CN105678778B - A kind of image matching method and device - Google Patents
A kind of image matching method and device Download PDFInfo
- Publication number
- CN105678778B CN105678778B CN201610022261.2A CN201610022261A CN105678778B CN 105678778 B CN105678778 B CN 105678778B CN 201610022261 A CN201610022261 A CN 201610022261A CN 105678778 B CN105678778 B CN 105678778B
- Authority
- CN
- China
- Prior art keywords
- image
- target
- template
- feature
- similarity
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000004364 calculation method Methods 0.000 claims abstract description 18
- 238000010586 diagram Methods 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 6
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 2
- 238000003384 imaging method Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of image matching method and devices.This method comprises: obtaining template image and target image；One group of template characteristic is obtained according to template image；One group of target signature is extracted according to target image；According to template characteristic and target signature, the image similarity of calculation template image and every frame target image, and be and the matched image of template image with the maximum target image of image similarity.In the image matching method and device of above-described embodiment, the image similarity between template image and each target image is calculated separately by the similarity between template characteristic and target signature to carry out images match, it can guarantee the not redundancy of feature during images match, the correct matching for guaranteeing image, improves the accuracy of images match.
Description
Technical field
The present invention relates to technical field of image matching, and in particular to a kind of image matching method and device.
Background technique
In computer vision research, images match is a very basic content.Since figure has extraordinary table
Existing power, and the important information in image can be saved, thus in recent years, figure matching is wide as a kind of method of images match
It is general to be applied to the fields such as social networks, data analysis, complex object identification and video analysis.
Non-rigid object can be coped with and the images match that deformation occurs has more extensive versatility.But due to
It is mathematically quadratic assignment problem, that is, NP difficult, therefore has many methods to attempt it.However, how to protect
The not redundancy on characteristic point and side and how correctly to be matched all be to need further to solve in card figure matching process.
Summary of the invention
The application provides a kind of correct for capable of guaranteeing the not redundancy of characteristic point during images match and image
The image matching method and device matched.
According in a first aspect, a kind of image matching method is provided in some embodiments, comprising steps of obtaining a frame Prototype drawing
Picture；Obtain multiframe target image；One group of template characteristic is obtained according to the template image, one group of template characteristic includes multiple
Template characteristic；Based on the pixel grey scale feature of every frame target image, one group of target signature is extracted from every frame target image, and
One group of target signature includes multiple target signatures；According to one group of template characteristic of template image and every frame target image
One group of target signature, the image similarity of calculation template image and every frame target image, obtain multiple images similarity；According to
Described multiple images similarity obtains the maximum value of described multiple images similarity；Obtain the corresponding target figure of the maximum value
Picture, with the corresponding target image of the maximum value be and the matched image of the template image.
According to second aspect, a kind of image matching apparatus is provided in a kind of embodiment, comprising: the first image acquisition unit,
The first image acquiring unit is for obtaining template image；Second image acquisition unit, second image acquisition unit are used
In acquisition multiframe target image；Template characteristic acquiring unit, the template characteristic acquiring unit are used for according to the template image
One group of template characteristic is obtained, wherein one group of template characteristic includes multiple template feature；Target's feature-extraction unit, the mesh
Mark feature extraction unit extracts one group of target for the pixel grey scale feature based on every frame target image from every frame target image
Feature, and one group of target signature includes multiple target signatures；Similarity calculated, the similarity calculated are used
According to one group of template characteristic of template image and one group of target signature of every frame target image, calculation template image and every
The image similarity of frame target image obtains multiple images similarity；Matching unit, the matching unit is according to the multiple figure
As similarity, the maximum value of described multiple images similarity is obtained, and obtains the corresponding target image of the maximum value, with described
The corresponding target image of maximum value be and the matched image of the template image.
In the image matching method and device of above-described embodiment, pass through the similarity between template characteristic and target signature point
Image similarity between other calculation template image and each target image carries out images match, it is ensured that images match mistake
The not redundancy of feature in journey, guarantees the correct matching of image, improves the accuracy of images match.
Detailed description of the invention
Fig. 1 is the flow diagram of the image matching method of some embodiments of the invention；
Fig. 2 is the template image of some embodiments of the invention and the schematic diagram of template characteristic point；
Fig. 3 is the schematic diagram of the target image of some embodiments of the invention；
Fig. 4 is image and the matching characteristic point to match with template characteristic point in Fig. 2 that is therefrom searching out in Fig. 3
Schematic diagram；
Fig. 5 is the structure of block diagram schematic diagram of the image matching apparatus of some embodiments of the invention.
Specific embodiment
Below by specific embodiment combination attached drawing, invention is further described in detail.
The embodiment of the present invention is related to the method and apparatus of images match.For example, in some embodiments, the images match side
Method and device can be used for from identifying in multiple image and the specific objective phase in another frame image or another frame image
Same or approximate target.Herein, will be referred to as the image of the target of identification (for example, " another frame image " above-mentioned)
" template image ", aforementioned specific objective is referred to as " interesting target ", and will need therefrom to identify and template image or sense
Targets of interest is identical or approximate image is referred to as " target image ".
In the embodiment of the present invention, generally, it can be obtained from template image using template image as template and characterize the mould
The feature (referred to herein as " template characteristic ") of plate image or interesting target therein, these template characteristics can be can
(herein with the point (referred to herein as " template characteristic point ") for characterizing the template image or the interesting target and/or side
Referred to as " template characteristic side ")；In addition, feature (referred to herein as " target signature ") is extracted from target image, the target
Feature is also possible to point (referred to herein as " target feature point ") and/or while (referred to herein as " when target signature ")；So
Afterwards, according to template characteristic and target signature, similarity between calculation template image and every frame target image, and compare acquisition
The size of similarity is the matching image to match with template image with the corresponding target image of the maximum value of similarity.This hair
In bright some embodiments, " side " mentioned here can be the line segment in image between two points.
Fig. 1 is the flow diagram of the image matching method of some embodiments of the invention.As shown in Figure 1, in step 10, it can
To obtain template image.In the embodiment of the present invention, template image, which can be, currently in real time to be obtained by various imaging devices
Image is also possible to obtain and be stored in the storage of the system using image matching method and device of the invention in advance
Image in device.Therefore, in step 10, template image can be obtained by various imaging devices, can also be read from memory
Template image out.
After obtaining template image, in a step 11, one group of template characteristic can be obtained according to the template image.These
Template characteristic, which can be in the image of the template image or the interesting target in the template image, special feature (position
Set, gray scale, angle etc.), by the information for containing template image or interesting target and characteristic, the template image can be characterized
Or point and/or the side of interesting target.This group of template characteristic may include multiple template feature, embody template image or
The feature of person's interesting target.
In some embodiments of the present invention, this group of template characteristic can be obtained by receiving the input of user.For example, with
Family can click the feature of selection template image or interesting target by input unit on template image.The present invention is implemented
The image matching apparatus of example receives the input of user, and obtains one group of template characteristic according to the input of user.
In other embodiments of the invention, this group of template characteristic can also be by image matching apparatus of the invention according to mould
Plate image or interesting target pixel grey scale feature (for example, gray average, gradient, variance, gray distribution features, etc.)
It is extracted from template image.
In the embodiment of the present invention, " one group of template characteristic " above-mentioned may include user's input or from template image
In the whole of feature that extracts, can also only comprising user's input or from one in the feature extracted in template image
Part.
Fig. 2 shows the template image in some embodiments of the invention, and wherein interesting target is in the template image
Automobile.In Fig. 2, point A1, A2, A3, A4, A5, A6, A7, A8, A9 and A10 are user's input or (are schemed according to interesting target
In 2 be automobile) pixel grey scale feature extraction go out template characteristic point.In Fig. 2, in order to clearly illustrate, template characteristic
Point A1-A10 is schematically illustrated as different size of circle.It should be appreciated that these circles are intended merely to schematically show template
Characteristic point, and the limitation not to the size of template characteristic point, position, shape etc..In Fig. 2, can also with point A1, A2, A3,
Line segment in A4, A5, A6, A7, A8, A9 and A10 between any two point is as template characteristic side.
In some embodiments of the present invention, template image is made with Fig. 2, using Fig. 3 (described below) as a target image,
For automobile in Fig. 2 as interesting target, image matching method and device according to an embodiment of the present invention can be by Fig. 3 and Fig. 2
In automobile matched.
In the embodiment of Fig. 1, in step 20, available multiframe target image.Similar with step 10, target image can
To be the image currently obtained by various imaging devices in real time, it is also possible to obtain and be stored in advance to use this hair
Image in the memory of the system of bright image matching method and device.Therefore, it in step 20, can be filled by various imagings
Acquisition target image is set, target image can also be read from memory.
Fig. 3 shows the target image obtained in some embodiments of the invention.
It, can be based on the pixel grey scale feature of target image, from every frame mesh in step 21 after obtaining target image
One group of target signature is extracted in logo image.It can be used from the method for extracting target signature in target image a variety of suitable in this field
The image characteristic extracting method of conjunction.For example, maximum stable extremal region method (MSER), scale can be used in some embodiments
Invariant features converter technique (SIFT), gloomy operator (Hess ian) method in sea, the affine method of Harris (Harris Affine) or histogram
Figure Attribute Association figure (HARG) method extracts one group of target signature from every frame target image.
In the embodiment of the present invention, " one group of target signature " above-mentioned may include the feature extracted from target image
Whole, can also only a part of feature comprising being extracted from target image.
It, in step 30, can be with after obtaining the template characteristic of template image and the target signature of every frame template image
According to this group of target signature of the one group of template characteristic and every frame target image of template image, template image and every frame are calculated
The image similarity of target image, to obtain multiple images similarity.
In some embodiments of the present invention, for each frame target image, calculating, it is similar to the image of template image
When spending, (at this point, participate in the frame target image that calculates be referred to as " current target image ") can be carried out according to the following steps:
Firstly, obtaining this group of mesh of each template characteristic and current target image in this group of template characteristic of template image
Mark the similarity (referred to herein as " similarity between feature ") between each target signature in feature.It is readily appreciated that, at this point,
When being characterized in here, similarity here is similarity between characteristic point；When being characterized in side here, phase here
Like similarity between degree as characteristic edge.In the embodiment of the present invention, the feature used be can be a little, be also possible to side, Huo Zheye
It can be the two while using.It is first between template characteristic and the target signature of every frame target image in the embodiment of the present invention
The calculation method that this field routine can be used in the characteristic similarity of beginningization obtains, and this will not be detailed here；
Then, it according to similarity between the feature of acquisition, is searched out from this group of target signature of current target image and mould
The feature (referred to herein as " matching target signature ") that each template characteristic in this group of template characteristic of plate image matches；
Then, according between the matching target signature in one group of template characteristic and current target image of template image
Feature between similarity (as it was noted above, similarity has obtained between feature between each template characteristic and each target signature
, therefore similarity is known at this time between each template characteristic and each feature matched between target signature), calculate current goal
The similarity of image and template image.
In some embodiments of the present invention, the similarity between the feature according to acquisition, from this group of mesh of current target image
When searching out the feature that each template characteristic in this group of template characteristic with template image matches in mark feature, for template
Each template characteristic in one group of template characteristic of image can be carried out according to the following steps (at this point, participating in the template calculated
Feature is referred to as " current template feature "):
Firstly, according to each target signature in current template feature and one group of target signature of current target image it
Between feature between similarity, selection and current template are special from the target signature in one group of target signature of current target image
Immediate multiple target signatures are levied, and here, multiple target signature of selection is only one group of current target image
A part in target signature, rather than all.It here, can be with multiple target signatures of current template feature " closest "
It is multiple target signatures of similarity relative maximum between the feature between current template feature, for example, in some embodiments, it will
The descending arrangement of similarity between feature between target signature and current template feature, it is immediate more with current template feature
A target signature can be the multiple target signatures of ranking up front, such as the first two, first three a or first four, etc.；
Then, similarity between multiple features between calculating current template feature and the multiple target signature selected
(as it was noted above, having similarity between a feature between each template characteristic and each target signature, therefore current mould here
With similarity between multiple features between plate features and multiple target signatures) weighted average, and from phase between multiple feature
Like the similarity between the immediate feature of the weighted average is obtained in degree, with the phase between the immediate feature of the weighted average
It is the matching target signature with current template characteristic matching like corresponding target signature is spent.
Each template characteristic point is scanned for according to aforementioned process, each template can be searched out from target signature
The matching target signature of feature.For example, schematically showed in Fig. 4 searched out in the image in Fig. 3 with the mould in Fig. 2
The matching target feature point (B1, B2, B3, B4, B5, B6, B7, B8, B9 and B10) that plate features point matches, these matching targets
Characteristic point symbolized in target image with interesting target (i.e. automobile in Fig. 2) approximate target in Fig. 2.
In these embodiments, in the matching target signature that search matches with template characteristic, surrounding multiple features
Calculating is taken part in, i.e., is referred to surrounding multiple features, the accuracy rate of search matching target signature is improved；Meanwhile
Target complete feature is not involved in calculating when calculating, but uses the weighting of surrounding a part of target signature neighbouring
The average value of value not only reduces calculation amount as measuring, and avoids the waste portion for calculating time and resource, and avoided
The interference of remote target signature improves matching accuracy.That is, the regionality for having comprehensively considered image is special in these embodiments
Property and rarefaction characteristic, that is, reduce calculation amount, also improve matching accuracy.
Target signature in some embodiments of the present invention, in one group of target signature above-mentioned from current target image
Middle selection can be three target signatures with the immediate multiple target signatures of current template feature, i.e., from current target image
Selection and immediate three target signatures of current template feature, calculate current mould in target signature in one group of target signature
The weighted average of similarity between feature between plate features and these three target signatures, and in these three target signatures and currently
Similarity and that immediate target signature of the average value match with current template feature between the feature of template characteristic
Match target signature.Inventor it has been investigated that, when taking three target signatures here, can obtain more preferably images match effect
Fruit.
It, can be as mentioned before according to template image after obtaining the matching target signature to match with template characteristic
One group of template characteristic matches similarity calculation current target image and template image between the feature between target signature with these
Similarity.For example, in some embodiments of the present invention, can calculate one group of template characteristic matched with these target signature it
Between feature between similarity sum, using this and as current target image and template image similarity.
It for each frame target image, is carried out according to aforementioned process, calculates every frame target image and template image
Similarity, to obtain multiple images similarity.Then, in the step 32, can be obtained according to the multiple images similarity of acquisition
The maximum value in multiple image similarity is obtained, i.e. acquisition maximum image similarity, and corresponding with the maximum image similarity
Target image be and the matched image of template image.
Correspondingly, as shown in figure 5, in some embodiments of the present invention, a kind of image matching apparatus is additionally provided, the image
Coalignment may include the first image acquisition unit 50, the second image acquisition unit 60, template characteristic acquiring unit 51, target
Feature extraction unit 61, similarity calculated 70 and matching unit 80.The image matching apparatus can execute described above
Each embodiment in image matching method.For example, the first image acquisition unit 50 is for obtaining template image；Second image
Acquiring unit 60 is for obtaining multiframe target image；Template characteristic acquiring unit 51 is used to obtain one group of template according to template image
Feature, which includes multiple template feature；Target's feature-extraction unit 61 is used for based on every frame target image
Pixel grey scale feature extracts one group of target signature from every frame target image, and one group of target signature includes that multiple targets are special
Sign；Similarity calculated 70 is used for according to one group of template characteristic of template image and one group of target of every frame target image
Feature, the image similarity of calculation template image and every frame target image obtain multiple images similarity；Matching unit 72 is used for
According to multiple image similarity, the maximum value of multiple image similarity is obtained, and obtains the corresponding target figure of the maximum value
Picture, with the corresponding target image of maximum value be and the matched image of template image.
In the embodiment of Fig. 5, template characteristic acquiring unit 51 can be used for receiving the input of user and according to the defeated of user
Enter to obtain one group of template characteristic and/or one group of template spy is extracted from template image according to the pixel grey scale feature of template image
Sign.
In the embodiment of Fig. 5, it is special that maximum stable extremal region method, Scale invariant can be used in target's feature-extraction unit 61
Levy converter technique, sub (Hessian) method of the gloomy detection in sea, the affine method of Harris or histogram properties associated diagram (HARG) method etc. side
Method extracts one group of described target signature from target image.
In the embodiment of Fig. 5, similarity calculated 70 can execute the following steps for every frame target image:
Obtain one group of mesh of each template characteristic and current target image in one group of template characteristic of template image
Similarity between the feature between each target signature in mark feature；
According to similarity between this feature, searched out and template image from one group of target signature of current target image
The matching target signature that each template characteristic in one group of template characteristic matches；
According to similarity meter between the feature between one group of template characteristic and the matching target signature of current target image
Calculate the similarity of current target image and template image.
In the embodiment of Fig. 5, similarity calculated 70 according to similarity between feature from current target image this one
The matching target that each template characteristic in one group of template characteristic with template image matches is searched out in group target signature
The following steps can be executed when feature:
According between each target signature in current template feature and one group of target signature of current target image
Similarity between feature, selection and the immediate multiple mesh of current template feature from one group of target signature of current target image
Mark feature.Here multiple target signatures are a part in one group of target signature of current target image；
The weighted average of similarity between multiple features between calculating current template feature and multiple target signature；
From obtaining the similarity between the immediate feature of the weighted average between multiple feature in similarity, and with this
The corresponding target signature of similarity is special with the matching target of current template characteristic matching between the immediate feature of weighted average
Sign.
In some embodiments, similarity calculated 70 selects and works as from one group of target signature of current target image
It is similar between the feature between these three target signatures to calculate current template feature for immediate three target signatures of front template feature
The weighted average of degree, and in these three target signatures between the feature of current template feature similarity and the weighted average
That immediate target signature is the matching target signature to match with current template feature.
In the embodiment of Fig. 5, matching unit 72 can be with one group of template characteristic of calculation template image and current goal figure
The sum of similarity between feature between these matching target signatures of picture, using this and as current target image and template image
Similarity.
In the image matching method and device of above-described embodiment, pass through the similarity between template characteristic and target signature point
Image similarity between other calculation template image and each target image carries out images match, it is ensured that images match mistake
The not redundancy of feature in journey, guarantees the correct matching of image, improves the accuracy of images match.
It will be understood by those skilled in the art that all or part of the steps of various methods can pass through in above embodiment
Program instructs related hardware to complete, which can be stored in a computer readable storage medium, storage medium can wrap
It includes: read-only memory, random access memory, disk or CD etc..
Use above specific case is illustrated the present invention, is merely used to help understand the present invention, not to limit
The system present invention.For those skilled in the art, according to the thought of the present invention, can also make several simple
It deduces, deform or replaces.
Claims (12)
1. a kind of image matching method, which is characterized in that comprising steps of
Obtain a frame template image；
Obtain multiframe target image；
One group of template characteristic is obtained according to the template image, one group of template characteristic includes multiple template feature, the mould
Plate features are template characteristic point and/or template characteristic side；
Based on the pixel grey scale feature of every frame target image, one group of target signature is extracted from every frame target image, and described
One group of target signature includes multiple target signatures, and the target signature is target feature point and/or target signature side；
According to one group of template characteristic of template image and one group of target signature of every frame target image, calculation template image with
The image similarity of every frame target image obtains multiple images similarity；
According to described multiple images similarity, the maximum value of described multiple images similarity is obtained；
The corresponding target image of the maximum value is obtained, is and the template image with the corresponding target image of the maximum value
The image matched；
Wherein, according to one group of template characteristic of template image and one group of target signature of every frame target image, calculation template
The image similarity of image and every frame target image, obtain multiple images similarity the step of include: for every frame target image,
The one group of target for obtaining each template characteristic and current target image in one group of template characteristic of template image is special
Similarity between the feature between each target signature in sign；According to similarity between the feature, from the institute of current target image
It states and searches out that each template characteristic in one group of template characteristic with template image matches in one group of target signature
With target signature；According to phase between the feature between one group of template characteristic and the matching target signature of current target image
The similarity of current target image and template image is calculated like degree；
Wherein, it according to similarity between the feature, is searched out from one group of target signature of current target image and template
The step of matching target signature that each template characteristic in one group of template characteristic of image matches includes: for template
Each template characteristic in one group of template characteristic of image, according to described the one of current template feature and current target image
Similarity between the feature between each target signature in group target signature, from one group of target signature of current target image
In target signature in selection with the immediate multiple target signatures of current template feature, wherein the multiple target signature be work as
A part in one group of target signature of preceding target image；It calculates between current template feature and the multiple target signature
Multiple features between similarity weighted average；From being obtained in similarity between the multiple feature with the weighted average most
Similarity between close feature, and between the immediate feature of the weighted average the corresponding target signature of similarity be with
The matching target signature of current template characteristic matching.
2. the method as described in claim 1, which is characterized in that according to the step for obtaining one group of template characteristic in the target image
Suddenly include:
It receives the input of user and one group of template characteristic is obtained according to the input of user；
And/or
One group of template characteristic is extracted from the template image according to the pixel grey scale feature of the template image.
3. the method as described in claim 1, which is characterized in that the step of extracting one group of target signature from every frame target image
Include: using maximum stable extremal region method, Scale invariant features transform method, the sub- method of the gloomy detection in sea, the affine method of Harris or
Histogram properties associated diagram method extracts one group of target signature from every frame target image.
4. the method as described in claim 1, which is characterized in that from one group of target feature point of current target image
The step of multiple target signatures immediate with current template feature are selected in target signature includes: the institute from current target image
State selection and immediate three target signatures of current template feature in the target signature in one group of target signature.
5. the method as described in claim 1, which is characterized in that according to the institute of one group of template characteristic and current target image
State matching target signature between feature between the similarity of similarity calculation current target image and template image the step of include:
Calculate similarity between the feature between one group of template characteristic and the matching target signature of current target image and, with
Similarity described and as current target image and template image.
6. the method as described in any one of claim 1-5, it is characterised in that:
Similarity between similarity is characterized a little between the feature, the matching target signature are matching target feature point；
And/or
Similarity is characterized similarity between side between the feature, and the matching target signature is matching target signature side.
7. a kind of image matching apparatus characterized by comprising
First image acquisition unit, the first image acquiring unit is for obtaining template image；
Second image acquisition unit, second image acquisition unit is for obtaining multiframe target image；
Template characteristic acquiring unit, the template characteristic acquiring unit are used to obtain one group of template according to the template image special
Sign, wherein one group of template characteristic includes multiple template feature, the template characteristic is template characteristic point and/or template characteristic
Side；
Target's feature-extraction unit, the target's feature-extraction unit for the pixel grey scale feature based on every frame target image from
One group of target signature is extracted in every frame target image, and one group of target signature includes multiple target signatures, the target
Feature is target feature point and/or target signature side；
Similarity calculated, the similarity calculated are used for according to one group of template characteristic of template image and every frame
One group of target signature of target image, the image similarity of calculation template image and every frame target image obtain multiple images phase
Like degree；
Matching unit, the matching unit obtain the maximum of described multiple images similarity according to described multiple images similarity
Value, and the corresponding target image of the maximum value is obtained, it is and the template image with the corresponding target image of the maximum value
Matched image；
Wherein, the similarity calculated executes step: obtaining one group of mould of template image for every frame target image
Between each target signature in each template characteristic in plate features and one group of target signature of current target image
Similarity between feature；According to similarity between the feature, searched out from one group of target signature of current target image with
The matching target signature that each template characteristic in one group of template characteristic of template image matches；According to one group of mould
Similarity calculation current target image and mould between feature between plate features and the matching target signature of current target image
The similarity of plate image；
Wherein, it according to similarity between the feature, is searched out from one group of target signature of current target image and template
When the matching target signature that each template characteristic in one group of template characteristic of image matches, the similarity calculation list
Member executes step: according to current template feature for each template characteristic in one group of template characteristic point of template image
Similarity between feature between each target signature in one group of target signature of current target image, from current goal
Selection and the immediate multiple target signatures of current template feature in one group of target signature of image, wherein the multiple mesh
Mark a part in one group of target signature that feature is current target image；Calculate current template feature and the multiple mesh
The weighted average of similarity between multiple features between mark feature；Add from being obtained in similarity between the multiple feature with described
Similarity between the immediate feature of weight average value, and with the corresponding mesh of similarity between the immediate feature of the weighted average
Marking feature is the matching target signature with current template characteristic matching.
8. device as claimed in claim 7, which is characterized in that the template characteristic acquiring unit is used to receive the input of user
And one group of template characteristic is obtained and/or according to the pixel grey scale feature of the template image from described according to the input of user
One group of template characteristic is extracted in template image.
9. device as claimed in claim 7, it is characterised in that: the target's feature-extraction unit uses maximum stable extremal area
Domain method, Scale invariant features transform method, gloomy sub- method, the affine method of Harris or the histogram properties associated diagram method of detecting in sea are from every frame
One group of target signature is extracted in target image.
10. device as claimed in claim 7, which is characterized in that institute of the similarity calculated from current target image
State selection and immediate three target signatures of current template feature in one group of target signature.
11. device as claimed in claim 7, it is characterised in that: it is special that the similarity calculated calculates one group of template
Levy between the feature between the matching target signature of current target image similarity and, using described and as current goal
The similarity of image and template image.
12. the device as described in any one of claim 7 to 11, it is characterised in that:
Similarity between similarity is characterized a little between the feature, the matching target signature are matching target feature point；
And/or
Similarity is characterized similarity between side between the feature, and the matching target signature is matching target signature side.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610022261.2A CN105678778B (en) | 2016-01-13 | 2016-01-13 | A kind of image matching method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610022261.2A CN105678778B (en) | 2016-01-13 | 2016-01-13 | A kind of image matching method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105678778A CN105678778A (en) | 2016-06-15 |
| CN105678778B true CN105678778B (en) | 2019-02-26 |
Family
ID=56300477
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610022261.2A Expired - Fee Related CN105678778B (en) | 2016-01-13 | 2016-01-13 | A kind of image matching method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105678778B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295644B (en) * | 2016-08-11 | 2020-06-09 | 海信集团有限公司 | Symbol identification method and device |
| WO2018068304A1 (en) * | 2016-10-14 | 2018-04-19 | 深圳配天智能技术研究院有限公司 | Image matching method and device |
| CN107516105B (en) * | 2017-07-20 | 2020-06-16 | 阿里巴巴集团控股有限公司 | Image processing method and device |
| CN107885835B (en) * | 2017-11-09 | 2020-08-28 | 东软集团股份有限公司 | Similar layer image searching method and device of tomography image |
| CN113128543B (en) * | 2020-01-15 | 2024-02-13 | 阿里巴巴集团控股有限公司 | Image matching method, application testing method, device and system |
| CN113760415A (en) * | 2020-06-04 | 2021-12-07 | Oppo广东移动通信有限公司 | Dial plate generation method and device, electronic equipment and computer readable storage medium |
| CN113486769B (en) * | 2021-07-01 | 2024-04-26 | 珍岛信息技术（上海）股份有限公司 | Quick image matching method in high-definition video |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101510257A (en) * | 2009-03-31 | 2009-08-19 | 华为技术有限公司 | Human face similarity degree matching method and device |
| CN102289681A (en) * | 2011-08-05 | 2011-12-21 | 上海邮政科学研究院 | Method for matching envelope images |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464951B (en) * | 2007-12-21 | 2012-05-30 | 北大方正集团有限公司 | Image recognition method and system |
| US8243988B1 (en) * | 2008-07-31 | 2012-08-14 | Google Inc. | Clustering images using an image region graph |
| CN101770582B (en) * | 2008-12-26 | 2013-05-08 | 鸿富锦精密工业（深圳）有限公司 | Image matching system and method |
| WO2013180660A1 (en) * | 2012-06-01 | 2013-12-05 | Agency For Science, Technology And Research | Robust graph representation and matching of retina images |
| CN105046684B (en) * | 2015-06-15 | 2017-09-29 | 华中科技大学 | A kind of image matching method based on polygon generalised Hough transform |
-
2016
- 2016-01-13 CN CN201610022261.2A patent/CN105678778B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101510257A (en) * | 2009-03-31 | 2009-08-19 | 华为技术有限公司 | Human face similarity degree matching method and device |
| CN102289681A (en) * | 2011-08-05 | 2011-12-21 | 上海邮政科学研究院 | Method for matching envelope images |
Non-Patent Citations (1)
| Title |
|---|
| 恒星光谱参数自动测量中不同模板匹配度量方法比较;刘杰 等;《光谱学与光谱分析》;20130228;第33卷(第2期);第559页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105678778A (en) | 2016-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105678778B (en) | A kind of image matching method and device | |
| CN109753928B (en) | Method and device for identifying illegal buildings | |
| US9275311B2 (en) | Efficient image matching for large sets of images | |
| CN110428399B (en) | Method, apparatus, device and storage medium for detecting image | |
| US8031947B2 (en) | Method and system for rapid matching of video streams | |
| CN109214238A (en) | Multi-object tracking method, device, equipment and storage medium | |
| CN108229301B (en) | Eyelid line detection method and device and electronic equipment | |
| CN105956059A (en) | Emotion recognition-based information recommendation method and apparatus | |
| KR102284096B1 (en) | System and method for estimating subject image quality using visual saliency and a recording medium having computer readable program for executing the method | |
| US10699156B2 (en) | Method and a device for image matching | |
| JP6071002B2 (en) | Reliability acquisition device, reliability acquisition method, and reliability acquisition program | |
| CN105243376A (en) | Living body detection method and device | |
| CN110751069A (en) | Face living body detection method and device | |
| CN109977824B (en) | Article taking and placing identification method, device and equipment | |
| US20160055627A1 (en) | Information processing device, image processing method and medium | |
| CN112818995B (en) | Image classification method, device, electronic equipment and storage medium | |
| CN109376717A (en) | Personal identification method, device, electronic equipment and the storage medium of face comparison | |
| US20190043168A1 (en) | An image processing device, an image processing method, and computer-readable recording medium | |
| WO2017131870A1 (en) | Decoy-based matching system for facial recognition | |
| CN105229700B (en) | Device and method for extracting peak figure picture from multiple continuously shot images | |
| CN103927529B (en) | The preparation method and application process, system of a kind of final classification device | |
| CN111191708A (en) | Automatic sample key point marking method, device and system | |
| Xu et al. | Detecting and classifying blurred image regions | |
| Schöning et al. | Semi-automatic ground truth annotation in videos: An interactive tool for polygon-based object annotation and segmentation | |
| Viacheslav et al. | Low-level features for inpainting quality assessment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190226 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |