CN110490112A - Football video segment detection method, device, system and storage medium - Google Patents
Football video segment detection method, device, system and storage medium Download PDFInfo
- Publication number
- CN110490112A CN110490112A CN201910744042.9A CN201910744042A CN110490112A CN 110490112 A CN110490112 A CN 110490112A CN 201910744042 A CN201910744042 A CN 201910744042A CN 110490112 A CN110490112 A CN 110490112A
- Authority
- CN
- China
- Prior art keywords
- video clip
- goal
- video
- football
- disaggregated model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
Abstract
The invention discloses a kind of football video segment detection methods, comprising: judges whether football video segment to be detected is goal video clip；It is goal video clip in response to determination football video segment to be detected, football video segment to be detected is input to four disaggregated model of goal video clip constructed in advance, obtains the first recognition result of four disaggregated model of goal video clip output；Four disaggregated model of goal video clip is by using each of soccer goal video clip training set video clip as input, and using the first marked content of corresponding goal video clip as output, the first disaggregated model based on convolutional neural networks is trained and is generated；First marked content is used to indicate the goal type of corresponding video clip.Technical solution provided by the invention can be improved the intelligence degree and accuracy of football video segment detection；The present invention also provides football video segment detection device, system and storage mediums simultaneously.
Description
Technical field
The present invention relates to technical field of computer vision more particularly to football video segment detection method, device, system and
Storage medium.
Background technique
Football is most popular one of sports, and major football game program is also by the pass of the plentiful fans
Note.Football game program generally includes the video clip of various football events, such as soccer goal video clip.However, at present
Still without scheme can distinguish the goal type in goal video clip be it is common score (the goal type i.e. other than positioning ball),
Corner-kick, penalty kick or free kick.
Summary of the invention
The present invention proposes football video segment detection method, device, system and storage medium, can be improved football video piece
The intelligence degree and accuracy of section detection.
The present invention provides a kind of football video segment detection methods, which comprises
Judge whether football video segment to be detected is goal video clip；
It is goal video clip in response to the determination football video segment to be detected, the football to be detected is regarded
Frequency segment is input to four disaggregated model of goal video clip constructed in advance, and it is defeated to obtain four disaggregated model of goal video clip
The first recognition result out；Four disaggregated model of goal video clip is by training soccer goal video clip in set
Each video clip as input, and using the first marked content of corresponding goal video clip as export, to base
It is trained and is generated in the first disaggregated model of convolutional neural networks；First marked content is used to indicate corresponding video
The goal type of segment.
It is described to judge whether football video segment to be detected is goal piece of video in a kind of optional embodiment
Section, comprising:
Football video segment to be detected is input to two disaggregated model of video clip constructed in advance, obtains the video
Second recognition result of two disaggregated model of segment output；Two disaggregated model of video clip is by instructing football video segment
Practice set each of video clip as input, and using the second marked content of corresponding video clip as export,
The second disaggregated model based on convolutional neural networks is trained and is generated；Second marked content is used to indicate corresponding
Whether video clip is goal video clip；
Then described in response to the determination football video segment to be detected is goal video clip, will be described to be detected
Football video segment is input to four disaggregated model of goal video clip constructed in advance, obtains the goal video clip four and classifies
First recognition result of model output, comprising:
In response to determining that the type in second recognition result is goal video clip, the football to be detected is regarded
Frequency segment is input to four disaggregated model of goal video clip constructed in advance, and it is defeated to obtain four disaggregated model of goal video clip
The first recognition result out.
In a kind of optional embodiment, first disaggregated model based on convolutional neural networks is TSN.
In a kind of optional embodiment, second disaggregated model based on convolutional neural networks is TSN.
In a kind of optional embodiment, the method also includes:
Before judging whether football video segment to be detected be goal video clip, soccer goal video clip is obtained
Training set；
It will be corresponding according to the goal type of each of soccer goal video clip training set video clip
Video clip is labeled, and obtains the first marked content of each video clip；
Using each of soccer goal video clip training set video clip as input, and will be corresponding
First marked content of goal video clip is trained the first disaggregated model based on convolutional neural networks as output,
Obtain four disaggregated model of goal video clip.
In a kind of optional embodiment, the method also includes: it is input to by football video segment to be detected
Before two disaggregated model of video clip constructed in advance, football video segment training set is obtained；
Judge whether each of football video segment training set video clip is goal video clip；
It is that corresponding video clip marks the second marked content according to judging result；
Using each of football video segment training set video clip as inputting, and by corresponding video clip
The second marked content as output, the second disaggregated model based on convolutional neural networks is trained, video clip is obtained
Two disaggregated models.
It is described to judge whether football video segment to be detected is goal piece of video in a kind of optional embodiment
Before section, soccer goal video clip training set is obtained, comprising:
Training the video clip in set in response to judging result for the football video segment is goal video clip, will
Corresponding video clip is determined as the video clip in soccer goal video clip training set.
The present invention also provides a kind of football video segment detection devices, comprising:
First judgment module, for judging whether football video segment to be detected is goal video clip；
First identification module will for being goal video clip in response to the determination football video segment to be detected
The football video segment to be detected is input to four disaggregated model of goal video clip constructed in advance, obtains the goal view
First recognition result of four disaggregated model of frequency segment output；Four disaggregated model of goal video clip is by by soccer goal
Each of video clip training set video clip is marked as input, and by the first of corresponding goal video clip
Content is trained the first disaggregated model based on convolutional neural networks and is generated as output；First marked content
It is used to indicate the goal type of corresponding video clip.
As an improvement of the above scheme,
The present invention is also corresponding to provide kind of a football video segment detection system, including processor, memory and is stored in
In the memory and it is configured as the computer program executed by the processor, the processor executes the computer journey
The football video segment detection method as described in above-mentioned any embodiment is realized when sequence.
The present invention is also corresponding to provide a kind of storage medium, and the storage medium includes the computer program of storage, wherein
The computer program controls football of the storage medium institute's Coupling device realization as described in above-mentioned any embodiment when running
Video clip detection method.
Compared with the existing technology, the present invention has following outstanding the utility model has the advantages that football video segment provided by the invention
Whether detection method, device, system and storage medium are goal video clip by first determining football video segment, then base
In goal of four disaggregated model of goal video clip that the first disaggregated model of convolutional neural networks generates to goal video clip
Type is classified, to identify the goal type of goal video clip, effectively improves the intelligence of football video segment detection
Degree.This programme first determines whether football video segment is goal video clip, is then just classified to goal video clip,
Reduce the categorical measure for carrying out being related to when disaggregated model training, is conducive to improve accuracy；This programme uses convolutional Neural net
First disaggregated model of network, does not need the intervention of the artificial design features such as subtitles appearances, and overcoming artificial design features can not be complete
The defect of all information in full response video image can be avoided because artificial design features are easy by different cameras, difference
The influences of factors such as illumination, different court, different angle and lead to the problem for identifying that accuracy is low；In addition, this programme is to football
The scene of video clip, angle requirement be not high, thus be not limited by can only handle live streaming camera lens in middle camera lens shooting into
Ball segment has stronger scene adaptability.
Detailed description of the invention
Fig. 1 is the flow diagram of the football video segment detection method of an embodiment provided by the invention；
Fig. 2 is the flow diagram of the football video segment detection method of another embodiment provided by the invention；
Fig. 3 is the part flow diagram of the football video segment detection method of another embodiment provided by the invention；
Fig. 4 is the part flow diagram of the football video segment detection method of another embodiment provided by the invention；
Fig. 5 is the structural schematic diagram of the football video segment detection device of an embodiment provided by the invention；
Fig. 6 is the structural schematic diagram of the football video segment detection device of another embodiment provided by the invention；
Fig. 7 is the structural schematic diagram of the football video segment detection system of an embodiment provided by the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
It is the flow diagram of the football video segment detection method of an embodiment provided by the invention, such as referring to Fig. 1
Shown in Fig. 1, the method includes the steps S110~step S120.
Step S110, judge whether football video segment to be detected is goal video clip.
It step S120, is goal video clip in response to the determination football video segment to be detected, it will be described to be checked
The football video segment of survey is input to four disaggregated model of goal video clip constructed in advance, obtains the goal video clip four
First recognition result of disaggregated model output.
Wherein, four disaggregated model of goal video clip be by will soccer goal video clip training set in it is every
One video clip is used as input, and using the first marked content of corresponding goal video clip as output, to based on volume
First disaggregated model of product neural network, which is trained, to be generated；First marked content is used to indicate corresponding video clip
Goal type.
Wherein, " corresponding video clip " refers to corresponding with the video clip of input, for example, by soccer goal video clip
Train the video clip 1 in set as input, then using the first marked content of video clip 1 as output, to based on convolution
First disaggregated model of neural network is trained.
Soccer goal video clip training set includes multiple soccer goal video clips corresponding to different goal types；
Specifically, soccer goal video clip training set includes corresponding to multiple soccer goal video clips, the correspondence commonly scored
In multiple soccer goal video clips of corner-kick, corresponding to multiple soccer goal video clips of free kick and corresponding to point
Multiple soccer goal video clips of ball.
First disaggregated model based on convolutional neural networks can be TSN (Temporal Segment
Networks), convolutional network such as AlexNet, VGG-16 that only spatial information is modeled based on 2D, VGG-19,
GoogleNet, ResNet, DenseNet etc.；C3D, I3D based on 3D convolution；The Non-Local that global information is modeled
Network；Or binary-flow network TSN, SlowFast Network based on two-stream etc..The present embodiment uses TSN network.
TSN network is two-stream (double fluid) structure.
Further, the method also includes: be not goal video in response to the determination football video segment to be detected
The football video segment to be detected is determined as video clip of not scoring by segment.
The present embodiment is by first determining whether football video segment is goal video clip, is then based on convolutional neural networks
The first disaggregated model generate four disaggregated model of goal video clip classify to the goal type of goal video clip, from
And identify the goal type of goal video clip, effectively improve the intelligence degree of football video segment detection.
The present embodiment first determines that whether football video segment is goal video clip, then just carries out goal video clip
Classification reduces the categorical measure for carrying out being related to when disaggregated model training, is conducive to improve accuracy；The present embodiment uses convolution
First disaggregated model of neural network, does not need the intervention of the artificial design features such as subtitles appearances, overcomes artificial design features
The defect that all information in video image can not be reacted completely can be avoided because artificial design features are easy by different camera shootings
The influences of factors such as machine, different illumination, different courts, different angle and lead to the problem for identifying that accuracy is low；In addition, this implementation
Example is not high to the scene of football video segment, angle requirement, therefore is not limited by the middle camera lens that can only be handled in live streaming camera lens
The goal segment of shooting has stronger scene adaptability.
Referring to fig. 2, be another embodiment provided by the invention football video segment detection method flow diagram,
As shown in Figure 1, the method includes the steps S210~step S230.
Step S210, football video segment to be detected is input to two disaggregated model of video clip constructed in advance, is obtained
The second recognition result exported to two disaggregated model of video clip.
Wherein, two disaggregated model of video clip is by the way that football video segment is trained each of set video
Segment is used as input, and using the second marked content of corresponding video clip as output, to based on convolutional neural networks
Second disaggregated model, which is trained, to be generated；Second marked content is used to indicate whether corresponding video clip is view of scoring
Frequency segment.
Here, " corresponding video clip " refer to it is corresponding with the video clip of input, for example, football video segment is trained
Video clip 2 in set is as input, then using the second marked content of video clip 2 as output, to based on convolutional Neural
Second disaggregated model of network is trained.
Second disaggregated model based on convolutional neural networks can be TSN (Temporal Segment
Networks), convolutional network such as AlexNet, VGG-16 that only spatial information is modeled based on 2D, VGG-19,
GoogleNet, ResNet, DenseNet etc.；C3D, I3D based on 3D convolution；The Non-Local that global information is modeled
Network；Or binary-flow network TSN, SlowFast Network based on two-stream etc..The present embodiment uses TSN network.
TSN network is two-stream (double fluid) structure.
Football video segment training set includes corresponding to different types of multiple football video segments；Specifically, football
Video clip training set includes multiple goal video clips and multiple video clips of not scoring.
Step S220, in response to determining that the type in second recognition result is not goal video clip, will it is described to
The football video segment of detection is determined as video clip of not scoring.
It step S230, will be described to be checked in response to determining that the type in second recognition result is goal video clip
The football video segment of survey is input to four disaggregated model of goal video clip constructed in advance, obtains the goal video clip four
First recognition result of disaggregated model output.
Wherein, four disaggregated model of goal video clip be by will soccer goal video clip training set in it is every
One video clip is used as input, and using the first marked content of corresponding goal video clip as output, to based on volume
First disaggregated model of product neural network, which is trained, to be generated；First marked content is used to indicate corresponding video clip
Goal type.
Further, the method also includes: before executing step S210, obtain football video segment training set；
Judge whether each of football video segment training set video clip is goal video clip；
It is that corresponding video clip marks the second marked content according to judging result；
Using each of football video segment training set video clip as inputting, and by corresponding video clip
The second marked content as output, the second disaggregated model based on convolutional neural networks is trained, video clip is obtained
Two disaggregated models.
In the present embodiment, football video segment training set is obtained, it specifically can be by intercepting video in normal play
Segment can also crawl a large amount of football video segment by web crawlers.It, can be by artificial when marking the second marked content
Mode is labeled the video clip in above-mentioned football video segment training set.
Specifically, the method also includes:
Before executing step S210, soccer goal video clip training set is obtained；
It will be corresponding according to the goal type of each of soccer goal video clip training set video clip
Video clip is labeled, and obtains the first marked content of each video clip；
Using each of soccer goal video clip training set video clip as input, and will be corresponding
First marked content of goal video clip is trained the first disaggregated model based on convolutional neural networks as output,
Obtain four disaggregated model of goal video clip.
Further, described before executing step S210, obtain soccer goal video clip training set, comprising:
Training the video clip in set in response to judging result for the football video segment is goal video clip, will
Corresponding video clip is determined as the video clip in soccer goal video clip training set.
In the present embodiment, when carrying out data mark, other than marking the second marked content (whether scoring), also mark
First marked content (classification of the goal segment), including commonly score, corner-kick, free kick, penalty kick.In the training stage, choosing
Take two disaggregated model of video clip training video segment that the second marked content is labelled in above-mentioned labeled data；Choose above-mentioned mark
The video clip of common goal, corner-kick, free kick, penalty kick, i.e., all goal segments are labeled as in note data, training is based on convolution
Four disaggregated model of goal video clip of neural network.
Method provided in this embodiment first passes through the second disaggregated model piece of video generated based on convolutional neural networks
Two disaggregated models of section determine whether football video segment is goal video clip, are then based on the first classification of convolutional neural networks
Four disaggregated model of goal video clip that model generates classifies to the goal type of goal video clip, to identify goal
The goal type of video clip effectively improves the intelligence degree of football video segment detection.The present embodiment uses convolutional Neural
The first disaggregated model and the second disaggregated model of network, do not need the intervention of the artificial design features such as subtitles appearances, overcome people
Work design feature can not react the defect of all information in video image completely, can be avoided because artificial design features be easy by
The influences of factors such as different cameras, different illumination, different courts, different angle and lead to the problem for identifying that accuracy is low；This
Outside, the present embodiment is not high to the scene of football video segment, angle requirement, therefore is not limited by handle and be broadcast live in camera lens
Middle camera lens shooting goal segment, have stronger scene adaptability.
Fig. 3 is the part flow diagram of the football video segment detection method of another embodiment provided by the invention.
In the present embodiment, football video segment detection method includes the steps that above-described embodiment S210~S250, further defines
The method includes the steps S202~step S204.Step S202~step S204 provides the goal video of an embodiment
The training method of four disaggregated model of segment.
Step S202, before judging whether football video segment to be detected be goal video clip, obtain football into
Ball video clip training set.
Further, as shown in connection with fig. 4, step S202 includes step S2021 and step S2022.Step S2021 and step
S2022 provides the generation method of the soccer goal video clip training set of an embodiment.
Step S2021, several section of football match video are obtained.
Specifically, several section of football match video include several conventional soccerball match videos and/or several positioning ball collection
Bright and beautiful video.
Conventional soccerball match video is the section of football match video including positioning ball video clip and other video clips.
Positioning ball collection of choice specimens video includes multiple positioning ball video clips.
Optionally, positioning ball collection of choice specimens video can be obtained by web crawlers.
Step S2022, it identifies the soccer goal video clip in several section of football match video, is recognized with basis
Soccer goal video clip generates soccer goal video clip training set.
Specifically, step S2021 can include: obtain several conventional soccerball match videos；
Then step S2022 includes:
For each conventional soccerball match video got, conventional soccerball match video is divided into multiple piece of video
Section；
Each video clip being divided into is input to two disaggregated model of video clip, obtains the video clip two
Disaggregated model output with the one-to-one third recognition result of each video clip being divided into；
Soccer goal is generated according to all video clips that the type in the third recognition result is goal video clip
Video clip training set.
It further, can be to soccer goal video clip training after generating soccer goal video clip training set
Set carries out manual examination and verification, to reduce noise.
Step S203, according to the goal type of each of soccer goal video clip training set video clip
Corresponding video clip is labeled, the first marked content of each video clip is obtained.
Step S204, using the soccer goal video clip training set each of video clip as input, with
And using the first marked content of corresponding goal video clip as output, to the first disaggregated model based on convolutional neural networks
It is trained, obtains four disaggregated model of goal video clip.
It is the structural schematic diagram of the football video segment detection device of an embodiment provided by the invention referring to Fig. 5.
As shown in figure 5, football video segment detection device 5 includes first judgment module 510 and the first identification module 520.
First judgment module 510 is for judging whether football video segment to be detected is goal video clip.
First identification module 520 is used in response to the determination football video segment to be detected be goal video clip,
The football video segment to be detected is input to four disaggregated model of goal video clip constructed in advance, obtains the goal
First recognition result of four disaggregated model of video clip output；Four disaggregated model of goal video clip be by by football into
Each of ball video clip training set video clip is marked as input, and by the first of corresponding goal video clip
Content is infused as output, the first disaggregated model based on convolutional neural networks is trained and is generated；In first mark
Hold the goal type for being used to indicate corresponding video clip.
Optionally, first disaggregated model based on convolutional neural networks is TSN.
It is the structural schematic diagram of the football video segment detection device of an embodiment provided by the invention referring to Fig. 6.
As shown in fig. 6, football video segment detection device 6 includes the second judgment module 610, the second identification module 620 and third identification
Module 630.
Second judgment module 610 divides for football video segment to be detected to be input to the video clip two constructed in advance
Class model obtains the second recognition result of two disaggregated model of the video clip output；Two disaggregated model of video clip is
By using each of football video segment training set video clip as inputting, and by the of corresponding video clip
Two marked contents are trained the second disaggregated model based on convolutional neural networks and are generated as output；Second mark
Note content is used to indicate whether corresponding video clip is goal video clip.
Optionally, second disaggregated model based on convolutional neural networks is TSN.
Second identification module 620 is used in response to determining that the type in second recognition result is not goal piece of video
Section, is determined as video clip of not scoring for the football video segment to be detected.
Third identification module 630 is used in response to determining that the type in second recognition result is goal video clip,
The football video segment to be detected is input to four disaggregated model of goal video clip constructed in advance, obtains the goal
First recognition result of four disaggregated model of video clip output.
Four disaggregated model of goal video clip is by the way that soccer goal video clip is trained each of set
Video clip is used as input, and using the first marked content of corresponding goal video clip as output, to based on convolution mind
The first disaggregated model through network, which is trained, to be generated；First marked content be used to indicate corresponding video clip into
Ball type.
Described device further include:
Module is obtained, for obtaining enough before judge whether football video segment to be detected be goal video clip
Ball goal video clip training set；
Labeling module, for the goal according to each of soccer goal video clip training set video clip
Corresponding video clip is labeled by type, obtains the first marked content of each video clip；
Training module, for using each of soccer goal video clip training set video clip as defeated
Enter, and using the first marked content of corresponding goal video clip as output, to first point based on convolutional neural networks
Class model is trained, and obtains four disaggregated model of goal video clip.
Optionally, the acquisition module includes:
Video acquisition unit, for obtaining several section of football match video；
Video identification unit, the soccer goal video clip in several section of football match video for identification, with basis
The soccer goal video clip recognized generates soccer goal video clip training set.
Further, video acquisition unit includes the first subelement, for obtaining several conventional soccerball match videos；Then video
Recognition unit includes:
Second subelement, for conventional soccerball match video being divided into multiple for each conventional soccerball match video
Video clip；
Third subelement is obtained for each video clip being divided into be input to two disaggregated model of video clip
What is exported to two disaggregated model of video clip ties with the one-to-one third identification of described each video clip being divided into
Fruit；
4th subelement, for being all piece of video of goal video clip according to the type in the third recognition result
Duan Shengcheng soccer goal video clip training set.
Present invention correspondence provides the football video segment detection system of an embodiment, is implementation of the present invention referring to Fig. 7
The structural schematic diagram of the football video segment detection system for the embodiment that example provides, football video segment detection system 3 are wrapped
It includes processor 301, memory 302 and storage in the memory and is configured as the computer executed by the processor
Program, the processor 301 realize the football video segment as described in above-mentioned any embodiment when executing the computer program
Detection method.Alternatively, the processor 301 realizes each module in each embodiment of above system when executing the computer program
Function, such as first judgment module 510 and the first identification module 520.
Illustratively, the computer program can be divided into one or more modules, one or more of modules
It is stored in the memory, and is executed by the processor, to complete the present invention.One or more of modules can be
The series of computation machine program instruction section of specific function can be completed, the instruction segment is for describing the computer program described
Implementation procedure in football video segment detection system.For example, the computer program can be divided into: first judgment module
510 for judging whether football video segment to be detected is goal video clip.First identification module 520 is used in response to true
The fixed football video segment to be detected is goal video clip, and the football video segment to be detected is input in advance
Four disaggregated model of goal video clip of building obtains the first identification knot of four disaggregated model of the goal video clip output
Fruit；Four disaggregated model of goal video clip is by the way that soccer goal video clip is trained each of set piece of video
Duan Zuowei input, and using the first marked content of corresponding goal video clip as output, to based on convolutional neural networks
The first disaggregated model be trained and generated；First marked content is used to indicate the goal class of corresponding video clip
Type.
The football video segment detection system can be desktop PC, notebook, palm PC and cloud service
Device etc. calculates equipment.The football video segment detection system may include, but be not limited only to, processor, memory.This field skill
Art personnel are appreciated that the schematic diagram is only the example of football video segment detection system, do not constitute to football video
The restriction of segment detection system may include perhaps combining certain components or different than illustrating more or fewer components
Component, such as the football video segment detection system can also include input-output equipment, network access equipment, bus etc..
Alleged processor can be central processing unit (Central Processing Unit, CPU), can also be it
His general processor, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor
Deng the processor is the control centre of the football video segment detection system, entire using various interfaces and connection
The various pieces of football video segment detection system.
The memory can be used for storing the computer program and/or module, and the processor is by operation or executes
Computer program in the memory and/or module are stored, and calls the data being stored in memory, described in realization
The various functions of football video segment detection system.The memory can mainly include storing program area and storage data area,
In, storing program area can application program needed for storage program area, at least one function (such as sound-playing function, image
Playing function etc.) etc.；Storage data area, which can be stored, uses created data (such as audio data, phone directory according to mobile phone
Deng) etc..In addition, memory may include high-speed random access memory, it can also include nonvolatile memory, such as firmly
Disk, memory, plug-in type hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital,
SD) block, flash card (Flash Card), at least one disk memory, flush memory device or other volatile solid-states
Part.
Wherein, if the module that the football video segment detection system integrates is realized in the form of SFU software functional unit simultaneously
When sold or used as an independent product, it can store in a computer readable storage medium.Based on such reason
Solution, the present invention realize all or part of the process in above-described embodiment method, can also instruct correlation by computer program
Hardware complete, the computer program can be stored in a storage medium, when which runs described in control
Equipment realizes the football video segment detection method as described in above-mentioned any embodiment where storage medium.Wherein, the calculating
Machine program includes computer program code, and the computer program code can be source code form, object identification code form, can hold
Style of writing part or certain intermediate forms etc..The computer-readable medium may include: that can carry the computer program code
Any entity or device, recording medium, USB flash disk, mobile hard disk, magnetic disk, CD, computer storage, read-only memory (ROM,
Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunications letter
Number and software distribution medium etc..It should be noted that the content that the computer-readable medium includes can be managed according to the administration of justice
Local legislation and the requirement of patent practice carry out increase and decrease appropriate, such as in certain jurisdictions, according to legislation and patent
Practice, computer-readable medium does not include electric carrier signal and telecommunication signal.
It should be noted that the apparatus embodiments described above are merely exemplary, wherein described be used as separation unit
The unit of explanation may or may not be physically separated, and component shown as a unit can be or can also be with
It is not physical unit, it can it is in one place, or may be distributed over multiple network units.It can be according to actual
It needs that some or all of the modules therein is selected to achieve the purpose of the solution of this embodiment.In addition, device provided by the invention
In embodiment attached drawing, user's trip relationship between module indicates there is communication connection between them, specifically can be implemented as one
Item or a plurality of communication bus or signal wire.Those of ordinary skill in the art are without creative efforts, it can
It understands and implements.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (10)
1. a kind of football video segment detection method characterized by comprising
Judge whether football video segment to be detected is goal video clip；
It is goal video clip in response to the determination football video segment to be detected, by the football video piece to be detected
Section is input to four disaggregated model of goal video clip constructed in advance, obtains four disaggregated model of the goal video clip output
First recognition result；Four disaggregated model of goal video clip be by will soccer goal video clip training set in it is every
One video clip is used as input, and using the first marked content of corresponding goal video clip as output, to based on volume
First disaggregated model of product neural network, which is trained, to be generated；First marked content is used to indicate corresponding video clip
Goal type.
2. football video segment detection method as described in claim 1, which is characterized in that the judgement football view to be detected
Whether frequency segment is goal video clip, comprising:
Football video segment to be detected is input to two disaggregated model of video clip constructed in advance, obtains the video clip
Second recognition result of two disaggregated models output；Two disaggregated model of video clip is by by football video segment training set
Each of conjunction video clip is used as input, and using the second marked content of corresponding video clip as output, to base
It is trained and is generated in the second disaggregated model of convolutional neural networks；Second marked content is used to indicate corresponding video
Whether segment is goal video clip；
Then described in response to the determination football video segment to be detected is goal video clip, by the football to be detected
Video clip is input to four disaggregated model of goal video clip constructed in advance, obtains four disaggregated model of goal video clip
First recognition result of output, comprising:
In response to determining that the type in second recognition result is goal video clip, by the football video piece to be detected
Section is input to four disaggregated model of goal video clip constructed in advance, obtains four disaggregated model of the goal video clip output
First recognition result.
3. football video segment detection method as claimed in claim 1 or 2, which is characterized in that described to be based on convolutional Neural net
First disaggregated model of network is TSN.
4. football video segment detection method as claimed in claim 2, which is characterized in that described based on convolutional neural networks
Second disaggregated model is TSN.
5. football video segment detection method as claimed in claim 2, which is characterized in that the method also includes:
Before judging whether football video segment to be detected be goal video clip, the training of soccer goal video clip is obtained
Set；
Train the goal type of each of set video clip by corresponding video according to the soccer goal video clip
Segment is labeled, and obtains the first marked content of each video clip；
Using each of soccer goal video clip training set video clip as inputting, and by corresponding goal
First marked content of video clip is trained the first disaggregated model based on convolutional neural networks, obtains as output
Four disaggregated model of goal video clip.
6. football video segment detection method as claimed in claim 5, which is characterized in that the method also includes:
Before football video segment to be detected to be input to two disaggregated model of video clip constructed in advance, football view is obtained
Frequency segment training set；
Judge whether each of football video segment training set video clip is goal video clip；
It is that corresponding video clip marks the second marked content according to judging result；
Using each of football video segment training set video clip as inputting, and by the of corresponding video clip
Two marked contents are trained the second disaggregated model based on convolutional neural networks, obtain video clip two and divide as output
Class model.
7. football video segment detection method as claimed in claim 6, which is characterized in that described to judge football to be detected
Before whether video clip is goal video clip, soccer goal video clip training set is obtained, comprising:
Training the video clip in set in response to judging result for the football video segment is goal video clip, will be corresponded to
Video clip be determined as soccer goal video clip training set in video clip.
8. a kind of football video segment detection device characterized by comprising
First judgment module, for judging whether football video segment to be detected is goal video clip；
First identification module will be described for being goal video clip in response to the determination football video segment to be detected
Football video segment to be detected is input to four disaggregated model of goal video clip constructed in advance, obtains the goal piece of video
First recognition result of section four disaggregated models output；Four disaggregated model of goal video clip is by by soccer goal video
Each of segment training set video clip is as input, and by the first marked content of corresponding goal video clip
As output, the first disaggregated model based on convolutional neural networks is trained and is generated；First marked content is used for
Indicate the goal type of corresponding video clip.
9. a kind of football video segment detection system, which is characterized in that including processor, memory and be stored in the storage
In device and it is configured as the computer program executed by the processor, the processor is realized when executing the computer program
Such as the described in any item football video segment detection methods of claim 1-7.
10. a kind of storage medium, which is characterized in that the storage medium includes the computer program of storage, the computer journey
Storage medium institute's Coupling device is controlled when sort run to realize such as the described in any item football video segment inspections of claim 1-7
Survey method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910744042.9A CN110490112A (en) | 2019-08-13 | 2019-08-13 | Football video segment detection method, device, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910744042.9A CN110490112A (en) | 2019-08-13 | 2019-08-13 | Football video segment detection method, device, system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110490112A true CN110490112A (en) | 2019-11-22 |
Family
ID=68550887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910744042.9A Pending CN110490112A (en) | 2019-08-13 | 2019-08-13 | Football video segment detection method, device, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110490112A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112380954A (en) * | 2020-11-10 | 2021-02-19 | 四川长虹电器股份有限公司 | Video classification intercepting system and method based on image recognition |
| CN113537168A (en) * | 2021-09-16 | 2021-10-22 | 中科人工智能创新技术研究院(青岛)有限公司 | Basketball goal detection method and system for rebroadcasting and court monitoring scene |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101127866A (en) * | 2007-08-10 | 2008-02-20 | 西安交通大学 | A method for detecting wonderful section of football match video |
| CN105641903A (en) * | 2016-03-23 | 2016-06-08 | 南京信息工程大学 | Basketball match recording system |
| CN105701460A (en) * | 2016-01-07 | 2016-06-22 | 王跃明 | Video-based basketball goal detection method and device |
| US20180304146A1 (en) * | 2017-04-24 | 2018-10-25 | Throwmotion, Inc. | System and method for digitally connecting a network of physical games |
| CN108959182A (en) * | 2018-05-21 | 2018-12-07 | 北京理工大学 | The small feature loss gravitational field modeling method returned based on Gaussian process |
| CN109145840A (en) * | 2018-08-29 | 2019-01-04 | 北京字节跳动网络技术有限公司 | video scene classification method, device, equipment and storage medium |
-
2019
- 2019-08-13 CN CN201910744042.9A patent/CN110490112A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101127866A (en) * | 2007-08-10 | 2008-02-20 | 西安交通大学 | A method for detecting wonderful section of football match video |
| CN105701460A (en) * | 2016-01-07 | 2016-06-22 | 王跃明 | Video-based basketball goal detection method and device |
| CN105641903A (en) * | 2016-03-23 | 2016-06-08 | 南京信息工程大学 | Basketball match recording system |
| US20180304146A1 (en) * | 2017-04-24 | 2018-10-25 | Throwmotion, Inc. | System and method for digitally connecting a network of physical games |
| CN108959182A (en) * | 2018-05-21 | 2018-12-07 | 北京理工大学 | The small feature loss gravitational field modeling method returned based on Gaussian process |
| CN109145840A (en) * | 2018-08-29 | 2019-01-04 | 北京字节跳动网络技术有限公司 | video scene classification method, device, equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 顾舒文: ""足球视频进球事件的分类或检索方法研究"", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112380954A (en) * | 2020-11-10 | 2021-02-19 | 四川长虹电器股份有限公司 | Video classification intercepting system and method based on image recognition |
| CN113537168A (en) * | 2021-09-16 | 2021-10-22 | 中科人工智能创新技术研究院(青岛)有限公司 | Basketball goal detection method and system for rebroadcasting and court monitoring scene |
| CN113537168B (en) * | 2021-09-16 | 2022-01-18 | 中科人工智能创新技术研究院(青岛)有限公司 | Basketball goal detection method and system for rebroadcasting and court monitoring scene |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109145840B (en) | Video scene classification method, device, equipment and storage medium | |
| CN109326310B (en) | Automatic editing method and device and electronic equipment | |
| CN113508604B (en) | System and method for generating trackable video frames from broadcast video | |
| CN109376603A (en) | A kind of video frequency identifying method, device, computer equipment and storage medium | |
| CN101639354B (en) | Method and apparatus for object tracking | |
| US20170255826A1 (en) | Methods and systems of spatiotemporal pattern recognition for video content development | |
| CN102819749B (en) | A kind of football offside automatic discrimination system and method based on video analysis | |
| EP3473016B1 (en) | Method and system for automatically producing video highlights | |
| Kwon et al. | First person action recognition via two-stream convnet with long-term fusion pooling | |
| US20150018990A1 (en) | Smart-court system and method for providing real-time debriefing and training services of sport games | |
| Liu et al. | Deep learning based basketball video analysis for intelligent arena application | |
| JP2008511186A (en) | Method for identifying highlight segments in a video containing a frame sequence | |
| CN105095884B (en) | A kind of pedestrian's identifying system and processing method based on random forest support vector machines | |
| CN107729809A (en) | A kind of method, apparatus and its readable storage medium storing program for executing of adaptive generation video frequency abstract | |
| Oskouie et al. | Multimodal feature extraction and fusion for semantic mining of soccer video: a survey | |
| CN110472561A (en) | Soccer goal kind identification method, device, system and storage medium | |
| CN110490112A (en) | Football video segment detection method, device, system and storage medium | |
| CN112330710B (en) | Moving target identification tracking method, device, server and readable storage medium | |
| CN103605666A (en) | Video copying detection method for advertisement detection | |
| CN105760844A (en) | Video stream data processing method, apparatus and system | |
| Javed et al. | A decision tree framework for shot classification of field sports videos | |
| CN115731266A (en) | Cross-camera multi-target tracking method, device and equipment and readable storage medium | |
| Lin et al. | Integrating color and spatial features for content-based video retrieval | |
| Kapela et al. | Real-time event detection in field sport videos | |
| CN110472562A (en) | Position ball video clip detection method, device, system and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191122 |
|
| RJ01 | Rejection of invention patent application after publication |