CN103020576A - Characteristic data compression device, multi-directional human face detection system and detection method thereof - Google Patents
Characteristic data compression device, multi-directional human face detection system and detection method thereof Download PDFInfo
- Publication number
- CN103020576A CN103020576A CN2011102796002A CN201110279600A CN103020576A CN 103020576 A CN103020576 A CN 103020576A CN 2011102796002 A CN2011102796002 A CN 2011102796002A CN 201110279600 A CN201110279600 A CN 201110279600A CN 103020576 A CN103020576 A CN 103020576A
- Authority
- CN
- China
- Prior art keywords
- transposition
- face detection
- mirror
- masterplate
- data
- 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.)
- Granted
Links
Images
Abstract
The invention discloses a characteristic data compression device, a multi-directional human face detection system and a detection method of the multi-directional human face detection system. The characteristic data compression device comprises a mirror reflection unit and a transposition unit. The mirror reflection unit is used for comparing a plurality of characteristic values to calculate to obtain mutually-symmetrical characteristic values, and getting rid of one of the two mutually-symmetrical characteristic values in the plurality of characteristic values to form first mirror reflection data. Or calculation is carried out to obtain a position of the characteristic value with the highest specific gravity value to form second mirror reflection data. The transposition unit is used for transposing at least one plate to an angle according to a transposition form to calculate to obtain transposition data.
Description
Technical field
The invention relates to a kind of face characteristic data compression device, but particularly relevant for a kind of mirror of face characteristic data and characteristic compression set that permutability is carried out data compression of utilizing, with multi-direction human face detecting system and the method for detecting of using this characteristic compression set.
Background technology
The human face detection method is that many pictures to be measured are inputed in the learning model at present.Learning model can learn how to tell whether include predefined plate feature in the picture to be measured according to the predefined plate feature of institute (Template Feature).No matter be active study framework for example: the study framework of class neural network (Neural Network), expert system, fuzzy system (Fuzzy) or sort-type for example: support vector machine (Support Vector Machine, SVM), main shaft component analysis (Principal Components Analysis, PCA), Snow method or Boosting method all need can learn according to the plate feature that sets.
In order to distinguish the difference of people's face and background in the picture to be measured, often utilize the Haar-like algorithm to carry out the acquisition of face characteristic.The Haar-like algorithm is the method that a kind of grain direction for block (Pattern) is carried out characteristic processing.Therefore the Haar-like algorithm can distinguish the different of people's face and complex background effectively.In addition, because the Haar-like algorithm is the grain direction that relies in the picture to be measured, therefore work as picture to be measured and be transposed to different directions, for example, with picture transposition 90 degree to be measured, when 180 degree or 270 are spent, because the grain direction of people's face is different in the picture to be measured, can cause the plate feature that is originally calculated by the Haar-like algorithm can't be used in by in the picture to be measured behind the transposition.
And be in the detecting picture to be measured by the people face of transposition at diverse location, need again to carry out the picture to be measured of different transposition position respectively by the Haar-like algorithm learning training of face characteristic, to produce respectively the plate feature of corresponding different transposition position.Therefore, said method not only can expend a large amount of memory body spaces, also can expend the operation time of several times.
Yet running gear such as camera, mobile phone or video camera, is inclined to compact design concept for being convenient for carrying at present, so, and the memory body space that also the limit running gear can be provided.For reaching the storage data amount that reduces the plate feature, can adopt the angle of dwindling human face detection, as only detecting with Plane Rotation angle or side face angle, perhaps, also can adopt the precision of simplifying the plate feature.Yet these methods may cause the detecting degree of people's face to reduce because of the factors such as restriction, low-light level or the dark colour of skin of angle.
Summary of the invention
Because the problem of above-mentioned existence, purpose of the present invention just is to provide multi-direction human face detecting system and the method for detecting thereof of a kind of characteristic compression set, application characteristic data compression device, and the people's face that is transposed to diverse location as detecting take solution need expend the operation time of numerical digit and expend the problem in a large amount of memory body spaces.
According to a wherein purpose of the present invention, the present invention by the following technical solutions: a kind of characteristic compression set, it comprises a converting unit, a mirror unit and a transposition unit.Converting unit is by a facial image, and computing obtains some eigenwerts of facial image, and forms at least one plate according to these some eigenwerts, and each eigenwert comprises a rate of specific gravity.The mirror unit connects converting unit, can compare one by one each eigenwert, and computing obtains the position of each symmetrical eigenwert, and gives up in these some eigenwerts symmetrical in twos one of them, forms one first mirror data.Perhaps, the mirror unit can be compared each eigenwert one by one, obtains a position with eigenwert of high specific gravity value with computing, forms one second mirror data.The transposition unit connects converting unit, and this at least one plate to an angle of foundation one transposition list transposition, and obtains some transposition eigenwerts by at least one plate computing of this transposition, forms transposition data.
Further, by described the first mirror data formation one positive face masterplate; By described the second mirror data formation one side face masterplate; By the positive face masterplate of described transposition data formation one transposition or a transposition side face masterplate.
According to another object of the present invention, the present invention is by the following technical solutions: a kind of multi-direction human face detecting system of application one characteristic compression set, it comprises an image capturing device, characteristic compression set and a human face detection device as mentioned above.Image capturing device captures a facial image, and the characteristic compression set connects image capturing device, the reception facial image.The human face detection device connects image capturing device and characteristic compression set, and the human face detection device carries out a positive face detection, a side face detection, the positive face detection of a transposition or a transposition side face detection according to the first mirror data, the second mirror data or transposition data.
Wherein, by the first mirror data formation one positive face masterplate, and the human face detection device carries out positive face detection according to described positive face masterplate; And by the second mirror data formation one side face masterplate, and the human face detection device carries out the side face detection according to described side face masterplate; And by the positive face masterplate of transposition data formation one transposition or a transposition side face masterplate, and the human face detection device carries out the positive face detection of transposition or transposition side face detection according to the positive face masterplate of described transposition or transposition side face masterplate.
Be to realize a further object of the present invention, the present invention by the following technical solutions: a kind of characteristic compression set comprises a modular converter, a mirror module and a transposition module.Modular converter is by a facial image, and some eigenwerts that computing obtains facial image are forming at least one plate, and each eigenwert comprises a rate of specific gravity.The mirror module is compared each eigenwert one by one, obtains the position of each symmetrical eigenwert with computing, and gives up in these some eigenwerts symmetrical in twos one of them, forms one first mirror data.Perhaps, the mirror unit is compared each eigenwert one by one, obtains a position with eigenwert of high specific gravity value with computing, forms one second mirror data.The transposition module is according to a transposition list transposition this at least one plate to an angle, and obtains some transposition eigenwerts by at least one plate computing of this transposition, to form transposition data.
In addition, for realizing purpose of the present invention, the present invention also proposes a kind of method for detecting, is applicable to a multi-direction human face detecting system.Multi-direction human face detecting system comprises an image capturing device, a converting unit, a mirror unit, a transposition unit and a human face detection device.And the human face detection method comprises the following step:
Capture a facial image by image capturing device; With converting unit computing facial image; Obtain some eigenwerts of facial image, and form at least one plate according to these some eigenwerts, and each eigenwert comprises a rate of specific gravity; Make the mirror unit compare one by one each eigenwert, obtain the position of each symmetrical eigenwert with computing, and give up in these some eigenwerts symmetrical in twos one of them, form one first mirror data; Perhaps, make the mirror unit compare one by one each eigenwert, obtain a position with eigenwert of high specific gravity value with computing, form one second mirror data; Make the transposition unit according to this at least one plate to an angle of a transposition list transposition, and obtain some transposition eigenwerts by at least one plate computing of this transposition, form transposition data; And according to the first mirror data, the second mirror data or transposition data, the human face detection device carries out a positive face detection, a side face detection, the positive face detection of a transposition or a transposition side face detection.
Described method also comprises the following steps: according to the first mirror data formation one positive face masterplate; And by positive face masterplate, the human face detection device carries out positive face detection.
Described method also comprises the following steps: according to the second mirror data formation one side face masterplate; And by side face masterplate, the human face detection device carries out the side face detection.
Described method also comprises the following steps: according to the positive face masterplate of transposition data formation one transposition or a transposition side face masterplate; And the human face detection device carries out the positive face detection of described transposition or described transposition side face detection.
From the above, according to multi-direction human face detecting system and the method for detecting of characteristic compression set of the present invention, application characteristic data compression device, it can have one or more following advantage:
(1) this characteristic compression set can be simplified initial characteristic data by the mirror of face characteristic, can reduce the data volume of characteristic with this, to reduce the amount of expending in memory body space.
(2) but this characteristic compression set can be by the permutability of face characteristic be transposed to a demand angle with primitive character money, can solve and to re-start the feature sampling with the picture to be measured behind the transposition with this, with the problem of the comparison plate that regenerates a correspondence, thereby reduce operation time.
(3) this characteristic compression set data volume that can compress the face characteristic data by mirror and the permutability of face characteristic, can avoid dwindling the human face detection angle or simplifying comparison plate precision, the problem that the detecting rate is reduced as reducing data volume take this.
Description of drawings
Fig. 1 is the calcspar of characteristic compression set of the present invention;
Fig. 2 is multi-direction human face detecting system's of the present invention calcspar;
Fig. 3 is multi-direction human face detecting system's of the present invention process flow diagram;
Fig. 4 is the formation process flow diagram of the masterplate of multi-direction human face detection method of the present invention;
Fig. 5 is the formation synoptic diagram of the first mirror data of multi-direction human face detection method of the present invention;
Fig. 6 is the formation synoptic diagram of the second mirror data of multi-direction human face detection method of the present invention;
Fig. 7 is the formation synoptic diagram of the transposition data of multi-direction human face detection method of the present invention.
Embodiment
Hereinafter with reference to correlative type, illustrate according to characteristic compression set of the present invention, the multi-direction human face detecting system of application characteristic data compression device and the embodiment of method for detecting, be convenient to understand for making, the similar elements among the following embodiment illustrates with identical symbology.
Fig. 1 is the calcspar of characteristic compression set of the present invention.As shown in the figure, the characteristic compression set 1 that is applicable among the multi-direction human face detecting system can comprise a converting unit 11, a mirror unit 12, a transposition unit 13 and a storage element 14.Converting unit 11 can be carried out the acquisition of face characteristic by a facial image 200, and it can be divided into some blocks with facial image 200, and obtains some eigenwerts 110 by these some block computings.And according to the texture otherness of people's face face, each eigenwert 110 can be provided with different rates of specific gravity, and in other words, each eigenwert 110 has the feature representativeness of varying strength.And converting unit 11 can form at least one plate 111 according to these some eigenwerts 110.Each plate 111 namely can be used as multi-direction human face detecting system when carrying out human face detection a face comparison pattern, for example the comparison pattern of forward face or side direction face.
Perhaps, after if mirror unit 12 is compared each eigenwert 110 one by one, when finding that these some eigenwerts 110 do not have symmetry, the position that mirror unit 12 can sort respectively and record each eigenwert 110 according to the rate of specific gravity of each eigenwert 110, to obtain the position of the eigenwert 110 with high specific gravity value, form one second mirror data 121.Briefly, because these some eigenwerts 110 do not have symmetrical in twos characteristic, therefore, the plate 111 that forms according to these some eigenwerts 110 can be face's comparison pattern of a left side face or right side face.Further, by the feature mirror of people's face left and right sides face, mirror unit 12 can only store the eigenwert 110 of left side face or a right side face side face wherein and form the second mirror data 121, can effectively dwindle data volume with this.Simultaneously, mirror unit 12 can have a mirror list 122, and can form the first mirror data 120 or the second mirror data 121 according to mirror list 122.
Transposition unit 13 can connect converting unit 11, and can this at least one plate 111 be transposed to an angle according to a transposition list 130, for example dextrorotation turn 90 degrees, 180 degree or 270 are spent, the plate 111 that union respectively has an angle obtains some transposition eigenwerts, to form transposition data 131.Briefly, but can make these some eigenwerts 110 have permutability according to transposition list 130, therefore, transposition unit 13 can obtain by the action of tabling look-up multidirectional transposition data 131, form further multidirectional transposition face comparison pattern, can effectively dwindle data volume with this.
Storage element 14 connects converting unit 11, mirror unit 12 and transposition unit 13, in order to store above-mentioned some eigenwerts 110, the first mirror data 120, the second mirror data 121 and transposition data 131.
In present embodiment, when multi-direction human face detecting system 2 received testing image 200, system can carry out lookup table mode and extract the first mirror data 120, the second mirror data 121 or transposition data 131 from storage element 14 by characteristic compression set 1.Make system's computing the first mirror data 120 compare pattern with the face that obtains a positive face, or make system's computing the second mirror data 121 to obtain face's comparison pattern of a side face.Or system also can obtain multidirectional transposition face comparison pattern by computing transposition data 131.Therefore, multi-direction human face detecting system 2 of the present invention can only store a sampling plate, table look-up by characteristic compression set 1 again and obtain the characteristic value data of mirror or transposition, can avoid because storing face's comparison pattern of multiple types with this, compare pattern such as positive face, left side face, right side face or rotation to positive face, rotation to the positive face of 270 degree of 90 degree, and make system need store a large amount of data volumes, cause expending of memory body.
From the above, see also Fig. 2, it is multi-direction human face detecting system's of the present invention calcspar.As shown in the figure, but the mirror that multi-direction human face detecting system 2 can utilize face characteristic by an above-mentioned characteristic compression set 1 with permutability and the compressive features data volume.Simultaneously, system also can table look-up by characteristic compression set 1 and obtain the characteristic value data of mirror or transposition, obtains poly-directional human face comparison pattern with this, and then detects the people's face in the picture to be measured.Multi-direction human face detecting system 2 also can comprise an image capturing device 20 and a human face detection device 21.Image capturing device 20 can be the capturing images elements such as camera, mobile phone camera, video camera or monitor, but its connection features data compression device 1 captures some testing images.
Converting unit 11 can connect image capturing device 20, and can carry out the acquisition of face characteristic by the facial image 200 that one in these some testing images comprise people's face.Converting unit 11 can be divided this person's face image 200 by some blocks, and obtains some eigenwerts 110 by these some block computings.And because the texture of people's face face can have different directivity or feature explicitly, therefore, each eigenwert 110 can be provided with different rates of specific gravity.And converting unit 11 can form at least one plate 111 according to these some eigenwerts 110.Each plate 111 namely can be multi-direction human face detecting system 2 when carrying out human face detection a face comparison pattern, for example the comparison pattern of forward face or side direction face.
But human face detection device 21 connection features data compression devices 1, and human face detection device 21 can read according to its detecting action some eigenwerts 110 of plate 111, perhaps can carry out lookup table mode by characteristic compression set 1 makes by extracting the first mirror data 120, the second mirror data 121 or transposition data 131 in the storage element 14, to carry out the positive face detection 24 of positive face detection 22, side face detection 23 or transposition or transposition side face detection 25.With this, human face detection device 21 can pass through the various people's face pattern in these some eigenwerts 110, the first mirror data 120, the second mirror data 121 or 131 comparisons of transposition data and the detecting testing image.
In other words, when having forward face in the testing image, human face detection device 21 can directly use above-mentioned this at least one plate 111 to form positive face masterplate 220, perhaps, can look into by characteristic compression set 1 and get the first mirror data 120 to form positive face masterplate 220.And human face detection device 21 can carry out by this positive face masterplate 220 the positive face detection 22 of forward face.And when having side direction face in the testing image, human face detection device 21 can directly use above-mentioned this at least one plate 111 to form side face masterplate 230, perhaps, can look into by characteristic compression set 1 and get the second mirror data 121 to form side face masterplate 230.And human face detection device 21 carries out the side face detection 23 of side direction face by this side face masterplate 230.And when the side direction face of the forward face that has rotation in the testing image or rotation, 21 in human face detection device can be looked into by characteristic compression set 1 and get transposition data 131 to form the positive face masterplate 240 of transposition or transposition side face masterplate 250.Then, human face detection device 21 carries out the positive face detection 24 of transposition of the forward face of transposition by the positive face masterplate 240 of transposition, again or human face detection device 21 carry out the transposition side face detection 25 of the side direction face of transposition by transposition side face masterplate 250.
In present embodiment, the first mirror data 120 can be the characteristic of left half face of forward face or right half face, and the second mirror data 121 then can be the characteristic of left side face or right side face.Transposition data 131 can be rotation to the positive face characteristic of 90 degree, 180 degree or 270 degree.Therefore, 220 of positive face masterplates can reduce the comparison masterplate that forms a forward face by the characteristic of this left half face of computing or right half face.And side face masterplate 230 also can be symmetrically formed the comparison masterplate of a right side face by the characteristic of computing left side face, or is symmetrically formed the comparison masterplate of a left side face by the characteristic of computing right side face.And the positive face masterplate 240 of transposition also can be different because of the direction of the characteristic behind the transposition, and can have the masterplate pattern of the directions such as 90 degree, 180 degree or 270 degree.
Like this, multi-direction human face detecting system 2 can only store a sampling plate, table look-up by characteristic compression set 1 again and obtain the characteristic value data of mirror or transposition, can avoid because storing face's comparison pattern of multiple types with this, such as positive face, left side face, right side face or the rotation comparison patterns such as positive faces to the positive faces of 90 degree, rotation to 270 degree, and make system need store a large amount of data volumes, cause expending of memory body.
Fig. 3 is the process flow diagram of multi-direction human face detection method of the present invention.As shown in the figure, multi-direction human face detection method can be a multi-direction human face detecting system human face detection method, and multi-direction human face detecting system can comprise an image capturing device, a characteristic compression set and a human face detection device, and the characteristic compression set also can have a converting unit, a mirror unit and a transposition unit.Its human face detection method can comprise the following step:
In step S30, capture some images by image capturing device, it is testing image, and enter step S300: the image frame that comprises people's face with one in these some testing images is as in order to capturing the facial image of face characteristic, and image capturing device inputs to this person's face image in the converting unit.
In step S31, after converting unit receives facial image, can divide facial image by some blocks, but obtain some eigenwerts of facial image by these some block computings.And according to the texture otherness of people's face face, each eigenwert can be provided with different rates of specific gravity, that is, each eigenwert can have varying strength feature representative.Therefore, execution in step S310: put down in writing the rate of specific gravity of each eigenwert, and form at least one plate by these some eigenwerts.
In step S32, the mirror unit receives and compares one by one this some eigenwerts.Enter step S320: if in comparison result, when finding that these some eigenwerts have symmetry, then this some positions with symmetric eigenwert will be put down in writing in the mirror unit, and give up in some eigenwerts symmetrical in twos one of them, to form the first mirror data.And after if the mirror unit compares each eigenwert one by one, when finding that these some eigenwerts do not have symmetry, the position that the mirror unit can sort respectively and record each eigenwert according to the rate of specific gravity of each eigenwert, to obtain the position of the eigenwert with high specific gravity value, form one second mirror data.
Simultaneously, in step S33, the transposition unit can receive this some eigenwerts, and according to a transposition list and with above-mentioned at least one plate rotation to an angle.Enter step S330: computing respectively has the plate of an angle and obtains some transposition eigenwerts, records these some transposition eigenwerts to form transposition data.
In step S34, eigenwert, plate, the first mirror data, the second mirror data and transposition data all can be stored in the storage element.
In step S35, when the forward face in the human face detection device wish detecting testing image, can form positive face masterplate according to the plate reduction, perhaps can form positive face masterplate by the first mirror data operation, carry out a positive face detection with this.
In step S36, when human face detection device wish was detected the forward face of the rotation in the testing image, the human face detection device can form the positive face masterplate of transposition according to the transposition characteristic, to carry out the positive face detection of a transposition.
In step S37, when the side direction face in the human face detection device wish detecting testing image, can form side face masterplate according to the plate reduction, perhaps can form side face masterplate by the second mirror data operation, carry out a side face detection with this.
The detailed description of multi-direction human face detection method of the present invention and embodiment were described when multi-direction human face detecting system of the present invention is narrated in the front, just no longer sewed for schematic illustration at this and stated.
Further, see also Fig. 4, Fig. 4 is the formation process flow diagram of the masterplate of multi-direction human face detection method of the present invention.As shown in the figure, the eigenwert that the human face detection device can be original, mirror data or transposition data and form respectively positive face masterplate, side face masterplate, the positive face masterplate of transposition or transposition side face masterplate, and carry out human face detection by these some masterplates.The flow process that forms masterplate also comprises the following step:
In step S40, when multi-direction human face detecting system captured some testing images by capture device, the human face detection device can read the eigenwert that is obtained by the computing of characteristic compression set institute.
In step S41, according to these some eigenwerts, the human face detection device reduces at least one plate by the characteristic compression set.
In step S42, by the action of tabling look-up of characteristic compression set, above-mentioned eigenwert can be carried out by the mirror unit simplification of characteristic, makes the plate of initial forward face be reduced to the plate of left half face or right half face.Perhaps, the plate that makes initial left side face is to being called the plate of right side face, and the plate of right side face is to being called the plate of left side face.In addition, above-mentioned eigenwert also can make the plate rotation of initial forward face to directions such as 90 degree, 180 degree or 270 degree, to form the plate that rotates positive face by the transposition unit.
In step S43, the converting unit computing above-mentioned about half face, left and right sides face or rotate the plate of positive face, simplify or mirror or the transposition eigenwert of transpose process in step S44, to obtain through mirror, be mirror data or transposition data.
Fig. 5 is the formation synoptic diagram of the first mirror data of multi-direction human face detection method of the present invention.As shown in the figure, behind the some testing images of image capturing device 20 acquisitions, the facial image 200 that converting unit 11 can comprise one in the testing image positive face is divided into some blocks, and carries out the eigenwert 110 that characteristic operation obtains some positive faces by these a little blocks and form at least one plate 111.Then, mirror unit 12 can be compared this a little eigenwerts 110 one by one, if these a little eigenwerts 110 have symmetry, then these a little plates 111 can be the comparison pattern of a forward face.Therefore, mirror by these a little eigenwerts 110, symmetrical in twos one of them can be given up in some eigenwerts 110 in mirror unit 12, and keep the wherein feature of arbitrary half face, form half face plate 3, for example, left side or right-hand part in the forward face can be kept, the effect of effectively simplifying the characteristic amount can be reached with this.
This has the position of symmetric eigenwert 110 12 records of mirror unit, by this half face plate 3 through the mirror simplification of converting unit 11 computings, obtains the mirror eigenwert, i.e. the first mirror data 120 again.
Then, when multi-direction human face detecting system 2 carried out human face detection, human face detection device 21 can form positive face masterplate 220 according to these the first mirror data 120 reduction, to carry out positive face detection 22.
Fig. 6 is the formation synoptic diagram of the second mirror data of multi-direction human face detection method of the present invention.As shown in the figure, behind the some testing images of image capturing device 20 acquisitions, the facial image 200 that converting unit 11 can comprise one in the testing image positive face is divided into some blocks, and carries out the eigenwert 110 that characteristic operation obtains some positive faces by these a little blocks and form at least one plate 111.Then, mirror unit 12 can be compared this a little eigenwerts 110 one by one, if when these a little eigenwerts 110 do not have symmetry, then these a little plates 111 can be a side face plate, are the comparison pattern of side direction face.
In the present embodiment, this at least one side face plate can be the comparison pattern of left side face, and by after 12 comparisons of mirror unit, finds that the eigenwert 110 that represents nose 4 has the highest proportion.Therefore, mirror unit 12 will sequentially record the position of eigenwert 110 of the comparison pattern of this left side face.And by tabling look-up, mirror unit 12 can this have the nose 4 of the most hyperbaric eigenwert 110 centered by point carry out mirror, and obtain the comparison pattern of right side face, also an i.e. symmetrical side face plate 40.Record this at least one right side to the position of the eigenwert 110 of the comparison pattern of face, by this symmetrical side face plate 40 through the mirror processing of converting unit 11 computings, obtain the mirror eigenwert, i.e. the second mirror data 121 again.
When multi-direction human face detecting system 2 carried out human face detection, human face detection device 21 can be symmetrically formed side face masterplate 230 according to these the second mirror data 121, to carry out side face detection 23.
Fig. 7 is the formation synoptic diagram of the transposition data of multi-direction human face detection method of the present invention.As shown in the figure, behind the some testing images of image capturing device 20 acquisitions, the facial image 200 that converting unit 11 can comprise one in the testing image positive face is divided into some blocks, and carries out the eigenwert 110 that characteristic operation obtains some positive faces by these a little blocks and form at least one plate 111.Then, transposition unit 13 can be transposed to an angle with this at least one plate 111 according to a transposition list 130, and for example dextrorotation turn 90 degrees, 180 degree or 270 are spent, and forms a rotation plate 5, also is a positive face comparison pattern with transposition angle.This rotates converting unit 11 computings plate 5 and can obtain some transposition eigenwerts, and record forms transposition data 131.
When multi-direction human face detecting system 2 carried out human face detection, human face detection device 21 can form the positive face masterplate 240 of transposition according to these transposition data 131, to carry out the positive face detection 24 of transposition.Briefly, transposition unit 13 can obtain multidirectional transposition data 131 by the action of tabling look-up, form further multidirectional transposition face comparison pattern, like this, multi-direction human face detecting system 2 can not must repeat to capture the face characteristic of different directions, also can not must store the positive face comparison pattern of different directions.Therefore, multi-direction human face detecting system 2 of the present invention can transposition be obtained the positive face masterplate 240 of the transposition of multiple direction by an identical positive face comparison pattern, effectively to reach the effect of dwindling data volume.
In sum, the multi-direction human face detecting system of characteristic compression set, application characteristic data compression device and method for detecting thereof can be simplified eigenwert by the mirror of face characteristic, can reduce the data volume of characteristic with this, to reduce the amount of expending in memory body space.And, but also can original plate be transposed to a demand angle by permutability, can reduce the plate quantity of system stores with this, or shorten the operation time from multiple plate acquisition feature.
The above only is illustrative, but not is restricted person.Anyly do not break away from spirit of the present invention and category, and to its equivalent modifications of carrying out or change, all should be contained in the accompanying scope of patent protection.
Claims (12)
1. characteristic compression set is characterized in that: comprise:
One converting unit, by a facial image, computing obtains some eigenwerts of described facial image, and forms at least one plate according to described some eigenwerts, and each described eigenwert comprises a rate of specific gravity;
One mirror unit, connect described converting unit, described mirror unit is compared each described eigenwert one by one, obtain the position of each described symmetrical eigenwert with computing, and give up in described some eigenwerts symmetrical in twos one of them, form one first mirror data, perhaps, described mirror unit is compared each described eigenwert one by one, obtains a position with eigenwert of high specific gravity value with computing, forms one second mirror data; And
One transposition unit connects described converting unit, and described transposition unit is according to a transposition list transposition described at least one plate to an angle, and obtains some transposition eigenwerts by at least one plate computing of described transposition, forms transposition data.
2. characteristic compression set as claimed in claim 1 is characterized in that, by described the first mirror data formation one positive face masterplate.
3. characteristic compression set as claimed in claim 1 is characterized in that, by described the second mirror data formation one side face masterplate.
4. characteristic compression set as claimed in claim 1 is characterized in that, by the positive face masterplate of described transposition data formation one transposition or a transposition side face masterplate.
5. a multi-direction human face detecting system who uses a characteristic compression set is characterized in that, comprises:
One image capturing device is used for acquisition one facial image;
One characteristic compression set, connect described image capturing device, described characteristic compression set comprises, one converting unit, by a facial image, computing obtains some eigenwerts of described facial image, and forms at least one plate according to described some eigenwerts, and each described eigenwert comprises a rate of specific gravity; One mirror unit, connect described converting unit, described mirror unit is compared each described eigenwert one by one, obtain the position of each described symmetrical eigenwert with computing, and give up in described some eigenwerts symmetrical in twos one of them, form one first mirror data, perhaps, described mirror unit is compared each described eigenwert one by one, obtains a position with eigenwert of high specific gravity value with computing, forms one second mirror data; And a transposition unit, connecting described converting unit, described transposition unit is according to described at least one plate to an angle of a transposition list transposition, and obtains some transposition eigenwerts by at least one plate computing of described transposition, forms transposition data; And
One human face detection device, connect described image capturing device and described characteristic compression set, and described human face detection device carries out a positive face detection, a side face detection, the positive face detection of a transposition or a transposition side face detection according to described the first mirror data, described the second mirror data or described transposition data.
6. multi-direction human face detecting system as claimed in claim 5 is characterized in that, by described the first mirror data formation one positive face masterplate, and described human face detection device carries out described positive face detection according to described positive face masterplate.
7. multi-direction human face detecting system as claimed in claim 5 is characterized in that, by described the second mirror data formation one side face masterplate, and described human face detection device carries out described side face detection according to described side face masterplate.
8. multi-direction human face detecting system as claimed in claim 5, it is characterized in that, by the positive face masterplate of described transposition data formation one transposition or a transposition side face masterplate, and described human face detection device carries out the positive face detection of described transposition or described transposition side face detection according to the positive face masterplate of described transposition or described transposition side face masterplate.
9. method for detecting, it is characterized in that, described method for detecting is applicable to a multi-direction human face detecting system, and described multi-direction human face detecting system comprises an image capturing device, a characteristic compression set and a human face detection device, and described characteristic compression set has a converting unit, a mirror unit and a transposition unit, and described method for detecting comprises the following step:
Capture a facial image by described image capturing device;
With the described facial image of described converting unit computing;
Obtain some eigenwerts of described facial image, and form at least one plate according to described some eigenwerts, and each described eigenwert comprises a rate of specific gravity;
Make described mirror unit compare one by one each described eigenwert, obtain the position of each described symmetrical eigenwert with computing, and give up in described some eigenwerts symmetrical in twos one of them, form one first mirror data;
Perhaps, make described mirror unit compare one by one each described eigenwert, obtain a position with eigenwert of high specific gravity value with computing, form one second mirror data;
Make described transposition unit according to described at least one plate to an angle of a transposition list transposition, and utilize at least one plate computing of described transposition to obtain some transposition eigenwerts, form transposition data; And
According to described the first mirror data, described the second mirror data or described transposition data, described human face detection device carries out a positive face detection, a side face detection, the positive face detection of a transposition or a transposition side face detection.
10. method for detecting as claimed in claim 9 is characterized in that, described method for detecting also comprises the following steps:
By described the first mirror data formation one positive face masterplate; And
By described positive face masterplate, described human face detection device carries out described positive face detection.
11. method for detecting as claimed in claim 9 is characterized in that, described method for detecting also comprises the following steps:
By described the second mirror data formation one side face masterplate; And
By described side face masterplate, described human face detection device carries out described side face detection.
12. method for detecting as claimed in claim 9 is characterized in that, described method for detecting also comprises the following steps:
By the positive face masterplate of described transposition data formation one transposition or a transposition side face masterplate; And
By the positive face masterplate of described transposition or described transposition side face masterplate, described human face detection device carries out the positive face detection of described transposition or described transposition side face detection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110279600.2A CN103020576B (en) | 2011-09-20 | 2011-09-20 | Characteristic data compression device, multi-direction human face detecting system and method for detecting thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110279600.2A CN103020576B (en) | 2011-09-20 | 2011-09-20 | Characteristic data compression device, multi-direction human face detecting system and method for detecting thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103020576A true CN103020576A (en) | 2013-04-03 |
| CN103020576B CN103020576B (en) | 2015-09-30 |
Family
ID=47969169
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110279600.2A Active CN103020576B (en) | 2011-09-20 | 2011-09-20 | Characteristic data compression device, multi-direction human face detecting system and method for detecting thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103020576B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1150282A (en) * | 1994-12-21 | 1997-05-21 | 伊斯曼柯达公司 | Method for compressing and decompressing standardized portrait images |
| CN1552041A (en) * | 2001-12-14 | 2004-12-01 | 日本电气株式会社 | Face meta-data creation and face similarity calculation |
| TW200725433A (en) * | 2005-12-29 | 2007-07-01 | Ind Tech Res Inst | Three-dimensional face recognition system and method thereof |
| US7254275B2 (en) * | 2002-12-17 | 2007-08-07 | Symbol Technologies, Inc. | Method and system for image compression using image symmetry |
-
2011
- 2011-09-20 CN CN201110279600.2A patent/CN103020576B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1150282A (en) * | 1994-12-21 | 1997-05-21 | 伊斯曼柯达公司 | Method for compressing and decompressing standardized portrait images |
| CN1552041A (en) * | 2001-12-14 | 2004-12-01 | 日本电气株式会社 | Face meta-data creation and face similarity calculation |
| US7254275B2 (en) * | 2002-12-17 | 2007-08-07 | Symbol Technologies, Inc. | Method and system for image compression using image symmetry |
| TW200725433A (en) * | 2005-12-29 | 2007-07-01 | Ind Tech Res Inst | Three-dimensional face recognition system and method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103020576B (en) | 2015-09-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11107232B2 (en) | Method and apparatus for determining object posture in image, device, and storage medium | |
| US8421877B2 (en) | Feature data compression device, multi-directional face detection system and detection method thereof | |
| CN107370942B (en) | Photographing method, photographing device, storage medium and terminal | |
| CN108805047B (en) | Living body detection method and device, electronic equipment and computer readable medium | |
| WO2021248733A1 (en) | Live face detection system applying two-branch three-dimensional convolutional model, terminal and storage medium | |
| CN111105352A (en) | Super-resolution image reconstruction method, system, computer device and storage medium | |
| CN111091075B (en) | Face recognition method and device, electronic equipment and storage medium | |
| US20150109416A1 (en) | Depth map generation | |
| Rehman et al. | SLNet: Stereo face liveness detection via dynamic disparity-maps and convolutional neural network | |
| CN111753782B (en) | False face detection method and device based on double-current network and electronic equipment | |
| CN111768336A (en) | Face image processing method and device, computer equipment and storage medium | |
| Makihara et al. | Gait collector: An automatic gait data collection system in conjunction with an experience-based long-run exhibition | |
| CN110532959B (en) | Real-time violent behavior detection system based on two-channel three-dimensional convolutional neural network | |
| CN110688950B (en) | Face living body detection method and device based on depth information | |
| CN109784230A (en) | A kind of facial video image quality optimization method, system and equipment | |
| CN112149517A (en) | Face attendance checking method and system, computer equipment and storage medium | |
| CN111222459A (en) | Visual angle-independent video three-dimensional human body posture identification method | |
| CN103020576B (en) | Characteristic data compression device, multi-direction human face detecting system and method for detecting thereof | |
| CN110633631A (en) | Pedestrian re-identification method based on component power set and multi-scale features | |
| CN112613457B (en) | Image acquisition mode detection method, device, computer equipment and storage medium | |
| CN114299569A (en) | Safe face authentication method based on eyeball motion | |
| CN112488076A (en) | Face image acquisition method, system and equipment | |
| CN111275183A (en) | Visual task processing method and device and electronic system | |
| CN117011945B (en) | Action capability assessment method, action capability assessment device, computer equipment and readable storage medium | |
| Ji et al. | A cross domain multi-modal dataset for robust face anti-spoofing |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20161222 Address after: China Taiwan Hsinchu Science Park Hsinchu City Road No. 12 5 floor Patentee after: Glomerocryst semiconductor limited company Address before: China Taiwan Hsinchu Science Park Hsinchu City Road No. 12 Patentee before: Huajing Science & Technology Co., Ltd. |
|
| TR01 | Transfer of patent right |