CN100589114C - Apparatus and method for tracking maximum face in image - Google Patents
Apparatus and method for tracking maximum face in image Download PDFInfo
- Publication number
- CN100589114C CN100589114C CN200710308520A CN200710308520A CN100589114C CN 100589114 C CN100589114 C CN 100589114C CN 200710308520 A CN200710308520 A CN 200710308520A CN 200710308520 A CN200710308520 A CN 200710308520A CN 100589114 C CN100589114 C CN 100589114C
- Authority
- CN
- China
- Prior art keywords
- face
- frame
- adult
- people
- tracking
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title abstract description 32
- 238000001514 detection method Methods 0.000 claims abstract description 49
- 230000015572 biosynthetic process Effects 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 8
- YBJHBAHKTGYVGT-ZKWXMUAHSA-N (+)-Biotin Chemical compound N1C(=O)N[C@@H]2[C@H](CCCCC(=O)O)SC[C@@H]21 YBJHBAHKTGYVGT-ZKWXMUAHSA-N 0.000 claims description 5
- FEPMHVLSLDOMQC-UHFFFAOYSA-N virginiamycin-S1 Natural products CC1OC(=O)C(C=2C=CC=CC=2)NC(=O)C2CC(=O)CCN2C(=O)C(CC=2C=CC=CC=2)N(C)C(=O)C2CCCN2C(=O)C(CC)NC(=O)C1NC(=O)C1=NC=CC=C1O FEPMHVLSLDOMQC-UHFFFAOYSA-N 0.000 claims description 5
- 230000007423 decrease Effects 0.000 abstract 2
- 238000005516 engineering process Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000006073 displacement reaction Methods 0.000 description 2
- 241000405217 Viola <butterfly> Species 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000011430 maximum method Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Abstract
The present invention discloses a tracking device and a method for the biggest face in images, which is applied to face identification and tracking, can decrease computational complexity and improve operation efficiency. The method of the present invention is characterized in that: according to the detector scale used in the detection of the biggest face in the prior frame and the face frame position of the biggest face detected, the face detector scale and detection range in the current frame are determined, and the biggest face detection is hereby performed to the current frame; then a faceframe corresponding to the detected biggest face is added to a biggest face tracking queue, so as to realize the tracking to the biggest face. The present invention utilizes the face frame of the biggest face detected in the prior frame and the face frame position to assist the biggest face detection for the current frame, which decreases computational complexity and improves tracking speed.
Description
Technical field
The present invention relates to image processing techniques, specifically, relate to the tracking means and the method for adult's face in a kind of image.
Background technology
In recent years, along with the development of mode identification technology, in computer vision and technical field of image processing, people's face information of obtaining in image or the video all has important use such as fields such as man-machine interaction, safety, amusements.Therefore, number, size, the positional information of obtaining people's face from the image automatically technology of line trace of going forward side by side has been subjected to great attention.
Existing face tracking technology generally all can at first adopt people's face to detect the people's face that exists in definite image, adopts the face tracking technology that target people face is followed the tracks of then.In the application that a lot of video images are handled; often only be concerned about people's face maximum in the video; such as carrying out AE (Auto-Exposure based on people's face information; automatic exposure), AF (Auto-Focus; automatically focus) and AWB (AutoWhite Balance; Automatic white balance is adjusted), sight protectio, numeral amplify human face region, and people's face of maximum in the video also often only is concerned about in applications such as recognition of face.And existing face tracking technology, some method can only be followed the tracks of people's face, nor be concerned about whether tracking target is maximum people's face, though other methods can be followed the tracks of a plurality of people's faces, determine and there is redundant operation in the application of following the tracks of adult's face for above-mentioned needs.
In sum, be necessary to provide the apparatus and method of adult's face in a kind of quick tracking image.
Summary of the invention
Technical matters to be solved by this invention is tracking means and the method that adult's face is provided to provide in a kind of image, to reduce computational complexity, improves operational efficiency.
In order to solve the problems of the technologies described above, the present invention at first provides the tracking means of adult's face in a kind of image, comprising:
Detection module is used for that present frame is carried out adult's face and detects;
The tracking queue memory module links to each other with described detection module, is used to store the detected pairing people's face of adult's face frame;
People's face frame record update module, certainly link to each other with described tracking queue storage mould, be used for according to the detected pairing people's face of the adult's face frame of present frame, the people's face frame record corresponding to same individual face that described tracking queue memory module is preserved mates renewal;
Tracking module links to each other with described tracking queue memory module, is used for carrying out maximum face tracking at people's face frame that described tracking queue memory module is preserved.
The tracking means of adult's face in the aforesaid image, described detection module can comprise:
Detecting device is preset submodule, is used for the yardstick of a default human-face detector;
Yardstick progression generates submodule, is used to obtain a series of yardstick progression;
Detecting device generates submodule, generates submodule with the default submodule of described detecting device with yardstick progression and links to each other, and according to the human-face detector of default yardstick, the mode of employing feature scaling obtains a series of human-face detectors of different scale;
Detect implementation sub-module, generate submodule with described detecting device and link to each other, according to by the order of large scale to small scale, adopt described a series of human-face detector that present frame is carried out adult's face detection successively, detected first people's face is the adult's face of present frame.
The present invention and then the tracking of adult's face in a kind of image is provided comprises step:
(1) according to human-face detector and a series of yardstick progression of default yardstick, the mode of employing feature scaling obtains a series of human-face detectors of different scale;
(2) if first frame of present frame for follow the tracks of handling, perhaps previous frame does not detect people's face, then adopts the described a series of human-face detectors that obtain in the step (1) to carry out adult's face in present frame full figure scope and detects;
If previous frame detects adult's face, then detect the employed detecting device yardstick of adult's face and detect people's face frame position of adult's face according to previous frame, determine present frame human-face detector range scale and sensing range, and, in the described sensing range of present frame, carry out adult's face and detect according to the human-face detector in the described present frame human-face detector range scale;
(3) the detected pairing people's face of adult's face frame is added in the maximum face tracking formation, and everyone face frame record in pairing people's face frame and the described tracking queue is mated,
If match people's face frame record, then adopt the described detected pairing people's face of adult's face frame to upgrade people's face frame record in the described tracking queue corresponding to same individual face corresponding to same individual face;
If do not match people's face frame record, then the detected pairing people's face of adult's face frame is added in the described tracking queue as new record corresponding to same individual face;
(4) in described tracking queue, determine adult's face of current tracking, and with the next frame of image as present frame, return step (2) and continue to carry out.
The tracking of adult's face in the aforesaid image obtains the step of a series of human-face detectors of different scale described in the step (1), can comprise:
Set the default yardstick MS of human-face detector, a series of yardstick progression Nk and detection yardstick scaling factor S S, the mode of employing feature scaling obtains a series of human-face detectors of different scale.
Further, set maximum yardstick MAXF, the minimum detection yardstick MINF of detecting, obtain described a series of yardstick progression Nk, can comprise step:
(1), obtains to satisfy MS*SS according to described MS, MAXF, MINF and SS
N1The maximum integer N1 of≤MINF satisfies MS*SS
N2The smallest positive integral N2 of 〉=MAXF;
(2) obtain described a series of yardstick progression N, wherein out to out progression is N2, and smallest dimension progression is M.
The tracking of adult's face in the aforesaid image carries out the step that adult's face detects in present frame full figure scope described in the step (2), can comprise:
To small scale, order adopts described a series of human-face detector to carry out adult's face in present frame full figure scope and detects by large scale, and detected first people's face is the adult's face of present frame.
The tracking of adult's face in the aforesaid image, detect the employed detecting device yardstick of adult's face according to previous frame described in the step (2) and detect people's face frame position of adult's face, determine the step of present frame human-face detector range scale and sensing range, can comprise:
Described previous frame is detected the employed detecting device yardstick of adult's face carry out proportional zoom, determine to comprise the described present frame human-face detector range scale that described previous frame detects the employed detecting device yardstick of adult's face;
Detect people's face frame position of adult's face according to previous frame, determine the horizontal ordinate and the ordinate of present frame sensing range central point;
Previous frame is detected people's face frame size of adult's face, and perhaps described previous frame detects the employed detecting device yardstick of adult's face and carries out proportional zoom, determines the width and the height of present frame sensing range.
Further, according to the human-face detector in the described present frame human-face detector range scale, in the described sensing range of present frame, carry out the step that adult's face detects in the step (2), can comprise:
Detecting the employed detecting device yardstick of adult's face with described previous frame is the expansions of middle mind-set both sides, adopt everyone face detector in the described present frame human-face detector range scale, described present frame sensing range is detected, and detected first people's face is the adult's face of present frame.
The tracking of adult's face in the aforesaid image, the step of in the step (3) everyone face frame record in pairing people's face frame and the described tracking queue being mated can comprise:
(31) calculate the centre distance of described pairing people's face frame and everyone face frame of described tracking queue and carry out normalization, obtain normalization centre distance;
(32) size of everyone face frame in described pairing people's face frame of calculating and the described tracking queue;
(33) if contain described normalization centre distance in the described tracking queue smaller or equal to default distance threshold and described size people's face frame record smaller or equal to default proportion threshold value, contain in the then described tracking queue and the people face frame record of described pairing people's face frame, otherwise do not have and the people face frame record of described pairing people's face frame in the described tracking queue corresponding to same individual face corresponding to same individual face.
The tracking of adult's face in the aforesaid image, step (3) may further include:
People's face frame table corresponding with the detected adult's face of present frame in the described tracking queue is shown that the detection frame number that detects adult's face adds 1, do not detect the omission frame number zero setting of people's face, and the current time that will detect adult's face is recorded as the concluding time;
If do not detect adult's face at present frame, the omission frame number that then everyone face frame record expression in the described tracking queue is not detected people's face adds 1.
Further, in described tracking queue, determine the step of adult's face of current tracking in the step (4), can comprise:
Detect frame number described in the described tracking queue greater than default detection frame number threshold value, described omission frame number is less than default omission frame number threshold value, and up-to-date people's face frame record of described concluding time is as adult's face frame of current tracking, and people's face of adult's face frame correspondence of described current tracking is adult's face of described current tracking.
Further, step (4) further can comprise:
Delete the people face frame record of omission frame number described in the described tracking queue greater than described omission frame number threshold value.
Compared with prior art, the present invention detects the people's face frame and the position of adult's face by previous frame, assists adult's face of present frame to detect, and has reduced computational complexity, has improved tracking velocity.
Description of drawings
Fig. 1 is the composition synoptic diagram of the maximum face tracking device of the present invention embodiment.
Fig. 2 is the composition synoptic diagram of detection module embodiment in apparatus of the present invention shown in Figure 1.
Fig. 3 is the step synoptic diagram of the maximum face tracking method embodiment of the present invention.
Embodiment
Describe embodiments of the present invention in detail below with reference to drawings and Examples, how the application technology means solve technical matters to the present invention whereby, and the implementation procedure of reaching technique effect can fully understand and implements according to this.
Basic thought of the present invention is the adult's face that adopts that maximum fast human face detection tech obtains to exist in the image, and the people's face that obtains in the different frame is mated, and obtains continuous position and the size of adult's face in video.In order further to reduce operand, employed detecting device yardstick in the time of can adopting previous frame to detect adult's face frame, and the position and the size of detected this adult's face frame, come to determine detect the detecting device range scale that the adult's face frame of present frame need use, and present frame detects the position range of adult's face.
Human body for low-speed motion, at adjacent two frames in pairing this short time, the displacement of people's face is limited, and in general in these adjacent two frames the variation of people's face size also not too large, based on such common sense knowledge, can assist people's face of present frame to detect by the corresponding information that detects people's face in the previous frame.
Based on basic thought of the present invention, Fig. 1 shows the structural representation of the maximum face tracking device of the present invention embodiment, comprises as lower module:
Tracking queue memory module 20 links to each other with detection module 10, is used to store the detected pairing people's face of adult's face frame;
People's face frame record update module 30, link to each other with tracking queue memory module 20, be used for according to the detected pairing people's face of the adult's face frame of present frame, the people's face frame record of preserving in the tracking queue memory module 20 corresponding to same individual face is mated renewal;
As shown in Figure 2, detection module 10 can be further divided into following submodule:
Detecting device is preset submodule 110, is used for the yardstick of a default human-face detector;
Yardstick progression generates submodule 120, is used to obtain a series of yardstick progression;
Detecting device generates submodule 130, generates submodule 120 with the default submodule 110 of detecting device with yardstick progression and links to each other, and according to the human-face detector of default yardstick, the mode of employing feature scaling obtains a series of human-face detectors of different scale;
Detect implementation sub-module 140, generate submodule 130 with detecting device and link to each other, according to by the order of large scale to small scale, adopt described a series of human-face detector that image is carried out adult's face detection successively, detected first people's face is the adult's face of present frame.
Below come understood in detail apparatus of the present invention embodiment by the inventive method the concrete operations mode.Fig. 3 shows an embodiment of the inventive method, comprises step:
Step 320 according to the detecting device in the detecting device range scale of the adult's face frame of the present frame of determining, in the sensing range of the adult's face frame of the present frame of determining, is carried out adult's face to present frame and is detected;
Step 330 if detect adult's face in the sensing range of the adult's face frame of the present frame of determining, is then changeed step 340, otherwise is changeed step 350;
Step 350 adopts the detecting device of all default yardsticks that present frame is carried out the full figure detection, avoids exporting incorrect result because sensing range is improper or the detecting device yardstick is improper;
Step 360 if detect adult's face in the full figure scope, is then changeed step 340, otherwise changes step 370;
Step 370 adds 1 with everyone the omission frame number of face frame record in the tracking queue, and keeps everyone the detection frame number of face frame record constant, changes step 380 and continues to carry out;
Finish after the detection and tracking of present frame, the next frame image as present frame, returned step 310, continue to detect and tracking image in adult's face.
In the inventive method when beginning operation, need be initialized as sky with maximum face tracking formation, also needs write down the detection frame number that successfully detects adult's face and to be used to write down the omission frame number that does not successfully detect adult's face.
In the foregoing description, need use maximum human face detection tech, promptly, adopt the human-face detector that limits yardstick, detect people's face of the out to out in this sensing range for limiting the image detection scope.The inventive method provides a kind ofly carries out the method that adult's face detects to limiting sensing range, comprises the steps:
(a) according to human-face detector and a series of yardstick progression of default yardstick, the mode of employing feature scaling obtains a series of human-face detectors of different scale;
(b) according to by the order of large scale to small scale, adopt described a series of human-face detector that described image is detected successively, retreat out processing detecting first people's face, this detected first people's face is adult's face.
A kind of more excellent implementation method that wherein adopts in the step (a) mode of feature scaling to obtain a series of human-face detectors of different scale is, the yardstick MS of an at first default human-face detector (comprising width and height), a series of yardstick progression Nk and a detection yardstick scaling factor S S, wherein a series of yardstick progression are mutual unequal a plurality of positive integers, can obtain a series of human-face detectors of different scale thus.For some yardstick progression n, corresponding detecting device yardstick is ROUND (MS*SS
n), wherein ROUND () is the operational symbol that rounds up.For different human-face detectors, the mode of feature scaling is also different, strengthen the embodiment of the method for detecting human face of (Adaboost) sorter below based on microstructure features (Haar-like Features) and level type self-adaptation, the method for detector feature scaling is described.
At document P.Viola and M.Jones.Robust real time object detection.IEEEICCV Workshop on Statistical and Computational Theories of Vision, Vancouver, Canada, Voila etc. has proposed a kind of human face detection tech that strengthens sorter based on microstructure features and level type self-adaptation among the July 13,2001.At first, adopt the method for training to obtain the human-face detector of a fixed size, such as the width of the human-face detector that trains with highly be respectively MW and the MH (MW=24 that Viola adopts, MH=24), in order to detect people's face of different sizes, diverse location, need carry out the feature scaling to the model that trains, to obtain a series of human-face detectors of different scale.Suppose that detecting the yardstick scaling factor still is SS, then adopt a series of different scales that the mode scaling of feature scaling obtains sorter width and highly be respectively ROUND (MW*SS
n) and ROUND (MH*SS
n).Wherein, n is the integer more than or equal to 0, the yardstick progression of expression human-face detector, and ROUND () expression is carried out the round computing to the numerical value in the bracket.Successively the microstructure features in the detecting device being carried out ratio is SS
nScaling, just obtained the human-face detector of corresponding scale.
Among the invention described above method embodiment, a kind of feasible program that everyone face frame in detected adult's face pairing people's face frame and the tracking queue is mated is:
Suppose that the detected people's face of present frame frame is R (cx ', cy ', w ', h '), wherein cx ' and cy ' represent horizontal ordinate and the ordinate of central point in image of this people's face frame respectively, and w ' and h ' represent the wide and high of this people's face frame rectangle respectively; People's face frame in the maximum face tracking formation be R (cxm, cym, wm, hm), wherein cxm and cym represent horizontal ordinate and the ordinate of central point in image of this people's face frame respectively, wm and hm represent the wide and high of this people's face frame rectangle respectively.
So, the centre distance cdis=sqrt ((cxm-cx ') of the people's face frame in present frame detected people's face frame and the maximum face tracking formation
2+ (cym-cy ')
2), normalization centre distance is
The size of the two is:
Wherein min is for getting the minimum value operation, and max is for getting maxima operation.
Based on aforesaid common sense knowledge, people's displacement is limited at short notice, and the variation of people's face size is not too large in adjacent two frames, therefore can draw:
If a certain individual face frame record satisfies in the tracking queue
And SR≤TSR, wherein TDIS is the threshold value of the normalization distance of setting, TSR is the threshold value of the dimension scale of setting, so:
Think that then the two is the people face of same individual in different frame, and this people's face frame information upgraded that TSR more preferably can be in value in the scope of (0,1), otherwise thinks that the two is not people's face frame of same individual face correspondence.
Among the invention described above method embodiment, TC and TM are the constant greater than zero, and the more excellent TC that for example gets is 4, and TM is 2, can certainly be worth for other, and the present invention is not as limit.If wherein the detection failure frame number of someone's face frame record is then deleted it greater than threshold value TM from maximum face tracking formation.
In an application example of the inventive method, be used for following the tracks of a series of human-face detectors of adult's face, be to follow the tracks of directly setting before the operation, according to yardstick descending be respectively M5, M4 ..., M1, totally five yardsticks of M5～M1 is all inequality.Suppose that previous frame detects adult's face, and what use is the detecting device of this yardstick of M3, the detected pairing people's face of adult's face frame is R (cx, cy, w, h), wherein cx and cy represent horizontal ordinate and the ordinate of central point in image of this people's face frame respectively, and w and h represent the wide and high of this people's face frame rectangle respectively.
When present frame is detected, determine that according to the M3 yardstick yardstick of present frame human-face detector is M2, M3 and M4, according to adult's face frame R (cx, cy, w h) determines that the present frame sensing range is RS (sx, sy, Ws, Hs), sx=cx wherein, sy=cy, be respectively the horizontal ordinate and the ordinate of present frame sensing range central point, Ws=w* α is the width of sensing range, and Hs=h* β is the height of sensing range, α and β are constant, for suitable amplification detection scope, α and β more preferably get the constant greater than 1, get 1.5 in the middle of this example.In addition, in the middle of other application examples of the present invention, determining the method for present frame sensing range, also can be according on the center with the detected adult's face frame of previous frame, and the yardstick of this detected adult's face frame is carried out various forms of suitable convergent-divergents and obtains.
In the middle of the Another application example of the inventive method, the a series of human-face detectors that are used for following the tracks of adult's face obtain according to parameter preset, particularly, at first default human-face detector that yardstick is MS, set maximum yardstick MAXF, minimum detection yardstick MINF and the detection yardstick scaling factor S S of detecting then, obtain to satisfy MS*SS more respectively
N1The maximum integer N1 of≤MINF and satisfy MS*SS
N2The smallest positive integral N2 of 〉=MAXF has so just obtained representing with N1≤N≤N2 that more than or equal to N1 and smaller or equal to a series of integers of N2 N wherein is an integer.These integers are corresponding human-face detector yardstick progression respectively, and wherein out to out progression is N2, and smallest dimension progression is N1.According to the yardstick MS of these yardstick progression and default human-face detector, just can obtain a series of human-face detectors of different scale, use MS*SS
NExpression.
Suppose that previous frame detects the yardstick MS*SS that the employed human-face detector of adult's face is
NkAlso be that the corresponding yardstick progression of employed human-face detector is Nk, be that expansion obtains several detecting device yardstick progression to both sides respectively at the center so with Nk, should be with being [N (k-a) in the middle of the example, N (k+a)], corresponding these several yardstick progression, the present frame detecting device range scale that obtains are from MS*SS
N (k-a)To MS*SS
N (k+a)
Determining the mode of present frame sensing range, also be to use the mode of determining sensing range in the application example with in the middle of the example.
Using from MS*SS
N (k-a)To MS*SS
N (k+a)Human-face detector when present frame is detected, should be with example according to being that Nk is that the center is gradually to the mode of both sides expansions with progression, respectively the determined sensing range of present frame is detected, if being determined, certain rectangle frame is people's face frame, then withdraw from, with its adult's face frame that is defined as that current detection arrives, otherwise continue to handle, finish up to all detecting to the pairing detecting device of all yardstick progression of N (k+a) from N (k-a).
In the middle of other embodiment of the inventive method, determine the method for present frame human-face detector range scale, can also realize according to the detected pairing people's face of the adult's face frame of previous frame.Still suppose that the detected pairing people's face of the adult's face frame of previous frame is R (cx, cy, w, h), can determine according to the width w or the height h of this people's face frame, such as according to w, determine that then present frame detecting device range scale is for being [w*SR0, w*SR1], wherein the SR0 relative previous frame of employed smallest dimension progression that is present frame in the sensing range of determining detects the proportionality constant of the employed yardstick progression of adult's face, more preferably such as getting 0.5, SR1 is the proportionality constant that present frame relative previous frame of employed out to out progression in the sensing range of determining detects the employed yardstick progression of adult's face, more preferably such as getting 2.
The detecting device range scale of above-mentioned present frame is that example is determined with w, in fact also can select to determine by h, also be in the tracking of the present invention, the detecting device range scale of present frame can be that the size of the detected adult's face frame of previous frame is carried out the convergent-divergent of proper proportion and obtained.In addition, because the size of the detected pairing people's face of adult's face frame (width and height), with the yardstick (width and height) of employed detecting device is corresponding, therefore carry out convergent-divergent according to detecting the pairing people's face of adult's face frame, with carry out convergent-divergent according to detecting the employed human-face detector of adult's face, principle is the same.
The present invention follows the tracks of the pairing adult's face frame of adult's face, and be not only the adult's face in the image, and have certain continuation, for adopting people's face to carry out AE, AWB, application such as AF all have use value.Be less than people's face of TC for occurrence number, think to be to disturb the mistake of avoiding the detection algorithm flase drop to cause.The setting of frequency of failure threshold value TM is avoided because once in a while failure and situation about can't follow the tracks of.Not only people's face in the maximum face tracking formation, when having guaranteed the adult's face disappearance when the front, other people face can be detected as adult's face.
When existing method is followed the tracks of adult's face, need each yardstick and position be detected, processing speed is very slow.The maximum method for detecting human face of the present invention at first on detection architecture, is handled large scale earlier, handles small scale then, and just withdraws from testing process after detecting people's face, has improved detection efficiency.Further, in the face tracking processing procedure, dwindle magnitude range and the position range of following the tracks of detection, quicken computing according to size that detects people's face and positional information.And, can also determine whether two people's faces are same individual face according to the position and the magnitude relationship of people's face between consecutive frame.
Though the disclosed embodiment of the present invention as above, the embodiment that described content just adopts for the ease of understanding the present invention is not in order to limit the present invention.Technician in any the technical field of the invention; under the prerequisite that does not break away from the disclosed spirit and scope of the present invention; can do a little change and retouching what implement in form and on the details; but scope of patent protection of the present invention, still must with appending claims the person of being defined be as the criterion.
Claims (12)
1, the tracking means of adult's face in a kind of image is characterized in that, comprising:
Detection module is used for that present frame is carried out adult's face and detects;
The tracking queue memory module links to each other with described detection module, is used to store the detected pairing people's face of adult's face frame;
People's face frame record update module, link to each other with described tracking queue memory module, be used for according to the detected pairing people's face of the adult's face frame of present frame, the people's face frame record corresponding to same individual face that described tracking queue memory module is preserved mates renewal;
Tracking module links to each other with described tracking queue memory module, is used for carrying out maximum face tracking at people's face frame that described tracking queue memory module is preserved.
2, the tracking means of adult's face in the image as claimed in claim 1 is characterized in that described detection module comprises:
Detecting device is preset submodule, is used for the yardstick of a default human-face detector;
Yardstick progression generates submodule, is used to obtain a series of yardstick progression;
Detecting device generates submodule, generates submodule with the default submodule of described detecting device with yardstick progression and links to each other, and according to human-face detector and described a series of yardstick progression of default yardstick, the mode of employing feature scaling obtains a series of human-face detectors of different scale;
Detect implementation sub-module, generate submodule with described detecting device and link to each other, according to by the order of large scale to small scale, adopt described a series of human-face detector that present frame is carried out adult's face detection successively, detected first people's face is the adult's face of present frame.
3, the tracking of adult's face in a kind of image is characterized in that, comprises step:
(1) according to human-face detector and a series of yardstick progression of default yardstick, the mode of employing feature scaling obtains a series of human-face detectors of different scale;
(2) if first frame of present frame for follow the tracks of handling, perhaps previous frame does not detect people's face, then adopts the described a series of human-face detectors that obtain in the step (1) to carry out adult's face in present frame full figure scope and detects;
If previous frame detects adult's face, then detect the employed detecting device yardstick of adult's face and detect people's face frame position of adult's face according to previous frame, determine present frame human-face detector range scale and sensing range, and, in the described sensing range of present frame, carry out adult's face and detect according to the human-face detector in the described present frame human-face detector range scale;
(3) the detected pairing people's face of adult's face frame is added in the maximum face tracking formation, and everyone face frame record in pairing people's face frame and the described tracking queue is mated,
If match people's face frame record, then adopt the described detected pairing people's face of adult's face frame to upgrade people's face frame record in the described tracking queue corresponding to same individual face corresponding to same individual face;
If do not match people's face frame record, then the detected pairing people's face of adult's face frame is added in the described tracking queue as new record corresponding to same individual face;
(4) in described tracking queue, determine adult's face of current tracking, and with the next frame of image as present frame, return step (2) and continue to carry out.
4, the tracking of adult's face in the image as claimed in claim 3 is characterized in that, obtains the step of a series of human-face detectors of different scale described in the step (1), comprising:
Set the default yardstick MS of human-face detector, a series of yardstick progression Nk and detection yardstick scaling factor S S, the mode of employing feature scaling obtains a series of human-face detectors of different scale;
Wherein, described human-face detector is one can judge whether the image window of certain yardstick is the model of human face region, and described yardstick refers to the width of judging window or/and height, and the value of described height and width is a positive integer, and unit is a pixel;
Described yardstick progression Nk is mutual unequal a plurality of positive integers;
Described detection yardstick scaling factor S S is an arithmetic number.
5, the tracking of adult's face in the image as claimed in claim 4 is characterized in that, sets maximum yardstick MAXF, the minimum detection yardstick MINF of detecting further, obtains described a series of yardstick progression Nk, comprises step:
(a), obtain to satisfy MS*SS according to described MS, MAXF, MINF and SS
N1The maximum integer N1 of≤MINF satisfies MS*SS
N2The smallest positive integral N2 of 〉=MAXF;
(b) obtain described a series of yardstick progression Nk, wherein out to out progression is N2, and smallest dimension progression is N1;
Wherein, the described maximum value that detects yardstick MAXF and minimum detection yardstick MINF is a positive integer, and unit is a pixel.
6, the tracking of adult's face in the image as claimed in claim 3 is characterized in that, carries out the step that adult's face detects described in the step (2) in present frame full figure scope, comprising:
To small scale, order adopts described a series of human-face detector to carry out adult's face in present frame full figure scope and detects by large scale, and detected first people's face is the adult's face of present frame.
7, the tracking of adult's face in the image as claimed in claim 3, it is characterized in that, detect the employed detecting device yardstick of adult's face according to previous frame described in the step (2) and detect people's face frame position of adult's face, determine the step of present frame human-face detector range scale and sensing range, comprising:
Described previous frame is detected the employed detecting device yardstick of adult's face carry out proportional zoom, determine to comprise the described present frame human-face detector range scale that described previous frame detects the employed detecting device yardstick of adult's face;
Detect people's face frame position of adult's face according to previous frame, determine the horizontal ordinate and the ordinate of present frame sensing range central point;
Previous frame is detected people's face frame size of adult's face, and perhaps described previous frame detects the employed detecting device yardstick of adult's face and carries out proportional zoom, determines the width and the height of present frame sensing range.
8, the tracking of adult's face in the image as claimed in claim 7, it is characterized in that, according to the human-face detector in the described present frame human-face detector range scale, in the described sensing range of present frame, carry out the step that adult's face detects in the step (2), comprising:
Obtain with described previous frame several detecting device yardstick progression that to detect the employed detecting device yardstick of adult's face be the center;
The human-face detector of the detecting device yardstick progression correspondence that employing obtains according to by the order of large scale to small scale, detects described present frame sensing range, and detected first people's face is the adult's face of present frame.
9, the tracking of adult's face in the image as claimed in claim 3 is characterized in that, the step of in the step (3) everyone face frame record in pairing people's face frame and the described tracking queue being mated comprises:
(31) calculate the normalization centre distance of described pairing people's face frame and everyone face frame of described tracking queue, described normalization centre distance be described pairing people's face frame with described tracking queue in the ratio of the width of people's face frame in centre distance and the described tracking queue of people's face frame;
(32) size of everyone face frame in described pairing people's face frame of calculating and the described tracking queue;
(33) if contain described normalization centre distance in the described tracking queue smaller or equal to default distance threshold and described size people's face frame record smaller or equal to default proportion threshold value, contain in the then described tracking queue and the people face frame record of described pairing people's face frame, otherwise do not have and the people face frame record of described pairing people's face frame in the described tracking queue corresponding to same individual face corresponding to same individual face.
10, the tracking of adult's face in the image as claimed in claim 3 is characterized in that step (3) further comprises:
People's face frame table corresponding with the detected adult's face of present frame in the described tracking queue is shown that the detection frame number that detects adult's face adds 1, do not detect the omission frame number zero setting of people's face, and the current time that will detect adult's face is recorded as the concluding time;
If do not detect adult's face at present frame, the omission frame number that then everyone face frame record expression in the described tracking queue is not detected people's face adds 1.
11, the tracking of adult's face in the image as claimed in claim 10 is characterized in that, determines the step of adult's face of current tracking in the step (4) in described tracking queue, comprising:
Detect frame number described in the described tracking queue greater than default detection frame number threshold value, described omission frame number is less than default omission frame number threshold value, and up-to-date people's face frame record of described concluding time is as adult's face frame of current tracking, and people's face of adult's face frame correspondence of described current tracking is adult's face of described current tracking.
12, the tracking of adult's face in the image as claimed in claim 11 is characterized in that step (4) further comprises:
Delete the people face frame record of omission frame number described in the described tracking queue greater than described omission frame number threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200710308520A CN100589114C (en) | 2007-12-29 | 2007-12-29 | Apparatus and method for tracking maximum face in image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200710308520A CN100589114C (en) | 2007-12-29 | 2007-12-29 | Apparatus and method for tracking maximum face in image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101196997A CN101196997A (en) | 2008-06-11 |
| CN100589114C true CN100589114C (en) | 2010-02-10 |
Family
ID=39547389
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200710308520A Expired - Fee Related CN100589114C (en) | 2007-12-29 | 2007-12-29 | Apparatus and method for tracking maximum face in image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100589114C (en) |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8988609B2 (en) | 2007-03-22 | 2015-03-24 | Sony Computer Entertainment America Llc | Scheme for determining the locations and timing of advertisements and other insertions in media |
| US8451380B2 (en) | 2007-03-22 | 2013-05-28 | Sony Computer Entertainment America Llc | Scheme for determining the locations and timing of advertisements and other insertions in media |
| CN102833486B (en) * | 2012-07-05 | 2016-09-21 | 深圳泰山体育科技股份有限公司 | The method and device of face displaying ratio in a kind of real-time adjusting video images |
| EP3103258B1 (en) * | 2014-02-07 | 2020-07-15 | Sony Interactive Entertainment America LLC | Scheme for determining the locations and timing of advertisements and other insertions in media |
| CN104680134B (en) * | 2015-02-04 | 2018-01-12 | 四川长虹电器股份有限公司 | Quick human body detecting method |
| CN104732210A (en) * | 2015-03-17 | 2015-06-24 | 深圳超多维光电子有限公司 | Target human face tracking method and electronic equipment |
| CN106156696B (en) * | 2015-03-30 | 2020-03-24 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN104794462B (en) * | 2015-05-11 | 2018-05-22 | 成都野望数码科技有限公司 | A kind of character image processing method and processing device |
| CN104866833B (en) * | 2015-05-29 | 2018-11-09 | 中国科学院上海高等研究院 | Video flowing method for detecting human face and its device |
| CN107103265A (en) * | 2016-02-19 | 2017-08-29 | 合肥君正科技有限公司 | A kind of method and device of evaluator face detection algorithm |
| CN105913028B (en) * | 2016-04-13 | 2020-12-25 | 华南师范大学 | Face + + platform-based face tracking method and device |
| CN106228112B (en) | 2016-07-08 | 2019-10-29 | 深圳市优必选科技有限公司 | Face datection tracking and robot head method for controlling rotation and robot |
| CN106250850B (en) * | 2016-07-29 | 2020-02-21 | 深圳市优必选科技有限公司 | Face detection tracking method and device, and robot head rotation control method and system |
| CN109033924A (en) * | 2017-06-08 | 2018-12-18 | 北京君正集成电路股份有限公司 | The method and device of humanoid detection in a kind of video |
| CN108171128B (en) * | 2017-12-14 | 2020-06-23 | 浙江大华技术股份有限公司 | Face detection method and device |
| CN109101949A (en) * | 2018-08-29 | 2018-12-28 | 广州洪荒智能科技有限公司 | A kind of human face in-vivo detection method based on colour-video signal frequency-domain analysis |
| CN109740516B (en) * | 2018-12-29 | 2021-05-14 | 深圳市商汤科技有限公司 | User identification method and device, electronic equipment and storage medium |
| CN111931677A (en) * | 2020-08-19 | 2020-11-13 | 北京影谱科技股份有限公司 | Face detection method and device and face expression detection method and device |
| CN112084992B (en) * | 2020-09-18 | 2021-04-13 | 北京中电兴发科技有限公司 | Face frame selection method in face key point detection module |
| CN112200174A (en) * | 2020-12-09 | 2021-01-08 | 上海齐感电子信息科技有限公司 | Face frame detection method and module and living body face verification method and system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6118887A (en) * | 1997-10-10 | 2000-09-12 | At&T Corp. | Robust multi-modal method for recognizing objects |
| US20020176609A1 (en) * | 2001-05-25 | 2002-11-28 | Industrial Technology Research Institute | System and method for rapidly tacking multiple faces |
| CN1523533A (en) * | 2002-12-06 | 2004-08-25 | ���ǵ�����ʽ���� | Human detection through face detection and motion detection |
| WO2005071614A1 (en) * | 2004-01-27 | 2005-08-04 | Seiko Epson Corporation | Human face detection position shift correction method, correction system, and correction program |
| JP2005309862A (en) * | 2004-04-22 | 2005-11-04 | Digital Fashion Ltd | Graphic extraction program, graphic extraction method and graphic extraction apparatus |
| CN1959701A (en) * | 2005-11-03 | 2007-05-09 | 中国科学院自动化研究所 | Method for tracking multiple human faces from video in real time |
-
2007
- 2007-12-29 CN CN200710308520A patent/CN100589114C/en not_active Expired - Fee Related
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6118887A (en) * | 1997-10-10 | 2000-09-12 | At&T Corp. | Robust multi-modal method for recognizing objects |
| US20020176609A1 (en) * | 2001-05-25 | 2002-11-28 | Industrial Technology Research Institute | System and method for rapidly tacking multiple faces |
| CN1523533A (en) * | 2002-12-06 | 2004-08-25 | ���ǵ�����ʽ���� | Human detection through face detection and motion detection |
| WO2005071614A1 (en) * | 2004-01-27 | 2005-08-04 | Seiko Epson Corporation | Human face detection position shift correction method, correction system, and correction program |
| JP2005309862A (en) * | 2004-04-22 | 2005-11-04 | Digital Fashion Ltd | Graphic extraction program, graphic extraction method and graphic extraction apparatus |
| CN1959701A (en) * | 2005-11-03 | 2007-05-09 | 中国科学院自动化研究所 | Method for tracking multiple human faces from video in real time |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101196997A (en) | 2008-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100589114C (en) | Apparatus and method for tracking maximum face in image | |
| US11062123B2 (en) | Method, terminal, and storage medium for tracking facial critical area | |
| CN106909888B (en) | Face key point tracking system and method applied to mobile equipment terminal | |
| US8792722B2 (en) | Hand gesture detection | |
| US20120027263A1 (en) | Hand gesture detection | |
| US20100232712A1 (en) | Detector, detection method, and integrated circuit for detection | |
| EP2336949B1 (en) | Apparatus and method for registering plurality of facial images for face recognition | |
| CN106650615B (en) | A kind of image processing method and terminal | |
| CN106934806B (en) | It is a kind of based on text structure without with reference to figure fuzzy region dividing method out of focus | |
| WO2006013913A1 (en) | Object image detection device, face image detection program, and face image detection method | |
| CN101196995B (en) | Method for detecting maximum face in image | |
| CN110633610A (en) | Student state detection algorithm based on YOLO | |
| CN104866805A (en) | Real-time face tracking method and device | |
| CN111524145A (en) | Intelligent picture clipping method and system, computer equipment and storage medium | |
| CN108986137B (en) | Human body tracking method, device and equipment | |
| CN104866833B (en) | Video flowing method for detecting human face and its device | |
| CN111104830A (en) | Deep learning model for image recognition, training device and method of deep learning model | |
| CN111652017A (en) | Dynamic gesture recognition method and system | |
| CN101320477B (en) | Human body tracing method and equipment thereof | |
| CN111382606A (en) | Tumble detection method, tumble detection device and electronic equipment | |
| CN106682669A (en) | Image processing method and mobile terminal | |
| CN107358621A (en) | Method for tracing object and device | |
| US20060126940A1 (en) | Apparatus and method for detecting eye position | |
| CN113221812A (en) | Training method of face key point detection model and face key point detection method | |
| EP2998928B1 (en) | Apparatus and method for extracting high watermark image from continuously photographed images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100210 Termination date: 20111229 |