CN104123396B - A kind of abstract of football video generation method and device based on cloud TV - Google Patents
A kind of abstract of football video generation method and device based on cloud TV Download PDFInfo
- Publication number
- CN104123396B CN104123396B CN201410401744.4A CN201410401744A CN104123396B CN 104123396 B CN104123396 B CN 104123396B CN 201410401744 A CN201410401744 A CN 201410401744A CN 104123396 B CN104123396 B CN 104123396B
- Authority
- CN
- China
- Prior art keywords
- video
- color
- key
- court
- audio
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/738—Presentation of query results
- G06F16/739—Presentation of query results in form of a video summary, e.g. the video summary being a video sequence, a composite still image or having synthesized frames
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention proposes a kind of abstract of football video generation method and device based on cloud TV, and wherein method includes:Excellent degree analysis in real time is carried out to football video, featured videos fragment is determined, featured videos fragment is uploaded to high in the clouds, form video frequency abstract.The present invention can combine real-time video summary with Cloud PVR technologies, mitigate network and high in the clouds pressure.
Description
Technical field
The present invention relates to cloud TV technology, more particularly to a kind of abstract of football video generation method based on cloud TV
And device.
Background technology
With market to possess individual video record a video (PVR, Personal Video Recorder) function intelligent television
The demand of product progressively strengthens, and the requirement to PVR functions is also no longer simple recording and plays back, and is needed more distinctive
Function.
Based on hard disk (HDD) though PVR functions welcome by user, it is limited as recorded video quality is higher
Hard drive space cannot the too many video of typing, by cloud (Cloud) service combine with PVR, then the limitation without this respect, and
And Cloud PVR can cause that user freely watches recorded program by other-end, such that it is able to bring more preferable user
Experience.
Football can all have all kinds of league matches in thousands of fields, Cup, often every year as one of most commonly used sports in the world
Field match all at least continues 90 minutes, if all of match recording got off, then this video library will be very huge,
People in most cases simply appreciate the wonderful of related match, therefore only excellent in storage football match simultaneously
Section, can greatly save carrying cost while people's demand is met;This wonderful is referred to as video frequency abstract.
But, existing video frequency abstract generation technique does not combine with Cloud PVR technologies, intelligent television
Cloud PVR functions are relatively easy, do not consider real-time video frequency abstract generate and be stored in high in the clouds, take at any time, real-time sharing
Application scenarios.
The content of the invention
The invention provides a kind of abstract of football video generation method based on cloud TV, can by real-time video summary with
Cloud PVR technologies combine, and mitigate network and high in the clouds pressure.
Present invention also offers a kind of abstract of football video generating means based on cloud TV, real-time video can be made a summary
Combine with Cloud PVR technologies, mitigate network and high in the clouds pressure.
The technical proposal of the invention is realized in this way:
A kind of abstract of football video generation method based on cloud TV, including:Excellent degree point in real time is carried out to football video
Analysis, determines featured videos fragment, and featured videos fragment is uploaded into high in the clouds, forms video frequency abstract.
A kind of abstract of football video generating means based on cloud TV, including:
Featured videos identification module, for carrying out excellent degree analysis in real time to football video, determines featured videos fragment;
Featured videos uploading module, for featured videos fragment to be uploaded into high in the clouds, forms video frequency abstract.
It can be seen that, abstract of football video generation method and device based on cloud TV proposed by the present invention determine foot in real time
The wonderful of ball video, and the wonderful that will be determined is uploaded to high in the clouds, the video frequency abstract of football video is formed, so as to subtract
Light network and high in the clouds pressure.
Brief description of the drawings
Fig. 1 is the overall flow schematic diagram of the method example of present invention generation abstract of football video;
Fig. 2 realizes flow chart for the embodiment of the present invention one;
Fig. 3 is the abstract of football video generating means structural representation based on cloud TV proposed by the present invention.
Specific embodiment
The present invention proposes a kind of abstract of football video generation method based on cloud TV, including:Reality is carried out to football video
Shi Jingcai degree is analyzed, and determines featured videos fragment, and featured videos fragment is uploaded into high in the clouds, forms video frequency abstract.
The present invention realizes the real-time generation of abstract of football video, to the video flowing recorded, in units of frame, carries out
Place detection, key color extraction and region segmentation, enter between picture group (GOP, Group of Pictures) on the basis of similitude
Row forbidden zone line judges;And audio extraction is carried out to the video segment, it is special to audio using the support vector machine classifier for training
Analysis is levied, to determine whether the audio is excellent audio section, determines whether the video is essence by audio and video highlight degree
Color video segment.
Such as the overall flow schematic diagram that Fig. 1 is the method example that the present invention generates abstract of football video.Overall flow includes:
Real-time reception video flowing, the analysis of video highlight degree, the excellent degree of audio are analyzed, generate video highlight fragment, upload high in the clouds in real time.
Wherein, the analysis phase is spent in video highlight, is mainly included the following steps that:
1) football match place screening:It is true according to specific rule by the image border of Canny operator extraction key frames
Whether the fixed image is arena map.
2) key color extraction in football match place is carried out:Lower-level vision feature extraction is carried out to key frame, by creating base
In the two-dimensional histogram of BGR (Blue-Green-Red) color space, its proportion BGR channel color values higher are taken;If by
The BGR passage color value merger of dry key frame obtains the court dominant color information of the match.
3) competition area region segmentation:On the basis of video mass-tone is obtained, the frame to needing analysis carries out sub-block segmentation,
The dominant color information that two-dimensional histogram analysis obtains sub-block is carried out to each sub-block, is then contrasted with video dominant color information, if
Information is unanimously then considered arena map;Otherwise it is assumed that be additional scene (such as auditorium or other camera lenses), by BGR (0,
0,0) color is filled to the block;Artwork after treatment is carried out into corrosion expansion denoising, then mask treatment is carried out with artwork, obtained
Take final place figure.
4) similitude between GOP is positioned based on hash algorithm is perceived:Floor area is divided using hash algorithm is perceived
Frame sequence after cutting carries out similarity analysis, if the similitude duration, in two more than GOP, can be prohibited on this basis
Area's line identification.
5) forbidden zone line identification:On the basis of Ground Split on the scene, rim detection is carried out using the Canny operators of self adaptation, obtained
To the obvious two-value figure in edge;Using all straight lines occurred in Hough linear transformation fitted figure, and by straight line and straight line
Between linear restriction relation determine whether to contain forbidden zone line in figure.
By said process, determine that video GOP is similar and video segment comprising forbidden zone line is to meet video highlight degree
Video segment;Afterwards, the audio to the video carries out excellent degree analysis, and concrete mode is:It is high and low not that excellent degree is collected in advance
Same audio, carries out feature extraction, and by training production SVM classifier, auxiliary video analysis generation featured videos are made a summary.
Further, because the video recorded is for TS flows encapsulation format, to ensure that the video frequency abstract of generation, without interim card, need to enter
Audio frequency and video PTS, DTS of capable each wonderful and TS layers PCR, CC amendment.
Based on the strategy while analyzing is recorded, to the video flowing of real-time reception, excellent degree is carried out according to above-mentioned steps
Analysis, generates video frequency abstract fragment and calls the SDK of cloud platform itself, carries out the real-time upload of video frequency abstract.
From said process, present invention reduces the complexity of football video analysis, real-time is high;Of the present invention
Cloud PVR combine with video frequency abstract, and the summary of real-time generation is transmitted high in the clouds, reduce network and high in the clouds pressure, favorably
In Consumer's Experience, product function is increased, improve product competitiveness.
Below in conjunction with accompanying drawing, lift specific embodiment and be discussed in detail.
Embodiment one:
The present embodiment introduction generates a specific example of abstract of football video, and such as Fig. 2 realizes flow for the present embodiment
Figure, for the football video of real-time reception, performs following steps:
Step 201:Complete the key color extraction of competition area.Including following 2 step:
1) the place detection of football video is carried out
The mass-tone of football video is football pitch ground colour, it is necessary to be filtered to target frame in key color extraction, removes and is not inconsistent
Frame normally.Specific filtering rule is that the marginal information of image is extracted using Canny algorithms, divides an image into N*N
Block, counts the marginal point number in each fritter;By the threshold value T of marginal point number in block1Empirical value is set to, if marginal point in figure
Number is less than threshold value T1Ratio be R, when R be more than threshold value T2When, it is believed that the figure contains the place figure of certain ratio, can be led
Color is extracted;Wherein T2It is adaptive threshold.
2) key color extraction is carried out to the competition area for detecting
To the arena map for detecting, carry out reducing dimension treatment, two-dimensional histogram is set up, in the histogram equalization
On the basis of, obtain the dominant color information of the figure;Screening is collected by the dominant color information in several courts, the ball of this match is obtained
Field mass-tone.
Step 202:The similarity analysis of video are proceeded by from current GOP, if similitude is high, step is continued executing with
203;Otherwise, using next GOP as current GOP, return and perform this step.
Specifically, the target of Video similarity analysis is the key frame in GOP and the N (usually fps/5) in the GOP
Individual non-key frame, after analyzing the non-key frame, follow-up B, P frame is not analyzed, and waits the arrival of next GOP data.
If the key frame of current GOP (i.e. Hamming distance much like with non-key frame<=3), and with the frame of next GOP
(key frame and n-th non-key frame) is much like, then it is assumed that similitude is high.
In addition, before similarity analysis are carried out, the frame of analysis can will be needed to carry out floor area segmentation, so as to reduce
Computation complexity.Floor area segmentation can be carried out according to court mass-tone, concrete mode is:
Frame to needing analysis carries out N*N segmentations, and the sub-block after segmentation equally carries out BGR two-dimensional histogram analyses, so that
Obtain the dominant color information of each sub-block;The mass-tone of sub-block is contrasted with the mass-tone sequence information of video, if meeting certain
Logical condition, then it is assumed that the sub-block is arena map, otherwise it is assumed that being interference information (pointing to auditorium or other scenes), makes
The sub-block is filled with BGR (0,0,0), reduces the interference of subsequent analysis.After region segmentation, carry out burn into expansion and go
Make an uproar, mask treatment is carried out with artwork, be just final place figure.
Step 203:Hough linear transformations are carried out in the I frames of the 3rd GOP, it is determined whether have forbidden zone line, if it has, then
Determine that the fragment of first GOP to the 3rd GOP meets video highlight degree, and the fragment for meeting video highlight degree will last till
Untill the heading line off of forbidden zone;Otherwise, using next GOP as current GOP, return and perform step 202.
Wherein, carrying out mode that forbidden zone line judges according to Hough linear transformations can be as:
Floor area image to having split does smoothing processing, and burn into expands to eliminate noise;The rudimentary of dependency graph picture regards
Feel that feature sets up the Canny boundary operators of self adaptation, obtain obvious bianry image;It is then based on the Hough transformation fitting of probability
The all straight lines occurred in figure.Because forbidden zone line may be truncated when place is split, need to be to the straight-line data after fitting
Merger is carried out, and determines whether to be forbidden zone line according to the mutual restriction relation between the line of forbidden zone;The restriction relation tool of forbidden zone line
Body surface is now with two groups of forbidden zone lines are parallel or less parallel, and slope meets certain logical relation.
Step 204:Judgement to meeting video highlight degree carries out the excellent degree analysis of audio, judges whether the fragment meets sound
Frequently excellent degree, if it is satisfied, then determining that the fragment is featured videos;Otherwise.Using next GOP as current GOP, return and perform
Step 202.
Wherein, the mode of the analysis of the excellent degree of audio can be:
Mel cepstrum coefficients (MFCC), color character (Chroma-based are extracted from ready audio repository in advance
Feature), the audio frequency characteristics such as zero-crossing rate, the excellent degree analysis of audio is carried out using support vector machine classifier.
Step 205:Featured videos fragment is merged into video frequency abstract, and is uploaded to high in the clouds.
Specifically, video frequency abstract is made up of featured videos fragment, and the start frame of each video segment is key frame, from
Two video segments start PTS, DTS information of CC, PCR and ES layer of TS layer of modification, to amended featured videos fragment tune
Uploaded in real time with the SDK of cloud platform.
The abstract of football video generation method based on cloud TV proposed by the present invention is described above, the present invention also proposes one
The abstract of football video generating means based on cloud TV are planted, such as Fig. 3 is the structural representation of the device, including:
Featured videos identification module 310, for carrying out excellent degree analysis in real time to football video, determines featured videos piece
Section;
Featured videos uploading module 320, for featured videos fragment to be uploaded into high in the clouds, forms video frequency abstract.
Above-mentioned featured videos identification module 310 can include:
Video highlight degree judge module 311, football video is carried out for the distributed intelligence based on edge algorithms and marginal point
Place detection, it is determined that can be used for the key frame of court key color extraction;To the key that can be used for court key color extraction
Frame collects according to color histogram and many frame informations carries out court key color extraction;Floor area point is carried out according to court dominant color information
Cut;On the basis of Performance Area regional partition, the GOP similitudes of football video are analyzed according to hash algorithm is perceived;Regarding
On the basis of frequency GOP is similar, forbidden zone line judgement is carried out according to the linear character of the conversion of Hough lines and forbidden zone line, be determined for compliance with regarding
Frequently the video segment of excellent degree;
The excellent degree judge module 312 of audio, for the video segment for meeting video highlight degree, judges corresponding audio
Whether audio excellent degree is met;If met, it is determined that the video segment is featured videos fragment.
Above-mentioned video highlight degree judge module 311 is based on edge algorithms and the distributed intelligence of marginal point and carries out football video
Place is detected, it is determined that can be used for the mode of the key frame of court key color extraction can be:
Using the marginal information of Canny operator extraction key frames, key frame is divided into N*N blocks, counts edge in each block
The number of point, wherein, N is integer set in advance;
When contained marginal point is more than ratio set in advance less than the ratio that the block of thresholding set in advance accounts for all pieces,
Judge that the key frame can be used for court key color extraction.
The key frame that above-mentioned video highlight degree judge module 311 pairs can be used for court key color extraction is according to color histogram
And many frame informations collect and carry out the mode of court key color extraction and can be:
For the key frame that can be used for court key color extraction, the two-dimensional histogram based on BGR color spaces is created, take ratio
Weight highest multiple BGR color values, and record the color proportion;Many key frames are carried out with key color extraction, merger BGR colors
Value and the color proportion;Proportion multiple BRG color values high are taken as court mass-tone.
Above-mentioned video highlight degree judge module 311 can be with according to the mode that court dominant color information carries out floor area segmentation
For:
It is divided into N*N blocks to frame of video, wherein, N is integer set in advance;
Each block is carried out into BGR two-dimensional histogram analyses, the dominant color information of each block is obtained;By the dominant color information of each block
The mass-tone sequence information with frame of video is contrasted respectively;If meeting logical condition set in advance, judge that the block is ratio
Competition field map;Otherwise, it is determined that the block is interference information;
It is filled to being judged to the block of interference information using BGR (0,0,0).
Above-mentioned video highlight degree judge module 311 is converted and forbidden zone on the basis of video GOP is similar according to Hough lines
The linear character of line carries out mode that forbidden zone line judges can be as:
Floor area image to having split does smoothing processing, and burn into expands to eliminate noise;The rudimentary of dependency graph picture regards
Feel that feature sets up the Canny boundary operators of self adaptation, obtain obvious bianry image;In Hough transformation fitted figure based on probability
The all straight lines for occurring;Merger is carried out to the straight-line data after fitting, and comes true according to the linear restriction relation between the line of forbidden zone
Whether fixed is forbidden zone line.
The above-mentioned excellent degree judge module 312 of audio judges that the mode whether corresponding audio meets the excellent degree of audio is:
Audio sample storehouse is pre-saved, audio frequency characteristics are extracted from audio sample storehouse, the audio frequency characteristics fall including Mel
Spectral coefficient MFCC, color character and zero-crossing rate;It is trained using support vector machine classifier and is classified, judges whether audio accords with
The excellent degree of synaeresis frequency.
As fully visible, abstract of football video generation method and device based on cloud TV proposed by the present invention, Cloud PVR
Combine with real-time video summarization generation, reduce network and high in the clouds pressure.Analyzed frame by frame for abstract of football video, calculate multiple
Miscellaneous degree shortcoming high, the present invention is only analyzed to the key frame in GOP and some non-key frames, is filled in limited information source
The excavation event information for dividing, and on the visual signature for extracting image, computation complexity is reduced by modes such as dimensionality reductions.The present invention
For the video flowing recorded, in units of frame, place detection, key color extraction and region segmentation are carried out, the similitude between GOP
On the basis of carry out forbidden zone line judgement;And audio extraction is carried out to the video segment, use the support vector cassification for training
Device, to determine whether the audio is excellent audio section, determines this and regards to Audio feature analysis by audio and video highlight degree
Whether frequency is featured videos fragment.
Presently preferred embodiments of the present invention is the foregoing is only, is not intended to limit the invention, it is all in essence of the invention
Within god and principle, any modification, equivalent substitution and improvements done etc. should be included within the scope of protection of the invention.
Claims (12)
1. a kind of abstract of football video generation method based on cloud TV, it is characterised in that the method includes:Football video is entered
Row excellent degree analysis in real time, determines featured videos fragment, and featured videos fragment is uploaded into high in the clouds, forms video frequency abstract;
Wherein, described that excellent degree analysis in real time is carried out to football video, the mode for determining featured videos fragment is:
Distributed intelligence based on edge algorithms and marginal point carries out the place detection of football video, it is determined that can be used for court mass-tone
The key frame of extraction;
The key frame that can be used for court key color extraction is collected according to color histogram and many frame informations carries out court master
Color is extracted;
Floor area segmentation is carried out according to court dominant color information;
On the basis of Performance Area regional partition, the picture group GOP similitudes of football video are divided according to hash algorithm is perceived
Analysis;On the basis of video GOP is similar, forbidden zone line judgement is carried out according to the linear character of the conversion of Hough lines and forbidden zone line, it is determined that
Meet the video segment of video highlight degree;
For the video segment for meeting video highlight degree, judge whether corresponding audio meets the excellent degree of audio;If met,
Determine that the video segment is featured videos fragment.
2. method according to claim 1, it is characterised in that the distributed intelligence based on edge algorithms and marginal point is entered
The place detection of row football video, it is determined that the mode that can be used for the key frame of court key color extraction is:
Using the marginal information of Canny operator extraction key frames, key frame is divided into N*N blocks, counts marginal point in each block
Number, wherein, N is integer set in advance;
When contained marginal point is more than ratio set in advance less than the ratio that the block of thresholding set in advance accounts for all pieces, judge
The key frame can be used for court key color extraction.
3. method according to claim 1, it is characterised in that the described pair of key frame that can be used for court key color extraction according to
Collecting according to color histogram and many frame informations carries out the mode of court key color extraction and is:
For the key frame that can be used for court key color extraction, the two-dimensional histogram based on BGR color spaces is created, take proportion most
Multiple BGR color values high, and record the color proportion;Many key frames are carried out with key color extraction, merger BGR color values and
The color proportion;Proportion multiple BRG color values high are taken as court mass-tone.
4. method according to claim 1, it is characterised in that described to carry out floor area segmentation according to court dominant color information
Mode be:
It is divided into N*N blocks to frame of video, wherein, N is integer set in advance;
Each block is carried out into BGR two-dimensional histogram analyses, the dominant color information of each block is obtained;The dominant color information of each block is distinguished
Mass-tone sequence information with frame of video is contrasted;If meeting logical condition set in advance, judge that the block is arena
Map;Otherwise, it is determined that the block is interference information;
It is filled to being judged to the block of interference information using BGR (0,0,0).
5. method according to claim 1, it is characterised in that described on the basis of video GOP is similar, according to Hough
Line convert and forbidden zone line linear character carry out mode that forbidden zone line judges as:
Floor area image to having split does smoothing processing, and burn into expands to eliminate noise;The lower-level vision of dependency graph picture is special
The Canny boundary operators for setting up self adaptation are levied, obvious bianry image is obtained;Occur in Hough transformation fitted figure based on probability
All straight lines;Merger is carried out to the straight-line data after fitting, and determines to be according to the linear restriction relation between the line of forbidden zone
No is forbidden zone line.
6. method according to claim 1, it is characterised in that described to judge whether corresponding audio meets the excellent degree of audio
Mode be:
Audio sample storehouse is pre-saved, audio frequency characteristics are extracted from audio sample storehouse, the audio frequency characteristics include mel cepstrum system
Number MFCC, color character and zero-crossing rate;It is trained using support vector machine classifier and is classified, judges whether audio meets sound
Frequently excellent degree.
7. a kind of abstract of football video generating means based on cloud TV, it is characterised in that the device includes:
Featured videos identification module, for carrying out excellent degree analysis in real time to football video, determines featured videos fragment;
Featured videos uploading module, for featured videos fragment to be uploaded into high in the clouds, forms video frequency abstract;
Wherein, the featured videos identification module includes:
Video highlight degree judge module, the place inspection of football video is carried out for the distributed intelligence based on edge algorithms and marginal point
Survey, it is determined that can be used for the key frame of court key color extraction;To the key frame that can be used for court key color extraction according to face
Color Histogram and many frame informations collect carries out court key color extraction;Floor area segmentation is carried out according to court dominant color information;It is on the scene
On the basis of ground region segmentation, the GOP similitudes of football video are analyzed according to hash algorithm is perceived;In video GOP phases
As on the basis of, carry out forbidden zone line judgement according to the linear character of the conversion of Hough lines and forbidden zone line, be determined for compliance with video highlight degree
Video segment;
The excellent degree judge module of audio, for the video segment for meeting video highlight degree, judges whether corresponding audio accords with
The excellent degree of synaeresis frequency;If met, it is determined that the video segment is featured videos fragment.
8. device according to claim 7, it is characterised in that the video highlight degree judge module be based on edge algorithms and
The distributed intelligence of marginal point carries out the place detection of football video, it is determined that can be used for the mode of the key frame of court key color extraction
For:
Using the marginal information of Canny operator extraction key frames, key frame is divided into N*N blocks, counts marginal point in each block
Number, wherein, N is integer set in advance;
When contained marginal point is more than ratio set in advance less than the ratio that the block of thresholding set in advance accounts for all pieces, judge
The key frame can be used for court key color extraction.
9. device according to claim 7, it is characterised in that the video highlight degree judge module is to can be used for court
The key frame of key color extraction collects according to color histogram and many frame informations to carry out the mode of court key color extraction and is:
For the key frame that can be used for court key color extraction, the two-dimensional histogram based on BGR color spaces is created, take proportion most
Multiple BGR color values high, and record the color proportion;Many key frames are carried out with key color extraction, merger BGR color values and
The color proportion;Proportion multiple BRG color values high are taken as court mass-tone.
10. device according to claim 7, it is characterised in that the video highlight degree judge module is according to court mass-tone
The mode that information carries out floor area segmentation is:
It is divided into N*N blocks to frame of video, wherein, N is integer set in advance;
Each block is carried out into BGR two-dimensional histogram analyses, the dominant color information of each block is obtained;The dominant color information of each block is distinguished
Mass-tone sequence information with frame of video is contrasted;If meeting logical condition set in advance, judge that the block is arena
Map;Otherwise, it is determined that the block is interference information;
It is filled to being judged to the block of interference information using BGR (0,0,0).
11. devices according to claim 7, it is characterised in that the video highlight degree judge module is similar in video GOP
On the basis of, according to the linear character of the conversion of Hough lines and forbidden zone line carry out mode that forbidden zone line judges as:
Floor area image to having split does smoothing processing, and burn into expands to eliminate noise;The lower-level vision of dependency graph picture is special
The Canny boundary operators for setting up self adaptation are levied, obvious bianry image is obtained;Occur in Hough transformation fitted figure based on probability
All straight lines;Merger is carried out to the straight-line data after fitting, and determines to be according to the linear restriction relation between the line of forbidden zone
No is forbidden zone line.
12. devices according to claim 7, it is characterised in that the excellent degree judge module of audio judges corresponding sound
Whether frequency meets the mode of the excellent degree of audio:
Audio sample storehouse is pre-saved, audio frequency characteristics are extracted from audio sample storehouse, the audio frequency characteristics include mel cepstrum system
Number MFCC, color character and zero-crossing rate;It is trained using support vector machine classifier and is classified, judges whether audio meets sound
Frequently excellent degree.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410401744.4A CN104123396B (en) | 2014-08-15 | 2014-08-15 | A kind of abstract of football video generation method and device based on cloud TV |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410401744.4A CN104123396B (en) | 2014-08-15 | 2014-08-15 | A kind of abstract of football video generation method and device based on cloud TV |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104123396A CN104123396A (en) | 2014-10-29 |
CN104123396B true CN104123396B (en) | 2017-07-07 |
Family
ID=51768807
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410401744.4A Active CN104123396B (en) | 2014-08-15 | 2014-08-15 | A kind of abstract of football video generation method and device based on cloud TV |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104123396B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019198951A1 (en) * | 2018-04-10 | 2019-10-17 | 삼성전자 주식회사 | Electronic device and operation method thereof |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106993209A (en) * | 2016-01-20 | 2017-07-28 | 上海慧体网络科技有限公司 | A kind of method that short video clip is carried out based on mobile terminal technology |
MX2018016323A (en) * | 2016-06-20 | 2019-09-05 | Pixellot Ltd | Method and system for automatically producing video highlights. |
CN108615043B (en) * | 2016-12-12 | 2021-05-25 | 中移(杭州)信息技术有限公司 | Video classification method and system |
CN106937160B (en) * | 2017-03-28 | 2019-12-24 | 青岛海信电器股份有限公司 | Display method and device for recorded file of terminal during fast forward and fast backward |
CN108093311B (en) * | 2017-12-28 | 2021-02-02 | Oppo广东移动通信有限公司 | Multimedia file processing method and device, storage medium and electronic equipment |
CN108683924B (en) * | 2018-05-30 | 2021-12-28 | 北京奇艺世纪科技有限公司 | Video processing method and device |
CN108882003A (en) * | 2018-07-25 | 2018-11-23 | 安徽新华学院 | A kind of electronic software control system that can detect excellent race automatically |
CN109344697B (en) * | 2018-08-16 | 2021-11-09 | 中国科学院信息工程研究所 | Method for identifying wonderful moment in antagonism competition |
CN110381367B (en) * | 2019-07-10 | 2022-01-25 | 咪咕文化科技有限公司 | Video processing method, video processing equipment and computer readable storage medium |
CN110505521B (en) * | 2019-08-28 | 2021-11-23 | 咪咕动漫有限公司 | Live broadcast competition interaction method, electronic equipment, storage medium and system |
CN113052033A (en) * | 2021-03-15 | 2021-06-29 | 深圳软牛科技有限公司 | Video wonderness measuring method, device, equipment and storage medium |
CN113052149B (en) * | 2021-05-20 | 2021-08-13 | 平安科技(深圳)有限公司 | Video abstract generation method and device, computer equipment and medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070109446A1 (en) * | 2005-11-15 | 2007-05-17 | Samsung Electronics Co., Ltd. | Method, medium, and system generating video abstract information |
CN101013444A (en) * | 2007-02-13 | 2007-08-08 | 华为技术有限公司 | Method and apparatus for adaptively generating abstract of football video |
CN101431689A (en) * | 2007-11-05 | 2009-05-13 | 华为技术有限公司 | Method and device for generating video abstract |
CN101778257A (en) * | 2010-03-05 | 2010-07-14 | 北京邮电大学 | Generation method of video abstract fragments for digital video on demand |
US20100284670A1 (en) * | 2008-06-30 | 2010-11-11 | Tencent Technology (Shenzhen) Company Ltd. | Method, system, and apparatus for extracting video abstract |
CN102801978A (en) * | 2012-08-08 | 2012-11-28 | 浪潮集团有限公司 | System for video data acquisition and management based on cloud storage |
-
2014
- 2014-08-15 CN CN201410401744.4A patent/CN104123396B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070109446A1 (en) * | 2005-11-15 | 2007-05-17 | Samsung Electronics Co., Ltd. | Method, medium, and system generating video abstract information |
CN101013444A (en) * | 2007-02-13 | 2007-08-08 | 华为技术有限公司 | Method and apparatus for adaptively generating abstract of football video |
CN101431689A (en) * | 2007-11-05 | 2009-05-13 | 华为技术有限公司 | Method and device for generating video abstract |
US20100284670A1 (en) * | 2008-06-30 | 2010-11-11 | Tencent Technology (Shenzhen) Company Ltd. | Method, system, and apparatus for extracting video abstract |
CN101778257A (en) * | 2010-03-05 | 2010-07-14 | 北京邮电大学 | Generation method of video abstract fragments for digital video on demand |
CN102801978A (en) * | 2012-08-08 | 2012-11-28 | 浪潮集团有限公司 | System for video data acquisition and management based on cloud storage |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019198951A1 (en) * | 2018-04-10 | 2019-10-17 | 삼성전자 주식회사 | Electronic device and operation method thereof |
Also Published As
Publication number | Publication date |
---|---|
CN104123396A (en) | 2014-10-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104123396B (en) | A kind of abstract of football video generation method and device based on cloud TV | |
CN106682108B (en) | Video retrieval method based on multi-mode convolutional neural network | |
KR100873638B1 (en) | Image processing method and apparatus | |
KR101369915B1 (en) | Video identifier extracting device | |
KR101993001B1 (en) | Apparatus and method for video highlight production | |
CN106162223B (en) | News video segmentation method and device | |
CN108600865B (en) | A kind of video abstraction generating method based on super-pixel segmentation | |
WO2018006834A1 (en) | Systems, processes and devices for occlusion detection for video-based object tracking | |
CN107087211B (en) | Method and device for detecting lens of host | |
EP2224357A1 (en) | Video segmentation | |
CN103942751B (en) | A kind of video key frame extracting method | |
WO2021109952A1 (en) | Video editing method, apparatus and server, and computer readable storage medium | |
CN103200463A (en) | Method and device for generating video summary | |
KR101341808B1 (en) | Video summary method and system using visual features in the video | |
CN107430780B (en) | Method for output creation based on video content characteristics | |
JP2008312215A (en) | Video-image analyzer, video-image analyzing method, automatic digest preparation system, and automatic highlight extraction system | |
JP2012533962A (en) | A method for adapting video processing by detecting scenes of distant views in sports video | |
Khorrami et al. | Multiple animal species detection using robust principal component analysis and large displacement optical flow | |
CN105139421A (en) | Video key frame extracting method of electric power system based on amount of mutual information | |
JP5644505B2 (en) | Collation weight information extraction device | |
CN113313635A (en) | Image processing method, model training method, device and equipment | |
CN106162092B (en) | A kind of method and system sport video acquisition and played | |
CN105989063A (en) | Video retrieval method and device | |
CN114500879A (en) | Video data processing method, device, equipment and storage medium | |
US20170061215A1 (en) | Clustering method using broadcast contents and broadcast related data and user terminal to perform the method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |