CN108763262A - A kind of brand logo search method - Google Patents
A kind of brand logo search method Download PDFInfo
- Publication number
- CN108763262A CN108763262A CN201810292281.0A CN201810292281A CN108763262A CN 108763262 A CN108763262 A CN 108763262A CN 201810292281 A CN201810292281 A CN 201810292281A CN 108763262 A CN108763262 A CN 108763262A
- Authority
- CN
- China
- Prior art keywords
- picture
- brand logo
- checked
- search method
- sequence
- 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
- 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
- G06V10/754—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 involving a deformation of the sample pattern or of the reference pattern; Elastic matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of brand logo search methods, the feature extraction and matching mode combined first according to various ways obtains accurate first ranking results, secondary or even multiple retrieval is carried out to first ranking results again, the set generated further according to secondary or even multiple retrieval finally sorts, this method has merged the ranking results of associated picture repeatedly retrieved, the more image of occurrence number has higher weight, the forward image that sorts has higher weight, the correlation between image is fully excavated, substantially increase the accuracy of brand logo similarity judgement；Set the brand logo database of the step S101 to the registered quotient of whole including at least the commodity identical and/or similar as the affiliated type of merchandize of trade mark picture to be checked, can not only meet the needs of retrieval trade mark similarity, database size and calculation amount can also effectively be reduced, it is cost-effective, improve efficiency.
Description
Technical field
The invention belongs to trade mark retrieval technique fields, and in particular to a kind of brand logo search method.
Background technology
Trade mark is factor indispensable in commercial economy, and the annual applications of trade mark are up to million ranks, and branding data is up to thousand
Ten thousand ranks, for so huge amount group, if differentiating or examining whether two trade marks are similar, similarity degree reaches more in people
When few, go to judge by human eye and subjective consciousness entirely, this has no matter from the objective stability of operation cycle or result
The larger space improved.
Nowadays how people's research is had become using computer and other electronic equipments come the retrieval and matching for carrying out trade mark
The hot issue of this field, people continuously attempt to various computerized algorithms and multimedia technology to realize automatically retrieval trademark image
Shape has done a large amount of experiment and improvement in the link of feature extraction and comparison, but in the output ring of pre-processing and anaphase
Section, has improved space, pre-processing, which must be got well, tends to effectively reduce calculation amount, cost-effective raising robustness, then
The output element of phase result often means that greatly improving for accuracy.
Invention content
The object of the present invention is to provide a kind of methods that can be improved brand logo similarity and judge accuracy.
Technical solution provided by the present invention is：A kind of brand logo search method, includes the following steps：
S101 inputs picture and merchandise classification information to be checked, establishes or transfer corresponding brand logo database；
S102 pre-processes picture to be checked, forms trade mark picture to be checked；
S103 carries out feature extraction to the trade mark in trademark image piece to be checked and brand logo database；
The feature extracted in the feature extracted in S103 and brand logo database is carried out characteristic matching by S104 respectively,
Similarity matching is obtained to set；
S105, by matching in S104 to screening, and the sequence by the result after screening according to similarity from high to low
It is ranked up, obtains first order result sequence of pictures { S0}=[I01,I02,…,I0k,…]；
S106, take the preceding k picture of first order result sequence of pictures in S105 respectively as new trade mark picture to be checked also into
Sequencing of similarity, wherein k are the integer more than or equal to 1 to row at least once, obtain second fruiting sequence of pictures { Sm}=[Im1,
Im2,…,Imk...], m=1 ..., k；
S107 obtains comprehensive knot by the second fruiting sequence of pictures in the first order result sequence of pictures and S106 in S105
Fruit picture set S={ S0∪S1∪…∪Sm∪ ... }={ st}；
S108 is ranked up each unit and is most terminated using identical picture in synthesis result picture set S as unit
Fruit sequence of pictures.
As an improvement of the present invention, in S108, the weight matrix w of relevance ranking is defined0=[wij], wherein wij=
Max (2- (i+j)/L, 1), L >=2k, L are weighting parameter, the weights superposition of unit in set SWhereinS according toIt is ranked up to obtain final result sequence of pictures from big to small.
As an improvement of the present invention, it refers to judging that the picture to be checked is to carry out pretreatment to picture to be checked in S102
The no picture specification requirement for meeting trade mark registration application simultaneously passes through picture set-up procedure to the undesirable picture to be checked
S201 is adjusted.
As an improvement of the present invention, the picture set-up procedure S201 includes at least conversion picture storage format, draws
One kind in stretching, compress, amplify, reduce, cut and rotating.
As an improvement of the present invention, the brand logo database of the S101 is included at least believes with the merchandise classification
Manner of breathing with and/or similar commodity whole registered trade marks.
As an improvement of the present invention, the brand logo database of the S101 includes and the merchandise classification information phase
With whole registered trade marks with similar commodity.
As an improvement of the present invention, using based on RANSAC algorithm debugs matching pair in the step S105
Mode is screened.
As an improvement of the present invention, the mode of the feature extraction in the step S103 is based on HOG (Histogram
Of Oriented Gradient, histograms of oriented gradients), LBP (Local Binary Pattern, local binary patterns),
Haar (Haar-like features), SIFT (Scale-invariant feature transform, scale invariant feature
Conversion), SURF (Speeded-Up Robust Features), ORB (Oriented FAST and Rotated BRIEF),
Hash (hash algorithm), CNN (Convolutional Neural Network, convolutional neural networks), LSD (Local
Statistical Distribution, partial statistics distribution characteristics) and GSD (GLobal Statistical
At least one of Distribution, global statistics distribution characteristics).
As an improvement of the present invention, the mode of the feature extraction in the step S103 is based on HOG (Histogram
Of Oriented Gradient, histograms of oriented gradients).
As an improvement of the present invention, the mode of the feature extraction in the step S103 is based on LSD (Local
Statistical Distribution, partial statistics distribution characteristics), Hash (hash algorithm) and GSD (GLobal
Statistical Distribution, global statistics distribution characteristics) combination.
Advantageous effect：
The method of the present invention novelty, practical function, the feature extraction and matching mode combined first according to various ways obtain
Accurate first ranking results, then secondary or even multiple retrieval is carried out to first ranking results, further according to secondary or even multiple inspection
The set that rope generates finally sorts, and this method has merged the ranking results of associated picture repeatedly retrieved, and occurrence number is more
Image have higher weight, the forward image that sorts have higher weight, fully excavated the correlation between image,
Substantially increase the accuracy of brand logo similarity judgement；
The brand logo database of the step S101 is set as including at least and the affiliated commodity kind of trade mark picture to be checked
The registered quotient of whole of the identical and/or similar commodity of class can not only meet the needs of retrieval trade mark similarity, additionally it is possible to
Database size and calculation amount are effectively reduced, it is cost-effective, improve efficiency；
Picture to be checked is pre-processed in S102, first determines whether the picture to be checked meets trade mark registration application
Picture specification requirement, to the undesirable picture to be checked by convert picture storage format, stretching, compression, amplification,
The modes such as diminution, cutting and rotation adjust accordingly；Pretreated picture is with brand logo database in picture specification
Distinctiveness be greatly reduced, be convenient for feature extraction and comparison, calculation amount is further decreased, to the prodigious side of subsequent step band
Just；
It is screened by the way of based on RANSAC algorithm debugs matching pair in the step S105, helps to drop
Low erroneous matching improves accuracy to the redundant computation amount brought to improve efficiency.
Description of the drawings
Fig. 1 is the flow diagram of the present invention；
Fig. 2 is the specification table of the multiple dimensioned sliding window in the embodiment of the present invention 3；
Fig. 3 is that the multiple dimensioned similar window weight in the embodiment of the present invention 3 is superimposed schematic diagram；
Fig. 4 is the initial threshold matrix T in the embodiment of the present invention 30Definition figure；
Fig. 5 is the region Similarity measures schematic diagram in the embodiment of the present invention 3；
Specific implementation mode
The embodiment further illustrated the present invention below in conjunction with the accompanying drawings.
Embodiment 1
As shown in Fig. 1, a kind of brand logo search method in the present embodiment, including：
S101 inputs picture and merchandise classification information to be checked, establishes or transfer corresponding brand logo database；
S102 pre-processes picture to be checked, forms trade mark picture to be checked；
S103 carries out feature extraction to the trade mark in trademark image piece to be checked and brand logo database；
The feature extracted in the feature extracted in S103 and brand logo database is carried out characteristic matching by S104 respectively,
Similarity matching is obtained to set；
S105, by matching in S104 to screening, and the sequence by the result after screening according to similarity from high to low
It is ranked up, obtains first order result sequence of pictures { S0}=[I01,I02,…,I0k,…]；
S106, take the preceding k picture of first order result sequence of pictures in S105 respectively as new trade mark picture to be checked also into
Sequencing of similarity, wherein k are the integer more than or equal to 1 to row at least once, obtain second fruiting sequence of pictures { Sm}=[Im1,
Im2,…,Imk...], m=1 ..., k；
S107 obtains comprehensive knot by the second fruiting sequence of pictures in the first order result sequence of pictures and S106 in S105
Fruit picture set S={ S0∪S1∪…∪Sm∪ ... }={ st}；
S108 is ranked up each unit and is most terminated using identical picture in synthesis result picture set S as unit
Fruit sequence of pictures.
In S108, the weight matrix w of relevance ranking is defined0=[wij], whereinL
For weighting parameter, the weights superposition of unit in set SWhereinS according toIt is arranged from big to small
Sequence obtains final result sequence of pictures.
In the present embodiment, the mode of the similarity of the feature extraction in the step S103 is examined based on existing angle point
It surveys, (GLobal Statistical Distribution, global statistics are distributed by LSD (partial statistics distribution characteristics) and GSD
Feature)；It is realized using existing corners Matching mode corresponding with the feature extraction in the step S104；Corners Matching
Carry out candidate similar area segmentation step later, then after being split based on HOG (Histogram of Oriented
Gradient, histograms of oriented gradients) and Hash (hash algorithm) feature extraction and corresponding characteristic matching, to obtain
Accurate first order result sequence of pictures.
Embodiment 2
The present embodiment 2 is on the basis of embodiment 1, using based on RANSAC algorithm debugs in the step S105
The mode of pairing is screened, and helps to reduce erroneous matching to the redundant computation amount brought, to improve efficiency, it is accurate to improve
Property.
Embodiment 3
The present embodiment 3 makes change on the basis of embodiment 2, for first order result sequence of pictures, specifically takes following
Mode obtains：
First, the database D={ D for including N width trademark images is established in S1011, D2..., DN, input trade mark to be retrieved
Q。
Then, Multi resolution feature extraction is carried out in S103,
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 attached drawing 2, (in the present embodiment, σ1=0.8, σ2=0.6, σ3=0.4), (μ takes sliding step parameter μ in the present embodiment
0.1), sliding window horizontal direction step-length stepx=w μ, vertical direction step-length stepy=h μ.
2. according to the size of multiple dimensioned sliding window defined above, by each sliding window with image Iw×hThe upper left corner is starting point,
According to sliding step stepx、stepyIt from left to right slides from top to bottom successively, obtains a series of local window images (total t
It is a) set R={ Ri, i=0,1 ..., t.
3. for each local window image R obtained in 2iExtract area image feature fi：
A, 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.
B quantifies gradient direction, obtains gradient orientation histogram：
One gradient direction is quantized in its two adjacent bin, i.e., by a direction projection to two neighboring side
To representation in components, if certain pixel (x, y) gradient direction be θ (x, y), adjacent two Bin are respectively θk、θk+1, then
The gradient direction point is quantized to θkComponent beIt is quantized to θk+1Component beThe gradient that will be obtained in a
Direction is quantified according to above-mentioned Fuzzy Quantifying, and the blur gradients direction of statistics all pixels point obtains gradient direction histogram
Figure.
Finally, RiGradient orientation histogram be
C calculates normalized gradient direction histogram：
RiGradient orientation histogramNormalization histogram is
D, histogram feature coding：
R is obtained by step ciNormalization histogramWherein
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.The code word of each bin is cascaded to obtain as a segment length after coding
For 4 × 8=32 binary stringsThat is fi。
Subsequently, similar area detection segmentation is carried out in S203：
E, characteristic window matches between global scale
To retrieve imageWith arbitrary image in databaseFor：To retrieving imageIn it is 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 about 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.,AndTogether
ReasonAnd
(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.
F, candidate similar area segmentation：
1, it is matched using the RANSAC algorithm debugs 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 It can
Solve L.
(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, similar area is gone out according to adaptivity Threshold segmentation.
Screening through the above steps excludes some erroneous matchings substantially, and correct match window is carried out quantitative weighting
Superposition, statistics cover the number of the similar window of each anchor point (central point of grid).More similar region overlay
The number of the similar window of region anchor point points is more.
The weight of each pair of window is determined that similarity distance is smaller by similarity distance, and weight is bigger, and similarity distance is bigger, power
Weight is smaller, and weight size is 1 or so.The meaning of weighting processing is：The higher match window of similarity degree has relatively high
Correct probability is overlapped with the weight more than 1, and the lower match window of similarity degree has relatively low correct probability, with
Weight less than 1 is overlapped.
After obtaining similar window stack result as shown in Fig. 3 (in schematic diagram color more deeply feel show superposition value more
It is small), according to adaptive threshold Factorization algorithm similar area, the initial threshold matrix of adaptivity threshold value and each position and on
State screening obtain similar window size and number it is related.
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
(3) willWithIt is reduced into CA by sampling10×10And CB10×10.
(4) as shown in Fig. 4, initial threshold matrix T is defined0,
T0Setting it is related with the specification of specific sliding window, be located at set I_best={ Ai：BjAll belong toInstitute
It is s to have the gross area of windowA, 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 matrixIn matrix not
The candidate similar area in image is indicated for 0 part.
Finally, in S105, first order result sequence is carried out with the similarity measurement of local similar area：
For CA obtained above10×10And CB10×10The similar area of middle expression is partitioned into the similar area ROI of A figuresAAnd
The similar area ROI of B figuresB, similar Window match in region is carried out according to the method in attached drawing 5, lookup method is local neighborhood
It searches.Steps are as follows：
To ROIAIn arbitrary sliding window Ai, image ROI in ergodic data libraryBIn all meet similar possible condition
Window Bj, j=k1,k2..., the similarity distance being calculated isFind out most like windowThis pair of similar window, i.e. d are marked if similarity distance is within the scope of similar threshold valuemin-i
< Tsim, TsimFor empirical value, value is about 0.4~0.6. in this example
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=0.2
(deviation range, it is proposed that value range 0.1~0.3), i.e.,AndSimilarlyAndHere AiAnd Bi-jPosition be the relative position in the regions roi.
(2) A is setiLength-width ratioBjLength-width ratioThen haveAndI.e. similar window is necessary
There is similar length-width ratio.
ROI is obtained by aforesaid operationsAAnd ROIBMatching set { the A of similar windowi：Bj}。
The similitude of sliding window is replaced with the similitude of sliding window central point in ROI region, and as shown in Fig. 5, pA (u, v) is
A central point for including window in the A of region, then the similitude of the point by all windows centered on the point correspondence phase
It is calculated like the mean value of property:
Then the similarity distance of two ROI regions is in AB：
Wherein, nA、nBRespectively ROIA、ROIBIn include the number of window center point, λ is similar area parameters and nA、nB
It is inversely proportional, the similar area gross area is bigger, and λ is smaller.
To image D={ D in retrieval image Q and database1, D2..., DNIn arbitrary image Di(i=1,2 ..., N)
Similarity distance d is calculatedi, it is ranked up return ranking results from small to large according to similarity distance.
Embodiment 4
The present embodiment 4 pre-processes picture to be checked in S102 on the basis of embodiment 3, first determines whether described
Whether picture to be checked meets the picture specification requirement of trade mark registration application, which is specially：Papery picture is little
In 10cm*10cm and it is not less than 5cm*5cm；Electronic picture format is JPG and less than 200KB, graphical pixel between " 400*400 "
To between " 1500*1500 "；To the undesirable picture to be checked by converting picture storage format, stretching, compressing, put
Greatly, the modes such as diminution, cutting and rotation adjust accordingly；Pretreated picture is advised with brand logo database in picture
Distinctiveness on lattice is greatly reduced, and is convenient for feature extraction and comparison, is further decreased calculation amount, very big to subsequent step band
Convenience.
Embodiment 5
The present embodiment 5 uses and the commodity class brand logo database of the S101 on the basis of embodiment 4
Whole registered trade marks of the other identical and similar commodity of information can not only meet the needs of retrieval trade mark similarity, also
Database size and calculation amount can be effectively reduced, it is cost-effective, improve efficiency；
The invention is not limited in above-mentioned specific implementation mode, those skilled in the art can also make a variety of variations accordingly,
But it is any all to cover within the scope of the claims with equivalent or similar variation of the invention.
Claims (10)
1. a kind of brand logo search method, which is characterized in that include the following steps：
S101 inputs picture and merchandise classification information to be checked, establishes or transfer corresponding brand logo database；
S102 pre-processes picture to be checked, forms trade mark picture to be checked；
S103 carries out feature extraction to the trade mark in trademark image piece to be checked and brand logo database；
The feature extracted in the feature extracted in S103 and brand logo database is carried out characteristic matching, obtained by S104 respectively
Similarity matching is to set；
S105 by matching in S104 to screening, and the result after screening is carried out according to the sequence of similarity from high to low
Sequence, obtains first order result sequence of pictures { S0}=[I01,I02,…,I0k,…]；
S106, take the preceding k picture of first order result sequence of pictures in S105 respectively as new trade mark picture to be checked also carry out to
A few sequencing of similarity, wherein k are the integer more than or equal to 1, obtain second fruiting sequence of pictures { Sm}=[Im1,Im2,…,
Imk...], m=1 ..., k；
S107 obtains synthesis result figure by the second fruiting sequence of pictures in the first order result sequence of pictures and S106 in S105
Piece set S={ S0∪S1∪…∪Sm∪ ... }={ st}；
S108 is ranked up each unit to obtain final result figure using identical picture in synthesis result picture set S as unit
Piece sequence.
2. a kind of brand logo search method as described in claim 1, which is characterized in that in S108, define relevance ranking
Weight matrix w0=[wij], wherein wij=max (2- (i+j)/L, 1), L >=2k, L are weighting parameter, the power of unit in set S
Value superpositionWhereinS according toIt is ranked up to obtain final result sequence of pictures from big to small.
3. a kind of brand logo search method as claimed in claim 1 or 2, which is characterized in that in S102 to picture to be checked into
Row pretreatment refers to judging whether the picture to be checked meets the picture specification requirement of trade mark registration application and to undesirable
The picture to be checked be adjusted by picture set-up procedure S201.
4. a kind of brand logo search method as claimed in claim 3, which is characterized in that the picture set-up procedure S201 is extremely
Few includes the one kind converted in picture storage format, stretching, compression, amplification, diminution, cutting and rotation.
5. a kind of brand logo search method as claimed in claim 4, which is characterized in that the brand logo data of the S101
Library includes at least whole registered trade marks of the commodity identical and/or similar as the merchandise classification information.
6. a kind of brand logo search method as claimed in claim 5, which is characterized in that the brand logo data of the S101
Library includes whole registered trade marks of the commodity identical and similar as the merchandise classification information.
7. a kind of brand logo search method as claimed in claim 6, which is characterized in that use and be based in the step S105
The mode of RANSAC algorithm debugs matching pair is screened.
8. a kind of brand logo search method as claimed in claim 7, which is characterized in that the feature in the step S103 carries
The mode taken is based at least one of HOG, LBP, Haar, SIFT, SURF, ORB, Hash, CNN, LSD and GSD.
9. a kind of brand logo search method as claimed in claim 8, which is characterized in that the feature in the step S103 carries
The mode taken is based on HOG.
10. a kind of brand logo search method as claimed in claim 8, which is characterized in that the feature in the step S103
Combination of the mode of extraction based on LSD, Hash and GSD.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810292281.0A CN108763262B (en) | 2018-04-03 | 2018-04-03 | Trademark graph retrieval method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810292281.0A CN108763262B (en) | 2018-04-03 | 2018-04-03 | Trademark graph retrieval method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108763262A true CN108763262A (en) | 2018-11-06 |
| CN108763262B CN108763262B (en) | 2022-04-29 |
Family
ID=63981013
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810292281.0A Active CN108763262B (en) | 2018-04-03 | 2018-04-03 | Trademark graph retrieval method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108763262B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109978067A (en) * | 2019-04-02 | 2019-07-05 | 北京市天元网络技术股份有限公司 | A kind of trade-mark searching method and device based on convolutional neural networks and Scale invariant features transform |
| CN110516412A (en) * | 2019-07-22 | 2019-11-29 | 平安科技（深圳）有限公司 | The original information acquisition method of picture, device, computer equipment and storage medium |
| CN112307239A (en) * | 2020-10-29 | 2021-02-02 | 泰康保险集团股份有限公司 | Image retrieval method, device, medium and equipment |
| CN112598093A (en) * | 2020-12-18 | 2021-04-02 | 湖南特能博世科技有限公司 | Legend complexity ordering method, legend matching method, device and computer equipment |
| CN115311477A (en) * | 2022-08-09 | 2022-11-08 | 北京惠朗时代科技有限公司 | Simulated trademark accurate detection method and system based on super-resolution reconstruction |
| CN116542818A (en) * | 2023-07-06 | 2023-08-04 | 图林科技(深圳)有限公司 | Trademark monitoring and analyzing method based on big data technology |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020138465A1 (en) * | 2001-03-21 | 2002-09-26 | Lee Eugene M. | Apparatus for and method of searching and organizing intellectual property information utilizing a classification system |
| US20030158743A1 (en) * | 2002-02-19 | 2003-08-21 | Havlick Scott Spenser | Method and apparatus for organizing, accessing and displaying data relating to trademark rights |
| CN104809142A (en) * | 2014-01-29 | 2015-07-29 | 北京瑞天科技有限公司 | Trademark inquiring system and method |
| CN105512175A (en) * | 2015-11-23 | 2016-04-20 | 东莞市凡豆信息科技有限公司 | Quick image retrieval method based on color features and texture characteristics |
| CN107391594A (en) * | 2017-06-29 | 2017-11-24 | 安徽睿极智能科技有限公司 | A kind of image search method based on the sequence of iteration vision |
| CN107506429A (en) * | 2017-08-22 | 2017-12-22 | 北京联合大学 | A kind of image rearrangement sequence method integrated based on marking area and similitude |
-
2018
- 2018-04-03 CN CN201810292281.0A patent/CN108763262B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020138465A1 (en) * | 2001-03-21 | 2002-09-26 | Lee Eugene M. | Apparatus for and method of searching and organizing intellectual property information utilizing a classification system |
| US20030158743A1 (en) * | 2002-02-19 | 2003-08-21 | Havlick Scott Spenser | Method and apparatus for organizing, accessing and displaying data relating to trademark rights |
| CN104809142A (en) * | 2014-01-29 | 2015-07-29 | 北京瑞天科技有限公司 | Trademark inquiring system and method |
| CN105512175A (en) * | 2015-11-23 | 2016-04-20 | 东莞市凡豆信息科技有限公司 | Quick image retrieval method based on color features and texture characteristics |
| CN107391594A (en) * | 2017-06-29 | 2017-11-24 | 安徽睿极智能科技有限公司 | A kind of image search method based on the sequence of iteration vision |
| CN107506429A (en) * | 2017-08-22 | 2017-12-22 | 北京联合大学 | A kind of image rearrangement sequence method integrated based on marking area and similitude |
Non-Patent Citations (2)
| Title |
|---|
| G.Y. TIAN: "Colour image retrieval using virtual reality", 《2000 IEEE CONFERENCE ON INFORMATION VISUALIZATION. AN INTERNATIONAL CONFERENCE ON COMPUTER VISUALIZATION AND GRAPHICS》 * |
| 孙会会: "基于反馈的现场鞋底痕迹花纹检索算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109978067A (en) * | 2019-04-02 | 2019-07-05 | 北京市天元网络技术股份有限公司 | A kind of trade-mark searching method and device based on convolutional neural networks and Scale invariant features transform |
| CN110516412A (en) * | 2019-07-22 | 2019-11-29 | 平安科技（深圳）有限公司 | The original information acquisition method of picture, device, computer equipment and storage medium |
| CN112307239A (en) * | 2020-10-29 | 2021-02-02 | 泰康保险集团股份有限公司 | Image retrieval method, device, medium and equipment |
| CN112307239B (en) * | 2020-10-29 | 2024-02-02 | 泰康保险集团股份有限公司 | Image retrieval method, device, medium and equipment |
| CN112598093A (en) * | 2020-12-18 | 2021-04-02 | 湖南特能博世科技有限公司 | Legend complexity ordering method, legend matching method, device and computer equipment |
| CN112598093B (en) * | 2020-12-18 | 2024-05-28 | 湖南特能博世科技有限公司 | Legend complexity ordering method, legend matching method, device and computer equipment |
| CN115311477A (en) * | 2022-08-09 | 2022-11-08 | 北京惠朗时代科技有限公司 | Simulated trademark accurate detection method and system based on super-resolution reconstruction |
| CN115311477B (en) * | 2022-08-09 | 2024-01-16 | 北京惠朗时代科技有限公司 | Super-resolution reconstruction-based simulated trademark accurate detection method and system |
| CN116542818A (en) * | 2023-07-06 | 2023-08-04 | 图林科技(深圳)有限公司 | Trademark monitoring and analyzing method based on big data technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108763262B (en) | 2022-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108763262A (en) | A kind of brand logo search method | |
| Parkhi et al. | Deep face recognition | |
| Kato et al. | Image reconstruction from bag-of-visual-words | |
| CN108596154B (en) | Remote sensing image classification method based on high-dimensional feature selection and multilevel fusion | |
| CN106126581A (en) | Cartographical sketching image search method based on degree of depth study | |
| Ahmad et al. | Saliency-weighted graphs for efficient visual content description and their applications in real-time image retrieval systems | |
| CN109101981B (en) | Loop detection method based on global image stripe code in streetscape scene | |
| CN104850822B (en) | Leaf identification method under simple background based on multi-feature fusion | |
| Jiang et al. | Joint kernel-based supervised hashing for scalable histopathological image analysis | |
| Tolias et al. | Towards large-scale geometry indexing by feature selection | |
| CN108319959A (en) | A kind of corps diseases image-recognizing method compressed based on characteristics of image with retrieval | |
| CN108764245A (en) | A method of it improving brand logo similarity and judges accuracy | |
| CN108897747A (en) | A kind of brand logo similarity comparison method | |
| CN108694411B (en) | Method for identifying similar images | |
| Bakić et al. | Inria IMEDIA2's participation at ImageCLEF 2012 plant identification task | |
| Chen et al. | Class-specific feature selection with local geometric structure and discriminative information based on sparse similar samples | |
| Singh et al. | Ensemble visual content based search and retrieval for natural scene images | |
| Sun et al. | Scene classification using multiple features in a two-stage probabilistic classification framework | |
| Parseh et al. | Semantic-aware visual scene representation | |
| Shanmugam et al. | An efficient perceptual of CBIR system using MIL-SVM classification and surf feature extraction | |
| Sinha et al. | Scene image classification using a wigner-based local binary patterns descriptor | |
| Lee et al. | Content-based image retrieval using spatial-color and Gabor texture on a mobile device | |
| Mahboubi et al. | Object classification based on graph kernels | |
| Liu et al. | Efficient logo recognition by local feature groups | |
| Nayef et al. | Efficient symbol retrieval by building a symbol index from a collection of line drawings |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |