CN108876672A - A kind of long-distance education teacher automatic identification image optimization tracking and system - Google Patents
A kind of long-distance education teacher automatic identification image optimization tracking and system Download PDFInfo
- Publication number
- CN108876672A CN108876672A CN201810576145.4A CN201810576145A CN108876672A CN 108876672 A CN108876672 A CN 108876672A CN 201810576145 A CN201810576145 A CN 201810576145A CN 108876672 A CN108876672 A CN 108876672A
- Authority
- CN
- China
- Prior art keywords
- target
- teacher
- image
- video
- module
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- 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
Abstract
The invention discloses a kind of long-distance education teacher automatic identification image optimization tracking and system, method includes that acquisition video data, target are confined, foreground extraction target identification, adjusting parameter are determining, video simplifies compression；System includes that video acquisition module, memory module, teacher's target confine processing module, foreground extracting module, data modeling module, pixel classifications module, teacher's target generation module, location position module, adjusting parameter generation module, video data compression processing module and communication module.The present invention can accurately extract teacher's target in video image, and according to the acquisition angles of the mobile synchronous adjustment video of target and direction, the image for collecting teacher's target of moment complete display, both the accurate and effective that data volume in transmission of video in turn ensures teacher's target information, practical value and broad application prospect with higher had been reduced.
Description
Technical field
The present invention relates to field of intelligent control technology, and in particular to a kind of long-distance education teacher automatic identification image optimization with
Track method and system.
Background technique
With the rapid development of social economy, scientific and technological rapid advances, people's lives mode have occurred that change, far
Journey video, video surveillance applications it is increasingly wider.Such as long-distance education, it is long-range anywhere, so that it may free selection
The content for needing to learn is learnt, and is installed on the more important field of the public safeties such as commercial square, supermarket and school for another example
The video monitoring system of conjunction.
And there is more problem for the extraction of teacher's target during existing video analysis, for example how effectively to determine
How teacher's target eliminates influence of the dynamic background in foreground extraction, how to mitigate influence of the target shadow to Objective extraction
And if reducing noise, usually common frame difference method is relatively specific for moving target and the gray scale difference of background is more, and
The video image processing that background varies less can not be widely applied, and other methods also can not effectively solve problem above.
On the other hand, there is also adjust shooting angle based on the coarse localization position of target in video in the prior art
Direct-seeding, but the Adjustment precision of this mode is inadequate, is easy to influence shooting picture bad student during adjustment, shooting figure
As may be smudgy, and be unable to satisfy the excessive situation of big visual field and goal activities section, it may appear that larger blind area or
By recording acquisition at a distance, reality is likely to occur when watching can not accurately track teacher's target.
Furthermore in the prior art, video carries out transmission because of the influence by factors such as network bandwidths, very by network
The phenomenon that being easy to appear off and on influences the effect played.Which kind of mode is used, either to obtain the religion of clear and smooth
The video display effect of teacher's target, the stream medium data for requiring to generate in dynamic to long-distance video are effectively compressed.But it is existing
There is technology still without providing to the targeted video compress solution of teacher's target.
Summary of the invention
In view of the deficiencies of the prior art, the present invention provides a kind of long-distance education teacher automatic identification image optimization tracking
And system, it can accurately extract teacher's target in video image, and the acquisition of the mobile synchronous adjustment video according to target
Angle and direction, the image for collecting teacher's target of moment complete display had not only reduced data volume in transmission of video but also had guaranteed
The accurate and effective of teacher's target information, practical value and broad application prospect with higher.
The present invention solves technical problem and adopts the following technical scheme that：
The present invention provides a kind of long-distance education teacher automatic identification image optimization trackings, include the following steps：
S1, the video data of acquisition training in real time, are decomposed into image data and audio data for video data, suitable according to the time
Sequence is stored separately, while being taken pretreatment measure mainly to image data and being included：Target shadow is alleviated using HSV color space
Influence to Objective extraction eliminates salt-pepper noise using median filter, so that image is become simple using image binaryzation, adopts
With iconology denoise in corrosion and expansion respectively extract eliminate picture noise and filling image cavity；
S2, each frame image data of the video data of step S1 acquisition is confined using Faster RCNN progress target,
Suggestion box is generated, it is to be processed according to prefixed time interval T stage extraction storage etc. that the image data in Suggestion box will be generated；
Image data mixed Gauss model in S3, the generation Suggestion box for storing stage extraction in step S2 carries out prospect
It extracts identification and obtains teacher's target data, including：
S31, any pixel point in the preceding M frame image in this section of image data is modeled to obtain the probability of the pixel
Distribution function：
X indicates the pixel gray value, and at the time of t is that the frame image corresponds to, K is the number for choosing the Gauss model of fitting
η is Gaussian probability-density function, and ω belongs to the weight of different functions, μ and ε be respectively n-th of Gauss model of t moment mean value to
Amount and covariance matrix；
S32, probability-distribution function P (X is being obtainedt) after, since M+1 frame, after being obtained to each frame image data more
New mixed Gauss model is matched with each pixel in present image with mixed Gauss model, if matching, then it is assumed that the picture
Vegetarian refreshments is background pixel；Otherwise foreground pixel data are extracted as；
S33, teacher's number of targets in each frame image is obtained according to the foreground pixel data combined treatment extracted in S32
According to；
S4, rectangular coordinate system is established as origin using the central point of every frame image, obtains the position of teacher's target in step S3
Parameter obtains the running track of teacher's target according to the difference of location parameter between consecutive frame image；
S5, according to the running track of teacher's target, obtain including adjustment time interval, angle information, elevation information, movement
Adjusting parameter including velocity information, if adjusting parameter is less than predetermined threshold, according to obtained adjusting parameter adjustment video
The angle and direction of acquisition is to guarantee that teacher's target is steadily restored at image center position；If adjusting parameter is more than
Predetermined threshold is simultaneously emitted by abnormal alarm information then by the angle and direction of max-thresholds adjustment video acquisition, closes road view
Defeated signal is kept pouring in, access includes the another road video transfer signal of teacher's target；
S6, the video data of acquisition is transmitted to after handling squeeze operation server storage wait user retrieve for examination, institute
Processing squeeze operation is stated to include the following steps：
S61, the teacher's target handled according to step 4 location parameter according to preset frame number Q, calculate adjacent Q
Teacher target position between frame image changes difference △ S；
S62, when target position variation difference △ S be more than preset threshold Y when, extremely by adjacent Q frame image whole compression transmission
Server；
When target position, variation difference △ S is less than preset threshold Y, adjacent Q frame image is removed according to every a frame
One frame image, the video image compression for obtaining scaled-down version are sent into server storage.
It preferably, include Kalman filtering and Morphological scale-space to the processing of foreground pixel data in the step S33.
Preferably, in the S4 distance representated by the per unit of rectangular coordinate system according in advance calibrated visual field
Minimum range representated by image minimum resolution converts to obtain.
The present invention also provides a kind of long-distance education teacher automatic identification image optimization tracking systems, including：
Video acquisition module, including multi-channel video capturing complexes, for acquiring video data in real time, and according to adjustment
Signal adjusts the angle and direction of every road video acquisition；
Memory module acquires video data for storing video acquisition in real time；
Teacher's target confines processing module, for extracting the video data of memory module storage, and by each frame picture number
It is confined according to using Faster RCNN to carry out target, generates Suggestion box, the image data in Suggestion box will be generated according to preset time
It is to be processed to be spaced T stage extraction storage etc.；
Foreground extracting module is carried out for the image data mixed Gauss model in the generation Suggestion box of stage extraction storage
Foreground extraction identifies to obtain teacher's target data, including：
Data modeling module, for being modeled to obtain to any pixel point in the preceding M frame image in this section of image data
The probability-distribution function of the pixel：
X indicates the pixel gray value, and at the time of t is that the frame image corresponds to, K is the number for choosing the Gauss model of fitting
η is Gaussian probability-density function, and ω belongs to the weight of different functions, μ and ε be respectively n-th of Gauss model of t moment mean value to
Amount and covariance matrix；
Pixel classifications module, for obtaining probability-distribution function P (Xt) after, since M+1 frame, to each frame image
Mixed Gauss model is updated after data acquisition, is matched with each pixel in present image with mixed Gauss model, if phase
Match, then it is assumed that the pixel is background pixel；Otherwise foreground pixel data are extracted as；
Teacher's target generation module, for being obtained in each frame image according to the foreground pixel data combined treatment of extraction
Teacher's target data；
Location position module obtains teacher's target for establishing rectangular coordinate system as origin using the central point of every frame image
Location parameter, the running track of teacher's target is obtained according to the difference of location parameter between consecutive frame image；
Adjusting parameter generation module obtains including adjustment time interval, angle for the running track according to teacher's target
Adjusting parameter including information, elevation information, motion velocity information, and adjusting parameter is fed back into video acquisition module, video
Acquisition module is according to the angle and direction of obtained adjusting parameter adjustment video acquisition to guarantee that teacher's target is steadily restored to
At image center position；
Video data compression processing module, for according to the obtained location parameter of teacher's target according to preset frame number Q,
Calculate the teacher target position variation difference △ S between adjacent Q frame image；When target position, variation difference △ S is more than default threshold
When value Y, by adjacent Q frame image whole compression transmission to server；When target position, variation difference △ S is less than preset threshold Y
When, adjacent Q frame image is removed into a frame image according to every a frame, obtains the video image data of scaled-down version and compression；
Communication module, video data transmission for acquiring video acquisition module to memory module, video counts will be passed through
Server storage is sent into according to compressing processing module treated video image data and adjusting parameter is transmitted to video acquisition
Module.
Preferably, teacher's target generation module further includes Kalman filtering module and Morphological scale-space module.
It preferably, further include abnormal parameters processing module in the adjusting parameter generation module, for surpassing when adjusting parameter
When crossing predetermined threshold, command signal is issued to video acquisition module, video acquisition module is by max-thresholds adjustment video acquisition
Angle and direction is simultaneously emitted by abnormal alarm information, closes the road video transfer signal, and access is regarded comprising the another road of teacher's target
Keep pouring in defeated signal.
It preferably, further include image pre-processing module, for being carried out before target is confined using Faster RCNN also to every
One frame image is pre-processed, including：Influence of the target shadow to Objective extraction is alleviated using HSV color space, is used
Median filter eliminates salt-pepper noise, using image binaryzation make image become it is simple, using iconology denoise in corrosion
It is extracted respectively with expansion and eliminates picture noise and filling image cavity.
Preferably, the video acquisition complexes include photographic device and adjustment device；
The photographic device includes main photographic device and second camera device, and the second camera device is two, respectively
There is 50% visual field that is superimposed with main photographic device, without superimposition between the second camera device；
The angle of the video acquisition for the adjusting parameter adjustment photographic device that the adjustment device is used to be transmitted according to communication module
Degree and direction.
Preferably, the photographic device be spherical panorama capture device, including ambient light sensor, optical filter, camera lens,
Lens bracket and spherical panorama high-definition camera.
Preferably, the size of the number K value of the Gauss model for choosing fitting is 4.
Compared with prior art, the present invention has following beneficial effect：
(1) a kind of long-distance education teacher automatic identification image optimization tracking of the invention passes through advanced to image data
Row target is confined, then carries out foreground extraction using mixed Gauss model, and then accurately obtain teacher's mesh target area, and innovation is adopted
Moving target can be picked out from the background of differing complexity with this combination extraction mode, additionally it is possible to effectively avoid by
Lead to background variation acutely in camera shake, and efficiently separate background, so as to complete track and identify Deng follow-up works.
(2) a kind of long-distance education teacher automatic identification image optimization tracking of the invention passes through teacher's mesh to extraction
Mark establish the position positioning of coordinate system formula, can automatic tracing teacher's target, do not need to manually adjust, can be accurate in real time
The pure and fresh video image information for taking teacher's target, and the adjustment photographic device that can continue, so that user retrieved for examination
It becomes apparent from, experience is more preferably.
(3) a kind of long-distance education teacher automatic identification image optimization tracking of the invention passes through data compression process,
The use of innovation is carried out simplifying frame number, can be filled during guaranteeing remote transmission according to the mobile degree in the position of teacher's target
While dividing displaying teacher's target information, the transmission quantity of teledata is effectively reduced, to be effectively reduced what development remotely monitored
Cost.
(4) a kind of long-distance education teacher automatic identification image optimization tracking of the invention, by fixed to teacher's target
Position output before carried out Kalman filtering and Morphological scale-space, effectively reduce noise light photograph influence, improve target and
The signal-to-noise ratio of image；Influence of the target shadow to Objective extraction is alleviated using HSV color space in image preprocessing simultaneously,
Salt-pepper noise is eliminated using median filter, so that image is become simple using image binaryzation, is denoised using iconology
Corrosion and expansion are extracted respectively eliminates picture noise and filling image cavity.
(5) a kind of long-distance education teacher automatic identification image optimization tracking system of the invention provides a kind of convenience, fast
It is prompt, efficiently, teacher's target lock-on of low cost operation with mode, system flexibility, the degree of automation are higher, recognition speed is fast,
Accuracy rate is high, at the same also have the advantages that it is versatile, open by force, scalability it is strong.
(6) a kind of long-distance education teacher automatic identification image optimization tracking system of the invention can accurately extract video
Teacher's target in image, and according to the acquisition angles of the mobile synchronous adjustment video of target and direction, moment complete display
The image of teacher's target is collected, the accurate and effective that data volume in transmission of video in turn ensures teacher's target information has both been reduced
Property, practical value and broad application prospect with higher.
Detailed description of the invention
Upon reading the detailed description of non-limiting embodiments with reference to the following drawings, other feature of the invention,
Objects and advantages will become more apparent upon：
Fig. 1 is a kind of long-distance education teacher automatic identification image optimization tracking flow chart of the invention；
Fig. 2 is that a kind of foreground extraction target of long-distance education teacher automatic identification image optimization tracking of the invention is known
The specific flow chart of other step；
Fig. 3 is that a kind of video of long-distance education teacher automatic identification image optimization tracking of the invention simplifies compression step
Rapid specific flow chart；
Fig. 4 is that a kind of image preprocessing of long-distance education teacher automatic identification image optimization tracking of the invention is specific
Flow chart；
Fig. 5 is a kind of structural schematic diagram of long-distance education teacher automatic identification image optimization tracking system of 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.
In addition, those skilled in the art will appreciate that, various aspects of the invention can be implemented as system, method or
Computer program product.Therefore, various aspects of the invention can be implemented as the form of software and hardware combining, can unite here
Referred to as circuit, " module " or " system ".In addition, in some embodiments, various aspects of the invention are also implemented as
The form of computer program product in one or more microprocessors readable medium, comprising micro- in the microprocessor readable medium
The readable program code of processor.
Below with reference to according to the method for the embodiment of the present invention, the flow chart of device (system) and computer program product
And/or the block diagram description present invention.It should be appreciated that each in each box and flowchart and or block diagram of flowchart and or block diagram
The combination of box can be realized by computer program instructions.These computer program instructions can be supplied to general purpose computer,
The processor of special purpose computer or other programmable data processing units, so that a kind of machine is produced, so that these computers
Program instruction when executed by a processor of a computer or other programmable data processing device, produces implementation flow chart
And/or the device of function action specified in one or more boxes in block diagram.It can also be these computer program instructions
Storage in computer-readable medium, these instruct so that computer, other programmable data processing units or other equipment with
Ad hoc fashion work, thus, the instruction stored in computer-readable medium is just produced including implementation flow chart and/or block diagram
In one or more boxes specified in function action instruction manufacture (article of manufacture).?
Computer program instructions can be loaded into computer, other programmable data processing units or other equipment make it is a series of
Operating procedure is executed in computer, other programmable devices or other equipment, is formed computer and is realized process, so as to
The instruction executed on a computer or other programmable device provides the one or more side realized in flowchart and or block diagram
The process of function action specified in frame.
As shown in Figs. 1-5, a kind of long-distance education teacher automatic identification image optimization tracking of the present embodiment, including with
Lower step：
S1, the video data of acquisition training in real time, are decomposed into image data and audio data for video data, suitable according to the time
Sequence is stored separately, while being taken pretreatment measure mainly to image data and being included：Target shadow is alleviated using HSV color space
Influence to Objective extraction eliminates salt-pepper noise using median filter, so that image is become simple using image binaryzation, adopts
With iconology denoise in corrosion and expansion respectively extract eliminate picture noise and filling image cavity；
S2, each frame image data of the video data of step S1 acquisition is confined using Faster RCNN progress target,
Suggestion box is generated, it is to be processed according to prefixed time interval T stage extraction storage etc. that the image data in Suggestion box will be generated；
Image data mixed Gauss model in S3, the generation Suggestion box for storing stage extraction in step S2 carries out prospect
It extracts identification and obtains teacher's target data, including：
S31, any pixel point in the preceding M frame image in this section of image data is modeled to obtain the probability of the pixel
Distribution function：
X indicates the pixel gray value, and at the time of t is that the frame image corresponds to, K is the number for choosing the Gauss model of fitting
η is Gaussian probability-density function, and ω belongs to the weight of different functions, μ and ε be respectively n-th of Gauss model of t moment mean value to
Amount and covariance matrix；
S32, probability-distribution function P (X is being obtainedt) after, since M+1 frame, after being obtained to each frame image data more
New mixed Gauss model is matched with each pixel in present image with mixed Gauss model, if matching, then it is assumed that the picture
Vegetarian refreshments is background pixel；Otherwise foreground pixel data are extracted as；
S33, teacher's number of targets in each frame image is obtained according to the foreground pixel data combined treatment extracted in S32
According to；
S4, rectangular coordinate system is established as origin using the central point of every frame image, obtains the position of teacher's target in step S3
Parameter obtains the running track of teacher's target according to the difference of location parameter between consecutive frame image；
S5, according to the running track of teacher's target, obtain including adjustment time interval, angle information, elevation information, movement
Adjusting parameter including velocity information, if adjusting parameter is less than predetermined threshold, according to obtained adjusting parameter adjustment video
The angle and direction of acquisition is to guarantee that teacher's target is steadily restored at image center position；If adjusting parameter is more than
Predetermined threshold is simultaneously emitted by abnormal alarm information then by the angle and direction of max-thresholds adjustment video acquisition, closes road view
Defeated signal is kept pouring in, access includes the another road video transfer signal of teacher's target；
S6, the video data of acquisition is transmitted to after handling squeeze operation server storage wait user retrieve for examination, institute
Processing squeeze operation is stated to include the following steps：
S61, the teacher's target handled according to step 4 location parameter according to preset frame number Q, calculate adjacent Q
Teacher target position between frame image changes difference △ S；
S62, when target position variation difference △ S be more than preset threshold Y when, extremely by adjacent Q frame image whole compression transmission
Server；
When target position, variation difference △ S is less than preset threshold Y, adjacent Q frame image is removed according to every a frame
One frame image, the video image compression for obtaining scaled-down version are sent into server storage.
It include Kalman filtering and Morphological scale-space to the processing of foreground pixel data in the present embodiment step S33.
Distance representated by the per unit of rectangular coordinate system is according in advance calibrated in S4 described in the present embodiment step
Minimum range representated by image minimum resolution converts to obtain in visual field.
One of the present embodiment long-distance education teacher's automatic identification image optimization tracking system, including：
Video acquisition module, including multi-channel video capturing complexes, for acquiring video data in real time, and according to adjustment
Signal adjusts the angle and direction of every road video acquisition；
Memory module acquires video data for storing video acquisition in real time；
Teacher's target confines processing module, for extracting the video data of memory module storage, and by each frame picture number
It is confined according to using Faster RCNN to carry out target, generates Suggestion box, the image data in Suggestion box will be generated according to preset time
It is to be processed to be spaced T stage extraction storage etc.；
Foreground extracting module is carried out for the image data mixed Gauss model in the generation Suggestion box of stage extraction storage
Foreground extraction identifies to obtain teacher's target data, including：
Data modeling module, for being modeled to obtain to any pixel point in the preceding M frame image in this section of image data
The probability-distribution function of the pixel：
X indicates the pixel gray value, and at the time of t is that the frame image corresponds to, K is for choosing the Gauss model of fitting
Number, η is Gaussian probability-density function, and ω belongs to the weight of different functions, and μ and ε are the equal of n-th of Gauss model of t moment respectively
It is worth vector sum covariance matrix；
Pixel classifications module, for obtaining probability-distribution function P (Xt) after, since M+1 frame, to each frame image
Mixed Gauss model is updated after data acquisition, is matched with each pixel in present image with mixed Gauss model, if phase
Match, then it is assumed that the pixel is background pixel；Otherwise foreground pixel data are extracted as；
Teacher's target generation module, for being obtained in each frame image according to the foreground pixel data combined treatment of extraction
Teacher's target data；
Location position module obtains teacher's target for establishing rectangular coordinate system as origin using the central point of every frame image
Location parameter, the running track of teacher's target is obtained according to the difference of location parameter between consecutive frame image；
Adjusting parameter generation module obtains including adjustment time interval, angle for the running track according to teacher's target
Adjusting parameter including information, elevation information, motion velocity information, and adjusting parameter is fed back into video acquisition module, video
Acquisition module is according to the angle and direction of obtained adjusting parameter adjustment video acquisition to guarantee that teacher's target is steadily restored to
At image center position；
Video data compression processing module, for according to the obtained location parameter of teacher's target according to preset frame number Q,
Calculate the teacher target position variation difference △ S between adjacent Q frame image；When target position, variation difference △ S is more than default threshold
When value Y, by adjacent Q frame image whole compression transmission to server；When target position, variation difference △ S is less than preset threshold Y
When, adjacent Q frame image is removed into a frame image according to every a frame, obtains the video image data of scaled-down version and compression；
Communication module, video data transmission for acquiring video acquisition module to memory module, video counts will be passed through
Server storage is sent into according to compressing processing module treated video image data and adjusting parameter is transmitted to video acquisition
Module.
Teacher's target generation module in the present embodiment further includes Kalman filtering module and Morphological scale-space module.
It further include abnormal parameters processing module in adjusting parameter generation module in the present embodiment, for surpassing when adjusting parameter
When crossing predetermined threshold, command signal is issued to video acquisition module, video acquisition module is by max-thresholds adjustment video acquisition
Angle and direction is simultaneously emitted by abnormal alarm information, closes the road video transfer signal, and access is regarded comprising the another road of teacher's target
Keep pouring in defeated signal.
Long-distance education teacher's automatic identification image optimization tracking system in the present embodiment further includes image pre-processing module,
For using Faster RCNN carry out target confine before also each frame image is pre-processed, including：Using HSV color
Color space alleviates influence of the target shadow to Objective extraction, eliminates salt-pepper noise using median filter, using image two
Value make image become it is simple, using iconology denoise in corrosion and expansion respectively extract eliminate picture noise and filling image
Cavity.
Video acquisition complexes in the present embodiment include photographic device and adjustment device；
The photographic device includes main photographic device and second camera device, and the second camera device is two, respectively
There is 50% visual field that is superimposed with main photographic device, without superimposition between the second camera device；
The angle of the video acquisition for the adjusting parameter adjustment photographic device that the adjustment device is used to be transmitted according to communication module
Degree and direction.
Photographic device in the present embodiment is spherical panorama capture device, including ambient light sensor, optical filter, mirror
Head, lens bracket and spherical panorama high-definition camera.
The size of the number K value of the Gauss model of fitting in this implementation is 4.
Significant progress and substantive distinguishing features in the present embodiment compared with the existing technology are analyzed as follows：
First video image is pre-processed in this implementation, including target shadow pair is alleviated using HSV color space
The influence of Objective extraction eliminates salt-pepper noise using median filter, so that image is become simple using image binaryzation, uses
Corrosion and expansion in iconology denoising are extracted respectively eliminates picture noise and filling image cavity, confine for subsequent target and
Foreground extraction reduces the bit error rate and difficulty, special to select these types operation that reach synergistic effect, at video image
The factor that may be influenced in reason is substantially checked one time, and the precision of subsequent processing is effectively raised；
Secondly creative in the application that first teacher's target is substantially confined using Faster RCNN, then using mixed
Close Gauss model carry out foreground extraction can effectively solve due to camera shake etc. caused by background variation acutely so as to cause
The bad problem of foreground extraction effect, after substantially being confined using Faster RCNN to teacher's target, frame is just used as one
The foundation that a noise is eliminated carries out foreground extraction using mixed Gauss model in subsequent progress foreground extraction, and uses picture
Vegetarian refreshments modeling, to newly comparing one by one into pixel, while each frame new images enter and all update mixed Gauss model, in this way can
The degree of fitting of Gauss model is increased as far as possible, by test of many times in the present embodiment, the Gauss model being fitted using 4 can be
It is achieved a better balance in speed and effect；
The application, using coordinate system is established, determines the position of teacher's target, obtains its variable quantity after teacher's target determines,
So as to adjust the angle and direction of video acquisition or even the transmission path of Switch Video, so that accurate quickly will clearly need
The target video image acquisition of tracking stores transmission, additionally by comparison moving distance, deletes picture frame width number to be spaced,
The data volume for effectively reducing transmission has saved resource and has improved rate, also ensures data accurate and effective；
Kalman filtering and Morphological scale-space also have been carried out to it before teacher's target final output in the present embodiment, it can
The smooth of teacher's target is realized, so that tracking to obtain the effect of better vision and sense organ in humanbody moving object.
In the overall process that the application tracks the extraction of teacher's target, whole optimization and design, so that Objective extraction is more
For accurate, tracking is more timely, feedback is more smoothly, practical effect impression is more preferable, have compared with the existing technology significant
Progress and substantive distinguishing features.
The flow chart and block diagram in the drawings show the system of multiple embodiments according to the present invention, method and computer journeys
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, section or code of table, a part of the module, section or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two continuous boxes can actually base
Originally it is performed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that
It is the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart, can uses and execute rule
The dedicated hardware based system of fixed function or movement is realized, or can use the group of specialized hardware and computer instruction
It closes to realize.
A kind of long-distance education teacher automatic identification image optimization tracking of the invention can accurately extract video figure
Teacher's target as in, and adopted according to the acquisition angles and direction, moment complete display of the mobile synchronous adjustment video of target
Collect the image of teacher's target, both reduce the accurate and effective that data volume in transmission of video in turn ensures teacher's target information,
Practical value and broad application prospect with higher.
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention.Any reference signs in the claims should not be construed as limiting the involved claims.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art
The other embodiments being understood that.
Claims (10)
1. a kind of long-distance education teacher automatic identification image optimization tracking, which is characterized in that include the following steps：
S1, the video data of acquisition training in real time, are decomposed into image data and audio data for video data, divide sequentially in time
Storage is opened, while pretreatment measure is taken mainly to image data and includes：Target shadow is alleviated to mesh using HSV color space
The influence extracted is marked, salt-pepper noise is eliminated using median filter, so that image is become simple using image binaryzation, using figure
Picture noise and filling image cavity are eliminated as learning the corrosion in denoising and expanding to extract respectively；
S2, each frame image data of the video data of step S1 acquisition is confined using Faster RCNN progress target, is generated
It is to be processed according to prefixed time interval T stage extraction storage etc. will to generate the image data in Suggestion box for Suggestion box；
Image data mixed Gauss model in S3, the generation Suggestion box for storing stage extraction in step S2 carries out foreground extraction
Identification obtains teacher's target data, including：
S31, any pixel point in the preceding M frame image in this section of image data is modeled to obtain the probability distribution of the pixel
Function：
X indicates the pixel gray value, and at the time of t is that the frame image corresponds to, K is the number for choosing the Gauss model of fitting, and η is
Gaussian probability-density function, ω belong to the weight of different functions, and μ and ε are the mean vector of n-th of Gauss model of t moment respectively
And covariance matrix；
S32, probability-distribution function P (X is being obtainedt) after, since M+1 frame, updates and mix after being obtained to each frame image data
Gauss model is matched with each pixel in present image with mixed Gauss model, if matching, then it is assumed that the pixel is
Background pixel；Otherwise foreground pixel data are extracted as；
S33, teacher's target data in each frame image is obtained according to the foreground pixel data combined treatment extracted in S32；
S4, rectangular coordinate system is established as origin using the central point of every frame image, obtains the location parameter of teacher's target in step S3,
The running track of teacher's target is obtained according to the difference of location parameter between consecutive frame image；
S5, according to the running track of teacher's target, obtain including adjustment time interval, angle information, elevation information, movement velocity
Adjusting parameter including information, if adjusting parameter is less than predetermined threshold, according to obtained adjusting parameter adjustment video acquisition
Angle and direction to guarantee that teacher's target is steadily restored at image center position；If adjusting parameter is more than predetermined
Threshold value is simultaneously emitted by abnormal alarm information then by the angle and direction of max-thresholds adjustment video acquisition, closes road video biography
Defeated signal, access include the another road video transfer signal of teacher's target；
S6, the video data of acquisition is transmitted to after handling squeeze operation server storage wait user retrieve for examination, the place
Reason squeeze operation includes the following steps：
S61, the teacher's target handled according to step 4 location parameter according to preset frame number Q, calculate adjacent Q frame figure
Teacher target position as between changes difference △ S；
S62, when target position variation difference △ S be more than preset threshold Y when, by adjacent Q frame image whole compression transmission to service
Device；
When variation difference △ S is less than preset threshold Y when target position, adjacent Q frame image is removed into a frame according to every a frame
Image, the video image compression for obtaining scaled-down version are sent into server storage.
2. a kind of long-distance education teacher automatic identification image optimization tracking according to claim 1, which is characterized in that
It include Kalman filtering and Morphological scale-space to the processing of foreground pixel data in the step S33.
3. a kind of long-distance education teacher automatic identification image optimization tracking according to claim 1, which is characterized in that
Distance representated by the per unit of rectangular coordinate system is according to image minimum resolution in advance calibrated visual field in the S4
Representative minimum range converts to obtain.
4. a kind of long-distance education teacher automatic identification image optimization tracking system, which is characterized in that including：
Video acquisition module, including multi-channel video capturing complexes, for acquiring video data in real time, and according to adjustment signal
Adjust the angle and direction of every road video acquisition；
Memory module acquires video data for storing video acquisition in real time；
Teacher's target confines processing module, for extracting the video data of memory module storage, and each frame image data is made
Target is carried out with Faster RCNN to confine, generates Suggestion box, will generate the image data in Suggestion box according to prefixed time interval
T stage extraction storage etc. is to be processed；
Foreground extracting module carries out prospect for the image data mixed Gauss model in the generation Suggestion box of stage extraction storage
It extracts identification and obtains teacher's target data, including：
Data modeling module, for being modeled to obtain the picture to any pixel point in the preceding M frame image in this section of image data
The probability-distribution function of vegetarian refreshments：
X indicates the pixel gray value, and at the time of t is that the frame image corresponds to, K is the number for choosing the Gauss model of fitting, and η is
Gaussian probability-density function, ω belong to the weight of different functions, and μ and ε are the mean vector of n-th of Gauss model of t moment respectively
And covariance matrix；
Pixel classifications module, for obtaining probability-distribution function P (Xt) after, since M+1 frame, to each frame image data
Mixed Gauss model is updated after acquisition, is matched with each pixel in present image with mixed Gauss model, if matching,
Think that the pixel is background pixel；Otherwise foreground pixel data are extracted as；
Teacher's target generation module, for obtaining the teacher in each frame image according to the foreground pixel data combined treatment of extraction
Target data；
Location position module obtains the position of teacher's target for establishing rectangular coordinate system as origin using the central point of every frame image
Parameter is set, the running track of teacher's target is obtained according to the difference of location parameter between consecutive frame image；
Adjusting parameter generation module obtains including adjustment time interval, angle letter for the running track according to teacher's target
Adjusting parameter including breath, elevation information, motion velocity information, and adjusting parameter is fed back into video acquisition module, video is adopted
Collection module is according to the angle and direction of obtained adjusting parameter adjustment video acquisition to guarantee that teacher's target is steadily restored in place
At image center position；
Video data compression processing module is calculated for the location parameter according to obtained teacher's target according to preset frame number Q
The teacher target position between adjacent Q frame image changes difference △ S out；When target position, variation difference △ S is more than preset threshold Y
When, by adjacent Q frame image whole compression transmission to server；When target position, variation difference △ S is less than preset threshold Y,
Adjacent Q frame image is removed into a frame image according to every a frame, obtains the video image data of scaled-down version and compression；
Communication module, video data transmission for acquiring video acquisition module to memory module, video data pressure will be passed through
Contracting processing module treated video image data is sent into server storage and adjusting parameter is transmitted to video acquisition module.
5. a kind of long-distance education teacher automatic identification image optimization tracking system according to claim 4, which is characterized in that
Teacher's target generation module further includes Kalman filtering module and Morphological scale-space module.
6. a kind of long-distance education teacher automatic identification image optimization tracking system according to claim 4, which is characterized in that
Further include abnormal parameters processing module in the adjusting parameter generation module, is used for when adjusting parameter is more than predetermined threshold, to
Video acquisition module issues command signal, and video acquisition module is adjusted the angle and direction of video acquisition by max-thresholds, simultaneously
Abnormal alarm information is issued, the road video transfer signal is closed, access includes the another road video transfer signal of teacher's target.
7. a kind of long-distance education teacher automatic identification image optimization tracking system according to claim 4, which is characterized in that
It further include image pre-processing module, for carrying out also carrying out each frame image before target is confined using Faster RCNN
Pretreatment, including：Influence of the target shadow to Objective extraction is alleviated using HSV color space, is removed using median filter
Salt-pepper noise makes image become simple using image binaryzation, using iconology denoise in corrosion and expansion respectively extract
Eliminate picture noise and filling image cavity.
8. a kind of long-distance education teacher automatic identification image optimization tracking system according to claim 4, which is characterized in that
The video acquisition complexes include photographic device and adjustment device；
The photographic device includes main photographic device and second camera device, and the second camera device is two, respectively with master
Photographic device has a 50% superposition visual field, without superimposition between the second camera device；
The angle of the video acquisition for the adjusting parameter adjustment photographic device that the adjustment device is used to transmit according to communication module with
Direction.
9. a kind of long-distance education teacher automatic identification image optimization tracking system according to claim 8, which is characterized in that
The photographic device is spherical panorama capture device, including ambient light sensor, optical filter, camera lens, lens bracket and spherical shape
Panoramic high-definition camera.
10. a kind of long-distance education teacher automatic identification image optimization tracking system according to claim 4, feature exist
In the size of the K value is 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810576145.4A CN108876672A (en) | 2018-06-06 | 2018-06-06 | A kind of long-distance education teacher automatic identification image optimization tracking and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810576145.4A CN108876672A (en) | 2018-06-06 | 2018-06-06 | A kind of long-distance education teacher automatic identification image optimization tracking and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108876672A true CN108876672A (en) | 2018-11-23 |
Family
ID=64337277
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810576145.4A Withdrawn CN108876672A (en) | 2018-06-06 | 2018-06-06 | A kind of long-distance education teacher automatic identification image optimization tracking and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108876672A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110631680A (en) * | 2019-04-26 | 2019-12-31 | 深圳市豪视智能科技有限公司 | Vibration detection system |
| CN111656275A (en) * | 2018-12-11 | 2020-09-11 | 华为技术有限公司 | Method and device for determining image focusing area |
| CN112734231A (en) * | 2021-01-09 | 2021-04-30 | 深圳市瑞驰文体发展有限公司 | Billiard event management platform |
| CN115880111A (en) * | 2023-02-22 | 2023-03-31 | 山东工程职业技术大学 | Virtual simulation training classroom teaching management method and system based on images |
-
2018
- 2018-06-06 CN CN201810576145.4A patent/CN108876672A/en not_active Withdrawn
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111656275A (en) * | 2018-12-11 | 2020-09-11 | 华为技术有限公司 | Method and device for determining image focusing area |
| CN110631680A (en) * | 2019-04-26 | 2019-12-31 | 深圳市豪视智能科技有限公司 | Vibration detection system |
| CN110631680B (en) * | 2019-04-26 | 2021-10-15 | 深圳市豪视智能科技有限公司 | Vibration detection system |
| CN112734231A (en) * | 2021-01-09 | 2021-04-30 | 深圳市瑞驰文体发展有限公司 | Billiard event management platform |
| CN112734231B (en) * | 2021-01-09 | 2021-08-24 | 深圳市瑞驰文体发展有限公司 | Billiard event management platform |
| CN115880111A (en) * | 2023-02-22 | 2023-03-31 | 山东工程职业技术大学 | Virtual simulation training classroom teaching management method and system based on images |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108765354A (en) | A kind of data-optimized transmission method of dynamic object recognition tracking image and system | |
| CN105069472B (en) | A kind of vehicle checking method adaptive based on convolutional neural networks | |
| CN106909888B (en) | Face key point tracking system and method applied to mobile equipment terminal | |
| EP2491530B1 (en) | Determining the pose of a camera | |
| CN108876672A (en) | A kind of long-distance education teacher automatic identification image optimization tracking and system | |
| CN104050449B (en) | A kind of face identification method and device | |
| US9317762B2 (en) | Face recognition using depth based tracking | |
| CN105160310A (en) | 3D (three-dimensional) convolutional neural network based human body behavior recognition method | |
| CN107481315A (en) | A kind of monocular vision three-dimensional environment method for reconstructing based on Harris SIFT BRIEF algorithms | |
| CN108806146A (en) | A kind of safety monitoring dynamic object track lock method and system | |
| CN103530599A (en) | Method and system for distinguishing real face and picture face | |
| CN110148223B (en) | Method and system for concentrating and expressing surveillance video target in three-dimensional geographic scene model | |
| CN111460968A (en) | Video-based unmanned aerial vehicle identification and tracking method and device | |
| CN108268864A (en) | Face identification method, system, electronic equipment and computer program product | |
| CN109145803A (en) | Gesture identification method and device, electronic equipment, computer readable storage medium | |
| CN108805837A (en) | A kind of dynamic object efficiently tracks image data transfer method and system | |
| CN110570457B (en) | Three-dimensional object detection and tracking method based on stream data | |
| CN110544301A (en) | Three-dimensional human body action reconstruction system, method and action training system | |
| CN110136144B (en) | Image segmentation method and device and terminal equipment | |
| CN107248174A (en) | A kind of method for tracking target based on TLD algorithms | |
| CN108112271A (en) | Movement in detection image | |
| CN108805073A (en) | A kind of safety monitoring dynamic object optimization track lock method and system | |
| CN103500452A (en) | Scenic spot scenery moving augmented reality method based on space relationship and image analysis | |
| CN108510520B (en) | A kind of image processing method, device and AR equipment | |
| CN111325051A (en) | Face recognition method and device based on face image ROI selection |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20181123 |