WO2013097078A1

WO2013097078A1 - Video search method and video search system

Info

Publication number: WO2013097078A1
Application number: PCT/CN2011/084666
Authority: WO
Inventors: 冀永楠
Original assignee: 华为技术有限公司
Priority date: 2011-12-26
Filing date: 2011-12-26
Publication date: 2013-07-04
Also published as: CN103548017A

Abstract

Provided are a video search method and a video search system. The method comprises: receiving a search request carrying a text data search criterion input by a user; finding metadata and/or composite metadata matching the text data from a video database, the composite metadata including a content of at least two pieces of metadata; and displaying a video feature corresponding to the metadata, and/or displaying a video feature corresponding to the composite metadata. The embodiments of the present invention are applied in video searching, and can avoid repeated displaying of videos having the same content, thereby enabling more videos having different contents to be displayed on a search result page.

Description

Video search method and video search system

The present invention relates to the field of search engines, and in particular, to a video search method and a video search system.

Background technique

With the advent of the Internet era, users' demand for the Internet is increasing. In the huge information system of the Internet, users need to use many search engines to find the information they want, but the search often returns a large amount of data with repeated content. What users often want is data that is related to the search problem but has different content.

Especially when performing video search, since the search of video data is mainly based on metadata, that is, a piece of information corresponding to the title, content description, keyword, duration, source, thumbnail connection, etc. of the corresponding video is saved based on the video. Data is searched. Therefore, when a user searches for a popular video on a website, there are often many similar or even identical content, but videos with different metadata such as title and uploader, for example, if the user wants to find as many different versions as possible. Or when the video of the same song sung by different singers, the most popular video will occupy a large number of pages, and it is necessary to read many pages to find the desired content, which brings inconvenience to the user's search.

Summary of the invention

Embodiments of the present invention provide a video search method and a video search system that cause as many different videos as possible to be displayed in a search result page.

In order to achieve the above, the embodiment of the present invention uses the following technical solutions: In one aspect, a video search method is provided, including:

Receiving a search request entered by the user and carrying a text data search condition;

Finding metadata and/or composite metadata matching the text data from the video database; wherein the composite metadata includes content of at least two metadata;

Displaying video features corresponding to the metadata, and/or displaying a video feature corresponding to the composite metadata. In another aspect, a video search system is provided, including:

a video database, configured to store metadata, composite metadata, and video features; wherein the composite metadata includes content of at least two metadata;

a user interface, configured to receive a search request by a user to carry a text data search condition; display a video feature corresponding to the metadata, and/or display a video feature corresponding to the composite metadata;

A search controller for finding metadata and/or composite metadata that matches the text data from a video database.

The video search method and the video search system provided by the embodiment of the present invention, after receiving a search request carrying text data, find metadata and/or composite metadata that matches the text data, where the composite metadata includes at least two metadata. Content; When displayed, only one video feature corresponding to the video feature and/or composite metadata corresponding to the metadata is displayed. Therefore, the searched repeated video content is not displayed, which avoids the phenomenon that the search result page is flooded with a large number of repeated videos in the prior art, so that the search result obtained by the user searching for the video can display the video of different content as much as possible, thereby improving the video. User experience.

DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below. Obviously, the drawings in the following description are only It is a certain embodiment of the present invention, and other drawings can be obtained from those skilled in the art without any creative work.

FIG. 1 is a schematic flowchart of a video search method according to an embodiment of the present invention;

2 is a schematic flowchart of another video search method according to an embodiment of the present invention; FIG. 3 is a schematic diagram showing a search result display of a video search method according to an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of a video search system according to an embodiment of the present invention.

detailed description

The technology in the embodiment of the present invention will be described below with reference to the accompanying drawings in the embodiments of the present invention. The present invention is described in a clear and complete manner, and it is obvious that the described embodiments are only a part of the embodiments of the invention, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

The video search method provided by the embodiment of the present invention, as shown in FIG. 1, the method steps include:

5 101. The video search system receives a search request entered by the user and carries a text data search condition.

5 102. The video search system finds metadata and/or composite metadata that matches the text data from the video database; wherein the composite metadata includes content of at least two metadata.

5 103. The video search system displays a video feature corresponding to the metadata, and/or displays a video feature corresponding to the composite metadata.

It should be noted that when the composite metadata is found to match the text data, only one of the plurality of video features corresponding to the composite metadata is displayed, and other video features may be hidden in the hidden list, and may be in a certain order. Arranged in a hidden list, such as according to the histogram, or the number of clicks of each video, or the rating of each video, or the popularity of the source site, etc., in a hidden list, so that not only can a search result be as much as possible in a page Display multiple videos with different content, and not miss different versions of the same video.

The video search method provided by the embodiment of the present invention, after receiving the search request carrying the text data, finds metadata and/or composite metadata that matches the text data, where the composite metadata includes content of at least two metadata; At the time, only one video feature corresponding to the video feature and/or the composite metadata corresponding to the metadata is displayed. Therefore, the repeated video content obtained by the search is not displayed, which avoids the phenomenon that the search result page is flooded with a large number of repeated videos in the prior art, so that the search result obtained by the user searching for the video can be displayed as the video of different content as much as possible, thereby improving The user feels.

A video search method according to another embodiment of the present invention is described by taking the video search system shown in FIG. 4 as an example. In the video search system shown in Figure 4, including the network The crawler 34, the video import controller 33, the video database 30 including the metadata sub-library 301 and the video feature sub-library 302, and the search controller 31 and the user interface 32.

Specifically, in the embodiment of the present invention, the video feature in the obtained new video data is an example of a key frame, but is not limited thereto. The method step is as shown in FIG. 2, and includes:

5201. The video search system crawls the new video through the web crawler 34.

5202. The video import controller 33 of the video search system acquires a key frame of the new video data, and records it as F _ia.

Exemplarily, the new video data is acquired, and the process of splitting and key frame extraction is performed by V, ·, and the key frame extracted from V,. is denoted as Fi.

5203. The video import controller 33 obtains metadata describing the new video data, and records it as Mi.

In this embodiment, step S202 is taken as an example before step S203. Actually, there is no chronological relationship between steps S202 and S203, which may be performed simultaneously or sequentially.

5204. The video import controller 33 searches the metadata sub-library 301 of the video database 30 for the metadata and the composite metadata that match Μ,·, if not, proceeds to step S205; if found, proceeds to step S206.

5205. The video import controller 33 does not find the metadata and the composite metadata matching the M in the metadata sub-library 301 of the video database 30, and imports the M into the metadata sub-library 301 of the video database 30, and imports the video database 30. The video feature sub-library 302 is then executed in step S210.

Exemplarily, the process of finding a matching result in the video database 30 of the video search system is: querying all the fields in each metadata and composite metadata stored in the matching video database 30 one by one according to the fields in Μ,. The matching score of each field of each metadata and composite metadata is added as a matching score of the metadata and the composite metadata. If the matching score of the primary data or the composite metadata is greater than the first threshold, the metadata or the composite Metadata is the matching result.

Further, the matching points can be obtained through the Lucene open source full-text search engine toolkit; The first threshold value can obtain a thousand matching scores for the text metadata of the existing one thousand sets of the same video data by the above method, and the matching statistics of the thousands of identical videos are arranged in descending order, and The matching score at the 5% position is set to the first threshold, which is only illustrated by way of example and is not limited thereto.

It is worth noting that when there is no metadata and composite metadata with a matching score higher than the first threshold in the video database, it will be imported into the metadata sub-library 30 1 of the video database and will be imported. The video feature sub-library 302 of the video database.

5206. Take out a matching metadata and/or composite metadata, denoted as m _k , and a corresponding key frame data set /.

5207. Determine whether the key frame data set corresponding to the same is the same. If Fi is not the same, step S208 is performed; if ^ is the same as one, step S209 is performed.

5208. Determine whether there are other metadata and/or composite metadata matching the Μ,. in the metadata sub-library 30 1 . If yes, execute step S206; if not, execute S 205. Here, step S205 is performed to explain that it is different from each other.

5209, adding Fi to f _k to obtain a new composite key frame ', will be imported into the video feature sub-library 302 of the video database 30; adding Μ,. to the corresponding m _k to obtain a new composite metadata m _k ', t' is imported into the metadata sub-library 301 of the video database 30.

Illustratively, the video search system finds a metadata and corresponding video feature data in the video database 30. _Fe can be a video metadata, then / ^ represents a key frame of a video; it can also be a composite metadata containing multiple video metadata, then Λ represents multiple key frames containing each, and these The video data corresponding to the key frame f _k has been judged as video data of the same content.

The process of determining whether Fi and Λ are the same is: When it is a video metadata, f _k represents a key frame of a video, and the corresponding key frame image is upsampled or downsampled to the same resolution, and the original image is recorded. The resolution is determined to be the same as Λ by comparing the sum of the ^ and ^ mean square values smaller than the second threshold and/or the sum of the mutual information values greater than the third threshold.

When it is a composite metadata containing multiple video metadata, then Λ represents the package When there are multiple key frames corresponding to each, the calculated mean square difference is calculated, and if the calculated maximum mean square difference is smaller than the second threshold; and/or the calculated mutual information value is calculated, If the minimum mutual information value is greater than the third value, then Fi is determined to be the same.

Otherwise, it is determined that ^ is different from Λ, and then it is necessary to continue to find another metadata in the video database 30 that meets the above matching score higher than the first threshold m), repeat the above-mentioned judging step, and so on.

It should be pointed out that the second threshold and the third interpretation value can use the key frames in the same video data to obtain the most accurate set value according to the statistical principle, such as obtaining 5 thousand pairs of the same key frame, and calculating their mean square. The difference values are arranged in descending order, taking the mean squared difference of the 10% position as the second threshold value, and the third threshold value is used, and the specific value depends on the usage. This embodiment is only an example, but Do not make any restrictions.

If none of the video databases 30 is the same as ^, the metadata sub-library 301 of the video database 30 will be imported into the video feature sub-library 302 of the video database 30.

If it is determined that there is one identical to ^, then ^ is added to /; to obtain a new composite key frame Λ ', and Μ,. is added to Λ corresponding m _k to obtain new composite metadata m _k ', and then m _k ' The metadata sub-library 301 of the video database 30 is imported into the video feature sub-library 302 of the video database 30.

Further, as shown in Table 1, adding Fi to f _k to obtain a new composite key frame f _k ', there will be many methods for adding corresponding new composite metadata m^t ', this embodiment uses a list method. For example, in the list to be included, get a new composite keyframe list'; add M _t to the corresponding m _k list to get a new composite metadata list ''. In the list formed by the composite metadata, there is a unique metadata ID, and some feature data of the video corresponding to each metadata, such as a title, a label, and the like, are all classified.

Metadata ID 101

Video ID 101a 101b 101c Title AABBCC AABB AACCDD Label Video Mix Video Video Dance Video MTV Link ht tp: //www.1. com http: / /www.1. com/ul. http: / /www. tudou. com / 1 html 1 2 Classified original music and music resolution 1 6 0 x 16 0 1 6 0 1 6 0 16 0 1 60 Table 1

In this way, the text-image two-level comparison method is adopted, and the text in the video metadata is very small, and the contrast is inaccurate. By using the image comparison, the video data with the same content is more accurately obtained, and they are Form a composite video data for the user to find.

5210. Determine whether the acquired new video data has been all imported into the video database 30. If all the imports have been imported, the import is completed; if not all are imported, step S202 is performed.

521 1. The video search system receives a search request by the user to carry a text data search condition through the user interface 32.

5212. The search controller 31 of the video search system finds metadata and/or composite metadata matching the text data from the metadata sub-library 301 of the video database 30; wherein the composite metadata includes content of at least two metadata .

Illustratively, the composite metadata that finds the matching text data is substantially the same as the general text-based metadata-based matching text data. After the user inputs the search request that carries the text data search condition, the video search system passes through the composite metadata. The fields of all video metadata are matched. For example, when the user searches for a title, if at least one of the composite metadata has a matching title, it is considered that the video search system finds the composite metadata matching the text data. The matching scores can be calculated in accordance with the open source LUCENE SCORE (Open Sources Full-Text Search Engine Toolkit to calculate scores). For example, a word that is repeated in all titles of the composite metadata and is not a stop word is calculated, and the word is used as the title of the composite metadata. Of course, the description of the composite metadata, the keyword, and the like are processed in the same manner. The video search system only finds the matching of the field of the composite metadata as a matching result, and does not need to calculate a matching score with each field corresponding to each individual video contained in the composite metadata to determine whether or not the matching is matched.

Further, it can be considered that the matching score exceeds the first value of the matching result, the use case For example, a thousand sets of the same video matching score statistics, all the matching scores are arranged in order of increasing to, and the matching score at the 10th position is set as the first threshold used for matching.

S2 13. The video search system displays a video key frame corresponding to the metadata through the user interface 32, and/or displays a video key frame corresponding to the composite metadata.

It should be noted that, in general, when displaying video key frames of multiple composite video data, these video key frames are sorted in the search result page according to the click rate or network source popularity.

Among them, each composite video data only displays one video key frame, and other video key frames are hidden in the drop-down list, such as only displaying the most abundant image content calculated by using the histogram, and other video key frames are also This order is hidden in the drop-down list. You can also select the video keyframes with the most balanced histogram, the highest click rate, or the highest network score. Other video keyframes are also arranged in a corresponding manner. As shown in Figure 3, when the user's mouse stops on a video keyframe displayed, the drop-down list is activated. When the user mouses over another video keyframe, the drop-down list is automatically closed, and the video keyframe where the mouse is located The drop-down list is activated; when the mouse clicks on the video keyframe corresponding to the metadata, no drop-down list is activated. This not only makes it possible to display as many different videos as possible in the pages of the search results, but also does not miss different versions of the same video.

Further, when the video search system receives the search request input by the user as a classified search, if the date search is received, the composite video data only displays one video key frame that meets the classification condition, and only the hidden drop-down list hides the matching condition. Video keyframes, equivalent to these video keyframes that meet the classification criteria, are recombined into a new temporary composite video data to display video features that satisfy the user's classification criteria.

The video search method provided by the embodiment of the present invention, after receiving the search request carrying the text data, finds metadata and/or composite metadata that matches the text data, wherein the composite metadata includes content of at least two metadata; When displayed, only one video feature corresponding to the video feature and/or composite metadata corresponding to the metadata is displayed. Therefore, the repeated video content obtained by the search is not displayed, and the search result page in the prior art is avoided. The phenomenon of flooding a large number of repeated videos enables the search results obtained by the user to search for videos to be displayed as videos of different contents as much as possible, thereby improving the user experience.

The video search system provided by the embodiment of the present invention, as shown in FIG. 4, includes: a video database 30, configured to store metadata, composite metadata, and video features, where the composite metadata includes at least two metadata. content.

The user interface 32 is configured to receive a search request for carrying a text data search condition input by the user, display a video feature corresponding to the metadata, and/or display a video feature corresponding to the composite metadata.

The search controller 31 is configured to find metadata and/or composite metadata matching the text data from the video database.

Further, the user interface 32 is further configured to display other video features corresponding to the composite metadata in a hidden list manner.

The video database 30 includes:

The metadata sub-library 301 is used for storing metadata and composite metadata.

The video feature sub-library 302 is configured to store video features.

In addition, the system also includes:

Web crawler 34, used to crawl video data.

Video import controller 33, used to acquire video features of new video data, recorded as

F, obtaining metadata describing the new video data, denoted as M, if the metadata and/or composite metadata matching the M is not found in the metadata sub-library 301, the Mi is The metadata sub-library 301 is imported, and the F is imported into the video feature sub-library 302.

The video import controller 34 is further configured to: if the metadata and/or composite metadata matching the Μ,· is found in the metadata sub-library 301, and recorded as m _k , determine the confusion Whether the video feature set corresponding to each of the fans is the same; if it is determined that the ^ is different from each of the UIs, the metadata sub-library 301 is imported, and the video feature sub-library is imported. 302. If it is determined that the Fi is the same as the one, the ^ is added to the new composite video feature set ,', and the 'imported into the video feature sub-library 302, the Μ _; · Add the corresponding one To the new composite metadata m _k ' , the metadata sub-library 30 1 is imported.

The video search system can use the video search method provided in the foregoing embodiment. The method has been described in detail in the foregoing embodiments, and details are not described herein again.

The video search system provided by the embodiment of the present invention, after receiving the search request carrying the text data, finds metadata and/or composite metadata that matches the text data, wherein the composite metadata includes content of at least two metadata; When displayed, only one video feature corresponding to the video feature and/or composite metadata corresponding to the metadata is displayed. Therefore, the repeated video content obtained by the search is not displayed, which avoids the phenomenon that the search result page is flooded with a large number of repeated videos in the prior art, so that the search result obtained by the user searching for the video can be displayed as the video of different content as much as possible, thereby improving The user feels.

A person skilled in the art can understand that all or part of the steps of implementing the above method embodiments may be completed by using hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, when executed, The foregoing steps include the steps of the foregoing method embodiments; and the foregoing storage medium includes: a medium that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.

The above is only the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope of the present invention. It should be covered by the scope of the present invention. Therefore, the scope of the invention should be determined by the scope of the claims.

Claims

Claim

A video search method, comprising:

Finding metadata and/or composite metadata that matches the text data from a video database; wherein the composite metadata includes content of at least two metadata;

Displaying video features corresponding to the metadata, and/or displaying a video feature corresponding to the composite metadata.

2. The method according to claim 1, wherein displaying a video feature corresponding to the composite metadata further comprises:

Display other video features corresponding to the composite metadata in a hidden list manner.

3. The method according to claim 2, wherein displaying other video features corresponding to the composite metadata in a hidden list manner comprises:

Other video features corresponding to the partial metadata of the composite metadata that match the search condition are displayed in a hidden list manner.

4. The method of claim 1, further comprising:

Get the video characteristics of the new video data, recorded as F,

Obtaining metadata describing the new video data, denoted as M,

If no metadata and/or composite metadata matching the Μ,. is found in the video database, the Mi and Fi are imported into the video database.

The method according to claim 1 or 4, wherein the matching comprises:

4. According to the input text data or the fields in the Μ,., query each field in the matching metadata database and all the fields in the composite metadata one by one, and each field of each metadata and composite metadata The matching score is added as a matching score of the metadata and the composite metadata. If the matching score of the unary data or the composite metadata is greater than the set first threshold, the metadata or the composite metadata is a matching result.

6. The method of claim 4, wherein

If the metadata and/or composite metadata matching the Mi is found in the video database, denoted as m _k , the video feature set corresponding to each of the videos is determined. Whether they are the same;

If it is determined that the said and the said each are not the same, the Mi and Fi are imported into the video database;

If it is determined that the same is the same as one, the ^ is added to the new composite video feature set Λ ', and the Μ is added to the corresponding _{t t of the} , to obtain new composite metadata m _k , And importing the Λ' and the into the video database.

The method according to claim 6, wherein determining whether the ^ is identical to each of the corresponding video feature sets comprises:

If the _t is metadata, if the mean value of the sum is less than the set second threshold, determining that the ^ is the same as the ;; if the mutual information value is greater than Setting the third noise value, determining that the Fi is the same as the ^;

If the m _fc is composite metadata, calculating a mean square difference between the ^ and each of the Λ, if the maximum mean square difference is less than the set second threshold, determining the Λ Alternatively, the mutual information value of the /A and each of the As is calculated. If the minimum mutual information value is greater than the set third threshold, it is determined that the ^ is the same as the threshold.

The method according to claim 6, wherein the new composite metadata includes a unique metadata identifier, and the feature data of the video corresponding to each metadata corresponds to the classification.

The method according to claim 2 or 3, further comprising: receiving a histogram of each video feature, or a number of clicks of each video, or each video, after receiving an instruction of the user to click the hidden list The rating, or the popularity of the source website, displays the other video features in a certain order.

A video search system, comprising:

a user interface, configured to receive a search request for carrying a text data search condition input by the user; display a video feature corresponding to the metadata, and/or display a video feature corresponding to the composite metadata; A search controller is configured to find metadata and/or composite metadata that matches the text data from a video database.

1 1. The system of claim 10, wherein:

The user interface is further configured to display other video features corresponding to the composite metadata in a hidden list manner.

2. The system according to claim 10, wherein the video data database comprises:

Metadata sub-library for storing metadata and composite metadata;

Video feature sub-library for storing video features.

The system according to claim 12, wherein the system further comprises:

a web crawler for crawling video data;

a video import controller, configured to acquire a video feature of the new video data, and recorded as acquiring metadata describing the new video data, denoted as M, if the metadata sub-database is not found to match the M; Metadata and/or composite metadata, then import the M _; into the metadata sub-library, and import the ^ into the video feature sub-library.

14. The system of claim 13 wherein:

The video import controller is further configured to: if the metadata and/or the composite metadata matching the Μ are found in the metadata sub-library, denoted as m _k , determine the corresponding to each of the Whether the video feature set is the same; if it is determined that the description is different from each of the foregoing, the UI is imported into the metadata sub-library, and the video feature sub-library is imported; Referring to the same as one, adding the said Λ, obtaining a new composite video feature set Λ ', importing the Λ ' into the video feature sub-library, adding the corresponding to the corresponding new The composite metadata _t ' is imported into the metadata sub-library.