CN106354738B - A kind of flat image typing and matched method and system - Google Patents
A kind of flat image typing and matched method and system Download PDFInfo
- Publication number
- CN106354738B CN106354738B CN201510442276.XA CN201510442276A CN106354738B CN 106354738 B CN106354738 B CN 106354738B CN 201510442276 A CN201510442276 A CN 201510442276A CN 106354738 B CN106354738 B CN 106354738B
- Authority
- CN
- China
- Prior art keywords
- image
- key point
- matched
- picture
- database
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/53—Querying
- G06F16/532—Query formulation, e.g. graphical querying
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- Mathematical Physics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of flat image typing and matched method and systems.A kind of flat image typing and matched system include image recording module, acquisition image module, image processing module, images match module and associated images display module.A kind of method of flat image typing is included: the processing of conventional images gray processing, binary conversion treatment, extremum extracting, with gaussian pyramid forms feature point set, association its eight directions, the description of key point is adaptively adjusted SIFT algorithm and is saved in the image handled in database in a manner of key point.A kind of method of images match include: existing image is entered into database, using the camera of electronic equipment shot obtained by picture or the image for calling electronic equipment to store, collected picture handled, the extraction of characteristic point carried out to image and carry out display output as the description of key point, to the matching of progress image inside database and by the associated picture matched.
Description
[technical field]
The present invention relates to a kind of Information Technology Agreement field, more particularly to a kind of flat image typing and matched method and
System.
[background technique]
It now, is usually all by the way of inputting search key in information retrieval.However, this retrieval mode
There are certain defects, such as: the retrieval mode of this keyword needs user to examine by the keyboard of computer or mobile phone
The input of rope keyword, therefore this just causes very big inconvenience to the use of user, as user not will pass through keyboard input
In the case where or some search information be difficult with text carry out accurate description when, this way of search is also difficult to ensure inspection
The real-time of rope information.
Therefore, when scanning for the information content in the form of inputting keyword, cause retrieval information inconvenient and
The low problem of search efficiency, there is great inconvenience.
[summary of the invention]
For overcome the problems, such as it is existing the inconvenient low with search efficiency of information is scanned in the form of inputting keyword,
The present invention provides a kind of higher flat image typing of convenient and search efficiency and matched method and systems.
The scheme that the present invention solves technical problem is to provide a kind of flat image typing and matched system, the system packet
It includes:
Image recording module, image recording module, for by extraction key point, that is, image of conventional images 360 degrees omnidirection
Characteristic point or verbal description, or be entered into the specified physical bit of database for the form of representative small pattern
In setting；The key point is to handle conventional images gray processing, then carry out binary conversion treatment, and further progress extremum extracting is used
Gaussian pyramid forms feature point set；It using the scaling of two dimensional image, stacks, is converted into three-dimensional processing, then pass through SIFT algorithm
It is adaptively adjusted and is associated with its eight directions；By the fitting of three-dimensional quadratic function, the accurate position for determining key point and scale, most
The description for forming key point afterwards is obtained；
Acquire image module, for using electronic equipment camera shot gained picture, or calling electronics set
The standby image stored；
Image processing module, for handling acquired image, by image with key point, that is, image characteristic point,
Perhaps verbal description or the form for representative small pattern go to describe；
Images match module, will acquire image and conventional images carry out the matching of key point, judge whether to meet condition；And
Associated images display module, according to matched similarity, to the matched image of institute according to degree of similarity to association
Image is shown.
The scheme that the present invention solves technical problem also provides a kind of method of image typing, this method comprises:
Step S1, the processing of conventional images gray processing；
Step S2, binary conversion treatment；
Step S3, extremum extracting form feature point set with gaussian pyramid；
Step S4 using the scaling of two dimensional image, is stacked, and is converted into three-dimensional processing, then pass through SIFT algorithm adaptability tune
Whole its eight directions of association；
Step S5, by the fitting of three-dimensional quadratic function, the accurate position for determining key point and scale eventually form key
The description of point；And
The image handled is saved in database by step S6 in a manner of key point.
Preferably, step S3 takes the detection in scale space local extremum, constructs the scale gaussian pyramid of image,
Local maximum and minimum are detected in gaussian pyramid.
Preferably, step S4 using two dimensional image scaling, stack, be converted into three-dimensional processing, then pass through SIFT algorithm
It is associated with eight directions of conventional images.
The scheme that the present invention solves technical problem also provides a kind of matched method of flat image, this method comprises:
Existing image is entered into database by step P1 using the method for flat image typing as described above；
Step P2, using the camera of electronic equipment shot obtained by picture, or call electronic equipment stored
Image；
Step P3 handles collected picture；
Step P4 carries out the extraction of characteristic point to image and as the description of key point；
Step P5 carries out the matching of image to database the inside；And
The associated picture matched is carried out display output by step P6.
Preferably, the image handled is entered into database physically by step P1 in a manner of key point, and
And it is corresponding physically that the specifying information typing of conventional images is saved in database.
Preferably, step P2 using electronic equipment camera shot gained picture, or calling electronic equipment
The image stored is sample of the user as the details for being used to search for this product or close product as acquisition image.
Preferably, step P3 handles collected picture, does gray processing processing, binaryzation to collected picture
Processing, extremum extracting and with gaussian pyramid formed feature point set and using the scaling of two dimensional image, stack, be converted into solid
Processing.
Preferably, step P5 is the matching that image is carried out inside collected picture to database, when the pass of original image
The description vectors of key point and collected picture key point generate after using at a distance from the two as the similitude between key point
Decision metric.
Compared with prior art, a kind of flat image typing of the invention and matched method and system are by by conventional images
Which extract invariant features to 360 ° omni-directional to be entered into database, to realize collected picture regardless of being in direction
It now can faster be matched to corresponding or similar image.Secondly, the present invention is by carrying out two dimensional image to conventional images
Scaling, stack, be converted into three-dimensional processing, realize collected picture and be more accurately matched to corresponding or similar figure
Picture.
[Detailed description of the invention]
Fig. 1 is a kind of flat image typing of the embodiment of the present invention one and matched system flow chart.
Fig. 2 is a kind of method flow diagram of the flat image typing of the embodiment of the present invention two.
Fig. 3 is a kind of three matched method flow diagram of flat image of the embodiment of the present invention.
[specific embodiment]
In order to make the purpose of the present invention, technical solution and advantage are more clearly understood, below in conjunction with attached drawing and embodiment,
The present invention will be described in further detail.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention,
It is not intended to limit the present invention.
Embodiment one:
Referring to Fig. 1, the invention proposes a kind of flat image typing and matched system, which includes:
Image recording module 101, for conventional images to be entered into the specified physical bit of database in the form of key point
In setting；
Acquire image module 102, for using electronic equipment camera shot gained picture, or calling electronics
The image that equipment has stored；
Image is gone to retouch in the form of key point by image processing module 103 for handling acquired image
It states；
Images match module 104, will acquire image and conventional images carry out the matching of key point, judge whether to meet item
Part；And
Associated images display module 105, according to matched similarity, to the matched image of institute according to degree of similarity to pass
Connection image carries out display output.
It wherein, is that conventional images are done to gray processing, binaryzation, extremum extracting and are mentioned in conventional images recording module 101
It being entered into database by way of key point again after taking, key point can be the characteristic point or verbal description of image,
It or is representative small pattern.
It wherein, is equally that will acquire image to do gray processing, binaryzation, extremum extracting in acquisition image processing module 103
And carry out subsequent match after extracting by way of key point again.
Wherein, in associated images display module 105, the key point of the key point for acquiring image and conventional images is carried out
Matching.
Embodiment two:
Referring to Fig. 2, the invention proposes a kind of methods of flat image typing, this method comprises:
Step S1, the processing of conventional images gray processing；
Step S2, binary conversion treatment；
Step S3, extremum extracting form feature point set with gaussian pyramid；
Step S4 is adaptively adjusted by SIFT algorithm and is associated with its eight directions；
Step S5, the description of key point；And
The image handled is saved in database by step S6 in a manner of key point.
In step sl, gray processing processing is carried out to conventional images, is that jpeg image is converted to gray image, gray scale
Being worth range is 0~255, and gray value is customized for 8 by this step.
In step s 2, binary conversion treatment is that whole image is showed to apparent black and white effect.I.e. by 256 brightness etc.
The gray level image of grade, which is chosen to obtain by threshold values appropriate, still can reflect the whole binary picture with local feature of image
Picture.The pixel that all gray scales belong to threshold range is judged as belonging to certain objects, and gray value is 255 to indicate, otherwise these
Pixel is excluded other than object area, gray value 0, indicates the object area of background or exception.Binarization makes
It is adjusted with Open C V institute band Threshold algorithm.
In step s3, the detection in scale space local extremum is taken, the scale gaussian pyramid of image is constructed, to ruler
Adjacent 2 layers of the degree every rank of gaussian pyramid successively make the difference, to obtain difference of Gaussian pyramid；It is detected in difference of Gaussian pyramid
Local maximum and minimum, part where each extreme value must be are adjacent (8 neighborhood points including same layer, one layer and next upper
Each corresponding 9 neighborhoods point of layer) in extreme point, using all extreme points as the candidate target of key point.Scale space pole
Value detection, exactly finds several candidate points on image, these are when the scale of image changes, that is, image
Become larger or become smaller, the relationship of they and surrounding others point remains unchanged.
In step s 4, using the scaling of two dimensional image, stack, be converted into three-dimensional processing, then closed by SIFT algorithm
Join conventional images eight directions, i.e., 360 ° omni-directional extract target invariant features, it is assumed that we to identify telephone set it is necessary to
A photo first is clapped to telephone set, then uses this photo as benchmark, extracts feature.After having extracted characteristic value, electricity is encountered in the future
The big figure of phone, small figure have rotated the figure of certain angle, can identify, if only size and all constant feelings of angle
Identify do not have real value under condition.Since small figure can identify in the case where rotation figure scheming greatly, then, it is inevitable to scheme
As inner " characteristic " is certainly existed, these features are being schemed greatly, small figure, rotate in figure all " constant ", or approximate constant.
In step s 5, key point accurate coordinates and scale determination and unstable candidate key point rejecting when,
By a kind of fitting of three-dimensional quadratic function, the accurate position for determining key point and scale (reaching sub- plain precision) are rejected simultaneously
The low key point of contrast rejects the key point with mobile rim response.And then the key point of selection is described, i.e.,
It is that conventional images are condensed into the multiple key points for being defined description, the form of key point is vector, such as TA=[a1,
a2...an] it is key point A.
In step s 6, the image handled is saved in database physically in a manner of key point.
Embodiment three:
Referring to Fig. 3, the invention proposes a kind of matched methods of flat image, this method comprises:
Existing image is entered into database by step P1；
Step P2, using the camera of electronic equipment shot obtained by picture, or call electronic equipment stored
Image；
Step P3, acquired image is handled；
Step P4 is carried out the extraction of characteristic point to image and is described with key point；
Step P5 carries out the matching of image to database the inside；And
The associated picture matched is carried out display output by step P6.
In step P1, use the method for image typing described in second embodiment by conventional images in a manner of key point
It is entered into database physically, and the specifying information typing of conventional images is saved in the corresponding physical bit of database
It sets.
In step P2, using the camera of electronic equipment shot obtained by picture, or call electronic equipment deposited
The image stored up is sample of the user as the details for being used to search for this product or close product.
In step P3, collected picture is handled, as processing method described in embodiment two, to acquisition
To picture do gray processing processing, binary conversion treatment, extremum extracting and with gaussian pyramid form feature point set and using two
It ties up the scaling of image, stack, be converted into three-dimensional processing.
In step P4, after the extraction that characteristic point is carried out to image, equally collected picture is carried out with key point
Description.
In step P5, the matching of image is carried out to database the inside, is utilized after the description vectors of key point generate:
Distance as between key point similarity determination measurement.Wherein TA=[a1,a2...an] and TB=[b1,
b2...bn] be respectively key point A and B description vectors.
In step P6, matching strategy is to find out description vectors distance according to SIFT algorithm and determine its phase from the near to the distant
Like degree, closer expression is more similar, and so on, the associated picture matched is subjected to display output.Such as it is therein some
2 the key points B and C that description vectors distance is nearest therewith are found in key point A, if nearest distance | | TA-TB| | with it is secondary close
Distance | | TA-TC| | ratio be less than some threshold values t, it may be assumed that
Then think that key point A is matched with apart from nearest key point B.
Compared with prior art, a kind of image typing of the invention and matched method and system pass through 360 ° of conventional images
Which extract invariant features in all directions to be entered into database, to realize collected picture regardless of being presented all with direction
Corresponding or similar image can be faster matched to.Secondly, the present invention passes through the contracting to conventional images progress two dimensional image
It puts, stack, be converted into three-dimensional processing, realize collected picture and be more accurately matched to corresponding or similar image.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in original of the invention
Made any modification within then, equivalent replacement and improvement etc. should all be comprising within protection scope of the present invention.
Claims (9)
1. a kind of flat image typing and matched system, it is characterised in that: the system includes:
Image recording module, for by extraction key point, that is, image characteristic point or text of conventional images 360 degrees omnidirection
Description, or be entered into the specified physical location of database for the form of representative small pattern；The key point is
Conventional images gray processing is handled, then carries out binary conversion treatment, further progress extremum extracting forms feature with gaussian pyramid
Point set；Using the scaling of two dimensional image, stack, be converted into three-dimensional processing, then by SIFT algorithm be adaptively adjusted association its eight
A direction；By the fitting of three-dimensional quadratic function, the accurate position for determining key point and scale eventually form retouching for key point
It states；
Acquire image module, for using electronic equipment camera shot gained picture, or calling electronic equipment
The image stored；
Image processing module, for handling acquired image, by image with key point, that is, image characteristic point or
Verbal description, or go to describe for the form of representative small pattern；
Images match module, will acquire image and conventional images carry out the matching of key point, judge whether to meet condition；And
Associated images display module, according to matched similarity, to the matched image of institute according to degree of similarity to associated images
It is shown.
2. a kind of method of flat image typing, which is characterized in that this method comprises:
Step S1, the processing of conventional images gray processing；
Step S2, binary conversion treatment；
Step S3, extremum extracting form feature point set with gaussian pyramid；
Step S4 using the scaling of two dimensional image, is stacked, and is converted into three-dimensional processing, then pass is adaptively adjusted by SIFT algorithm
Join its eight directions；
Step S5, by the fitting of three-dimensional quadratic function, the accurate position for determining key point and scale eventually form key point
Description；And
The image handled is saved in database by step S6 in a manner of key point.
3. a kind of method of flat image typing according to claim 2, it is characterised in that: step S3 takes in scale sky
Between local extremum detection, construct the scale gaussian pyramid of image, local maximum and minimum detected in gaussian pyramid
Value.
4. a kind of method of flat image typing according to claim 2, it is characterised in that: step S4 utilizes two dimensional image
Scaling, stack, be converted into three-dimensional processing, then by SIFT algorithm be associated with conventional images eight directions.
5. a kind of matched method of flat image, which is characterized in that this method comprises:
Existing image is entered into database by step P1 using the method for flat image typing as claimed in claim 2；
Step P2, using the camera of electronic equipment shot obtained by picture, or call the figure that has stored of electronic equipment
Picture；
Step P3 handles collected picture；
Step P4 carries out the extraction of characteristic point to image and as the description of key point；
Step P5 carries out the matching of image to database the inside；And
The associated picture matched is carried out display output by step P6.
6. a kind of matched method of flat image according to claim 5, it is characterised in that: the figure that step P1 will have been handled
As being entered into database in a manner of key point physically, and the specifying information typing of conventional images is saved in number
It is corresponding physically according to library.
7. a kind of matched method of flat image according to claim 5, it is characterised in that: step P2 uses electronic equipment
Camera shot obtained by picture, or call the image that has stored of electronic equipment as image is acquired, be that user does
For the sample for searching for the details of this product or close product.
8. a kind of matched method of flat image according to claim 5, it is characterised in that: step P3 is by collected figure
Piece is handled, and is done gray processing processing, binary conversion treatment, extremum extracting to collected picture and with gaussian pyramid is formed spy
Sign point set and using the scaling of two dimensional image, stack, be converted into three-dimensional processing.
9. a kind of matched method of flat image according to claim 5, it is characterised in that: step P5 is collected figure
The matching that image is carried out inside piece to database, when the key point of original image and the description vectors of collected picture key point
It is measured after generation using the distance of the two as the similarity determination between key point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510442276.XA CN106354738B (en) | 2015-07-18 | 2015-07-18 | A kind of flat image typing and matched method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510442276.XA CN106354738B (en) | 2015-07-18 | 2015-07-18 | A kind of flat image typing and matched method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106354738A CN106354738A (en) | 2017-01-25 |
| CN106354738B true CN106354738B (en) | 2019-09-10 |
Family
ID=57843238
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510442276.XA Active CN106354738B (en) | 2015-07-18 | 2015-07-18 | A kind of flat image typing and matched method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106354738B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108694405A (en) * | 2017-04-08 | 2018-10-23 | 钟少童 | Body trace intelligent identifying system and method |
| TWI779836B (en) * | 2021-09-15 | 2022-10-01 | 英業達股份有限公司 | Keyboard file verification method based on image processing |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20080073933A (en) * | 2007-02-07 | 2008-08-12 | 삼성전자주식회사 | Object tracking method and apparatus, and object pose information calculating method and apparatus |
| CN100458832C (en) * | 2007-06-21 | 2009-02-04 | 中国科学院合肥物质科学研究院 | Palm grain identification method based on direction character |
| CN104077569A (en) * | 2014-06-24 | 2014-10-01 | 纵横壹旅游科技（成都）有限公司 | Image recognizing method and system |
| CN104537376B (en) * | 2014-11-25 | 2018-04-27 | 深圳创维数字技术有限公司 | One kind identification platform calibration method and relevant device, system |
| CN104615676B (en) * | 2015-01-20 | 2018-08-24 | 同济大学 | One kind being based on the matched picture retrieval method of maximum similarity |
-
2015
- 2015-07-18 CN CN201510442276.XA patent/CN106354738B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106354738A (en) | 2017-01-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111126360B (en) | Cross-domain pedestrian re-identification method based on unsupervised combined multi-loss model | |
| US10095951B2 (en) | Describing objects using edge-pixel-feature descriptors | |
| Zamir et al. | Accurate image localization based on google maps street view | |
| Fan et al. | Unsupervised feature learning for land-use scene recognition | |
| CN102782708A (en) | Fast subspace projection of descriptor patches for image recognition | |
| CN108596010B (en) | Implementation method of pedestrian re-identification system | |
| JP5563494B2 (en) | Corresponding reference image search device and method, content superimposing device, system and method, and computer program | |
| CN102460475A (en) | Histogram methods and systems for object recognition | |
| KR20130142191A (en) | Robust feature matching for visual search | |
| Srivastava et al. | Image classification using SURF and bag of LBP features constructed by clustering with fixed centers | |
| Logoglu et al. | Cospair: colored histograms of spatial concentric surflet-pairs for 3d object recognition | |
| Soni et al. | Keypoints based enhanced multiple copy‐move forgeries detection system using density‐based spatial clustering of application with noise clustering algorithm | |
| CN106354738B (en) | A kind of flat image typing and matched method and system | |
| Ma | Improving SAR target recognition performance using multiple preprocessing techniques | |
| Tu et al. | Detection of damaged rooftop areas from high-resolution aerial images based on visual bag-of-words model | |
| CN110704667B (en) | Rapid similarity graph detection method based on semantic information | |
| Wang et al. | An efficient method to detect mutual overlap of a large set of unordered images for structure-from-motion | |
| Santos et al. | RECOGNIZING AND EXPLORING AZULEJOS ON HISTORIC BUILDINGS’FACADES BY COMBINING COMPUTER VISION AND GEOLOCATION IN MOBILE AUGMENTED REALITY APPLICATIONS | |
| Kumar et al. | Cross domain descriptor for sketch based image retrieval using siamese network | |
| Jalilvand et al. | CH-SIFT: A local kernel color histogram SIFT based descriptor | |
| Zeng et al. | Cloud detection of remote sensing images on Landsat-8 by deep learning | |
| Chen et al. | Geometric and non-linear radiometric distortion robust multimodal image matching via exploiting deep feature maps | |
| Fan et al. | Vegetation coverage detection from very high resolution satellite imagery | |
| Abadi et al. | Evaluation of feature points descriptors' performance for visual finger printing localization of smartphones | |
| Yang et al. | Aurora image search with a saliency-weighted region network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210107 Address after: No. 028, Zone E, 6 / F, Trading Plaza, No. 1 South China City, No. 1 South China Avenue, Pinghu street, Longgang District, Shenzhen City, Guangdong Province, 518000 Patentee after: Yan Zhihong Address before: No. 028, Zone E, 6 / F, Trading Plaza, No. 1 South China City, No. 1 South China Avenue, Pinghu street, Longgang District, Shenzhen City, Guangdong Province, 518000 Patentee before: Yan Zhihong Patentee before: Huang Zhisheng |