CN101458695A - Mixed picture index construct and enquiry method based on key word and content characteristic and use thereof - Google Patents
Mixed picture index construct and enquiry method based on key word and content characteristic and use thereof Download PDFInfo
- Publication number
- CN101458695A CN101458695A CNA2008101366221A CN200810136622A CN101458695A CN 101458695 A CN101458695 A CN 101458695A CN A2008101366221 A CNA2008101366221 A CN A2008101366221A CN 200810136622 A CN200810136622 A CN 200810136622A CN 101458695 A CN101458695 A CN 101458695A
- Authority
- CN
- China
- Prior art keywords
- picture
- index
- keyword
- user
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
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
- G06F16/58—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
- G06F16/5838—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using colour
Abstract
The invention discloses a picture index construction based on keywords and contents and corresponding searching method, wherein the index construction comprises: constructing the keywords inverted index based on keywords according to the picture describing explanation and the name describing explanation thereof; constructing the picture characteristic index based on content by extracting the picture characteristic vector. The searching method comprises: searching the keywords and the keywords set to perform mode matching to obtain a semantic-related picture set based on the searching request submitted by the user; searching in the picture characteristic vector index by the Hash method with characteristic vector sensitive on the applied position for sampling the picture characteristic to obtain a similar picture characteristic vector; returning the picture with high similarity based on the integrative result to the user. The invention can compromise the keyword index and the picture index, not only can use the keywords to improve the searching speed, but also can use the picture index to improve the searching result correlation degree, so that the precision ratio can be improved. The technical scheme can be applied to the searching field of the picture image.
Description
Technical field
The invention belongs to the picture retrieval field of information retrieval technique, related to a kind of mixing picture index construct and querying method and application thereof more specifically based on keyword and content.
Background technology
Current along with the rapid of network popularized, the widespread use of digital photographing, the quantity of picture has been the trend of explosive growth on the internet.Also usually there is a large amount of pictures to need to handle in fields such as space flight exploration, biomedicine, pharmacy.In the picture that like this quantity is extremely various, how can search fast and accurately, locate required picture and just become problem demanding prompt solution.
Traditional picture retrieval has been used the technology based on keyword that often is used in literature search.This technology is set up keyword index, so that find corresponding picture by keyword query for extraction and analysis that the filename and the explanatory note of picture carries out keyword.Main flow photographic search engine in the prior art, as, Google (Google), Baidu etc. adopt this technology.This technology implements fairly simple, and promptly the logical schema coupling by keyword obtains Query Result, and this method can be finished user's query demand more fast.But the drawback of this method also is conspicuous.The content of describing as the title of picture file or explanatory note and picture not directly, necessary relation.Inquire about the picture that obtains the demand frequent and user by this method and have a certain distance, promptly the Query Result and the user's request degree of correlation are lower.
In order to overcome the weak point of traditional picture retrieval based on keyword, rise in the content-based picture retrieval of the nineties in last century and paid attention to by people gradually.This content-based picture retrieval is not to adopt the explanatory note conduct of picture to set up the basis of index, but analyzes extraction for the feature of picture itself, and sets up index on the feature of picture itself.Can submit a width of cloth master drawing to system during user inquiring; System just can relatively return to the similar picture of user with the feature of building in the index again by master drawing being carried out feature extraction.The feature of the picture that extracts in this technology own comprises color, texture, graphics shape, pixel space relation of picture itself etc.Though this content retrieval based on picture feature is intended to overcome the deficiency of traditional search method based on keyword, has also brought a lot of new problems.
The first, though information such as color, texture, shape can objectively reflect the feature of picture, can not express the picture loaded information fully.
The second,, also just need more computing time to the analysis and the extraction of picture color, texture, shape facility and set up index and on the calculated amount of information, substantially exceed merely for the extraction of keyword.In the wait that usually needs the long period of searching, usually can not be fit to user's requirement at present for aspect indexing.
Have above-mentioned variety of problems and defective based on keyword and content-based picture retrieval technology in the prior art, the demand that can find a kind of picture retrieval technology to solve these problems, defective rapidly is of long duration.The present invention therefore.
Summary of the invention
Fundamental purpose of the present invention is to provide a kind of mixing picture index construct and querying method based on keyword and content, lower, the time-consuming defective such as very long of search efficiency during lower, the content-based picture indices of the Query Result and the user's request degree of correlation when having solved in the prior art picture indices based on keyword.
In order to address the above problem, technical scheme provided by the invention is as follows:
A kind of mixing picture index structuring method based on keyword and content may further comprise the steps:
Description explanation or its title with picture are illustrated as the keyword index of the structure on basis based on keyword;
Picture is carried out feature extraction; The picture feature that obtains with extraction makes up content-based picture feature index.
Preferably, the keyword index that makes up in the described method based on keyword is based on Lucene index engine structure; Its construction step comprises that the description of picture explanation or its title explanation full-text index are resolved into indexing key words sets up the indexing key words set; Described indexing key words set is deposited in the Lucene index.
Preferably, describedly make up keyword index based on keyword based on the Lucene index engine and also comprise the indexing key words that deposits in the Lucene index is set up inverted index.
Preferably, make up content-based picture feature index in the described method and be based on Lucene index engine structure; Described construction method comprises according to the different abstracting methods of picture feature handles the proper vector set that generation is made up of real number, and described proper vector set is deposited in the Lucene index.
Preferably, described Lucene index comprises described keyword index, picture feature index related with the picture file path respectively.
Preferably, describedly be associated as the chained list association.
Preferably, described feature extraction step comprises that color, texture, graphics shape, pixel space relation to picture extract; Described abstracting method comprises edge histogram (MPEG-7EdgeHistrogram) or auto color contingency table (Auto Colour Correlogram) method.
Another object of the present invention is to provide a kind of mixing picture inquiry method, may further comprise the steps based on keyword and content:
The user proposes to comprise the query demand of keyword and master drawing;
Keyword set in user's keyword and the Lucene index is carried out pattern match, calculates the keyword similarity that matching files is described explanation and user;
The master drawing that the user is provided obtains the set of master drawing proper vector by feature extraction, and the picture feature index in master drawing proper vector and the Lucene index is carried out distance relatively and calculate similarity;
The user is presented in comprehensive keyword matching result and the ordering of master drawing matching result.
Preferably, described Hash (LocalitySensitive Hashing) method apart from comparative approach employing position-based sensitivity compares.
Preferably, described user's request comprises keyword, master drawing and relevant weights thereof.
Technical solution of the present invention combines based on the picture retrieval of literal and content-based picture retrieval advantage separately, has not only used text index but also use characteristic index so that improve retrieval precision to greatest extent, satisfies user's needs.Index provided by the invention in addition preferably uses the Lucene index, and this index engine framework is the framework of an opening, can support the various features extraction algorithm.And (Locality Sensitive Hashing, LSH) technology make the time of searching be lower than pure linear search to the hash method of position-based sensitivity used in the present invention, and query responding time is short, real-time when making large-scale application.
In the technical scheme of this case invention, each picture file can have a textual description file of unique correspondence with it.Storing the text description and the explanation of corresponding picture in the text supporting paper.These textual description files can generate by Automatic Program, also can generate by the conclusion to information such as picture periphery literal in the process of network crawl.
The textual description file that the hybrid index algorithm meeting Treatment Analysis that adopts in this case invention in setting up the process of index is all also utilizes Full-text Indexing Technology that these textual description files are set up keyword index in full.
Meanwhile, feature extraction in mixed index part can be handled all picture files, utilizes single or the various features extracting method extracts picture feature.Feature extracting method can be edge histogram (MPEG-7 Edge Histogram) and auto color contingency table (Auto Colour Correlogram) method.Full-text index of setting up in text and the feature of extracting from picture all can be stored in unified Lucene index.Use by these methods makes user inquiring result's response time shorten greatly, has satisfied user's real-time requirement.
The advantage of this invention is that technical scheme provided by the invention can take into account the advantage of keyword index and picture indices, can utilize keyword retrieval speed fast, utilizes picture indices to improve the Query Result degree of correlation again, improves precision ratio.This technical scheme can be applied in the searching field of picture.
Description of drawings
Below in conjunction with drawings and Examples the present invention is further described:
Fig. 1 is the process flow diagram of the embodiment of the invention based on the open picture indices structure of keyword and image content;
Fig. 2 is the process flow diagram of application examples user's picture retrieval inquiry of the present invention.
Embodiment
For the technical scheme of more detailed statement foregoing invention, the following inventor lists specific embodiment and comes bright technique effect; It is emphasized that these embodiment are used to the present invention is described and are not limited to limit the scope of the invention.
Embodiment
Open picture indices based on keyword and image content makes up
Each picture file all has a textual description file of unique correspondence with it, the text description and the explanation of storing corresponding picture in the text supporting paper.These textual description files can generate by Automatic Program, also can generate by the conclusion to information such as picture periphery literal in the process of network crawl.
Come the picture in the presentation graphs valut to gather I={I with I
1, I
2, K, I
n, wherein n is the number of picture.D is the set with the corresponding description document of picture file, D={d
1, d
2, K, d
n, d wherein
iPicture I for correspondence
iDescription.Adopted k picture feature abstracting method in the supposing the system altogether.Represent the set of picture abstracting method, V={V with V
1, V
2, K, V
k, V wherein
j, j ∈ [1, k] is a kind of feature extraction method.Feature extraction method V
j, j ∈ [1, k] acts on picture I
i, the result of i ∈ [1..n] produces a proper vector of being made up of real number, uses
Represent.And, use
Represent picture I
iWith the set of the proper vector of generation after all k the feature extraction methods processing,
The step that index is set up is as follows:
For each picture I
iWith its corresponding description document d
i, with document d
iResolve into the set of indexing key words
Will
Deposit among the Lucene index L.
For each picture feature abstracting method V
j, j ∈ [1, k] calculates
Will
Deposit among the Lucene index L.
Deposited on the indexing key words of L in going into and to be set up inverted index.
The textual description file that the hybrid index algorithm meeting Treatment Analysis that adopts in setting up the process of index is all also utilizes Full-text Indexing Technology that these textual description files are set up keyword index in full.Meanwhile, feature extraction in mixed index part can be handled all picture files, utilizes single or the various features extracting method extracts picture feature.Feature extracting method can be edge histogram (MPEG-7Edge Histogram) and auto color contingency table (Auto Colour Correlogram) method.Full-text index of setting up in text and the feature of extracting from picture all can be stored in unified Lucene index.Use by these methods makes user inquiring result's response time shorten greatly, has satisfied user's real-time requirement.
Like this, by input picture set I and description document set D, deposited in
With
And
On set up the Lucene index L of inverted index.Whole based on keyword and image content, support quick, large-scale open picture retrieval framework just to make up to finish.
Set up corresponding user's picture retrieval inquiry with index
The user submits a searching keyword and a width of cloth master drawing to system in the process of inquiry.Use the matching inquiry algorithm in the Lucene full-text index, to find the textual description file of coupling according to user's keyword.Next from the Lucene index, read textual description file correspondence picture file feature and utilize Locality Sensitive Hashing technology to create the suitable data structure in the set of this feature, and compare and return similar result according to the feature that from master drawing, extracts.The result who returns has like this guaranteed not only close with searching of user on the semantic level but also close with searching of user on the feature aspect.
The query requests Q that the user submits in native system can represent Q=(T with one polynary group
q, I
q, w
t, W
V).Wherein, T
qBeing search key, can be single speech, also can be a plurality of speech, or the logical expression of keyword.I
qBe the user submit to search master drawing, the user wishes that system returns and I at last
qFeature similarity again and T
qSemantic relevant picture.w
tBeing the character search weights, is the importance degree that is used for assessing the result who obtains by keyword retrieval.Require 0≤w
t≤ 1.
W
VIt is a weights set.Weights wherein are the importance degrees that are used for assessing the result who finds by picture feature.
Wherein
Be suitable for and assess by feature extraction method V
j, the result's that j ∈ [1, k] retrieves importance degree.
And requirement
System promptly begins inquiry behind the submit queries request Q.Whole query script is divided into semantic query and characteristic query.
At semantic query stage, T
qIn search key be used to Lucene index L in added inverted index keyword relatively.According to the method for document retrieval vector space model, document d is described for each
iAnd T
q, calculate a similarity
D is described
iAnd T
qDegree of correlation.By abandoning the lower description document of similarity, can obtain the set D of the higher description document of similarity
SSuppose wherein to have s file, make D
S={ d
S1, d
S2, K, d
Ss.All Files in this set all is and T
qWhat semantically be correlated with.Pass through D
SIn document can obtain the set I of its corresponding picture file
SMake I
S={ I
S1, I
S2, K, I
Ss.So can think that this gathers I
SIn all pictures all relevant with semanteme.
In the characteristic query stage, calculate master drawing I
qUse V={V
1, V
2, K, V
kIn the feature extraction method extract vector set after the feature
And taking out the index establishment stage deposits among the Lucene index L
Here each
It all is the set of a vector
Use each then
With each set
In proper vector carry out distance relatively and calculate similarity
J ∈ [1, k], i ∈ [1, s] describes master drawing I
qWith picture I
SiAt feature extraction method V
jSimilarity under the j ∈ [1, k].The characteristic distance comparison can be used the technology of Locality Sensitive Hashing.
When the characteristic query stage finishes for each width of cloth at I
S={ I
S1, I
S2, K, I
SsIn picture I
SiI ∈ [1, s] has obtained following data:
A similarity
I is described
Si, i ∈ [1, s] and T
qIn semantically similar degree.A weight w
tAssess the importance degree of semantic similarity.
K similarity
I is described
qWith picture I
SiAt feature extraction method V
j, the similarity degree of j ∈ [1, k].K characteristic similarity weights
Assess the importance degree of characteristic similarity.
At this moment, for each at I
S={ I
S1, I
S2, K, I
SsIn picture I
Si, i ∈ [1, s] and upload master drawing I
qCan calculate one and mix similarity
I is described
qWith picture I
SiCharacteristic aspect and semantic aspect similarity.Computing method are as follows:
All similarities
Peaked result divided by similarity is final similarity, uses
Represent.All I
S={ I
S1, I
S2, K, I
SsIn the picture basis
Sort.Abandon the lower picture of similarity, remaining picture is the lookup result that returns to the user.
Like this, use T during user inquiring
qRelatively, finally obtain the set I of semantic picture of being correlated with in the keyword inverted index of in Lucene index L, setting up
S, I
S={ I
S1, I
S2, K, I
Ss.
Calculate master drawing I
qAt feature extraction method set V={V
1, V
2, K, V
kIn the effect of feature extraction method under proper vector set
For gathering V={V in the feature extraction method
1, V
2, K, V
kIn feature extraction method V
j, j ∈ [1, k], proper vector set
Use Locality Sensitive Hashing technology and set up the Hash structure.Use the master drawing proper vector
On the Hash structure of having set up, make distance relatively, obtain similarity
For master drawing I
qWith each I
Si, i ∈ [1, s] calculates
The result according to
Sort, and return to the user.
To sum up, user input query request Q, by the Lucene index L that the index establishment stage is set up, can obtain with query requests Q semantically with feature on the output result of all relevant picture, and sort by degree of correlation.
Adopted Locality Sensitive Hashing technology among the present invention, a kind of Hash table technology of position-based sensitivity can realize high position data is searched fast.Because the proper vector of extracting from picture is usually up to the hundreds of dimension, traditional tree-shaped index structure (as the B tree) can not obtain to be lower than the speed of linear search usually.
Comparing with traditional picture retrieval based on literal has increased for the picture internal characteristics, as the analysis and the comparison of color, texture, shape, makes the user can find the picture of feature similarity by master drawing, has enriched the inquiry means.
Compare index and the inquiry that has added based on literal with pure content-based picture retrieval, alleviated simple dependence, thereby increased the accuracy rate of returning picture for picture feature.Make inquiry velocity be greatly improved owing to use Locality Sensitive Hashing technology, for extensive, real-time application provides the basis of widely applying.
Test case
Pictures have amounted to 1511 JPEG pictures.Each picture file all has the corresponding text description document.Index construct program and polling routine all adopt C Plus Plus to write, and optimize option with g++4.3.2 compiling and unlatching-O3.Executable file after the compiling moves on the machine of operation Linux (Kernel2.6.27-generic, 64).Routine call C Lucene functional packet (http://sourceforge.net/projects/clucene/) and E
2The LSH0.1 functional packet (function that provides among the http://www.mit.edu/~andoni/LSH/), and for E
2Code among the LSH0.1 has been done modification, has saved and has wherein estimated k, m, the part of L parameter value and use instead at the beginning of program run and to be k, m, the method for L parameter setting fixed value.(about k, m, the implication of L parameter is referring to E
2Among the LSH0.1 with user manual.)
The machine that is used to test is equipped with Intel Core 2 Cuo P8600@2.4GHz CPU, although have only a core to move in test.The main memory of machine has 4GB, but having only 1GB to be set to polling routine can use.
The nearly clock more than 10 of the process of whole index building.The proper vector of picture, edge histogram (MPEG-7 Edge Histogram) and auto color contingency table (Auto ColourCorrelogram), and the description document keyword is all deposited in the Lucene index.And on keyword, set up inverted index.
The inquiry test has 13 groups.Each group test subscriber submits searching keyword to, master drawing and correlation parameter.Query requests Q each time
j, j ∈ [1,13] represents.Master drawing in 13 group pollings and correlation parameter are all just the same.The radius of searching for the Hash (Locality SensitiveHashing) of the position sensing of edge histogram (MPEG-7 Edge Histogram) and auto color contingency table (Auto Colour Correlogram) all is set to 1.0.Every group of keyword difference of selecting for use in testing.Having 6 groups, what select for use is single keyword, uses t
i, i ∈ [1,6] represents.Remaining 7 groups is the logical OR expression formula of keyword.Write down following data in each group test:
| I
S| the quantity of the picture that the semanteme that retrieves by keyword query is relevant.
T
tThe time that the key word semantic query is used.
The pretreated time that is used for the Hash (Locality Sensitive Hashing) of the position sensing of edge histogram (MPEG-7Edge Histogram).
The query time of Hash (Locality Sensitive Hashing) that is used for the position sensing of edge histogram (MPEG-7 Edge Histogram).
The pretreated time of Hash (Locality Sensitive Hashing) that is used for the position sensing of auto color contingency table (Auto Colour Correlogram).
The query time of Hash (Locality Sensitive Hashing) that is used for the position sensing of auto color contingency table (Auto Colour Correlogram).
T
T+vThe summation that is used for semantic query and picture feature query time.Approximate T
t, T
vAnd.
Wherein all time all uses millisecond (ms) to measure, i.e. thousand minutes one second.Test data is summarized in following table 1.
Table 1 test data
Even if the data that test obtains from table the longest inquiry of time spent as can be seen also only need finish about 6 minutes one second, can reach in real time requirement fast basically.
Above-mentioned example only is explanation technical conceive of the present invention and characteristics, and its purpose is to allow the people who is familiar with this technology can understand content of the present invention and enforcement according to this, can not limit protection scope of the present invention with this.All equivalent transformations that spirit is done according to the present invention or modification all should be encompassed within protection scope of the present invention.
Claims (10)
1. mixing picture index structuring method based on keyword and content may further comprise the steps:
Description explanation or its title with picture are illustrated as the keyword index of the structure on basis based on keyword;
Picture is carried out feature extraction; The picture feature that obtains with extraction makes up content-based picture feature index.
2, the mixing picture index structuring method based on keyword and content according to claim 1 is characterized in that the keyword index that makes up in the described method based on keyword is based on Lucene index engine structure; Its construction step comprises that the description of picture explanation or its title explanation full-text index are resolved into indexing key words sets up the indexing key words set; Described indexing key words set is deposited in the Lucene index.
3, the mixing picture index structuring method based on keyword and content according to claim 2 is characterized in that describedly making up keyword index based on keyword based on the Lucene index engine and also comprising the indexing key words that deposits in the Lucene index is set up inverted index.
4, the mixing picture index structuring method based on keyword and content according to claim 1 is characterized in that making up in the described method content-based picture feature index and is based on Lucene index engine structure; Described construction method comprises according to the different abstracting methods of picture feature handles the proper vector set that generation is made up of real number, and described proper vector set is deposited in the Lucene index.
5,, it is characterized in that described Lucene index comprises described keyword index, picture feature index related with the picture file path respectively according to claim 2 or 4 described mixing picture index structuring methods based on keyword and content.
6, the mixing picture index structuring method based on keyword and content according to claim 5 is characterized in that the described chained list association that is associated as.
7, the mixing picture index structuring method based on keyword and content according to claim 1 is characterized in that described feature extraction step comprises that color, texture, graphics shape, pixel space relation to picture extract; Described abstracting method comprises edge histogram or auto color contingency table method.
8, a kind of mixing picture inquiry method based on keyword and content may further comprise the steps:
The user proposes to comprise the query demand of keyword and master drawing;
Keyword set in user's keyword and the Lucene index is carried out pattern match, calculates the keyword similarity that matching files is described explanation and user;
The master drawing that the user is provided obtains the set of master drawing proper vector by feature extraction, and the picture feature index in master drawing proper vector and the Lucene index is carried out distance relatively and calculate similarity;
The user is presented in comprehensive keyword matching result and the ordering of master drawing matching result.
9, the mixing picture inquiry method based on keyword and content according to claim 8 is characterized in that described hash method apart from comparative approach employing position-based sensitivity compares.
10, the mixing picture inquiry method based on keyword and content according to claim 8 is characterized in that described user's request comprises keyword, master drawing and relevant weights thereof.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2008101366221A CN101458695A (en) | 2008-12-18 | 2008-12-18 | Mixed picture index construct and enquiry method based on key word and content characteristic and use thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2008101366221A CN101458695A (en) | 2008-12-18 | 2008-12-18 | Mixed picture index construct and enquiry method based on key word and content characteristic and use thereof |
Publications (1)
Publication Number | Publication Date |
---|---|
CN101458695A true CN101458695A (en) | 2009-06-17 |
Family
ID=40769560
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA2008101366221A Pending CN101458695A (en) | 2008-12-18 | 2008-12-18 | Mixed picture index construct and enquiry method based on key word and content characteristic and use thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101458695A (en) |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102117324A (en) * | 2011-02-24 | 2011-07-06 | 上海北大方正科技电脑系统有限公司 | File management method and management system applying fuzzy matrice |
CN102262670A (en) * | 2011-07-29 | 2011-11-30 | 中山大学 | Cross-media information retrieval system and method based on mobile visual equipment |
CN102411582A (en) * | 2010-09-21 | 2012-04-11 | 腾讯科技(深圳)有限公司 | Picture searching method, device and client |
CN102693231A (en) * | 2011-03-23 | 2012-09-26 | 百度在线网络技术(北京)有限公司 | Method, device and device for confirming atlas according to images from network |
CN103136228A (en) * | 2011-11-25 | 2013-06-05 | 阿里巴巴集团控股有限公司 | Image search method and image search device |
CN103246701A (en) * | 2013-04-02 | 2013-08-14 | 百度在线网络技术(北京)有限公司 | Quick search method, system and device |
CN103440261A (en) * | 2013-07-31 | 2013-12-11 | 东莞中山大学研究院 | System and method for searching biomedical flow chart basing on content and structure |
CN103714096A (en) * | 2012-10-09 | 2014-04-09 | 阿里巴巴集团控股有限公司 | Lucene-based inverted index system construction method and device, and Lucene-based inverted index system data processing method and device |
CN103744860A (en) * | 2013-12-12 | 2014-04-23 | 中国科学院深圳先进技术研究院 | Method, device and mobile phone for searching for picture |
CN103902595A (en) * | 2012-12-27 | 2014-07-02 | 北京新媒传信科技有限公司 | Method and device for establishing database |
CN103955514A (en) * | 2014-05-05 | 2014-07-30 | 陈浩 | Image feature indexing method based on Lucene inverted index |
CN104050204A (en) * | 2013-03-11 | 2014-09-17 | 宏达国际电子股份有限公司 | Method of combining network data and mobile device using the same |
CN104239535A (en) * | 2014-09-22 | 2014-12-24 | 重庆邮电大学 | Method and system for matching pictures with characters, server and terminal |
CN104424257A (en) * | 2013-08-28 | 2015-03-18 | 北大方正集团有限公司 | Information indexing unit and information indexing method |
CN106407450A (en) * | 2016-09-30 | 2017-02-15 | 网易(杭州)网络有限公司 | File searching method and apparatus |
CN106649545A (en) * | 2016-11-03 | 2017-05-10 | 广州凯耀资产管理有限公司 | Retrieval method and retrieval server for traffic video |
CN106708940A (en) * | 2016-11-11 | 2017-05-24 | 百度在线网络技术(北京)有限公司 | Method and device used for processing pictures |
CN108027816A (en) * | 2015-10-28 | 2018-05-11 | 株式会社东芝 | Data management system, data managing method and program |
CN108124478A (en) * | 2017-12-05 | 2018-06-05 | 深圳前海达闼云端智能科技有限公司 | Picture searching method and apparatus |
CN108416026A (en) * | 2018-03-09 | 2018-08-17 | 腾讯科技(深圳)有限公司 | Index generation method, content search method, device and equipment |
CN109388649A (en) * | 2018-09-28 | 2019-02-26 | 湖南土流信息有限公司 | Intelligent land recommendation method and system |
CN109446363A (en) * | 2018-10-22 | 2019-03-08 | 镇江悦乐网络科技有限公司 | A kind of method for quickly retrieving based on three-dimensional CG animation |
CN110046062A (en) * | 2019-03-07 | 2019-07-23 | 佳都新太科技股份有限公司 | Distributed data processing method and system |
CN110390030A (en) * | 2019-06-28 | 2019-10-29 | 中山大学 | The storage method and device of pictorial information |
CN110399515A (en) * | 2019-06-28 | 2019-11-01 | 中山大学 | Picture retrieval method, device and picture retrieval system |
CN110674329A (en) * | 2019-09-23 | 2020-01-10 | 北京百度网讯科技有限公司 | Picture filtering method and device, electronic equipment and storage medium |
CN110737794A (en) * | 2019-10-16 | 2020-01-31 | 北京锐安科技有限公司 | Image query method, system, server and storage medium |
CN111460088A (en) * | 2019-01-22 | 2020-07-28 | 阿里巴巴集团控股有限公司 | Similar text retrieval method, device and system |
CN112148831A (en) * | 2020-11-26 | 2020-12-29 | 广州华多网络科技有限公司 | Image-text mixed retrieval method and device, storage medium and computer equipment |
CN112836069A (en) * | 2021-04-22 | 2021-05-25 | 泰德网聚(北京)科技股份有限公司 | Self-recognition retrieval system based on image keywords |
WO2021160100A1 (en) * | 2020-02-12 | 2021-08-19 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Methods for searching images and for indexing images, and electronic device |
-
2008
- 2008-12-18 CN CNA2008101366221A patent/CN101458695A/en active Pending
Cited By (45)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102411582A (en) * | 2010-09-21 | 2012-04-11 | 腾讯科技(深圳)有限公司 | Picture searching method, device and client |
CN102411582B (en) * | 2010-09-21 | 2016-04-27 | 腾讯科技(深圳)有限公司 | Image searching method, device and client |
CN102117324B (en) * | 2011-02-24 | 2012-09-05 | 上海北大方正科技电脑系统有限公司 | File management method and management system applying fuzzy matrice |
CN102117324A (en) * | 2011-02-24 | 2011-07-06 | 上海北大方正科技电脑系统有限公司 | File management method and management system applying fuzzy matrice |
CN102693231B (en) * | 2011-03-23 | 2019-07-26 | 百度在线网络技术(北京)有限公司 | For determining the method, apparatus and equipment of atlas according to the image for carrying out automatic network |
CN102693231A (en) * | 2011-03-23 | 2012-09-26 | 百度在线网络技术(北京)有限公司 | Method, device and device for confirming atlas according to images from network |
CN102262670A (en) * | 2011-07-29 | 2011-11-30 | 中山大学 | Cross-media information retrieval system and method based on mobile visual equipment |
CN103136228A (en) * | 2011-11-25 | 2013-06-05 | 阿里巴巴集团控股有限公司 | Image search method and image search device |
US9460117B2 (en) | 2011-11-25 | 2016-10-04 | Alibaba Group Holding Limited | Image searching |
CN103714096A (en) * | 2012-10-09 | 2014-04-09 | 阿里巴巴集团控股有限公司 | Lucene-based inverted index system construction method and device, and Lucene-based inverted index system data processing method and device |
CN103714096B (en) * | 2012-10-09 | 2018-02-13 | 阿里巴巴集团控股有限公司 | Inverted index system constructing, data processing method and device based on Lucene |
CN103902595A (en) * | 2012-12-27 | 2014-07-02 | 北京新媒传信科技有限公司 | Method and device for establishing database |
CN103902595B (en) * | 2012-12-27 | 2017-02-08 | 北京新媒传信科技有限公司 | Method and device for establishing database |
CN104050204A (en) * | 2013-03-11 | 2014-09-17 | 宏达国际电子股份有限公司 | Method of combining network data and mobile device using the same |
CN104050204B (en) * | 2013-03-11 | 2019-02-22 | 宏达国际电子股份有限公司 | Merge the method and mobile device of network data |
CN103246701A (en) * | 2013-04-02 | 2013-08-14 | 百度在线网络技术(北京)有限公司 | Quick search method, system and device |
CN103440261A (en) * | 2013-07-31 | 2013-12-11 | 东莞中山大学研究院 | System and method for searching biomedical flow chart basing on content and structure |
CN104424257A (en) * | 2013-08-28 | 2015-03-18 | 北大方正集团有限公司 | Information indexing unit and information indexing method |
CN103744860A (en) * | 2013-12-12 | 2014-04-23 | 中国科学院深圳先进技术研究院 | Method, device and mobile phone for searching for picture |
CN103955514A (en) * | 2014-05-05 | 2014-07-30 | 陈浩 | Image feature indexing method based on Lucene inverted index |
CN104239535B (en) * | 2014-09-22 | 2018-04-27 | 重庆邮电大学 | A kind of method, server, terminal and system for word figure |
CN104239535A (en) * | 2014-09-22 | 2014-12-24 | 重庆邮电大学 | Method and system for matching pictures with characters, server and terminal |
CN108027816A (en) * | 2015-10-28 | 2018-05-11 | 株式会社东芝 | Data management system, data managing method and program |
CN108027816B (en) * | 2015-10-28 | 2021-10-26 | 株式会社东芝 | Data management system, data management method, and recording medium |
CN106407450A (en) * | 2016-09-30 | 2017-02-15 | 网易(杭州)网络有限公司 | File searching method and apparatus |
CN106649545A (en) * | 2016-11-03 | 2017-05-10 | 广州凯耀资产管理有限公司 | Retrieval method and retrieval server for traffic video |
CN106708940B (en) * | 2016-11-11 | 2020-06-30 | 百度在线网络技术(北京)有限公司 | Method and device for processing pictures |
CN106708940A (en) * | 2016-11-11 | 2017-05-24 | 百度在线网络技术(北京)有限公司 | Method and device used for processing pictures |
CN108124478A (en) * | 2017-12-05 | 2018-06-05 | 深圳前海达闼云端智能科技有限公司 | Picture searching method and apparatus |
CN108416026A (en) * | 2018-03-09 | 2018-08-17 | 腾讯科技(深圳)有限公司 | Index generation method, content search method, device and equipment |
CN109388649A (en) * | 2018-09-28 | 2019-02-26 | 湖南土流信息有限公司 | Intelligent land recommendation method and system |
CN109446363A (en) * | 2018-10-22 | 2019-03-08 | 镇江悦乐网络科技有限公司 | A kind of method for quickly retrieving based on three-dimensional CG animation |
CN111460088A (en) * | 2019-01-22 | 2020-07-28 | 阿里巴巴集团控股有限公司 | Similar text retrieval method, device and system |
CN110046062A (en) * | 2019-03-07 | 2019-07-23 | 佳都新太科技股份有限公司 | Distributed data processing method and system |
CN110046062B (en) * | 2019-03-07 | 2021-03-19 | 佳都新太科技股份有限公司 | Distributed data processing method and system |
CN110390030A (en) * | 2019-06-28 | 2019-10-29 | 中山大学 | The storage method and device of pictorial information |
CN110399515A (en) * | 2019-06-28 | 2019-11-01 | 中山大学 | Picture retrieval method, device and picture retrieval system |
CN110399515B (en) * | 2019-06-28 | 2022-05-17 | 中山大学 | Picture retrieval method, device and system |
CN110674329A (en) * | 2019-09-23 | 2020-01-10 | 北京百度网讯科技有限公司 | Picture filtering method and device, electronic equipment and storage medium |
CN110737794A (en) * | 2019-10-16 | 2020-01-31 | 北京锐安科技有限公司 | Image query method, system, server and storage medium |
WO2021160100A1 (en) * | 2020-02-12 | 2021-08-19 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Methods for searching images and for indexing images, and electronic device |
CN112148831B (en) * | 2020-11-26 | 2021-03-19 | 广州华多网络科技有限公司 | Image-text mixed retrieval method and device, storage medium and computer equipment |
CN112148831A (en) * | 2020-11-26 | 2020-12-29 | 广州华多网络科技有限公司 | Image-text mixed retrieval method and device, storage medium and computer equipment |
CN112836069A (en) * | 2021-04-22 | 2021-05-25 | 泰德网聚(北京)科技股份有限公司 | Self-recognition retrieval system based on image keywords |
CN112836069B (en) * | 2021-04-22 | 2021-09-10 | 泰德网聚(北京)科技股份有限公司 | Self-recognition retrieval system based on image keywords |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101458695A (en) | Mixed picture index construct and enquiry method based on key word and content characteristic and use thereof | |
US8489589B2 (en) | Visual search reranking | |
CN105393263B (en) | Feature in compuman's interactive learning is completed | |
AU2011210535B2 (en) | Joint embedding for item association | |
CN102243647B (en) | Higher-order knowledge is extracted from structural data | |
CN104991905B (en) | A kind of mathematic(al) representation search method based on level index | |
CN104834693A (en) | Depth-search-based visual image searching method and system thereof | |
CN101398832A (en) | Image searching method and system by utilizing human face detection | |
CN107103016A (en) | Represent to make the method for image and content matching based on keyword | |
CN105426529A (en) | Image retrieval method and system based on user search intention positioning | |
CN101393565A (en) | Facing virtual museum searching method based on noumenon | |
CN111522905A (en) | Document searching method and device based on database | |
Kennedy et al. | Query-adaptive fusion for multimodal search | |
CN111061828B (en) | Digital library knowledge retrieval method and device | |
CN101957825A (en) | Method for searching image based on image and video content in webpage | |
CN111061954B (en) | Search result sorting method and device and storage medium | |
CN102314464B (en) | Lyrics searching method and lyrics searching engine | |
Radu et al. | A hybrid machine-crowd approach to photo retrieval result diversification | |
Fu et al. | Enhancing Semantic Search of Crowdsourcing IT Services using Knowledge Graph. | |
CN102222067A (en) | Searching method for accurately querying information according to IP (Internet Protocol) address of keyword | |
CN111625570B (en) | List data resource retrieval method and device | |
Bolettieri et al. | An image retrieval system for video | |
Mitsui et al. | Query generation as result aggregation for knowledge representation | |
Springmann et al. | QbS: searching for known images using user-drawn sketches | |
Budikova et al. | Multi-modal image retrieval for search-based image annotation with RF |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20090617 |