895A (en) | A kind of sensation target search method and system based on target detection | |
CN104850822B (en) | Leaf identification method under simple background based on multi-feature fusion | |
Chatfield et al. | Efficient retrieval of deformable shape classes using local self-similarities | |
KR20080088778A (en) | System and method for image identifier based on singular value decomposition and feature point | |
US8027978B2 (en) | Image search method, apparatus, and program | |
CN103473545A (en) | Text-image similarity-degree measurement method based on multiple features | |
CN112926592B (en) | Trademark retrieval method and device based on improved Fast algorithm | |
US20130243341A1 (en) | Frequency domain interest point descriptor | |
CN108763262A (en) | A kind of brand logo search method | |
CN108694411B (en) | Method for identifying similar images | |
CN108763266B (en) | Trademark retrieval method based on image feature extraction | |
CN108764245B (en) | Method for improving similarity judgment accuracy of trademark graphs | |
CN108763263A (en) | A kind of trade-mark searching method | |
CN108897747A (en) | A kind of brand logo similarity comparison method | |
CN101118544A (en) | Method for constructing picture shape contour outline descriptor | |
Nigam et al. | Content based trademark retrieval by integrating shape with colour and texture information | |
CN108763265B (en) | Image identification method based on block retrieval | |
CN108763261B (en) | Graph retrieval method | |
Rusinol et al. | Perceptual image retrieval by adding color information to the shape context descriptor | |
CN108804499B (en) | Trademark image retrieval method | |
CN108846843A (en) | A kind of image characteristic extracting method | |
Shinde et al. | Content based image retrieval using multi-sequential search | |
CN108897746B (en) | Image retrieval method | |
Manandhar et al. | Feature repetitiveness similarity metrics in visual search |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181106 |