CN104581437B - A kind of video frequency abstract generates and the method and system of video backtracking - Google Patents
A kind of video frequency abstract generates and the method and system of video backtracking Download PDFInfo
- Publication number
- CN104581437B CN104581437B CN201410830140.1A CN201410830140A CN104581437B CN 104581437 B CN104581437 B CN 104581437B CN 201410830140 A CN201410830140 A CN 201410830140A CN 104581437 B CN104581437 B CN 104581437B
- Authority
- CN
- China
- Prior art keywords
- target
- video
- frequency abstract
- information
- video frequency
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/80—Generation or processing of content or additional data by content creator independently of the distribution process; Content per se
- H04N21/85—Assembly of content; Generation of multimedia applications
- H04N21/854—Content authoring
- H04N21/8549—Creating video summaries, e.g. movie trailer
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Computer Security & Cryptography (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention provides the method and system of a kind of generation of video frequency abstract and video backtracking, moving target in original video is detached and the moving target of separation is identified and is classified by the present invention, and compact, the highly concentrated video frequency abstract file of a content is generated using the method that telescopic image integration technology and multi-mode information merge.In video trace-back process, there is the number of target and the interested object of user in the picture selected first according to user, flexibly controls the type of target occurs in video pictures number and target.When user clicks object therein, system will trace back to rapidly the original video on server, and the whole process that the object occurs is presented to the user.
Description
Technical field
It is generated the present invention relates to field of video monitoring more particularly to a kind of video frequency abstract and the method for video backtracking and is
System.
Background technology
In field of storage, with the rapid development of society, people are more and more stronger to safety precaution, while demand is also more next
Higher, effective means of the video monitoring as safety precaution field will be used wider and wider, and application demand is also constantly carrying
It is high.For surveillance video as the carrier in safety precaution field, it can allow people more efficiently to look back reduction past tense
Between the thing that is occurred, but the storage data quantity of video record is big, and storage time is long, passes through to record a video and searches a certain time in the past
Just seem very time-consuming when the people, vehicle or the certain events that are occurred, considerably increases human and material resources.In order to rapidly and efficiently
Ground browses these surveillance videos, and video summarization technique is just particularly important.
Video summarization technique be main information in video is extracted by the analysis to video content and structure, and by they
The brief video or video image frame of video semanteme content can be given full expression to by being merged into some way.The purpose of video frequency abstract
It is the redundancy removed in original video, to improve the utilization rate of video.
Video frequency abstract plays key player in video analysis and content based video retrieval system, passes through video frequency abstract skill
The brief video frequency abstract that art generates, contains all important activities in original video.Video summarization technique by image recognition,
Multiple events that different time in original video occurs are compressed into a letter by the multinomial technological means such as redundancy removal, video merging
Short video frequency abstract.By video frequency abstract can with the thing of personage, vehicle and the generation occurred in fast browsing special time period,
It spends the short period to understand the behavior of the people passed through in the monitoring range for grasping a few houres or even several days, vehicle and generation, occurs
When case can quick lock in target object, solve a case speed for public security quickening, improve major case, the efficiency of solving a case of important case has weight
Want directive significance.
The purpose of video frequency abstract is that user quickly checks that video, the quality of the video frequency abstract of generation directly affect for convenience
The experience effect of user.The problems such as generally existing target is imperfect in video frequency abstract at present, ghost.Further, since video is plucked
Upset the sequential logic of target in original video, if user it is to be understood that some target truth, it is also necessary to borrow
Help original video to check, make troubles for the use of user, video frequency searching it is inefficient.
Invention content
In view of this, the present invention proposes the method and system that a kind of video frequency abstract generates and video is recalled, for solving such as
The technical issues of what quickly traces back in original video the original case for checking the target from video frequency abstract file.
In order to achieve the above objectives, the technical solution of the embodiment of the present invention is realized in：
A kind of video frequency abstract generates and the method for video backtracking, this method include：
Video background image is extracted from original video files；
Moving target is isolated from original video files using the video background image of extraction, and carry out foreground target with
Track is identified and classifies to the moving target separated；
Storage background image, the moving target separated and corresponding target information, the target information include：Target
Temporal information, spatial information, the target type occurred in original video files；
The moving target separated is fused in the background image extracted, video frequency abstract is generated, the video is plucked
To include moving target and corresponding target information；
The graphical user interface provided by video frequency abstract directly recalls target in original video files according to target information
The video pictures of appearance.
Further, the described pair of moving target separated is identified and classification is specially：
Histograms of oriented gradients HOG features are extracted from the positive negative sample of all kinds of targets；
The HOG features of obtained positive negative sample are put into support vector machines and are trained, each target type is obtained
Feature templates；
The video frame for inputting moving target to be detected extracts the HOG features of moving target to be detected；
It is matched with the HOG features of the moving target to be detected of extraction using the obtained characteristic module, determines fortune
Moving-target type.
Further, the target letter corresponding with moving target is stored by way of index file or database table
Breath has one-to-one incidence relation between the video frequency abstract, index file or database table, original video files；It is described
At the time of temporal information in target information includes that target occurs in original video, the time span of appearance；The time letter
Cease the position occurred for positioning the moving target in the original video files.
Further, the method that the moving target separated is fused in the background image extracted is specific
For：
Video frequency abstract playout software is realized by the program implementation of object-oriented, when showing the video frequency abstract,
Video file, background image file and the target information of the moving target are read, establishes target object for each target, each
Target object includes the target information of the target；
The video frequency abstract playout software provides display target number and configures interface one, and the interface is for realizing control video
The number of the target occurred simultaneously in abstract picture；
The video frequency abstract playout software provides target number and configures interface two, and the interface is for realizing according to target object
Temporal information, spatial information in associated target information judge that target accounts for the overlapping between the ratio of video pictures and target
Degree, and the number for appearing in the target in video frequency abstract picture is determined according to preset parameter；
The video frequency abstract playout software provides type option and installment interface, and the interface is for realizing according to target object institute
Target type in associated target information determines the target object shown by video frequency abstract picture；
The video frequency abstract playout software provides scalable configuration interface, which is closed for realizing according to target object
Spatial information in the target information of connection determines whether to zoom in or out target.
Further, the graphical user interface provided by video frequency abstract directly recalls original regard according to target information
The video pictures of target appearance are specially in frequency file：
User is responded in the window events of video frequency abstract picture, the corresponding target information of target object is based on, directly transfers
Original video files simultaneously trace back to the position progress video that the corresponding moving target of target object occurs in original video files
Playback.
Based on the embodiment of the present invention, the present invention also provides the system of a kind of generation of video frequency abstract and video backtracking, the systems
Including：
Background extracting device, for extracting video background image from original video files；
Target separator, for isolating movement mesh from original video files using the video background image of extraction
Mark, and carry out foreground target tracking；
Target classification device, for being identified to the moving target separated and generating target information after classifying；
Storage device, it is sorted for storing the background image extracted, the moving target separated and identification
Corresponding target information, target information include：Temporal information that target occurs in original video files, spatial information, target
Type；
Video frequency abstract generating means, it is raw for the moving target separated to be fused in the background image extracted
It at video frequency abstract file and stores, the video frequency abstract includes moving target and corresponding target information；
Video frequency abstract playing device, the display for video frequency abstract and video backtracking, by graphical user interface according to mesh
Mark information directly recalls the video pictures that target occurs in original video files.
Further, the sorter includes：
Target's feature-extraction unit, for extracting histograms of oriented gradients HOG features from the positive negative sample of all kinds of targets；
And for from the video frame of the moving target to be detected of input, extracting the HOG features of moving target to be detected；
Training unit is trained for the HOG features of obtained positive negative sample to be put into support vector machines, obtains
To the feature templates of each target type；
Taxon, the HOG features of the moving target to be detected of the characteristic module and extraction for using into
Row matching, determines moving target type.
Further, the storage device stores described and moving target pair by way of index file or database table
The target information answered has one-to-one association between the video frequency abstract, index file or database table, original video files
Relationship；At the time of temporal information in the target information includes that target occurs in original video, the time span of appearance；Institute
State the position that temporal information occurs for positioning the moving target in the original video files.
Further, the program of the video frequency abstract generating means and video frequency abstract playing device based on object-oriented is realized
Mode is realized；
The video frequency abstract playing device when showing the video frequency abstract, read the moving target video file,
Background image file and target information establish target object for each target, and each target object includes that the target of the target is believed
Breath；
The video frequency abstract playing device provides display target number and configures interface one, and the interface is for realizing control video
The number of the target occurred simultaneously in abstract picture；
The video frequency abstract playing device provides target number and configures interface two, and the interface is for realizing according to target object
Temporal information, spatial information in associated target information judge that target accounts for the overlapping between the ratio of video pictures and target
Degree, and the number for appearing in the target in video frequency abstract picture is determined according to preset parameter；
The video frequency abstract playing device provides type option and installment interface, and the interface is for realizing according to target object institute
Target type in associated target information determines the target object shown by video frequency abstract picture；
The video frequency abstract playing device provides scalable configuration interface, which is closed for realizing according to target object
Spatial information in the target information of connection determines whether to zoom in or out target.
Further, the video frequency abstract playing device response user is based on mesh in the window events of video frequency abstract picture
The corresponding target information of object is marked, original video files is directly transferred and traces back to the corresponding moving target of target object original
The position occurred in video file carries out video playback.
Moving target in original video is detached and the moving target of separation is identified and is classified by the present invention program,
Compact, the highly concentrated video of a content is generated using the method that telescopic image integration technology and multi-mode information merge
Summary file.In video trace-back process, the picture selected first according to user the number of target occurs and user is interested
Object flexibly controls the type of target occurs in video pictures number and target.When user clicks object therein, system
The original video on server will be traced back to rapidly, and the whole process that the object occurs is presented to the user.
Description of the drawings
Fig. 1 is a kind of video frequency abstract based on scalable integration technology provided in an embodiment of the present invention and retrogressive method flow
Schematic diagram；
Fig. 2 is the schematic diagram of background image provided in an embodiment of the present invention and sport foreground separation；
Fig. 3 is the schematic diagram that the embodiment of the present invention is provided as foreground object classification process；
Fig. 4 is the knot of a kind of video frequency abstract and backtracking system based on scalable integration technology provided in an embodiment of the present invention
Structure schematic diagram.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, below by way of specific embodiment and join
See attached drawing, the present invention is described in detail.
Existing video frequency abstract generation technique can realize the concentration of original video content, can will be different under same video background
Target is spliced and combined in video content in different time periods in common background image, however the various types in video frequency abstract
Target, target object in different time periods simply mix that be likely to result in abstract image chaotic, be unfavorable for quickly
Browsing and lock onto target object, further, since between target object and original video in video frequency abstract without direct
Incidence relation needs user to position target object in original video by hand, also causes location difficulty and recall precision not high.
It is an object of the invention to propose a kind of video frequency abstract generate with video retrogressive method and system, believed by multi-mode
The method of breath fusion extracts the Hot Contents of visualization, such as target object, event from original video, and passes through time-domain information
Be concentrated in specific spatial domain compactly to express magnanimity vision content by the video frequency abstract that is generated, finally by video frequency abstract,
Video low-level image feature is combined together with high-level semantics features to facilitate the interested video content of user search.Obtaining height
After the video frequency abstract of the concentration of saturation, user may browse through concentration video frequency abstract picture, and click wherein interested hot spot,
It directly traces back in original video frames, realizes the high speed retrieval of picture interested.
Fig. 1 is that a kind of video frequency abstract provided in an embodiment of the present invention generates and the step flow of the method for video backtracking is illustrated
Figure, includes the following steps：
Step 100 extracts video background image from original video files；
The step is used to carry out video background reconstruction to original video files to make after reading in an original video files
With Background Reconstruction algorithm, such as Gaussian Background algorithm for reconstructing/mixture Gaussian background model extraction algorithm, extracted from original video
Video background image；
Step 102 isolates moving target using the video background image of extraction from original video files, and before progress
Scape target following is identified and classifies to the moving target separated；
Refering to what is shown in Fig. 2, in an embodiment of the present invention, the method that moving target is separated from original video
For：The frame sequence for inputting original video files, using Gaussian Background difference algorithm, with the background image extracted and current original
The video frame of beginning video carries out calculus of differences, to separate moving target, that is, foreground target in original video.
In an embodiment of the present invention, it can be karr to carry out the algorithm of foreground target tracking to the moving target isolated
Graceful filtering algorithm, the present invention are not specifically limited.
In an embodiment of the present invention, the moving target separated is identified and is classified, so that it is determined that movement mesh
Target type, the target type include but not limited to：People, motor vehicle, non-motor vehicle, animal etc., pedestrian can also be into one
Step is divided into man, woman, and motor vehicle can also further discriminate between as truck, car, motorcycle etc., and non-motor vehicle can also be into one
Step divides into bicycle, rickshaw etc..
Refering to what is shown in Fig. 3, in an embodiment of the present invention, moving target being identified and being classified in the following way：
First, from the positive negative sample of all kinds of targets extract histograms of oriented gradients (Histogram of Oriented Gradient,
HOG) feature, then by the HOG features of obtained positive negative sample be put into support vector machines (Support Vector Machine,
SVM, a kind of trainable machine learning method) in be trained, obtain the feature templates of each target type；It is being identified
When classification, the video frame of moving target to be detected is inputted first, is extracted the HOG features of moving target to be detected, is then used
To characteristic module matched with the HOG features of the moving target to be detected of extraction, according to preset matching degree thresholding come really
Determine the type of moving target, such as when the matching degree of the feature templates of the moving target and pedestrian isolated is met or exceeded
When 90%, then the type of the moving target is determined as pedestrian.
Step 104, storage background image, the moving target separated and corresponding target information, the target information
Including：Temporal information that target occurs in original video files, spatial information, target type；
In one embodiment of the invention, the video for the moving target separated is classified by the target type determined
Storage, and accordingly store the target information of each target.The target information includes but not limited to：Target is in original video text
The temporal information occurred in part, spatial information, the target type etc. that occur in picture.The temporal information includes that target exists
(frame position that corresponding target occurs), the time span (survival of target in video occurred at the time of appearance in original video
Time span) etc..The spatial information includes the information such as coordinates regional, target sizes that target occurs in background image.
In one embodiment of the invention, target information corresponding with moving target is stored by index file, for example, being every
One original video establishes an index file and stores the correspondence of original video files and index file, and index file is used
In storing temporal information, spatial information, the target type of each moving target etc. occurred in original video, pass through index text
Target information corresponding with moving target in part can quickly navigate to original video files and moving target in original video text
The position of appearance in part.In addition, can also store original video totalframes, total number of targets, total target in the index file
The information such as Information Number are used for the initialization and display of video frequency abstract.
In an alternative embodiment of the invention, video frequency abstract, original video text can be also stored by way of relational database
The correspondence of part and target information, such as video frequency abstract table is established, each video frequency abstract corresponds to one in video frequency abstract table
A list item, access address information, the title of original video and visit comprising video frequency abstract title and video frequency abstract file in list item
Ask address information, the target information etc. for the moving target isolated from original video files.
The moving target separated is fused in the background image extracted by step 106, generates video frequency abstract, institute
It includes moving target and corresponding target information to state video frequency abstract；
Each foreground target is fused in the video background picture established front by the step, generates compact, concentration, table
Up to the video frequency abstract file of massive video content.
In one embodiment of the invention, the moving target separated is merged with background image to be stretched using the present invention is distinctive
Contracting image fusion technology is realized video frequency abstract playout software by the program implementation of object-oriented, is regarded described in display
When frequency is made a summary, video file, background image file and the target information of the moving target are read, target is established for each target
Object, each target object contain the target information of the target.The target object passes through the index established from preceding step
In file or video frequency abstract table corresponding target information is read to be instantiated in a program, by the data of target object at
Member, event functions and interface realize the control and interaction to target object, to realize following function：
(1) the video frequency abstract playout software provides display target number and configures interface one, and the interface is for realizing control
The number of the target occurred simultaneously in video frequency abstract picture；
(2) the video frequency abstract playout software provides target number and configures interface two, and the interface is for realizing according to target
The temporal information in target information associated by object, spatial information judge that target accounts between the ratio of video pictures and target
Overlapping degree, and the number for appearing in the target in video frequency abstract picture is determined according to preset parameter；
(3) the video frequency abstract playout software provides type option and installment interface, and the interface is for realizing according to target pair
The target object shown by video frequency abstract picture is determined as the target type in associated target information, you can passes through software
The configuration interface selection of human-computer interaction interface shows the target object of which target type；
Such as, it is desirable that only focus on is pedestrian, then only shows pedestrian when playing summarized radio, can also select concern institute
The target type occurred, i.e., all types of targets are all shown in video pictures.
(4) the video frequency abstract playout software provides scalable configuration interface, and the interface is for realizing according to target object
Spatial information (size for including the position of target appearance, target) in associated target information determines whether to carry out target
It zooms in or out so that target obtains better visual effect when appearing in video frequency abstract picture, can remain regard in this way
Frequency content it is complete, and obtain high information saturation degree in abstract frame.
Step 108, the graphical user interface provided by video frequency abstract directly recall original video text according to target information
The video pictures that target occurs in part.
In one embodiment of the invention, after obtaining video frequency abstract that is HI SA highly saturated, concentrating, regarded based on what is generated
Frequency is made a summary, and user may browse through video frequency abstract picture, and the graphical user interface provided by video frequency abstract playout software
(Graphic User Interface, GUI) responds the interface operation of user, such as wherein interested target when the user clicks
It after object, can directly trace back in original video frame, realize the high speed retrieval of picture interested.
Based on the program implementation of foregoing object-oriented, responds user by the event functions of target object and exist
The window events of video frequency abstract picture, such as double click event, it is soft based on video frequency abstract broadcasting in the receptance function of event
The calling interface that part provides, directly transfers original video files and traces back to the corresponding moving target of target object in original video
The position occurred in file carries out video playback, and the input parameter of wherein calling interface includes the access position of original video files
It sets, the parameters such as target object mark, the corresponding target information of target object.
In one embodiment of the invention, in order to improve the speed that summarized radio is generated and accessed, the video separated is plucked
It wants file, the video data of moving target and index file to be stored in local storage, original video files is stored in tool
The data center for having mass storage capacity can directly and quickly trace back to data center services by streaming media on demand technology
Original video on device is locally preserving original video without user, is greatly reducing carrying cost and the storage of user in this way
Space.
Fig. 4 is that a kind of video frequency abstract provided in an embodiment of the present invention generates and the system of video backtracking, the system 400 are wrapped
It includes：Background extracting device 410, target separator 420, target classification device 430, storage device 440, video frequency abstract generate dress
Set 450, video frequency abstract playing device 460.
Background extracting device 410 extracts video background image from original video files, and target separator 420 is utilized and carried
The video background image taken isolates moving target from original video files, and carries out foreground target tracking.Target classification fills
The moving target that 430 pairs are separated is set to be identified and generate target information after classifying.The storage of storage device 440 extracts
Background image, the moving target separated and the sorted corresponding target information of identification, target information include：Target exists
Temporal information, spatial information, the target type occurred in original video files；
Video frequency abstract generating means 450 obtain background image, the moving target separated and right from storage device 440
The moving target separated is fused in the background image extracted by the target information answered, and generates video frequency abstract file simultaneously
It is stored in storage device 440, video frequency abstract includes moving target and corresponding target information.Wherein, storage device 440 should be wide
Being interpreted as justice includes memory, magnetic storage medium, storing network or the cloud storage equipment provided by network etc., all can be carried
For interim and permanently store the storage device of function.
Display and video backtracking, graphical user interface of the video frequency abstract playing device 450 for video frequency abstract allow to use
Directly recall mesh in original video files by window operations such as clicks, based on moving target object and associated target information in family
Mark existing video pictures.
Further, sorter 430 includes：Target's feature-extraction unit 431, training unit 432, taxon 433.
Target's feature-extraction unit 431 extracts histograms of oriented gradients HOG features from the positive negative sample of all kinds of targets, from waiting for for input
In the video frame for detecting moving target, the HOG features of moving target to be detected are extracted.The positive and negative sample that training unit 432 will obtain
This HOG features, which are put into support vector machines, to be trained, and the feature templates of each target type are obtained.Taxon 433
It is matched with the HOG features of the moving target to be detected of extraction using the obtained characteristic module, determines moving target class
Type.
Further, storage device 440 is stored corresponding with moving target by way of index file or database table
Target information between the video frequency abstract, index file or database table, original video files there is one-to-one association to close
System；
Further, at the time of the temporal information in the target information includes that target occurs in original video, occur
Time span；The temporal information is used to position the position that the moving target occurs in the original video files.
Further, the program of video frequency abstract generating means 450 and video frequency abstract playing device 460 based on object-oriented is real
Existing mode realizes that video frequency abstract playing device 460 reads the video text of the moving target when showing the video frequency abstract
Part, background image file and target information establish target object for each target, and each target object includes the target of the target
Information；Video frequency abstract playing device 460 provides display target number and configures interface one, and the interface is for realizing control video frequency abstract
The number of the target occurred simultaneously in picture.Video frequency abstract playing device 460 also provides target number configuration interface two, the interface
Judge that target accounts for video pictures for realizing the temporal information in the target information associated by target object, spatial information
Overlapping degree between ratio and target, and of the target appeared in video frequency abstract picture is determined according to preset parameter
Number.Video frequency abstract playing device 460 provides type option and installment interface, and the interface is for realizing according to associated by target object
Target type in target information determines the target object shown by video frequency abstract picture.Video frequency abstract playing device 460 carries
For scalable configuration interface, which is for realizing the spatial information determination in the target information associated by target object
It is no that target is zoomed in or out.
Further, the realization method based on above-mentioned object-oriented, video frequency abstract playing device 460 can respond user regarding
The window events of frequency abstract picture, are based on the corresponding target information of target object, directly transfer original video files and trace back to
The position that the corresponding moving target of target object occurs in original video files carries out video playback.
The Hot Contents that the embodiment of the present invention is visualized by the method extraction that multi-mode information merges, it is proposed that establish one
Video frequency abstract model multi-modal kind MMAR (Multiple Mode Abstract Retrival), in the video frequency abstract of concentration
The medias content such as video, audio, text, case index has been merged in picture, forms the video frequency abstract that can be recalled at a high speed
The subject object and case index of information layered display, classification, can be logical to arbitrary object of interest in frame or case index
It crosses click and returns to original video frames, to realize that the high speed to interested target picture is retrieved.
Video frequency abstract proposed by the present invention based on scalable integration technology and retrogressive method and system, ensure that generation
Video frequency abstract file can completely include the object of interest and event that all original videos occur.In video playback
When, telescopic image integration technology is used, can be according to the needs of user, elastic controls what target in video pictures occurred
The type of number and target can not only be such that target completely appears in video pictures, but also can obtain preferable visual effect.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
With within principle, any modification, equivalent substitution, improvement and etc. done should be included within the scope of protection of the invention god.
Claims (8)
1. a kind of video frequency abstract generates and the method for video backtracking, which is characterized in that this method includes：
Video background image is extracted from original video files；
Moving target is isolated from original video files using the video background image of extraction, and carries out foreground target tracking,
The moving target separated is identified and is classified；
Storage background image, the moving target separated and corresponding target information, the target information include：Target is in original
Temporal information, spatial information, the target type occurred in beginning video file；
The moving target separated is fused in the background image extracted, video frequency abstract, the video frequency abstract packet are generated
Containing moving target and corresponding target information；
The graphical user interface provided by video frequency abstract is directly recalled target in original video files according to target information and is occurred
Video pictures；
The method that the moving target separated is fused in the background image extracted is specially：
Video frequency abstract playout software is realized by the program implementation of object-oriented, when showing the video frequency abstract, is read
Video file, background image file and the target information of the moving target establish target object, each target for each target
Object includes the target information of the target；
The video frequency abstract playout software provides display target number and configures interface one, and the interface is for realizing control video frequency abstract
The number of the target occurred simultaneously in picture；
The video frequency abstract playout software provides target number and configures interface two, which is closed for realizing according to target object
Temporal information, spatial information in the target information of connection judge that target accounts for the overlapping journey between the ratio of video pictures and target
Degree, and the number for appearing in the target in video frequency abstract picture is determined according to preset parameter；
The video frequency abstract playout software provides type option and installment interface, and the interface is for realizing according to associated by target object
Target information in target type determine the target object shown by video frequency abstract picture；
The video frequency abstract playout software provides scalable configuration interface, and the interface is for realizing according to associated by target object
Spatial information in target information determines whether to zoom in or out target.
2. the method as described in claim 1, which is characterized in that the described pair of moving target separated is identified and classifies
Specially：
Histograms of oriented gradients HOG features are extracted from the positive negative sample of all kinds of targets；
The HOG features of obtained positive negative sample are put into support vector machines and are trained, the spy of each target type is obtained
Levy template；
The video frame for inputting moving target to be detected extracts the HOG features of moving target to be detected；
It is matched with the HOG features of the moving target to be detected of extraction using the obtained feature templates, determines movement mesh
Mark type.
3. the method as described in claim 1, which is characterized in that
The target information corresponding with moving target is stored by way of index file or database table, the video is plucked
It wants, there is one-to-one incidence relation between index file or database table, original video files；
At the time of temporal information in the target information includes that target occurs in original video, the time span of appearance；Institute
State the position that temporal information occurs for positioning the moving target in the original video files.
4. the method as described in claim 1, which is characterized in that it is described by video frequency abstract provide graphical user interface according to
Target information directly recalls the video pictures that target occurs in original video files：
User is responded in the window events of video frequency abstract picture, the corresponding target information of target object is based on, directly transfers original
Video file simultaneously traces back to the position progress video playback that the corresponding moving target of target object occurs in original video files.
5. a kind of video frequency abstract generates and the system of video backtracking, which is characterized in that the system includes：
Background extracting device, for extracting video background image from original video files；
Target separator, for isolating moving target from original video files using the video background image of extraction, and
Carry out foreground target tracking；
Target classification device, for being identified to the moving target separated and generating target information after classifying；
Storage device, for storing the background image extracted, the moving target separated and the sorted correspondence of identification
Target information, target information includes：Temporal information that target occurs in original video files, spatial information, target type；
Video frequency abstract generating means, for the moving target separated to be fused in the background image extracted, generation regards
Frequency Summary file simultaneously stores, and the video frequency abstract includes moving target and corresponding target information；
Video frequency abstract playing device, the display for video frequency abstract and video backtracking, are believed by graphical user interface according to target
Breath directly recalls the video pictures that target occurs in original video files；
The program implementation realization of the video frequency abstract generating means and video frequency abstract playing device based on object-oriented,
The video frequency abstract playing device reads video file, the background of the moving target when showing the video frequency abstract
Image file and target information establish target object for each target, and each target object includes the target information of the target；
The video frequency abstract playing device provides display target number and configures interface one, and the interface is for realizing control video frequency abstract
The number of the target occurred simultaneously in picture；
The video frequency abstract playing device provides target number and configures interface two, which is closed for realizing according to target object
Temporal information, spatial information in the target information of connection judge that target accounts for the overlapping journey between the ratio of video pictures and target
Degree, and the number for appearing in the target in video frequency abstract picture is determined according to preset parameter；
The video frequency abstract playing device provides type option and installment interface, and the interface is for realizing according to associated by target object
Target information in target type determine the target object shown by video frequency abstract picture；
The video frequency abstract playing device provides scalable configuration interface, and the interface is for realizing according to associated by target object
Spatial information in target information determines whether to zoom in or out target.
6. system according to claim 5, which is characterized in that the sorter includes：
Target's feature-extraction unit, for extracting histograms of oriented gradients HOG features from the positive negative sample of all kinds of targets；And
For from the video frame of the moving target to be detected of input, extracting the HOG features of moving target to be detected；
Training unit is trained for the HOG features of obtained positive negative sample to be put into support vector machines, is obtained each
The feature templates of a target type；
Taxon, the HOG features progress of the moving target to be detected of the feature templates and extraction for using
Match, determines moving target type.
7. system according to claim 5, which is characterized in that
The storage device stores the target information corresponding with moving target by way of index file or database table,
There is one-to-one incidence relation between the video frequency abstract, index file or database table, original video files；
At the time of temporal information in the target information includes that target occurs in original video, the time span of appearance；Institute
State the position that temporal information occurs for positioning the moving target in the original video files.
8. system according to claim 5, which is characterized in that
The video frequency abstract playing device response user is based on the corresponding mesh of target object in the window events of video frequency abstract picture
Information is marked, original video files is directly transferred and traces back to the corresponding moving target of target object and occur in original video files
Position carry out video playback.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410830140.1A CN104581437B (en) | 2014-12-26 | 2014-12-26 | A kind of video frequency abstract generates and the method and system of video backtracking |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410830140.1A CN104581437B (en) | 2014-12-26 | 2014-12-26 | A kind of video frequency abstract generates and the method and system of video backtracking |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104581437A CN104581437A (en) | 2015-04-29 |
| CN104581437B true CN104581437B (en) | 2018-11-06 |
Family
ID=53096473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410830140.1A Active CN104581437B (en) | 2014-12-26 | 2014-12-26 | A kind of video frequency abstract generates and the method and system of video backtracking |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104581437B (en) |
Families Citing this family (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102396036B1 (en) | 2015-05-18 | 2022-05-10 | 엘지전자 주식회사 | Display device and controlling method thereof |
| CN105007464A (en) * | 2015-07-20 | 2015-10-28 | 江西洪都航空工业集团有限责任公司 | Method for concentrating video |
| TWI616763B (en) * | 2015-09-25 | 2018-03-01 | 財團法人工業技術研究院 | Method for video indexing and device using the same |
| US10708673B2 (en) | 2015-09-25 | 2020-07-07 | Qualcomm Incorporated | Systems and methods for video processing |
| CN106708890A (en) * | 2015-11-17 | 2017-05-24 | 创意引晴股份有限公司 | Intelligent high fault-tolerant video identification system based on multimoding fusion and identification method thereof |
| CN105469425A (en) * | 2015-11-24 | 2016-04-06 | 上海君是信息科技有限公司 | Video condensation method |
| CN107396165B (en) * | 2016-05-16 | 2019-11-22 | 杭州海康威视数字技术股份有限公司 | A kind of video broadcasting method and device |
| EP3249651B1 (en) * | 2016-05-23 | 2018-08-29 | Axis AB | Generating a summary video sequence from a source video sequence |
| EP3465673B1 (en) * | 2016-05-27 | 2020-07-29 | Dolby Laboratories Licensing Corporation | Transitioning between video priority and graphics priority |
| CN105872859A (en) * | 2016-06-01 | 2016-08-17 | 深圳市唯特视科技有限公司 | Video compression method based on moving target trajectory extraction of object |
| CN107493441B (en) * | 2016-06-12 | 2020-03-06 | 杭州海康威视数字技术股份有限公司 | Abstract video generation method and device |
| KR101805018B1 (en) * | 2016-07-08 | 2017-12-06 | 한양대학교 산학협력단 | Apparatus, method and computer readable medium having computer program for compact video |
| CN106780664A (en) * | 2016-11-17 | 2017-05-31 | 温州医科大学 | A kind of technical journal figure summary editing system based on vector graphics element |
| CN106714007A (en) * | 2016-12-15 | 2017-05-24 | 重庆凯泽科技股份有限公司 | Video abstract method and apparatus |
| CN108881119B (en) * | 2017-05-12 | 2021-02-12 | 华为技术有限公司 | Method, device and system for video concentration |
| CN109309868B (en) * | 2018-08-19 | 2019-06-18 | 上海极链网络科技有限公司 | Video file Command Line Parsing system |
| CN110166851B (en) * | 2018-08-21 | 2022-01-04 | 腾讯科技（深圳）有限公司 | Video abstract generation method and device and storage medium |
| CN109783688A (en) * | 2018-12-28 | 2019-05-21 | 广州烽火众智数字技术有限公司 | A kind of distributed video abstract processing system |
| CN112711966B (en) * | 2019-10-24 | 2024-03-01 | 阿里巴巴集团控股有限公司 | Video file processing method and device and electronic equipment |
| CN111526413A (en) * | 2020-04-29 | 2020-08-11 | 江苏加信智慧大数据研究院有限公司 | Course video playback system and playback method |
| WO2021248432A1 (en) * | 2020-06-12 | 2021-12-16 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for performing motion transfer using a learning model |
| CN112004117B (en) * | 2020-09-02 | 2023-03-24 | 维沃移动通信有限公司 | Video playing method and device |
| CN115455275B (en) * | 2022-11-08 | 2023-02-03 | 广东卓维网络有限公司 | Video processing system integrated with inspection equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102156707A (en) * | 2011-02-01 | 2011-08-17 | 刘中华 | Video abstract forming and searching method and system |
| CN102930061A (en) * | 2012-11-28 | 2013-02-13 | 安徽水天信息科技有限公司 | Video abstraction method and system based on moving target detection |
| CN103150319A (en) * | 2012-11-16 | 2013-06-12 | 佳都新太科技股份有限公司 | CS (Client Server) framework-based feature retrieval rear video abstraction retrieval system |
| CN103617234A (en) * | 2013-11-26 | 2014-03-05 | 公安部第三研究所 | Device and method for active video concentration |
-
2014
- 2014-12-26 CN CN201410830140.1A patent/CN104581437B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102156707A (en) * | 2011-02-01 | 2011-08-17 | 刘中华 | Video abstract forming and searching method and system |
| CN103150319A (en) * | 2012-11-16 | 2013-06-12 | 佳都新太科技股份有限公司 | CS (Client Server) framework-based feature retrieval rear video abstraction retrieval system |
| CN102930061A (en) * | 2012-11-28 | 2013-02-13 | 安徽水天信息科技有限公司 | Video abstraction method and system based on moving target detection |
| CN103617234A (en) * | 2013-11-26 | 2014-03-05 | 公安部第三研究所 | Device and method for active video concentration |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104581437A (en) | 2015-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104581437B (en) | A kind of video frequency abstract generates and the method and system of video backtracking | |
| Huang et al. | Movienet: A holistic dataset for movie understanding | |
| KR102290419B1 (en) | Method and Appratus For Creating Photo Story based on Visual Context Analysis of Digital Contents | |
| US10970334B2 (en) | Navigating video scenes using cognitive insights | |
| Rao et al. | A unified framework for shot type classification based on subject centric lens | |
| Reddy et al. | Recognizing 50 human action categories of web videos | |
| WO2021120818A1 (en) | Methods and systems for managing image collection | |
| US12051209B2 (en) | Automated generation of training data for contextually generated perceptions | |
| CN105611382A (en) | Electronic apparatus of generating summary content and method thereof | |
| Zhang et al. | Multiple adverse weather conditions adaptation for object detection via causal intervention | |
| Ul Haq et al. | An effective video summarization framework based on the object of interest using deep learning | |
| CN113301360A (en) | Information prompting method, computing device and storage medium | |
| EP3918489A1 (en) | Contextually generated perceptions | |
| Husa et al. | HOST-ATS: automatic thumbnail selection with dashboard-controlled ML pipeline and dynamic user survey | |
| JP6909657B2 (en) | Video recognition system | |
| CN111797175B (en) | Data storage method and device, storage medium and electronic equipment | |
| Manju et al. | Organizing multimedia big data using semantic based video content extraction technique | |
| CN115687692A (en) | Video processing method and device, computer storage medium and intelligent interactive panel | |
| Sharma et al. | Analyzing the Need for Video Summarization for Online Classes Conducted During Covid-19 Lockdown | |
| YM et al. | Analysis on Exposition of Speech Type Video Using SSD and CNN Techniques for Face Detection | |
| Sun et al. | Lecture video automatic summarization system based on DBNet and Kalman filtering | |
| Dilber et al. | A new video synopsis based approach using stereo camera | |
| Sharma et al. | Prediction of Criminal Activities Forecasting System and Analysis Using Machine Learning | |
| Sharma et al. | An effective video surveillance framework for ragging/violence recognition | |
| GM | Insights to Video Analytic Modelling Approach with Future Line of Research |
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 |