CN101404032A - Video retrieval method and system based on contents - Google Patents
Video retrieval method and system based on contents Download PDFInfo
- Publication number
- CN101404032A CN101404032A CNA2008102262681A CN200810226268A CN101404032A CN 101404032 A CN101404032 A CN 101404032A CN A2008102262681 A CNA2008102262681 A CN A2008102262681A CN 200810226268 A CN200810226268 A CN 200810226268A CN 101404032 A CN101404032 A CN 101404032A
- Authority
- CN
- China
- Prior art keywords
- proper vector
- hash
- video
- video frequency
- frame
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
The invention discloses a video retrieval method based on contents and a system thereof which are used for solving the problem of lower efficiency of the video retrieval based on contents. The method comprises the following steps: a characteristic vector of a video frame to be detected is acquired; a corresponding cartesian coordinate is searched according to the index number of the characteristic vector of a sensitive video frame to be compared in a video fingerprint database and a predefined transformation rule; the characteristic vector of the video frame to be detected is sent to a retrieval module of a corresponding server according to the searched cartesian coordinate; and the retrieval module judges the similarity between the characteristic vector of the video frame to be detected and the characteristic vector of the sensitive video frame to be compared. As a content addressable network is employed to reasonably organize the video fingerprint database, create indexes, and acquire the video example to be detected, the method utilizes a certain retrieval algorithm to search the best matched characteristic vector of the sensitive video frame from the fingerprint database, which improves the video retrieval efficiency.
Description
Technical field
The invention belongs to the multimedia application field, particularly a kind of Content-based Video Retrieval method and system.
Background technology
Content-based Video Retrieval (CBID) is can be according to the technology of the quick search video object of video content.By feature extraction object video is mapped as point (promptly extracting video finger print) in the high dimension vector space, so just the arest neighbors that the similarity searching problem of object video is converted in the higher dimensional space searches problem.
For mass data, how setting up effective index structure is the key issue that improves retrieval precision and efficient.Popular multi-dimensional indexing technology comprises gridfile, k-d-B tree, quaternary tree, hB tree, R tree and mutation R+ tree and R* tree etc. now, these all are based on the space or based on the division methods of DATA DISTRIBUTION, have good performance under the situation of dimension not too high (below 10 dimensions).
The multi-dimensional indexing technology comprises that also some are suitable for the more indexing means of higher-dimension, as vector approximation method (VA-file), LSH (Locality Sensitive Hashing, local sensitivity Hash table), space filling curve (space-filling curve) etc.
Adopting which kind of multi-dimensional indexing technology in the prior art all is to finish relevant treatment in a station server, and the efficient of Content-based Video Retrieval is lower like this.
Summary of the invention
For the lower problem of efficient that solves Content-based Video Retrieval, the embodiment of the invention provides a kind of Content-based Video Retrieval method, this method is applied to content addressed network, is provided with a plurality of servers on the node in content addressed network cartesian coordinate space, comprising:
Acquisition module obtains the proper vector of frame of video to be detected;
Search the call number of module according to sensitive video frequency frame proper vector to be compared in the video finger print storehouse, and predefined transformation rule, from content addressed network, find the Cartesian coordinates that correspondence is preserved the server of sensitive video frequency frame proper vector to be compared in the Cartesian coordinates of each server;
Sending module sends to the proper vector of frame of video to be detected the retrieval module of corresponding with service device according to the Cartesian coordinates that finds;
Retrieval module is judged the proper vector of frame of video to be detected and the similarity of sensitive video frequency frame proper vector to be compared, and determines the sensitive video frequency proper vector of coupling.
Simultaneously the embodiment of the invention also provides a kind of Content-based Video Retrieval system, comprising:
The video finger print storehouse: be used to preserve the sensitive video frequency frame proper vector with call number, described video finger print storehouse is evenly distributed in a plurality of servers on the content addressed network cartesian coordinate space node;
Acquisition module: the proper vector that is used to obtain frame of video to be detected;
Search module: be used for call number according to video finger print storehouse sensitive video frequency frame to be compared proper vector, and predefined transformation rule, from the Cartesian coordinates of each server, find the Cartesian coordinates that correspondence is preserved the server of sensitive video frequency frame proper vector to be compared;
Sending module: be used for the proper vector of frame of video to be detected being sent to the retrieval module of corresponding with service device according to the Cartesian coordinates that finds;
Retrieval module: be used to judge the proper vector of frame of video to be detected and the similarity of sensitive video frequency frame proper vector to be compared, and determine the sensitive video frequency proper vector of coupling.
The specific embodiments that is provided by the invention described above as can be seen, just because of application content addressing network rationalization is carried out in the video finger print storehouse, set up index, after obtaining video example to be detected, adopt certain searching algorithm from fingerprint base, to search the sensitive video frequency proper vector of mating most, make the efficient of video frequency searching be improved.
Description of drawings
Fig. 1 is the first embodiment method flow diagram provided by the invention;
Fig. 2 is a content addressed network diagram provided by the invention;
Fig. 3 is the second embodiment system construction drawing provided by the invention.
Embodiment
For the lower problem of efficient that solves Content-based Video Retrieval, the embodiment of the invention provides a kind of Content-based Video Retrieval method, to improve recall precision, adapts to the needs of large scale network video frequency searching.When the user imports video example to be measured, adopt certain searching algorithm from fingerprint base, to search the sensitive video frequency frame and the corresponding sensitive video frequency segment of coupling, and pairing similar sensitive video frequency fragment is returned to the user.The video finger print here is meant the frame of video proper vector of extracting from original video data, can represent the content of this video.
Wherein, the vector approximation method can solve the problem of accurate arest neighbors retrieval, and additive method is then only at approximate arest neighbors retrieval.Because video finger print itself is exactly the approximate representation of video content, the arest neighbors of spatial signature vectors does not also mean that arest neighbors on the video content, so even accurately arest neighbors retrieval does not guarantee to obtain Query Result the most accurately yet.And, under many circumstances, select suitable approximate query algorithm can return and the accurate identical result of search algorithm, and have higher efficient.What video frequency searching needed is the balance of a precision and efficient.Particularly in big, higher to the response time requirement occasion of data scale, approximate arest neighbors retrieval will be brought into play more importantly effect, therefore adopt LSH (LocalitySensitive Hashing) algorithm as preferred version.
The LSH algorithm is at first proposed by Indyk and Motwani, utilizes statistical theory, can solve the k-neighbour fast and inquire about problem under the prerequisite that guarantees certain accuracy (with probabilistic manner).Paper " SimilaritySearch in High Dimensions via Hashing " has provided the specific implementation step of this algorithm, its basic thought is, for the point data collection, utilize one group of hash function to set up a plurality of Hash tables with certain constraint condition, make under certain similarity measure condition, the probability that similar point clashes is bigger, and the probability that dissimilar point clashes is less relatively.
First embodiment provided by the invention is a kind of Content-based Video Retrieval method, adopts the LSH algorithm to carry out index in the present embodiment and sets up, and the LSH function definition is: one group of hash function H={h1, ..., hm}, m are positive integer, for data point p, if q is p, distance D (p between the q, q)<R, then P[hi (q)=hi (p)]>P1, if D (p, q)>cR, then P[hi (q)=hi (p)]<P2.Wherein function P (.) is a probability function, and P1, P2 are given probability, and P1>P2, i are random number, i ∈ 1 ..., m}.This group hash function is called as so that (P1 P2) is the LSH group of functions of parameter for R, cR.Data point p wherein, different sensitive video frequency frame proper vector in the corresponding present embodiment of q
The LSH function is:
Wherein vectorial
Satisfy normal distribution (Gaussian distribution), w is any real number, and b is any real number between [0, w].Method flow comprises as shown in Figure 1:
Step 101: n sensitive video frequency frame proper vector is mapped among L the Hash table g.
N sensitive video frequency frame proper vector arranged in the video finger print storehouse
Adopt the LSH algorithm to carry out index and set up, use L hash function g (.) n sensitive video frequency frame proper vector
Be mapped among L the Hash table g, for example: 10 sensitive video frequency frame proper vectors are arranged
With
Wherein
With
Be mapped among the Hash table g1 all the other 5
With
Be mapped among the Hash table g2, adopt different hash function g (.) to be mapped to the sensitive video frequency frame proper vector of Hash table
Can be different, as
With
Be mapped among the Hash table g1 all the other 5
With
Be mapped among the Hash table g2.
Step 102: the sensitive video frequency frame proper vector among each Hash table g is carried out hash by the LSH function, and the gained result is carried out the secondary hash again, and the sensitive video frequency frame proper vector in each Hash table is mapped in a plurality of hash buckets.
Hash table gj=[h1 wherein
(j)..., hk
(j)] (j=1,2 ..., L), hi (.) ∈ H (i=2 ... k), wherein H is a LSH family of functions, i.e. one group of hash function H={h1 ..., hm}.As in the g1 Hash table
With
Carry out the secondary hash, will
With
Be mapped in 7 hash buckets.Hash table g1=[h1
(1), h2
(1), h3
(1), h4
(1)] expression employing h1
(1), h2
(1), h3
(1), h4
(1)Function will
With
Be mapped in 7 hash buckets.As pass through h1
(1)Will
With
Be mapped to first hash bucket, will
With
Be mapped to second hash bucket, pass through h2
(1)Will
With
Be mapped to the 3rd hash bucket, pass through h3
(1)Will
With
Be mapped to the 4th hash bucket, will
With
Be mapped to the 5th hash bucket, pass through h4
(1)Will
With
Be mapped to the 6th hash bucket, will
With
Be mapped to the 7th hash bucket.
Step 103:, determine the Cartesian coordinates of L Hash table numbering correspondence in content addressed network according to the numbering and the predefined transformation rule of L Hash table.
The number table of Hash table g1 is shown binary sequence x, as 0100001010.If
Wherein d is the dimension such as the d=2 of Virtual Space, and m is the figure place 10 of binary sequence x.X is divided into groups to a high position from low level, and per 8 is one group, and being divided into is 2 groups (last group can be discontented with 8), and first group is 00001010, the second group is 10, the dimension of 2 in the corresponding Virtual Space.Calculate every group decimal value xi (i=1 ..., d), first group is 10, the second groups is 1, xi%2
dBe the i dimension coordinate of pairing node, promptly the Cartesian coordinates of 0100001010 correspondence is (2,1).
Adopt the call number of the numbering of Hash table g in the present embodiment as sensitive video frequency frame proper vector.
Step 104: according to the corresponding relation of numbering and the Cartesian coordinates of Hash table g, L Hash table (including the sensitive video frequency frame proper vector that is mapped to wherein) be distributed in N the server in the content addressed network preserve, each server all has corresponding Cartesian coordinates, wherein N≤L in content addressed network.
The Cartesian coordinates that No. 8 servers (Hash table is numbered 0100001010) are corresponding is (2,1), preserves Hash table g1 in this server.Wherein have Cartesian coordinates each server content addressed network diagram as shown in Figure 2.
Step 105: the acquisition module of arbitrary server obtains the proper vector of frame of video to be detected in the content addressed network
One section video to be detected is arranged from network, and No. 4 servers therefrom obtain the proper vector of the frame of video to be detected of this video in the content addressed network
Step 106: obtain
Server search module, determine the Cartesian coordinates of this Hash table place server successively according to the numbering of sensitive video frequency frame proper vector to be compared place Hash table.
As sensitive video frequency frame proper vector to be compared in Hash table g1, No. 4 servers search the Cartesian coordinates that module at first will be determined No. 8 servers in Hash table g1 place, determine that according to the numbering of Hash table g1 the Cartesian coordinates of No. 8 server correspondences is (2,1).
Step 107: obtain
Server sending module will
Send to the retrieval module of the server of storage sensitive video frequency frame proper vector to be compared place Hash table.
The Cartesian coordinates of No. 4 server correspondences is (1,1), is that (2,1) are adjacent because Cartesian coordinates is (1,1) and Cartesian coordinates, and the sending module of No. 4 servers directly will
Send to the retrieval module of No. 8 servers.
If acquisition module obtains the proper vector of frame of video to be detected
Server be not No. 4 servers but No. 8 servers, then the sending module of No. 8 servers directly sends
Retrieval module to book server.
Step 108: the retrieval module of server of storing sensitive video frequency frame proper vector to be compared place Hash table is right
Carry out the secondary hash, it is mapped in the hash bucket of Hash table.
Retrieval module is right
Carrying out the secondary hash is mapped to it in first hash bucket of Hash table g1.
Step 109: take out with
Fall into the sensitive video frequency frame proper vector of same hash bucket, and calculate wherein each sensitive video frequency frame proper vector and
Euclidean distance, judge similarity each other, and determine the sensitive video frequency proper vector of coupling.
In first hash bucket of taking-up Hash table g1
With
And calculate respectively with
Euclidean distance, judge similarity each other, and determine
For with
(similar) sensitive video frequency proper vector of coupling.Up to obtaining abundant similar sensitive video frequency frame proper vector, or relatively finish with whole sensitive video frequency frame proper vectors.
Wherein in the step 102 in order to guarantee the performance of LSH algorithm, need to consider two important parameters here---the number k of LSH function h (.) in the number L of Hash table g and each Hash table.The value of L and k can directly have influence on the performance of this algorithm.Consider following performance index: index Time Created: O (nLkt), wherein t is for calculating required time of each h (.), space: O (the nL)+required space of preservation data point, query time: O (L (kt+dnP
2 k)), should guarantee that L and k have following relation:
Wherein P1 is a given probability in the LSH function as previously mentioned.
Also can adopt in the present embodiment as the multi-dimensional indexing technology as: gridfile, k-d-B tree, quaternary tree, hB tree, R tree and mutation R+ tree and R* tree etc., these all are based on the space or based on the division methods of DATA DISTRIBUTION, generate the index of each sensitive video frequency frame proper vector in the video finger print storehouse by said method, sensitive video frequency frame proper vector with different index number (as 1-10000), be distributed in 10 servers in the content addressed network and preserve, each server all has corresponding Cartesian coordinates in content addressed network.When Content-based Video Retrieval, after wherein the acquisition module of 3# server obtains the proper vector of frame of video to be detected, the 3# server search the call number 1000 of module according to sensitive video frequency frame proper vector to be compared, and predefined transformation rule, from the Cartesian coordinates of 10 servers, find correspondence and preserve the Cartesian coordinates of server of sensitive video frequency frame proper vector to be compared for (0,0); The sending module of 3# server sends to the proper vector of frame of video to be detected the retrieval module of 2# server (Cartesian coordinates is (0,0)) according to the Cartesian coordinates (0,0) that finds;
The retrieval module of 2# server is judged the proper vector of frame of video to be detected and the similarity of sensitive video frequency frame proper vector to be compared.
Second embodiment provided by the invention is an a kind of Content-based Video Retrieval system, and its structure comprises as shown in Figure 3:
Video finger print storehouse 201: be used to preserve the sensitive video frequency frame proper vector with call number, described video finger print storehouse is evenly distributed in a plurality of servers on the content addressed network cartesian coordinate space node;
Acquisition module 202: the proper vector that is used to obtain frame of video to be detected;
Search module 203: be used for call number according to video finger print storehouse sensitive video frequency frame to be compared proper vector, and predefined transformation rule, from the Cartesian coordinates of each server, find the Cartesian coordinates that correspondence is preserved the server of sensitive video frequency frame proper vector to be compared;
Sending module 204: be used for the proper vector of frame of video to be detected being sent to the retrieval module of corresponding with service device according to the Cartesian coordinates that finds;
Retrieval module 205: be used to judge the proper vector of frame of video to be detected and the similarity of sensitive video frequency frame proper vector to be compared, and determine the sensitive video frequency proper vector of coupling.
Further, the video finger print storehouse 201 of each server comprises Hash table 2011: be used to preserve sensitive video frequency frame proper vector, the numbering of described Hash table is as the call number of sensitive video frequency frame proper vector;
Described system also comprises:
Secondary Hash module 206: be used for using the LSH algorithm that each the sensitive video frequency frame proper vector that is kept at each server Hash table is carried out hash, again the gained result carried out the secondary hash, obtain a plurality of hash buckets;
Hash table 2011 comprises a plurality of hash buckets 20111: each the sensitive video frequency frame proper vector that is used for Hash table is preserved and is carried out the sensitive video frequency frame proper vector after the hash twice after carrying out twice hash;
Retrieval submodule 2051: be used for the proper vector of frame of video to be detected is carried out hash twice, obtain the hash bucket of the correspondence that the proper vector of frame of video to be detected is mapped to;
Judge the similarity of sensitive video frequency frame proper vector to be compared in the proper vector of frame of video to be detected and the corresponding hash bucket.
Further, the secondary Hash module 206: also be used for L hash function g (.) with each sensitive video frequency frame proper vector
After being mapped to L the Hash table gj that is kept on each server, by k LSH function hi (.) to each the sensitive video frequency frame proper vector among the Hash table gj
Carry out hash, i.e. gj=[h1 (j) ..., hk (j)] (j=1,2 ..., L), hi (.) ∈ H (i=1,2 ... k), H is a LSH family of functions.
Further, retrieval module 205: also be used for the proper vector by calculating frame of video to be detected and the Euclidean distance of sensitive video frequency frame proper vector to be compared and judge similarity each other.
Obviously, those skilled in the art can carry out various changes and modification to the present invention and not break away from the spirit and scope of the present invention.Like this, if of the present invention these are revised and modification belongs within the scope of claim of the present invention and equivalent technologies thereof, then the present invention also is intended to comprise these changes and modification interior.
Claims (10)
1, a kind of Content-based Video Retrieval method is characterized in that, is provided with a plurality of servers on the node in content addressed network cartesian coordinate space, and this method comprises:
Acquisition module obtains the proper vector of frame of video to be detected;
Search the call number of module according to sensitive video frequency frame proper vector to be compared in the video finger print storehouse, and predefined transformation rule, from content addressed network, find the Cartesian coordinates that correspondence is preserved the server of sensitive video frequency frame proper vector to be compared in the Cartesian coordinates of each server;
Sending module sends to the proper vector of frame of video to be detected the retrieval module of corresponding with service device according to the Cartesian coordinates that finds;
Retrieval module is judged the proper vector of frame of video to be detected and the similarity of sensitive video frequency frame proper vector to be compared, and determines the sensitive video frequency proper vector of coupling.
2, the method for claim 1 is characterized in that, also comprises before finding corresponding Cartesian coordinates step:
Use local sensitivity Hash table LSH algorithm that each the sensitive video frequency frame proper vector that is kept in each server Hash table is carried out hash, obtain a plurality of hash buckets, each sensitive video frequency frame proper vector is mapped in each hash bucket, simultaneously with the numbering of each Hash table call number as sensitive video frequency frame proper vector;
Also comprise after sending the retrieval module step of proper vector to the corresponding with service device of frame of video to be detected:
The proper vector of frame of video to be detected is carried out hash twice, obtain the hash bucket of the correspondence that the proper vector of frame of video to be detected is mapped to;
The step of judging characteristic vector similarity is specially:
Judge the similarity of sensitive video frequency frame proper vector to be compared in the proper vector of frame of video to be detected and the corresponding hash bucket.
3, method as claimed in claim 2 is characterized in that, uses the LSH algorithm to carry out hash and be specially being kept at each sensitive video frequency frame proper vector in each server Hash table:
With L hash function g (.) with each sensitive video frequency frame proper vector
Be mapped among L the Hash table gj that is kept on each server;
4, method as claimed in claim 3, it is characterized in that, use the LSH algorithm that each the sensitive video frequency frame proper vector that is kept in each server Hash table is carried out hash, the number k of LSH function hi (.) has following relation in the number L of Hash table and each Hash table:
Or
At the LSH of LSH algorithm group of functions H={h1 ..., among the hm}, 2 sensitive video frequency frame proper vectors wherein
Between distance
Then
Wherein R is predefined distance, P
1Be predefined probable value.
6, method as claimed in claim 2 is characterized in that, the step of comparative feature vector similarity is specially:
The Euclidean distance of sensitive video frequency frame proper vector to be compared is judged similarity each other in proper vector by calculating frame of video to be detected and the corresponding hash bucket.
7, a kind of Content-based Video Retrieval system is characterized in that, comprising:
The video finger print storehouse: be used to preserve the sensitive video frequency frame proper vector with call number, described video finger print storehouse is evenly distributed in a plurality of servers on the content addressed network cartesian coordinate space node;
Acquisition module: the proper vector that is used to obtain frame of video to be detected;
Search module: be used for call number according to video finger print storehouse sensitive video frequency frame to be compared proper vector, and predefined transformation rule, from the Cartesian coordinates of each server, find the Cartesian coordinates that correspondence is preserved the server of sensitive video frequency frame proper vector to be compared;
Sending module: be used for the proper vector of frame of video to be detected being sent to the retrieval module of corresponding with service device according to the Cartesian coordinates that finds;
Retrieval module: be used to judge the proper vector of frame of video to be detected and the similarity of sensitive video frequency frame proper vector to be compared, and determine the sensitive video frequency proper vector of coupling.
8, system as claimed in claim 7 is characterized in that,
The video finger print storehouse of each server comprises Hash table: be used to preserve sensitive video frequency frame proper vector, the numbering of described Hash table is as the call number of sensitive video frequency frame proper vector;
Described system also comprises:
Secondary Hash module: be used for using the LSH algorithm that each the sensitive video frequency frame proper vector that is kept at each server Hash table is carried out hash, again the gained result carried out the secondary hash, obtain a plurality of hash buckets;
Hash table comprises a plurality of hash buckets: each the sensitive video frequency frame proper vector that is used for Hash table is preserved and is carried out the sensitive video frequency frame proper vector after the hash twice after carrying out twice hash;
Described retrieval module also comprises:
Retrieval submodule: be used for the proper vector of frame of video to be detected is carried out hash twice, obtain the hash bucket of the correspondence that the proper vector of frame of video to be detected is mapped to;
Judge the similarity of sensitive video frequency frame proper vector to be compared in the proper vector of frame of video to be detected and the corresponding hash bucket.
9, system as claimed in claim 8 is characterized in that,
Secondary Hash module: also be used for L hash function g (.) with each sensitive video frequency frame proper vector
After being mapped to L the Hash table gj that is kept on each server, by k LSH function hi (.) to each the sensitive video frequency frame proper vector among the Hash table gj
Carry out hash, i.e. gj=[h1 (j) ..., hk (j)] (j=1,2 ..., L), hi (.) ∈ H (i=1,2 ... k), H is a LSH family of functions.
10, system as claimed in claim 7 is characterized in that,
Retrieval module: also be used for the proper vector by calculating frame of video to be detected and the Euclidean distance of sensitive video frequency frame proper vector to be compared and judge similarity each other.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008102262681A CN101404032B (en) | 2008-11-11 | 2008-11-11 | Video retrieval method and system based on contents |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008102262681A CN101404032B (en) | 2008-11-11 | 2008-11-11 | Video retrieval method and system based on contents |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101404032A true CN101404032A (en) | 2009-04-08 |
CN101404032B CN101404032B (en) | 2011-09-28 |
Family
ID=40538044
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2008102262681A Active CN101404032B (en) | 2008-11-11 | 2008-11-11 | Video retrieval method and system based on contents |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101404032B (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102117313A (en) * | 2010-12-29 | 2011-07-06 | 天脉聚源(北京)传媒科技有限公司 | Video retrieval method and system |
CN102722554A (en) * | 2012-05-28 | 2012-10-10 | 中国人民解放军信息工程大学 | Randomness weakening method of location-sensitive hash |
CN102760169A (en) * | 2012-06-13 | 2012-10-31 | 天脉聚源(北京)传媒科技有限公司 | Method for detecting advertising slots in television direct transmission streams |
CN103020094A (en) * | 2011-12-19 | 2013-04-03 | 北京捷成世纪科技股份有限公司 | Method for counting video playing times |
CN103093761A (en) * | 2011-11-01 | 2013-05-08 | 腾讯科技(深圳)有限公司 | Audio fingerprint retrieval method and retrieval device |
CN103294696A (en) * | 2012-02-27 | 2013-09-11 | 盛乐信息技术(上海)有限公司 | Audio and video content retrieval method and system |
CN103594083A (en) * | 2012-08-14 | 2014-02-19 | 韩凯 | Technology of television program automatic identification through television accompanying sound |
CN104142984A (en) * | 2014-07-18 | 2014-11-12 | 电子科技大学 | Video fingerprint retrieval method based on coarse and fine granularity |
CN104915403A (en) * | 2015-06-01 | 2015-09-16 | 腾讯科技(北京)有限公司 | Information processing method and server |
CN106126619A (en) * | 2016-06-20 | 2016-11-16 | 中山大学 | A kind of video retrieval method based on video content and system |
CN106156284A (en) * | 2016-06-24 | 2016-11-23 | 合肥工业大学 | Video retrieval method is closely repeated based on random the extensive of various visual angles Hash |
CN106331768A (en) * | 2016-09-19 | 2017-01-11 | 杭州当虹科技有限公司 | Efficient video content compliance detection mechanism |
CN106557765A (en) * | 2015-09-29 | 2017-04-05 | 欧姆龙株式会社 | Note detection means and note detection method |
CN106570166A (en) * | 2016-11-07 | 2017-04-19 | 北京航空航天大学 | Video retrieval method and apparatus based on multiple partial sensitive hash tables |
CN108027816A (en) * | 2015-10-28 | 2018-05-11 | 株式会社东芝 | Data management system, data managing method and program |
CN114064948A (en) * | 2021-10-15 | 2022-02-18 | 西安深信科创信息技术有限公司 | Hash image retrieval method and device based on generalized average pooling strategy |
-
2008
- 2008-11-11 CN CN2008102262681A patent/CN101404032B/en active Active
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102117313A (en) * | 2010-12-29 | 2011-07-06 | 天脉聚源(北京)传媒科技有限公司 | Video retrieval method and system |
CN103093761A (en) * | 2011-11-01 | 2013-05-08 | 腾讯科技(深圳)有限公司 | Audio fingerprint retrieval method and retrieval device |
CN103020094B (en) * | 2011-12-19 | 2016-03-16 | 北京捷成世纪科技股份有限公司 | Video playback number of times statistical method |
CN103020094A (en) * | 2011-12-19 | 2013-04-03 | 北京捷成世纪科技股份有限公司 | Method for counting video playing times |
CN103294696A (en) * | 2012-02-27 | 2013-09-11 | 盛乐信息技术(上海)有限公司 | Audio and video content retrieval method and system |
CN103294696B (en) * | 2012-02-27 | 2018-01-19 | 上海果壳电子有限公司 | Audio-video frequency content search method and system |
CN102722554B (en) * | 2012-05-28 | 2014-07-02 | 中国人民解放军信息工程大学 | Randomness weakening method of location-sensitive hash |
CN102722554A (en) * | 2012-05-28 | 2012-10-10 | 中国人民解放军信息工程大学 | Randomness weakening method of location-sensitive hash |
CN102760169A (en) * | 2012-06-13 | 2012-10-31 | 天脉聚源(北京)传媒科技有限公司 | Method for detecting advertising slots in television direct transmission streams |
CN103594083A (en) * | 2012-08-14 | 2014-02-19 | 韩凯 | Technology of television program automatic identification through television accompanying sound |
CN104142984B (en) * | 2014-07-18 | 2017-04-05 | 电子科技大学 | It is a kind of to be based on thick fine-grained video fingerprint retrieval method |
CN104142984A (en) * | 2014-07-18 | 2014-11-12 | 电子科技大学 | Video fingerprint retrieval method based on coarse and fine granularity |
CN104915403A (en) * | 2015-06-01 | 2015-09-16 | 腾讯科技(北京)有限公司 | Information processing method and server |
CN104915403B (en) * | 2015-06-01 | 2018-07-27 | 腾讯科技(北京)有限公司 | A kind of information processing method and server |
CN106557765A (en) * | 2015-09-29 | 2017-04-05 | 欧姆龙株式会社 | Note detection means and note detection method |
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 |
CN106126619A (en) * | 2016-06-20 | 2016-11-16 | 中山大学 | A kind of video retrieval method based on video content and system |
CN106156284A (en) * | 2016-06-24 | 2016-11-23 | 合肥工业大学 | Video retrieval method is closely repeated based on random the extensive of various visual angles Hash |
CN106331768A (en) * | 2016-09-19 | 2017-01-11 | 杭州当虹科技有限公司 | Efficient video content compliance detection mechanism |
CN106570166A (en) * | 2016-11-07 | 2017-04-19 | 北京航空航天大学 | Video retrieval method and apparatus based on multiple partial sensitive hash tables |
CN114064948A (en) * | 2021-10-15 | 2022-02-18 | 西安深信科创信息技术有限公司 | Hash image retrieval method and device based on generalized average pooling strategy |
Also Published As
Publication number | Publication date |
---|---|
CN101404032B (en) | 2011-09-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101404032B (en) | Video retrieval method and system based on contents | |
CN102368237B (en) | Image retrieval method, device and system | |
CN105022840A (en) | News information processing method, news recommendation method and related devices | |
KR100903961B1 (en) | Indexing And Searching Method For High-Demensional Data Using Signature File And The System Thereof | |
CN108304409B (en) | Carry-based data frequency estimation method of Sketch data structure | |
CN106326475A (en) | High-efficiency static hash table implement method and system | |
CN101276361A (en) | Method and system for displaying related key words | |
CN111868710A (en) | Random extraction forest index structure for searching large-scale unstructured data | |
WO2019165546A1 (en) | Layered locality sensitive hashing (lsh) partition indexing for big data applications | |
CN108182242A (en) | A kind of indexing means for the inquiry of magnanimity multi dimensional numerical data area | |
CN102890719B (en) | A kind of method that license plate number is searched for generally and device | |
CN103218443A (en) | Blogging webpage retrieval system and retrieval method | |
CN105404675A (en) | Ranked reverse nearest neighbor space keyword query method and apparatus | |
CN109885640A (en) | A kind of multi-key word ciphertext ordering searching method based on α fork index tree | |
CN102375813A (en) | Duplicate detection system and method for search engines | |
CN104636349A (en) | Method and equipment for compression and searching of index data | |
CN105357247A (en) | Multi-dimensional cloud resource interval finding method based on hierarchical cloud peer-to-peer network | |
CN107451302A (en) | Modeling method and system based on position top k keyword queries under sliding window | |
CN106874788A (en) | A kind of method for secret protection in sensitive data issue | |
CN105302833A (en) | Content based video retrieval mathematic model establishment method | |
CN104636368A (en) | Data retrieval method and device and server | |
CN103455491B (en) | To the method and device of query word classification | |
CN109446293A (en) | A kind of parallel higher-dimension nearest Neighbor | |
CN113434511B (en) | Clustering index method based on Hilbert curve | |
CN110941831B (en) | Vulnerability matching method based on slicing technology |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |