CN106375781A - Method and device for judging duplicate video - Google Patents
Method and device for judging duplicate video Download PDFInfo
- Publication number
- CN106375781A CN106375781A CN201510438517.3A CN201510438517A CN106375781A CN 106375781 A CN106375781 A CN 106375781A CN 201510438517 A CN201510438517 A CN 201510438517A CN 106375781 A CN106375781 A CN 106375781A
- Authority
- CN
- China
- Prior art keywords
- video
- template
- section
- measured
- vector
- 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
- 238000000034 method Methods 0.000 title claims abstract description 44
- 239000013598 vector Substances 0.000 claims abstract description 373
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 3
- 230000002194 synthesizing effect Effects 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 15
- 238000010586 diagram Methods 0.000 description 10
- 230000008569 process Effects 0.000 description 10
- 238000004590 computer program Methods 0.000 description 6
- 230000003252 repetitive effect Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 241001269238 Data Species 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 206010008190 Cerebrovascular accident Diseases 0.000 description 1
- 208000006011 Stroke Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 210000003746 feather Anatomy 0.000 description 1
- 244000144992 flock Species 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/234—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs
- H04N21/23418—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs
- H04N21/44008—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics in the video stream
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a device for judging a duplicate video. The method comprises steps: multiple sections of to-be-detected videos are acquired from a real-time video stream; a target feature vector of each section of to-be-detected video in the multiple sections of to-be-detected videos is acquired; multiple sections of template videos are extracted from a template video library; M template pictures are synthesized into a target picture; the feature vector of the target picture is acquired, and the feature vector of the target picture serves as a template feature vector of each section of template video; the vector distance between the target feature vector of each section of to-be-detected video and the template feature vector of each section of template video is calculated; and when the vector distance between the target feature vector of a section of to-be-detected video and the template feature vector of a section of template video is smaller than or equal to a preset vector distance, at least part of video in the section of to-be-detected video is duplicate with the section of template video. Through the technical scheme of the invention, video duplication detection can be carried out accurately, and program prediction can be carried out accurately.
Description
Technical field
The present invention relates to technical field of video processing, particularly to a kind of determination methods repeating video and device.
Background technology
At present, in electric program menu (epg, electronic program guide) processing system,
Need to carry out program real-time estimate.And predict that program actually will carry out video repeatability detection.But people
The task amount that work carries out repeatability detection is very big, less efficient, and the accuracy rate of repeatability detection and precision
All ratio is relatively low, therefore, it is impossible to carry out program real-time estimate exactly, this brings very big inconvenience to user.
Content of the invention
The present invention provides a kind of determination methods repeating video and device, in order to be entered exactly using template video
The repeatability detection of row video, and guarantee precision and the efficiency of video repeatability detection, thus realizing to regarding in real time
The program that frequency is play in flowing carries out accurate programming predictions.
The present invention provides a kind of determination methods repeating video, comprising: from live video stream, obtains multistage
Video to be measured;
Obtain the target feature vector of every section of video to be measured in multistage video to be measured, wherein, described target is special
Levy vector and be used for the described video to be measured of every section of unique mark;
Extract multistage template video from template video storehouse, wherein, every section of template in multistage template video regards
M is had to open template picture in frequency;
M in every section of described template video is opened the picture of the target pixel points of the same position in template picture
Plain value is weighted suing for peace, and described m is opened template picture and synthesizes a Target Photo, wherein, often
The pixel of the target pixel points in described Target Photo for the pixel value of the target pixel points of Zhang Suoshu template picture
In value shared weight and be equal to 1;
Obtain the characteristic vector of described Target Photo, and using the characteristic vector of described Target Photo as every section of institute
State the template characteristic vector of template video;
Calculate the target feature vector of every section of described video to be measured and the template characteristic of every section of described template video
Vector distance between vector;
Between the target feature vector and the template characteristic vector of one section of template video of one section of video to be measured
When vector distance is less than or equal to default vector distance, judge at least partly regarding in described one section of video to be measured
Frequency is repeated with described one section of template video;Otherwise, it is determined that described one section of video to be measured is regarded with described one section of template
Frequency does not repeat completely.
In one embodiment, in the target feature vector calculating every section of described video to be measured and every section of described mould
Before vector distance between the template characteristic vector of plate video, methods described also includes:
The template characteristic vector of template video described in multistage is clustered, with obtain multiclass template characteristic to
Amount;
At least part of video in judging described one section of video to be measured is repeated with described one section of template video;No
Then, before judging that described one section of video to be measured and described one section of template video do not repeat completely, methods described is also
Including:
Vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From during less than or equal to default vector distance, judge at least part of video in described one section of video to be measured and institute
State the corresponding To Template video of a class template characteristic vector to repeat, and start to judge described one section of video to be measured
In at least part of video whether repeat with described one section of template video;Otherwise, it is determined that described one section to be measured regards
Frequency is not repeated completely with described To Template video, and no longer judges at least partly regarding in one section of video to be measured
Whether frequency is repeated with described one section of template video.
In one embodiment, at least part of video in judging described one section of video to be measured and a described class
Before the corresponding To Template video of template characteristic vector repeats, methods described also includes:
Arrow between the target feature vector of described one section of video to be measured and a described class template characteristic vector
Span, from during less than or equal to described default vector distance, each in described one section of video to be measured is treated mapping
Piece is mated successively with each the To Template picture in described To Template video;
Arrow between the described target feature vector when described one section of video to be measured and a class template characteristic vector
Span, from during less than or equal to default vector distance, judges at least part of video in described one section of video to be measured
To Template video corresponding with a described class template characteristic vector repeats;Otherwise, it is determined that described one section to be measured
Video is not repeated completely with described To Template video, comprising:
Vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From closest to described default vector distance and described less than or equal to default vector distance, described vector distance
In one section of video to be measured with described To Template video in the To Template picture picture to be measured that matches
When number is higher than preset number, judge at least part of video in described one section of video to be measured and described target mould
Plate video repeats;Otherwise, it is determined that at least part of video in described one section of video to be measured and described To Template
Video does not repeat completely;
Methods described also includes:
At least part of video in judging described one section of video to be measured is with described To Template video completely not
After repetition, find the target feature vector with described one section of video to be measured successively in multiclass template characteristic vector
Between vector distance closest to described default vector distance other class template characteristic vectors;
Judge to regard with other corresponding To Templates of class template characteristic vector described in described one section of video to be measured
Whether the number of the picture to be measured that the To Template picture in frequency matches is higher than described preset number, and successively
Circulation, is regarded with other corresponding To Templates of class template characteristic vector described in described one section of video to be measured
The number of the picture to be measured that the Target Photo in frequency matches is higher than described preset number, or multistage video to be measured
In every section of video to be measured in the mesh in other class template characteristic vectors described corresponding To Template video
The number of the picture to be measured that mark template picture matches is below described preset number.
In one embodiment, in the characteristic vector obtaining described Target Photo, and by described Target Photo
Before the template characteristic vector as template video every section described for the characteristic vector, methods described also includes:
Described target icon is carried out dct conversion, obtains the horizontal coefficient of pixel coordinate after dct conversion
With longitudinal coefficient;
And
The described characteristic vector obtaining described Target Photo, and using the characteristic vector of described Target Photo as every
The template characteristic vector of Duan Suoshu template video, comprising:
Choose the front q2 potential coefficient in the front q1 potential coefficient and described longitudinal coefficient in described horizontal coefficient,
Generate the template characteristic of every section of described template video according to described front q1 potential coefficient and described front q2 potential coefficient
Vector.
In one embodiment, the m in every section of described template video is being opened the same position in template picture
The pixel value of the target pixel points put is weighted suing for peace, and described m is opened template picture and synthesizes one
During target icon, the kth that described m opens in model picture opens the pixel value of the target pixel points of model picture
Shared weight α in the pixel value of the target pixel points of described Target Photok=γ∧k, wherein, γ is little
In 1 constant.
The present invention also provides a kind of judgment means repeating video, comprising: the first acquisition module, for from reality
When video flowing in, obtain multistage video to be measured;
Second acquisition module, for obtain the target characteristic of every section of video to be measured in multistage video to be measured to
Amount, wherein, described target feature vector is used for the described video to be measured of every section of unique mark;
Processing module, for extracting multistage template video, wherein, multistage template video from template video storehouse
In every section of template video in have m to open template picture;
Summation module, for opening the same position in template picture by the m in every section of described template video
The pixel value of target pixel points is weighted suing for peace, and described m is opened template picture and synthesizes a target
Picture, wherein, the pixel value of the target pixel points of every described template picture is in the target of described Target Photo
In the pixel value of pixel shared weight and be less than or equal to 1;
3rd acquisition module, for obtaining the characteristic vector of described Target Photo, and by described Target Photo
Characteristic vector is as the template characteristic vector of template video every section described;
Computing module, target feature vector and every section of described template for calculating every section of described video to be measured regard
Vector distance between the template characteristic vector of frequency;
Determination module, for as the target feature vector of one section of video to be measured and the template spy of one section of template video
When levying vector distance between vector and being less than or equal to default vector distance, judge in described one section of video to be measured
At least part of video repeat with described one section of template video;Otherwise, it is determined that described one section of video to be measured and institute
State one section of template video not repeat completely.
In one embodiment, described device also includes:
Cluster module, in the target feature vector calculating every section of described video to be measured and every section of described template
Video template characteristic vector between vector distance before, by the template characteristic of template video described in multistage to
Amount is clustered, to obtain multiclass template characteristic vector;
Described determination module includes:
Decision sub-module, at least part of video in the described one section of video to be measured of judgement and described one section
Template video repeats;Otherwise, it is determined that described one section of video to be measured is not repeated completely with described one section of template video
Before, the vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From during less than or equal to default vector distance, judge at least part of video in described one section of video to be measured and institute
State the corresponding To Template video of a class template characteristic vector to repeat, and start to judge described one section of video to be measured
In at least part of video whether repeat with described one section of template video;Otherwise, it is determined that described one section to be measured regards
Frequency is not repeated completely with described To Template video, and no longer judges at least partly regarding in one section of video to be measured
Whether frequency is repeated with described one section of template video.
In one embodiment, described device also includes:
Matching module, for judging at least part of video in described one section of video to be measured and a described class mould
Before the plate features corresponding To Template video of vector repeats, when described one section of video to be measured target characteristic to
When vector distance between amount and a described class template characteristic vector is less than or equal to described default vector distance,
By each the target mould in each picture to be measured in described one section of video to be measured and described To Template video
Plate picture is mated successively;
Described decision sub-module is additionally operable to:
Vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From closest to described default vector distance and described less than or equal to default vector distance, described vector distance
In one section of video to be measured with described To Template video in the To Template picture picture to be measured that matches
When number is higher than preset number, judge at least part of video in described one section of video to be measured and described target mould
Plate video repeats;Otherwise, it is determined that at least part of video in described one section of video to be measured and described To Template
Video does not repeat completely;
Described device also includes:
Find module, for judging at least part of video in described one section of video to be measured and described target mould
After plate video does not repeat completely, find and described one section of video to be measured in multiclass template characteristic vector successively
Vector distance between target feature vector closest to described default vector distance other class template features to
Amount;
Judge module is corresponding with other class template characteristic vectors described in described one section of video to be measured for judging
To Template video in the number of picture to be measured that matches of To Template picture whether higher than described pre-
If number, and circulate successively, with other class template characteristic vectors pair described in described one section of video to be measured
The number of the picture to be measured that the Target Photo in the To Template video answered matches is higher than described preset number,
Or in every section of video to be measured in multistage video to be measured with other corresponding targets of class template characteristic vector described
The number of the picture to be measured that the To Template picture in template video matches is below described preset number.
In one embodiment, described device also includes:
Modular converter, for the characteristic vector in the described Target Photo of acquisition, and the spy by described Target Photo
Before levying the template characteristic vector as template video every section described for the vector, described target icon is carried out dct
Conversion, obtains the horizontal coefficient of pixel coordinate and longitudinal coefficient after dct changes;
Described 3rd acquisition module includes:
Generate submodule, for choosing in front q1 potential coefficient and described longitudinal coefficient in described horizontal coefficient
Front q2 potential coefficient, every section of described template is generated according to described front q1 potential coefficient and described front q2 potential coefficient
The template characteristic vector of video.
In one embodiment, the m in every section of described template video is being opened the same position in template picture
The pixel value of the target pixel points put is weighted suing for peace, and described m is opened template picture and synthesizes one
During target icon, the kth that described m opens in model picture opens the pixel value of the target pixel points of model picture
Shared weight α in the pixel value of the target pixel points of described Target Photok=γ∧k, wherein, γ is little
In 1 constant.
The technical scheme that embodiment of the disclosure provides can include following beneficial effect:
Can be according to the template characteristic vector of the target feature vector of every section of video to be measured and every section of template video
Between vector distance, determine the similarity of every section of video to be measured and every section of template video, and then determine this section
Whether at least part of video in video to be measured is repeated with this section of template video, and then realizes carrying out exactly
Video repeatability detection, and guarantee precision and the efficiency of video repeatability detection, and finally realize true exactly
Concrete broadcast start time in video to be measured for the solid plate video, thus realize carrying out programming predictions exactly.
Other features and advantages of the present invention will illustrate in the following description, and, partly from explanation
Become apparent in book, or understood by implementing the present invention.The purpose of the present invention and other advantages can
Realized by specifically noted structure in the description write, claims and accompanying drawing and obtain
?.
Below by drawings and Examples, technical scheme is described in further detail.
Brief description
Accompanying drawing is used for providing a further understanding of the present invention, and constitutes a part for description, with this
Bright embodiment is used for explaining the present invention together, is not construed as limiting the invention.In the accompanying drawings:
Fig. 1 is a kind of flow chart of the determination methods of repetition video according to an exemplary embodiment.
Fig. 2 is a kind of block diagram of the judgment means of repetition video according to an exemplary embodiment.
Fig. 3 is the block diagram of another kind of judgment means repeating video according to an exemplary embodiment.
Fig. 4 is the block diagram of the judgment means of another repetition video according to an exemplary embodiment.
Fig. 5 is the block diagram of the judgment means of another repetition video according to an exemplary embodiment.
Specific embodiment
Below in conjunction with accompanying drawing, the preferred embodiments of the present invention are illustrated it will be appreciated that described herein
Preferred embodiment is merely to illustrate and explains the present invention, is not intended to limit the present invention.
In correlation technique, in electric program menu (epg, electronic program guide) processing system
In system, need to carry out program real-time estimate.And predict that program actually will carry out video repeatability detection.
But manually carry out repeatability detection task amount very big, less efficient, and repeatability detection accuracy rate and
Precision all ratios are relatively low, therefore, it is impossible to carry out program real-time estimate exactly, this brings very very much not to user
Just.
In order to solve above-mentioned technical problem, the embodiment of the present disclosure provides a kind of determination methods repeating video,
The method be applied to the determining program of repetition video, system or device, as shown in figure 1,
In step s101, from live video stream, obtain during multistage video to be measured it is preferable that making to appoint
Anticipate and there is between two sections of adjacent videos to be measured the frame of video of part coupling, for example: treat for two sections of arbitrary neighborhood
Survey the coincidence factor having 50% between video, such as: one whole section of live video stream is mnpqrsdpddqc, permissible
Intercept into mnpqrs, tri- sections of videos to be measured of qrsdpd, dpddqc, so can avoid identical with template video
Video to be measured be divided into two sections and to contrast with template video respectively, so that it is guaranteed that the seriality of video to be measured,
Certainly, every section of video to be measured can include an independent program or advertisement.
In step s102, obtain the target feature vector of every section of video to be measured in multistage video to be measured,
Wherein, target feature vector is used for every section of unique mark video to be measured;Target feature vector is used for unique mark
Every section of video to be measured, be every section of video to be measured fingerprint be every section of video to be measured video finger print.
In step s103, extract multistage template video from template video storehouse, wherein, multistage template regards
M is had to open template picture in every section of template video in frequency;
Template video storehouse can be several customized channel deriving, that is, interior history program, wide
The video data such as accuse, be stored in video template storehouse, for example: can be interior several
The video datas such as the program of customized channel, advertisement, specifically: the institute from program start time to finish time
There are the pictures such as program, advertisement, therefore, the use of this known template video storehouse (can be multiple programs or wide
Accuse, and every section of template video can include an independent advertisement or program) can be in the reality of all channels
When program stream in search for this program or advertisement, and when searching this program or advertisement in real time programme stream,
Can determine that this program or advertisement concrete time started in the real time programme stream play, thus realizing
Programming predictions.
Secondly, when the m in obtaining every section of template video opens template picture, can be from template video storehouse
Extracted every 5 (or 2) pictures using down-sampled mode in all template picture of every section of template video
Once, so both decrease the process of same template picture, decrease the data processing amount of template picture again,
Rational sample rate does not result in the loss of important information yet.
In addition, when extraction m opens template picture, every template picture can be carried out using Gaussian function
Smooth treatment, and picture unification is reduced into size for w*h=720*480, to be uniformly processed,
Furthermore it is preferred that there being 50% coincidence factor in multistage template video between two sections of template video of arbitrary neighborhood,
As: template video storehouse is abcdefghigklmn, can intercept into abcdef, defghi, ghigkl, gklmn
Four sections of template video, in order to guarantee seriality with template video when comparing for the video to be measured, certainly, by
Represent a program or advertisement in every section of video, and the duration of different programs or advertisement is different, so every section
The template picture that template video includes possible different, namely the corresponding m of every section of template video is different;
And if prediction programme information, every section of template video can be the video-frequency band of program head;If prediction advertisement letter
Breath, every section of template video can choose the video-frequency band of advertisement.
In step s104, the m in every section of template video is opened the target of the same position in template picture
The pixel value of pixel is weighted suing for peace, and m is opened template picture and synthesizes a Target Photo, its
In, the pixel value of the target pixel points of every template picture is in the pixel value of the target pixel points of Target Photo
Shared weight and be equal to 1;
By the m in every section of template video being opened the picture of the target pixel points of the same position in template picture
Plain value is weighted suing for peace, and m can be opened template picture and synthesize a representative Target Photo,
Make the characteristic synthetic feature of m pictures of this Target Photo, so that the feature of this Target Photo
Vector can represent the template characteristic vector of this section of template video;And the target pixel points of every template picture
Pixel value in the pixel value of the target pixel points of Target Photo shared weight and be equal to 1, then be to ensure that this
M open template picture can synthesize pictures essential condition so that synthesis after each target pixel points
Pixel value still open the maximum in this target pixel points for the every template picture in template picture less than m
Pixel value, wherein, weighted sum is exactly that the weight numeral of m pictures adds and is equal to 1,
A1*p1+a2*p2 ...=p, wherein p1, p2 ... is the pixel value of the target pixel points that m opens every figure of in figure, p
It is the pixel value of this target pixel points in this Target Photo, a1, a2, a3 ... are corresponding to be in m pictures
Fig. 1, the pixel value of 2,3 ... certain target pixel points accounts for the weight of this target pixel points in this Target Photo,
Weighted value is all nonnegative number, a1+a2+ ...=1.
In step s105, obtain the characteristic vector of Target Photo, and the characteristic vector of Target Photo is made
Template characteristic vector for every section of template video;Template characteristic vector is used for every section of template video of unique mark,
For every section of template video fingerprint be every section of template video video finger print.
In step s106, calculate the target feature vector of every section of video to be measured and the mould of every section of template video
Vector distance between plate features vector;
Wherein, vector distance is used for characterizing the similarity between every section of video to be measured and every section of template video, and
Vector distance is less, and similarity is higher;By calculating the target feature vector of every section of video to be measured and every successively
Section template video template characteristic vector between vector distance, can accurately determine every section of video to be measured with
Similarity between every section of template video, accurately locking is to be measured with every section of template video repetitive rate highest regards
Frequency range is it is ensured that the precision of video repeatability detection and efficiency, and prevents detection and omit, and vector distance
Less, similarity is higher, illustrates that this section of video to be measured is higher with the repetitive rate of this section of template video.
In step s107, when the target feature vector of one section of video to be measured and the template of one section of template video
When vector distance between characteristic vector is less than or equal to default vector distance, judge in one section of video to be measured
At least partly video and one section of template video repeat;Otherwise, it is determined that one section of video to be measured and one section of template video
Do not repeat completely.
Between the target feature vector and the template characteristic vector of this section of template video of this section of video to be measured
When vector distance is less than or equal to default vector distance, the phase of this section of video to be measured and this section of template video is described
Seemingly spend high, it is high with the repetitive rate of this section of template video to further relate to this section of video to be measured, then it is considered that
At least part of video in this section of video to be measured is repeated with this section of template video, that is, in this section of video to be measured extremely
Small part video is the data in this section of template video, and then is easy to determine that this section of template video is to be measured at this section
Starting position in video, thus to the program play in real-time stream, (i.e. this template regards exactly
Frequently) it is predicted.Certainly, when this vector distance is more than and generally presets vector distance, illustrate that this section to be measured regards
Frequency is extremely low with the similarity of this section of template video, and then the repetitive rate of this section of video to be measured and this section of template video
Extremely low, then it is considered that this section of video to be measured is not repeated completely with this section of template video;In addition, should determining
After at least part of video in section video to be measured is repeated with this section of template video, should in order to accurately determine further
Which picture to be measured in section video to be measured is mated with which pictures in this section of template video (repeating), or
Person, in order to further accurately determine repetitive rate, by every picture to be measured in this section of video to be measured and can be somebody's turn to do
Every pictures in section template video mate one by one.
In one embodiment, the target feature vector of every section of video to be measured and every section of template video are being calculated
Before vector distance between template characteristic vector, method also includes:
The template characteristic vector of multistage template video is clustered, to obtain many class templates in template video storehouse
Characteristic vector, wherein, every class template characteristic vector is all a cluster centre;
Because every section of template video all corresponds to a template characteristic vector, and the hop count mistake including in template video
When many, the number of template characteristic vector will be excessive, or even has much incoherent template characteristic vectors,
Therefore, if directly calculate the template characteristic of every section of video object characteristic vector to be measured and every section of template video to
Vector distance between amount, will necessarily make due to template characteristic vector excessively, and the meter of impact vector distance
Calculate speed, increase the computation burden of system, therefore, by by the mould of template video essentially identical for similarity
Plate features vector is clustered, so that the number of vectors of the template characteristic vector after cluster significantly subtracts
Few (for example: the quantity of the vector in template characteristic vector after cluster is 5% before clustering).
At least part of video in judging one section of video to be measured and one section of template video repeat;Otherwise, it is determined that
Before one section of video to be measured and one section of template video do not repeat completely, method also includes:
Target feature vector and a class template characteristic vector (i.e. a certain apoplexy due to endogenous wind of birdsing of the same feather flock together when one section of video to be measured
The heart) between vector distance when being less than or equal to default vector distance, judge in one section of video to be measured at least
Partial video To Template video corresponding with a class template characteristic vector repeats, and start to judge one section to be measured
Whether at least part of video in video is repeated with one section of template video;Otherwise, it is determined that one section of video to be measured with
To Template video does not repeat completely, and no longer judge at least part of video in one section of video to be measured whether with
One section of template video repeats, and wherein, the corresponding To Template video of every class cluster centre is higher by similarity
Some sections of template video composition, that is, every class cluster centre all have several higher template characteristic of similarity to
Amount composition, wherein, this several corresponding video of template characteristic vector is this some sections of template video, and should
One section of video to be measured is one section of video to be measured in this corresponding To Template video of class template characteristic vector.
Due to judging whether at least part of video in every section of video to be measured is repeated with every section of template video one by one
Workload very big, the processing load of system is very heavy, therefore, by mould essentially identical for similarity
After the template characteristic vector of plate video is clustered, can first calculate the target characteristic of every section of video to be measured to
Vector distance between amount and every class template characteristic vector, if this vector distance is less than default vector distance,
Then illustrate that this section of video to be measured is high with the similarity of this To Template video, and then can tentatively judge this section
At least part of video in video to be measured is repeated with such corresponding To Template video of template characteristic vector, and
For the concrete airtime of template video library in further accurate lock live video stream, can sentence further
Whether vectorial with such template characteristic corresponding To Template of at least part of video in disconnected this section of video to be measured
A certain section of template video in video repeats, and then is easy to lock every section of template in template video storehouse exactly
Particular location in live video stream for the video namely concrete airtime;Certainly, if this vector distance is big
In or be equal to default vector distance, then the similarity pole of this section of video to be measured and this To Template video is described
Low, basic can consider is not repeated completely, then be completely unnecessary and determine whether in this section of video to be measured
At least partly whether video is repeated with certain section of template video in this To Template video, thus alleviating system
Video duplicate detection and prediction burden.In addition, k-means clustering algorithm can be used during cluster.
In one embodiment, at least part of video in judging one section of video to be measured and a class template feature
Before the corresponding To Template video of vector repeats, method also includes:
Vector distance between the target feature vector of one section of video to be measured and a class template characteristic vector is little
In or when being equal to default vector distance, by each picture to be measured in one section of video to be measured and To Template video
In each To Template picture mated successively;
If this vector distance is less than default vector distance, can only tentatively judge in this section of video to be measured at least
Partial video is repeated with such corresponding To Template video of template characteristic vector, and in order to further accurately
Determine that at least part of video in this section of video to be measured is regarded with such corresponding To Template of template characteristic vector
Whether the similarity of frequency really repeats, and needs each picture to be measured deeply comparing in this section of video to be measured and is somebody's turn to do
Whether each the To Template picture in To Template video mates, to guarantee the precision of video duplicate detection,
And then guarantee the programming predictions precision in live video stream.
Vector distance between the target feature vector of one section of video to be measured and a class template characteristic vector is little
In or when being equal to default vector distance, judge at least part of video in one section of video to be measured and class template spy
Levy vectorial corresponding To Template video to repeat;Otherwise, it is determined that one section of video to be measured is complete with To Template video
Entirely do not repeat, comprising:
Vector distance between the target feature vector of one section of video to be measured and a class template characteristic vector is little
In or be equal to default vector distance, described vector distance closest to described default vector distance (i.e. this class mould
Plate features vector be immediate cluster centre) and one section of video to be measured in To Template video in mesh
When the number of the picture to be measured that mark template picture matches is higher than preset number, judge in one section of video to be measured
At least partly video and To Template video repeat;Otherwise, it is determined that at least partly regarding in one section of video to be measured
Frequency is not repeated completely with To Template video;
Vector distance between the target feature vector and such template characteristic vector of this section of video to be measured is little
In or be equal to default vector distance and such template characteristic vector closest to this section of video to be measured target characteristic
When vectorial, if picture to be measured matches number higher than pre- with corresponding To Template picture in this section of video to be measured
If during number, illustrate that this section of video to be measured is higher with the similarity of To Template video, and then can be determined that this
At least part of video in video to be measured is repeated with this To Template video, otherwise, this section of video to be measured is described
Relatively low with the similarity of this To Template video, and then can be determined that at least part of video in this video to be measured
Do not repeat completely with this To Template video, certainly, in the target feature vector determining a certain section of video to be measured
Vector distance and a certain class template characteristic vector between whether this default vector distance of proximity when, need count
Calculate the every class template characteristic vector in the target feature vector and multiclass template characteristic vector of this section of video to be measured
Between vector distance, then each vector distance is compared and can therefrom select with this default vector distance
Select out the immediate vector distance with this default vector distance, immediate with this default vector distance from determining
The To Template video of a certain class template characteristic vector.
Method also includes:
After at least part of video in judging one section of video to be measured and To Template video do not repeat completely, according to
Secondary multiclass template characteristic vector in find and the target feature vector of one section of video to be measured between vector away from
From other class template characteristic vectors (being cluster centre close to other times) closest to default vector distance;
After at least part of video in judging this section of video to be measured and this To Template video do not repeat completely,
The vector and target feature vector of this section of video to be measured between can be found in multiclass template characteristic vector
Distance closest to other near class template characteristic vectors of default vector distance, with judge secondary near-lying mode plate features to
Measure whether corresponding To Template video is repeated with this section of video to be measured.
Judge in one section of video to be measured with the mesh in other class template characteristic vectors corresponding To Template video
Whether the number of the picture to be measured that mark template picture matches is higher than preset number, and circulate successively, Zhi Daoyi
With the Target Photo phase in other class template characteristic vectors corresponding To Template video in section video to be measured
The number of the picture to be measured joined be higher than preset number, or multistage video to be measured in every section of video to be measured in its
The picture to be measured that To Template picture in his class template characteristic vector corresponding To Template video matches
Number be below preset number.
Arrow between the target feature vector and certain other near-lying mode plate features vector of this section of video to be measured
Span from less than or equal to default vector distance, if picture to be measured and this other near-lying mode in this section of video to be measured
The plate features corresponding To Template picture of vector matches number when being higher than preset number, illustrates that this section to be measured regards
The similarity of frequency and this corresponding To Template video of other near-lying mode plate features vector is higher, and then can be determined that
At least part of video in this video to be measured and this corresponding To Template video of other near-lying mode plate features vector
Repeat, otherwise, this section of video to be measured and this corresponding To Template video of other near-lying mode plate features vector is described
Similarity relatively low, and then can determine at least part of video in this video to be measured and this other near-lying mode plate
The corresponding To Template video of characteristic vector does not repeat completely;Certainly, if in this section of video to be measured with all its
What the To Template picture in his time corresponding To Template video of near-lying mode plate features vector matched treats mapping
The number of piece is below preset number, then illustrate that this section of video to be measured is not repeated with whole template video storehouse,
Circulation then can be jumped out, the video to be measured of other sections of reselection is mated with To Template video, with accurate
Determine in this live video stream, whether to have play this template video storehouse, and have this template video if play
Storehouse, it specifically plays the time of this template video.
In one embodiment, in the characteristic vector obtaining Target Photo, and the characteristic vector by Target Photo
Before the template characteristic vector of every section of template video, method also includes:
Target icon is carried out dct conversion, obtains after dct conversion the horizontal coefficient of pixel coordinate and vertical
To coefficient;
Above-mentioned steps s105 can be performed as:
Choose the front q2 potential coefficient in the front q1 potential coefficient and longitudinal coefficient in horizontal coefficient, according to front q1
Potential coefficient and front q2 potential coefficient generate the template characteristic vector of every section of template video.
When generating template characteristic vector, in horizontal coefficient and longitudinal coefficient, more forward several potential coefficients are characterizing
During this Target Photo, shared weight is bigger, and the template characteristic vector of generation more can accurately represent this target figure
Piece, and if the coefficient digit selecting is more, the amount of calculation of system will be bigger, and the computation burden of system is just
Can be heavier, the real-time of the Target Photo characteristic vector calculating will be lower, therefore, is generating target figure
During piece characteristic vector, can be with integration objective picture feature vector permissible accuracy, the computing capability of system and spy
Front q2 position in front q1 potential coefficient and longitudinal coefficient that the requirement of real-time levying vector is chosen in horizontal coefficient
Coefficient, to meet the real-time of the required precision of characteristic vector, the computing capability of system and characteristic vector simultaneously
Require, wherein, q1 and q2 is preset value it is preferable that the value of q1 and q2 is 1.
In one embodiment, the m in every section of template video is being opened the same position in template picture
The pixel value of target pixel points is weighted suing for peace, and m is opened template picture and synthesizes a target icon
When, the kth that m opens in model picture opens the mesh in Target Photo for the pixel value of the target pixel points of model picture
Shared weight α in the pixel value of mark pixelk=γ∧k, wherein, γ is the constant less than 1.
When m being opened template picture synthesizing a target icon, the kth that this m opens in model picture is opened
The object pixel of correspondence position in this Target Photo for the pixel value of each target pixel points of model picture
Shared weight α in the pixel value of pointk=γ∧k, wherein, γ is preferably 0.65, certainly, for certain
For one target pixel points, the pixel value of this target pixel points of every template picture is in this Target Photo
Shared weight α in the pixel value of this target pixel pointskAnd should be less than or equal to 1, certainly, due to
αkIt is an exponential form, therefore, m opens the corresponding α of template picturekBe likely larger than 1, at this
Only need to be to each weight α in the case of kindkIt is normalized.
As shown in Fig. 2 the present invention also provides a kind of judgment means repeating video, comprising: the first acquisition mould
Block 201, is configured to from live video stream, obtains multistage video to be measured;
Second acquisition module 202, is configured to obtain the target of every section of video to be measured in multistage video to be measured
Characteristic vector, wherein, described target feature vector is configured to the described video to be measured of every section of unique mark;
Processing module 203, is configured to extract multistage template video, wherein, multistage from template video storehouse
M is had to open template picture in every section of template video in template video;
Summation module 204, is configured to open the m in every section of described template video same in template picture
The pixel value of the target pixel points of position is weighted suing for peace, and described m is opened template picture and synthesizes one
Open Target Photo, wherein, the pixel value of the target pixel points of every described template picture is in described Target Photo
The pixel value of target pixel points in shared weight and be equal to 1;
3rd acquisition module 205, is configured to obtain the characteristic vector of described Target Photo, and by described mesh
Mark on a map piece characteristic vector as every section described template video template characteristic vector;
Computing module 206, is configured to calculate the target feature vector of every section of described video to be measured and every section of institute
State the vector distance between the template characteristic vector of template video;
Determination module 207, is configured as the target feature vector of one section of video to be measured and one section of template video
Template characteristic vector between vector distance when being less than or equal to default vector distance, judge described one section and treat
At least part of video surveyed in video is repeated with described one section of template video;Otherwise, it is determined that described one section to be measured
Video is not repeated completely with described one section of template video.
As shown in figure 3, in one embodiment, described device also includes:
Cluster module 301, is configured to calculating the target feature vector of every section of described video to be measured and every section
Before vector distance between the template characteristic vector of described template video, by the mould of template video described in multistage
Plate features vector is clustered, to obtain multiclass template characteristic vector;
Described determination module 207 includes:
Decision sub-module 2071, be configured to judge described one section of video to be measured at least part of video with
Described one section of template video repeats;Otherwise, it is determined that described one section of video to be measured is complete with described one section of template video
Before entirely not repeating, between the target feature vector of described one section of video to be measured and a class template characteristic vector
Vector distance when being less than or equal to default vector distance, judge at least part of in described one section of video to be measured
Video To Template video corresponding with a described class template characteristic vector repeats, and starts to judge described one section
Whether at least part of video in video to be measured is repeated with described one section of template video;Otherwise, it is determined that described one
Section video to be measured is not repeated completely with described To Template video, and no longer judges in described one section of video to be measured
At least part of video whether repeat with described one section of template video.
As shown in figure 4, in one embodiment, described device also includes:
Matching module 401, is configured at least part of video in judging described one section of video to be measured and institute
Before stating the repetition of a class template characteristic vector corresponding To Template video, when the mesh of described one section of video to be measured
Vector distance between mark characteristic vector and a described class template characteristic vector is less than or equal to described default arrow
Span from when, will be every in each picture to be measured in described one section of video to be measured and described To Template video
Individual To Template picture is mated successively;
Described decision sub-module 207 is also configured to
Vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From closest to described default vector distance and described less than or equal to default vector distance, described vector distance
In one section of video to be measured with described To Template video in the To Template picture picture to be measured that matches
When number is higher than preset number, judge at least part of video in described one section of video to be measured and described target mould
Plate video repeats;Otherwise, it is determined that at least part of video in described one section of video to be measured and described To Template
Video does not repeat completely;
Described device also includes:
Find module 402, be configured at least part of video in judging described one section of video to be measured and institute
State after To Template video do not repeat completely, find in multiclass template characteristic vector successively and treat with described one section
Survey the vector distance between the target feature vector of video closest to other class moulds of described default vector distance
Plate features vector;
Judge module 403, be configured to judge in described one section of video to be measured with other class template features described
Whether the number of the picture to be measured that the To Template picture in the corresponding To Template video of vector matches is high
In described preset number, and circulate successively, special with other class templates described in described one section of video to be measured
The number levying the picture to be measured that the Target Photo in vectorial corresponding To Template video matches is higher than described
With other class template characteristic vectors pair described in every section of video to be measured in preset number, or multistage video to be measured
The number of the picture to be measured that the To Template picture in the To Template video answered matches is below described pre-
If number.
As shown in figure 5, in one embodiment, described device also includes:
Modular converter 501, is configured in the characteristic vector obtaining described Target Photo, and by described target
Before the template characteristic vector as template video every section described for the characteristic vector of picture, by described target icon
Carry out dct conversion, obtain the horizontal coefficient of pixel coordinate and longitudinal coefficient after dct changes;
3rd acquisition module 205 includes:
Generate submodule 2051, be configured to choose front q1 potential coefficient in described horizontal coefficient and described vertical
To the front q2 potential coefficient in coefficient, generate every section according to described front q1 potential coefficient and described front q2 potential coefficient
The template characteristic vector of described template video.
In one embodiment, the m in every section of described template video is being opened the same position in template picture
The pixel value of the target pixel points put is weighted suing for peace, and described m is opened template picture and synthesizes one
During target icon, the kth that described m opens in model picture opens the pixel value of the target pixel points of model picture
Shared weight α in the pixel value of the target pixel points of described Target Photok=γ∧k, wherein, γ is little
In 1 constant.
With regard to the device in above-described embodiment, wherein the concrete mode of unit execution operation is relevant
It has been described in detail in the embodiment of the method, explanation will be not set forth in detail herein.
Finally, the judgment means of the repetition video in the present invention are applied to terminal unit.For example, it may be moving
Mobile phone, computer, digital broadcast terminal, messaging devices, game console, tablet device, doctor
Treatment equipment, body-building equipment, personal digital assistant etc..
Those skilled in the art are it should be appreciated that embodiments of the invention can be provided as method, system or meter
Calculation machine program product.Therefore, the present invention can be using complete hardware embodiment, complete software embodiment or knot
Close the form of the embodiment of software and hardware aspect.And, the present invention can adopt and wherein wrap one or more
Computer-usable storage medium containing computer usable program code (including but not limited to disk memory and
Optical memory etc.) the upper computer program implemented form.
The present invention is to produce with reference to method according to embodiments of the present invention, equipment (system) and computer program
The flow chart of product and/or block diagram are describing.It should be understood that can by computer program instructions flowchart and
/ or block diagram in each flow process and/or the flow process in square frame and flow chart and/or block diagram and/
Or the combination of square frame.These computer program instructions can be provided to general purpose computer, special-purpose computer, embed
The processor of formula datatron or other programmable data processing device is to produce a machine so that passing through to calculate
The instruction of the computing device of machine or other programmable data processing device produces for realizing in flow chart one
The device of the function of specifying in individual flow process or multiple flow process and/or one square frame of block diagram or multiple square frame.
These computer program instructions may be alternatively stored in and computer or other programmable datas can be guided to process and set
So that being stored in this computer-readable memory in the standby computer-readable memory working in a specific way
Instruction produce and include the manufacture of command device, the realization of this command device is in one flow process or multiple of flow chart
The function of specifying in flow process and/or one square frame of block diagram or multiple square frame.
These computer program instructions also can be loaded in computer or other programmable data processing device, makes
Obtain and series of operation steps is executed on computer or other programmable devices to produce computer implemented place
Reason, thus the instruction of execution is provided for realizing in flow chart one on computer or other programmable devices
The step of the function of specifying in flow process or multiple flow process and/or one square frame of block diagram or multiple square frame.
Obviously, those skilled in the art can carry out various changes and modification without deviating from this to the present invention
Bright spirit and scope.So, if the present invention these modification and modification belong to the claims in the present invention and
Within the scope of its equivalent technologies, then the present invention is also intended to comprise these changes and modification.
Claims (10)
1. a kind of determination methods repeating video are it is characterised in that include:
From live video stream, obtain multistage video to be measured;
Obtain the target feature vector of every section of video to be measured in multistage video to be measured, wherein, described target is special
Levy vector and be used for the described video to be measured of every section of unique mark;
Extract multistage template video from template video storehouse, wherein, every section of template in multistage template video regards
M is had to open template picture in frequency;
M in every section of described template video is opened the picture of the target pixel points of the same position in template picture
Plain value is weighted suing for peace, and described m is opened template picture and synthesizes a Target Photo, wherein, often
The pixel of the target pixel points in described Target Photo for the pixel value of the target pixel points of Zhang Suoshu template picture
In value shared weight and be equal to 1;
Obtain the characteristic vector of described Target Photo, and using the characteristic vector of described Target Photo as every section of institute
State the template characteristic vector of template video;
Calculate the target feature vector of every section of described video to be measured and the template characteristic of every section of described template video
Vector distance between vector;
Between the target feature vector and the template characteristic vector of one section of template video of one section of video to be measured
When vector distance is less than or equal to default vector distance, judge at least partly regarding in described one section of video to be measured
Frequency is repeated with described one section of template video;Otherwise, it is determined that described one section of video to be measured is regarded with described one section of template
Frequency does not repeat completely.
2. method described in 1 is wanted according to right it is characterised in that
Special in the template of the target feature vector calculating every section of described video to be measured and every section of described template video
Before levying the vector distance between vector, methods described also includes:
The template characteristic vector of template video described in multistage is clustered, with obtain multiclass template characteristic to
Amount;
At least part of video in judging described one section of video to be measured is repeated with described one section of template video;No
Then, before judging that described one section of video to be measured and described one section of template video do not repeat completely, methods described is also
Including:
Vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From during less than or equal to default vector distance, judge at least part of video in described one section of video to be measured and institute
State the corresponding To Template video of a class template characteristic vector to repeat, and start to judge described one section of video to be measured
In at least part of video whether repeat with described one section of template video;Otherwise, it is determined that described one section to be measured regards
Frequency is not repeated completely with described To Template video, and no longer judges at least partly regarding in one section of video to be measured
Whether frequency is repeated with described one section of template video.
3. method described in 2 is wanted it is characterised in that in judging described one section of video to be measured according to right
At least part of video To Template video corresponding with a described class template characteristic vector repeat before, described
Method also includes:
Arrow between the target feature vector of described one section of video to be measured and a described class template characteristic vector
Span, from during less than or equal to described default vector distance, each in described one section of video to be measured is treated mapping
Piece is mated successively with each the To Template picture in described To Template video;
Arrow between the described target feature vector when described one section of video to be measured and a class template characteristic vector
Span, from during less than or equal to default vector distance, judges at least part of video in described one section of video to be measured
To Template video corresponding with a described class template characteristic vector repeats;Otherwise, it is determined that described one section to be measured
Video is not repeated completely with described To Template video, comprising:
Vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From closest to described default vector distance and described less than or equal to default vector distance, described vector distance
In one section of video to be measured with described To Template video in the To Template picture picture to be measured that matches
When number is higher than preset number, judge at least part of video in described one section of video to be measured and described target mould
Plate video repeats;Otherwise, it is determined that at least part of video in described one section of video to be measured and described To Template
Video does not repeat completely;
Methods described also includes:
At least part of video in judging described one section of video to be measured is with described To Template video completely not
After repetition, find the target feature vector with described one section of video to be measured successively in multiclass template characteristic vector
Between vector distance closest to described default vector distance other class template characteristic vectors;
Judge to regard with other corresponding To Templates of class template characteristic vector described in described one section of video to be measured
Whether the number of the picture to be measured that the To Template picture in frequency matches is higher than described preset number, and successively
Circulation, is regarded with other corresponding To Templates of class template characteristic vector described in described one section of video to be measured
The number of the picture to be measured that the Target Photo in frequency matches is higher than described preset number, or multistage video to be measured
In every section of video to be measured in the mesh in other class template characteristic vectors described corresponding To Template video
The number of the picture to be measured that mark template picture matches is below described preset number.
4. method described in 1 is wanted it is characterised in that in the feature obtaining described Target Photo according to right
Vector, and using the characteristic vector of described Target Photo as the template characteristic vector of template video every section described
Before, methods described also includes:
Described target icon is carried out dct conversion, obtains the horizontal coefficient of pixel coordinate after dct conversion
With longitudinal coefficient;And
The described characteristic vector obtaining described Target Photo, and using the characteristic vector of described Target Photo as every
The template characteristic vector of Duan Suoshu template video, comprising:
Choose the front q2 potential coefficient in the front q1 potential coefficient and described longitudinal coefficient in described horizontal coefficient,
Generate the template characteristic of every section of described template video according to described front q1 potential coefficient and described front q2 potential coefficient
Vector.
5. method any one of 1 to 4 is wanted according to right it is characterised in that
In the target pixel points that the m in every section of described template video is opened the same position in template picture
Pixel value is weighted suing for peace, when described m being opened template picture synthesizing a target icon, described
The kth that m opens in model picture opens the mesh in described Target Photo for the pixel value of the target pixel points of model picture
Shared weight α in the pixel value of mark pixelk=γ∧k, wherein, γ is the constant less than 1.
6. a kind of judgment means repeating video are it is characterised in that include:
First acquisition module, for, from live video stream, obtaining multistage video to be measured;
Second acquisition module, for obtain the target characteristic of every section of video to be measured in multistage video to be measured to
Amount, wherein, described target feature vector is used for the described video to be measured of every section of unique mark;
Processing module, for extracting multistage template video, wherein, multistage template video from template video storehouse
In every section of template video in have m to open template picture;
Summation module, for opening the same position in template picture by the m in every section of described template video
The pixel value of target pixel points is weighted suing for peace, and described m is opened template picture and synthesizes a target
Picture, wherein, the pixel value of the target pixel points of every described template picture is in the target of described Target Photo
In the pixel value of pixel shared weight and be less than or equal to 1;
3rd acquisition module, for obtaining the characteristic vector of described Target Photo, and by described Target Photo
Characteristic vector is as the template characteristic vector of template video every section described;
Computing module, target feature vector and every section of described template for calculating every section of described video to be measured regard
Vector distance between the template characteristic vector of frequency;
Determination module, for as the target feature vector of one section of video to be measured and the template spy of one section of template video
When levying vector distance between vector and being less than or equal to default vector distance, judge in described one section of video to be measured
At least part of video repeat with described one section of template video;Otherwise, it is determined that described one section of video to be measured and institute
State one section of template video not repeat completely.
7. the device described in 6 is wanted it is characterised in that described device also includes according to right:
Cluster module, in the target feature vector calculating every section of described video to be measured and every section of described template
Video template characteristic vector between vector distance before, by the template characteristic of template video described in multistage to
Amount is clustered, to obtain multiclass template characteristic vector;
Described determination module includes:
Decision sub-module, at least part of video in the described one section of video to be measured of judgement and described one section
Template video repeats;Otherwise, it is determined that described one section of video to be measured is not repeated completely with described one section of template video
Before, the vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From during less than or equal to default vector distance, judge at least part of video in described one section of video to be measured and institute
State the corresponding To Template video of a class template characteristic vector to repeat, and start to judge described one section of video to be measured
In at least part of video whether repeat with described one section of template video;Otherwise, it is determined that described one section to be measured regards
Frequency is not repeated completely with described To Template video, and no longer judges at least partly regarding in one section of video to be measured
Whether frequency is repeated with described one section of template video.
8. the device described in 7 is wanted it is characterised in that described device also includes according to right:
Matching module, for judging at least part of video in described one section of video to be measured and a described class mould
Before the plate features corresponding To Template video of vector repeats, when described one section of video to be measured target characteristic to
When vector distance between amount and a described class template characteristic vector is less than or equal to described default vector distance,
By each the target mould in each picture to be measured in described one section of video to be measured and described To Template video
Plate picture is mated successively;
Described decision sub-module is additionally operable to:
Vector between the target feature vector of described one section of video to be measured and a class template characteristic vector away from
From less than or equal to default vector distance, described vector distance is closest to described default vector distance and institute
State in one section of video to be measured with described To Template video in the picture to be measured that matches of To Template picture
Number when being higher than preset number, judge at least part of video and the described target in described one section of video to be measured
Template video repeats;Otherwise, it is determined that at least part of video in described one section of video to be measured and described target mould
Plate video does not repeat completely;
Described device also includes:
Find module, for judging at least part of video in described one section of video to be measured and described target mould
After plate video does not repeat completely, find and described one section of video to be measured in multiclass template characteristic vector successively
Vector distance between target feature vector closest to described default vector distance other class template features to
Amount;
Judge module is corresponding with other class template characteristic vectors described in described one section of video to be measured for judging
To Template video in the number of picture to be measured that matches of To Template picture whether higher than described pre-
If number, and circulate successively, with other class template characteristic vectors pair described in described one section of video to be measured
The number of the picture to be measured that the Target Photo in the To Template video answered matches is higher than described preset number,
Or in every section of video to be measured in multistage video to be measured with other corresponding targets of class template characteristic vector described
The number of the picture to be measured that the To Template picture in template video matches is below described preset number.
9. the device described in 6 is wanted it is characterised in that described device also includes according to right:
Modular converter, for the characteristic vector in the described Target Photo of acquisition, and the spy by described Target Photo
Before levying the template characteristic vector as template video every section described for the vector, described target icon is carried out dct
Conversion, obtains the horizontal coefficient of pixel coordinate and longitudinal coefficient after dct changes;
Described 3rd acquisition module includes:
Generate submodule, for choosing in front q1 potential coefficient and described longitudinal coefficient in described horizontal coefficient
Front q2 potential coefficient, every section of described template is generated according to described front q1 potential coefficient and described front q2 potential coefficient
The template characteristic vector of video.
10. device any one of 6 to 9 is wanted according to right it is characterised in that
In the target pixel points that the m in every section of described template video is opened the same position in template picture
Pixel value is weighted suing for peace, when described m being opened template picture synthesizing a target icon, described
The kth that m opens in model picture opens the mesh in described Target Photo for the pixel value of the target pixel points of model picture
Shared weight α in the pixel value of mark pixelk=γ∧k, wherein, γ is the constant less than 1.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510438517.3A CN106375781B (en) | 2015-07-23 | 2015-07-23 | A kind of judgment method and device repeating video |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510438517.3A CN106375781B (en) | 2015-07-23 | 2015-07-23 | A kind of judgment method and device repeating video |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106375781A true CN106375781A (en) | 2017-02-01 |
CN106375781B CN106375781B (en) | 2019-09-13 |
Family
ID=57880180
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510438517.3A Expired - Fee Related CN106375781B (en) | 2015-07-23 | 2015-07-23 | A kind of judgment method and device repeating video |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106375781B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108959492A (en) * | 2018-06-25 | 2018-12-07 | 江苏大学 | A kind of nearly repetition video detecting method based on Teoplitz core offset minimum binary |
CN109934142A (en) * | 2019-03-04 | 2019-06-25 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating the feature vector of video |
WO2019184522A1 (en) * | 2018-03-29 | 2019-10-03 | 北京字节跳动网络技术有限公司 | Method and apparatus for determining duplicate video |
CN110413603A (en) * | 2019-08-06 | 2019-11-05 | 北京字节跳动网络技术有限公司 | Determination method, apparatus, electronic equipment and the computer storage medium of repeated data |
CN111426693A (en) * | 2020-04-26 | 2020-07-17 | 湖南恒岳重钢钢结构工程有限公司 | Quality defect detection system and detection method thereof |
CN114422841A (en) * | 2021-12-17 | 2022-04-29 | 北京达佳互联信息技术有限公司 | Subtitle generating method, device, electronic equipment and storage medium |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1938714A (en) * | 2004-03-23 | 2007-03-28 | 英国电讯有限公司 | Method and system for semantically segmenting scenes of a video sequence |
US20080059287A1 (en) * | 2002-10-03 | 2008-03-06 | Polyphonic Human Media Interface S.L. | Method and system for video and film recommendation |
CN101359368A (en) * | 2008-09-09 | 2009-02-04 | 华为技术有限公司 | Video image clustering method and system |
CN101833650A (en) * | 2009-03-13 | 2010-09-15 | 清华大学 | Video copy detection method based on contents |
CN101853377A (en) * | 2010-05-13 | 2010-10-06 | 复旦大学 | Method for identifying content of digital video |
US20100318515A1 (en) * | 2009-06-10 | 2010-12-16 | Zeitera, Llc | Media Fingerprinting and Identification System |
CN101976258A (en) * | 2010-11-03 | 2011-02-16 | 上海交通大学 | Video semantic extraction method by combining object segmentation and feature weighing |
CN102693299A (en) * | 2012-05-17 | 2012-09-26 | 西安交通大学 | System and method for parallel video copy detection |
CN103593464A (en) * | 2013-11-25 | 2014-02-19 | 华中科技大学 | Video fingerprint detecting and video sequence matching method and system based on visual features |
-
2015
- 2015-07-23 CN CN201510438517.3A patent/CN106375781B/en not_active Expired - Fee Related
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080059287A1 (en) * | 2002-10-03 | 2008-03-06 | Polyphonic Human Media Interface S.L. | Method and system for video and film recommendation |
CN1938714A (en) * | 2004-03-23 | 2007-03-28 | 英国电讯有限公司 | Method and system for semantically segmenting scenes of a video sequence |
CN101359368A (en) * | 2008-09-09 | 2009-02-04 | 华为技术有限公司 | Video image clustering method and system |
CN101833650A (en) * | 2009-03-13 | 2010-09-15 | 清华大学 | Video copy detection method based on contents |
US20100318515A1 (en) * | 2009-06-10 | 2010-12-16 | Zeitera, Llc | Media Fingerprinting and Identification System |
CN101853377A (en) * | 2010-05-13 | 2010-10-06 | 复旦大学 | Method for identifying content of digital video |
CN101976258A (en) * | 2010-11-03 | 2011-02-16 | 上海交通大学 | Video semantic extraction method by combining object segmentation and feature weighing |
CN102693299A (en) * | 2012-05-17 | 2012-09-26 | 西安交通大学 | System and method for parallel video copy detection |
CN103593464A (en) * | 2013-11-25 | 2014-02-19 | 华中科技大学 | Video fingerprint detecting and video sequence matching method and system based on visual features |
Non-Patent Citations (1)
Title |
---|
焦淑红: "《多媒体信息***》", 31 August 2007 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019184522A1 (en) * | 2018-03-29 | 2019-10-03 | 北京字节跳动网络技术有限公司 | Method and apparatus for determining duplicate video |
CN110324660A (en) * | 2018-03-29 | 2019-10-11 | 北京字节跳动网络技术有限公司 | A kind of judgment method and device repeating video |
US11265598B2 (en) | 2018-03-29 | 2022-03-01 | Seijing Bytedance Network Technology Co., Ltd. | Method and device for determining duplicate video |
CN108959492A (en) * | 2018-06-25 | 2018-12-07 | 江苏大学 | A kind of nearly repetition video detecting method based on Teoplitz core offset minimum binary |
CN109934142A (en) * | 2019-03-04 | 2019-06-25 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating the feature vector of video |
CN109934142B (en) * | 2019-03-04 | 2021-07-06 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating feature vectors of video |
CN110413603A (en) * | 2019-08-06 | 2019-11-05 | 北京字节跳动网络技术有限公司 | Determination method, apparatus, electronic equipment and the computer storage medium of repeated data |
CN110413603B (en) * | 2019-08-06 | 2023-02-24 | 北京字节跳动网络技术有限公司 | Method and device for determining repeated data, electronic equipment and computer storage medium |
CN111426693A (en) * | 2020-04-26 | 2020-07-17 | 湖南恒岳重钢钢结构工程有限公司 | Quality defect detection system and detection method thereof |
CN114422841A (en) * | 2021-12-17 | 2022-04-29 | 北京达佳互联信息技术有限公司 | Subtitle generating method, device, electronic equipment and storage medium |
CN114422841B (en) * | 2021-12-17 | 2024-01-02 | 北京达佳互联信息技术有限公司 | Subtitle generation method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106375781B (en) | 2019-09-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106375781A (en) | Method and device for judging duplicate video | |
CN110716648B (en) | Gesture control method and device | |
CN111179419B (en) | Three-dimensional key point prediction and deep learning model training method, device and equipment | |
CN108234870A (en) | Image processing method, device, terminal and storage medium | |
CN107423306B (en) | Image retrieval method and device | |
US8355062B2 (en) | Method and apparatus for processing image | |
CN108154086A (en) | A kind of image extraction method, device and electronic equipment | |
CN105528078B (en) | The method and device of controlling electronic devices | |
CN108156384A (en) | Image processing method, device, electronic equipment and medium | |
CN112508974B (en) | Training method and device for image segmentation model, electronic equipment and storage medium | |
CN106375849B (en) | A kind of method, apparatus, the update method of video and device generating template | |
CN105229700B (en) | Device and method for extracting peak figure picture from multiple continuously shot images | |
US11886499B2 (en) | Apparatus for training recognition model, apparatus for analyzing video, and apparatus for providing video search service | |
CN106375850A (en) | Method and device for judging matched videos | |
CN111428125B (en) | Ordering method, ordering device, electronic equipment and readable storage medium | |
CN104573085A (en) | Image retrieval method, image retrieval device and terminal | |
Zhu et al. | Mask guided knowledge distillation for single shot detector | |
CN110288668B (en) | Image generation method, device, computer equipment and storage medium | |
CN106354736A (en) | Judgment method and device of repetitive video | |
Fiaz et al. | Ps-arm: An end-to-end attention-aware relation mixer network for person search | |
JP2016181181A (en) | Image processing apparatus, image processing method, and program | |
Kasprowski et al. | Implicit calibration using predicted gaze targets | |
CN112770063B (en) | Image generation method and device | |
CN106372092A (en) | Method and device for generating template, and video update method and device | |
CN106412690A (en) | Determination method and device of video playing |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A judgment method and device for repeated video Effective date of registration: 20210104 Granted publication date: 20190913 Pledgee: Inner Mongolia Huipu Energy Co.,Ltd. Pledgor: WUXI TVMINING MEDIA SCIENCE & TECHNOLOGY Co.,Ltd. Registration number: Y2020990001517 |
|
PE01 | Entry into force of the registration of the contract for pledge of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190913 Termination date: 20210723 |
|
CF01 | Termination of patent right due to non-payment of annual fee |