CN106570445A - Feature detection method and apparatus - Google Patents
Feature detection method and apparatus Download PDFInfo
- Publication number
- CN106570445A CN106570445A CN201510657140.0A CN201510657140A CN106570445A CN 106570445 A CN106570445 A CN 106570445A CN 201510657140 A CN201510657140 A CN 201510657140A CN 106570445 A CN106570445 A CN 106570445A
- Authority
- CN
- China
- Prior art keywords
- similarity
- result
- feature
- regression
- regression result
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention discloses a feature detection method and apparatus. The feature detection method with high accuracy comprises: face position and size detection and face feature detection are carried out on a first target image to obtain a first key point feature; according to N datum dimensions, regression calculation is carried out on the first key point feature to obtain N first regression results, wherein the N is an integer larger than 1 and the datum dimensions are ones of a face feature detection region; a similarity degree of the N first regression results is calculated; and if the similarity degree of the N first regression results meets a similarity condition, a mean value of the N first regression results is calculated and a first feature detection result is obtained.
Description
Technical field
The present invention relates to image processing field, more particularly to a kind of characteristic detection method and device.
Background technology
The method that face critical point detection is generally basede on recurrence at present, because the method effect for returning is relatively
It is good, and speed is faster.But it is not know last based on the shortcoming of the face critical point detection for returning
The effect of recurrence how, and in the high application scenario of accuracy requirement, the result of the method is often inadequate
It is preferable.
In the existing face critical point detection technology based on recurrence, a kind of accuracy to regression algorithm is commented
The processing method of valency is that multiple textural characteristics are extracted around key point, and based on these features in addition
One grader of training.The effect of regression algorithm is judged using the grader, take the recurrence by grader
The result output of algorithm.The program the disadvantage is that, just for the regression effect of single frames figure, and in reality
Using in, due to the attitude of face, expression is blocked, and the factor such as illumination varies so that classification
Precision be extremely difficult to it is very high, it is when the situation that its training set is not included is run into, easy to make mistake
Conclusion.
Also, during face critical point detection, it is necessary first to according to the yardstick of facial features localization frame
Target image is normalized into unified mark yardstick by size, then according to the parameter of facial features localization frame
One average face of initialization, then on target image after normalization, from the beginning of the average face, foundation
The textural characteristics of each key point peripheral region, carry out cascade recurrence, and final knot is obtained after multiple regression
Really.The method is sensitive to initial key point shape matching, and original shape is limited by facial features localization
The size of frame.So, it may also be said to this technology is affected very big by the size of facial features localization frame, is returned
The reason for effect is bad is greatly because that facial features localization frame is too big or too little.
The content of the invention
Embodiments provide a kind of characteristic detection method and device, for provide a kind of accuracy compared with
High characteristic detection method.
Characteristic detection method provided in an embodiment of the present invention, including：
Facial features localization is carried out to first object image, the first key point feature is obtained；
Respectively according to N number of benchmark, regressing calculation is carried out to the first key point feature, N is obtained
Individual first regression result, the N are the integer more than 1, and the benchmark is facial features localization area
The benchmark in domain；
Calculate the similarity of N number of first regression result；
If the similarity of N number of first regression result meets condition of similarity, described N number of first is calculated
The average of regression result, obtains fisrt feature testing result.
Feature detection device provided in an embodiment of the present invention, including：
Critical point detection unit, for carrying out facial features localization to first object image, obtains the first pass
Key point feature；
Unit is returned, according to N number of benchmark, the first key point feature is carried out back for respectively
Return computing, obtain N number of first regression result, the N is the integer more than 1, the benchmark is
The benchmark in facial features localization region；
Similarity calculated, for calculating the similarity of N number of first regression result；
Average calculation unit, if the similarity for N number of first regression result meets condition of similarity,
The average of N number of first regression result is then calculated, fisrt feature testing result is obtained.
As can be seen from the above technical solutions, the embodiment of the present invention has advantages below：
In embodiments of the present invention, after face characteristic is completed, can be examined according to N number of face characteristic respectively
The benchmark in region is surveyed, regressing calculation is carried out to the first key point feature, and is based on multiple faces
Feature detection region can reduce facial features localization frame and initially set inaccurate brought impact, also,
After N number of first regression result is obtained, the similarity of N number of first regression result can be calculated, if phase
It is higher like degree, then can determine that the setting of N number of benchmark is more accurate, can be by described N number of the
The average of one regression result is exported as the result of facial features localization, improves the standard of facial features localization
True property.
Description of the drawings
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to reality
Apply accompanying drawing to be used needed for example to be briefly described, it should be apparent that, drawings in the following description are only
Only it is some embodiments of the present invention, for those of ordinary skill in the art, is not paying creativeness
On the premise of work, can be with according to these other accompanying drawings of accompanying drawings acquisition.
Fig. 1 is a schematic flow sheet of characteristic detection method in the embodiment of the present invention；
Fig. 2 is a signaling process schematic diagram of characteristic detection method in the embodiment of the present invention；
Fig. 3 is a logical construction schematic diagram of feature detection device in the embodiment of the present invention.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out
Clearly and completely describe, it is clear that described embodiment is only a part of embodiment of the invention, and
It is not all, of embodiment.Based on the embodiment in the present invention, those of ordinary skill in the art are without work
The every other embodiment obtained under the premise of going out creative work, belongs to the scope of protection of the invention.
It is described in detail individually below.
Term " first ", " second " in description and claims of this specification and above-mentioned accompanying drawing, "
Three " (if present) such as " 4th " is for distinguishing similar object, without specific suitable for describing
Sequence or precedence.It should be appreciated that the data for so using can be exchanged in the appropriate case, so as to here
The embodiments of the invention of description can be with the order reality in addition to those for illustrating here or describing
Apply.Additionally, term " comprising " and " having " and their any deformation, it is intended that cover non-exclusive
Comprising for example, containing process, method, system, product or the equipment of series of steps or unit not
Be necessarily limited to those steps or the unit clearly listed, but may include clearly not list or for
Other intrinsic steps of these processes, method, product or equipment or unit.
The characteristic detection method in the embodiment of the present invention is described below by detailed embodiment, please
Refering to Fig. 1, in the embodiment of the present invention, one embodiment of characteristic detection method includes：
101st, facial features localization is carried out to first object image；
Feature detection device carries out facial features localization to first object image, obtains the first key point feature.
The first object image is the relatively front image in continuous two picture frames.Described
The feature of 1 key point feature initial key point；Specifically, key point feature can be face in eyes,
The coordinate data in the target image of the features such as nose, face.
Specifically, facial features localization method can for Scale invariant features transform (SIFT,
Scale-invariant feature transform), or other feature extraction algorithms, this time specifically not
It is construed as limiting.
In embodiments of the present invention, the first object image can be the first image in continuous picture frame.
102nd, regressing calculation is carried out to the first key point feature according to N number of benchmark respectively；
Feature detection device according to N number of benchmark, carries out recurrence fortune to the first key point feature respectively
Calculate, obtain N number of first regression result, the N is the integer more than 1, and the benchmark is that face is special
Levy the benchmark of detection zone.
Specifically, regressing calculation can be linear regression, or nonlinear regression.
103rd, calculate the similarity of N number of first regression result；
Feature detection device calculates the similarity of N number of first regression result, and first regression result is
The feature of key point after regressing calculation.
Specifically, the calculating of similarity can be：
First, calculate the similarity of corresponding first regression result of two neighboring benchmark；Such as：N is 3,3
Individual benchmark is respectively 1.1S0, S0And 0.9S0, then calculate 1.1S0And S0Difference corresponding first returns knot
Similarity between fruit, and 0.9S0And S0Similarity between corresponding first regression result of difference；
In practical application, the setting of N number of benchmark is incremented by successively or successively decreases successively；The S0It is more than zero
Real number
2nd, the similarity between calculating in N number of first regression result two-by-two；Such as：N is 3,3 benchmarks
Respectively 1.1S0, S0And 0.9S0, then calculate 1.1S0And S0Phase between corresponding first regression result of difference
Like degree, 0.9S0And S0Similarity between corresponding first regression result of difference, and 1.1S0And 0.9S0
Similarity between corresponding first regression result of difference.
It is understood that in actual applications, the calculating of similarity is not limited to both the above calculation.
104th, judge whether the similarity of N number of first regression result meets condition of similarity；
Feature detection device judges whether the similarity of N number of first regression result meets condition of similarity, if
The similarity of N number of first regression result meets condition of similarity, then execution step 105；If described N number of
The similarity of the first regression result is unsatisfactory for condition of similarity, then execution step 106.
Specifically, in actual applications, meet the parameter phase knowledge and magnanimity ratio that condition of similarity refers to participate in comparing
It is higher；Exemplary, a threshold value can be set, when the similarity of N number of first regression result reaches
During the threshold value, then it is considered that the similarity of N number of first regression result meets condition of similarity；When described
When the similarity of N number of first regression result is not reaching to the threshold value, then condition of similarity is unsatisfactory for.
It is understood that in actual applications, whether the similarity of N number of first regression result is judged
The method for meeting condition of similarity can have various, and the embodiment of the present invention only enumerates one way in which.
105th, obtain feature detection result；
If the similarity of N number of first regression result meets condition of similarity, described N number of first time is calculated
Sum up the average of fruit, obtain fisrt feature testing result, the fisrt feature testing result is cached.
106th, the flow process of feature detection is performed again.
Specifically, in actual applications, if step 101 is to before 104 execution, special in the embodiment of the present invention
Levy the feature detection result that detection means does not obtain previous frame target image, and current first object figure
As the similarity of corresponding N number of first regression result is unsatisfactory for condition of similarity, then it is right not export and do not retain
The testing result of the current first object image, and to the target image of next frame execution step again
101 to 104 flow process, until the similarity of corresponding N number of first regression result of target image for detecting is full
Sufficient condition of similarity.
If before in the embodiment of the present invention, step 101 is performed to 104, feature detection device has been obtained for
The feature detection result of one frame target image, and current first object image corresponding N number of first returns knot
The similarity of fruit is unsatisfactory for condition of similarity, then by the feature detection result of previous frame target image, respectively with
N number of first regression result does Characteristic Contrast, exports the feature detection knot with the previous frame target image
One the first regression result of characteristic similarity highest of fruit, as fisrt feature testing result.
In embodiments of the present invention, after face characteristic is completed, can respectively according to N number of facial features localization
The benchmark in region, carries out regressing calculation, and it is special to be based on multiple faces to the first key point feature
Levy detection zone and can reduce facial features localization frame and initially set inaccurate brought impact, also,
After N number of first regression result is obtained, the similarity of N number of first regression result can be calculated, if similar
Degree is higher, then can determine that the setting of N number of benchmark is more accurate, can be by described N number of first time
The average for summing up fruit is exported as the result of facial features localization, improves the accuracy of facial features localization.
Further below the characteristic detection method in the embodiment of the present invention is described, Fig. 2 is referred to, this
In inventive embodiments, one embodiment of characteristic detection method includes：
201st, facial features localization is carried out to first object image；
Feature detection device carries out facial features localization to first object image, obtains the first key point feature.
In embodiments of the present invention, the first object image is the first frame target image in consecutive image.
The feature of the first key point feature initial key point；Specifically, key point feature can be in face
Eyes, nose, the coordinate data in the target image of the feature such as face.
Specifically, feature detection device can be to carrying out face location and size inspection to the first object image
Survey, based on the face location and the size detection result, normalizing is carried out to the first object image
Change computing；It is special that the first object image after to normalization computing carries out Scale invariant features transform SIFT
Extraction is levied, the first key point feature is obtained.
202nd, regressing calculation is carried out to the first key point feature according to N number of benchmark respectively；
Feature detection device according to N number of benchmark, carries out recurrence fortune to the first key point feature respectively
Calculate, obtain N number of first regression result, the N is the integer more than 1, and the benchmark is that face is special
Levy the benchmark of detection zone.Specifically, the N can be 3.
203rd, calculate the similarity of N number of first regression result；
Feature detection device calculates the similarity of N number of first regression result, and first regression result is
The feature of key point after regressing calculation.
Specifically, the calculating of similarity can be：
First, calculate the Euclidean distance of corresponding first regression result of two neighboring benchmark；Such as：N is 3,
3 benchmarks are respectively 1.1S0, S0And 0.9S0, it is exemplary, in the chi of the Face datection frame of input
Spend for standard scale S0On the basis of, respectively with more than standard scale SbigWith less than standard scale Ssmall, three
Carry out the detection of face key point, S on individual yardstick respectivelybig=1.10*S0, Ssmall=0.9*S0.Return
The evaluation methodology of effect adopts Euclidean distance：Here xi1 and xi2 distinguishes table to d=sqrt (∑ (xi1-xi2) ^2)
Show two shapes comprising all key points, wherein xi={ k1, k2, k3 ..., kn }, k1, k2 ..., kn
Represent that n face key point is put.
2nd, the Euclidean distance between calculating in N number of first regression result two-by-two；Such as：N is 3,3 station meters
Degree is respectively 1.1S0, S0And 0.9S0, then calculate 1.1S0And S0Between corresponding first regression result of difference
Similarity, 0.9S0And S0Similarity between corresponding first regression result of difference, and 1.1S0And 0.9S0
Similarity between corresponding first regression result of difference.
It is understood that in actual applications, the calculating of similarity is not limited to both the above calculation.
204th, judge whether the similarity of N number of first regression result meets condition of similarity；
Feature detection device judges whether the similarity of N number of first regression result meets condition of similarity, if
The similarity of N number of first regression result meets condition of similarity, then execution step 205；If described N number of
The similarity of the first regression result is unsatisfactory for condition of similarity, then update the first object image, hold again
Row step 201.
Specifically, in actual applications, meet the parameter phase knowledge and magnanimity ratio that condition of similarity refers to participate in comparing
It is higher；Exemplary, a threshold value can be set, when the similarity of N number of first regression result reaches
During the threshold value, then it is considered that the similarity of N number of first regression result meets condition of similarity；When described
When the similarity of N number of first regression result is not reaching to the threshold value, then condition of similarity is unsatisfactory for.
It is understood that in actual applications, whether the similarity of N number of first regression result is judged
The method for meeting condition of similarity can have various, and the embodiment of the present invention only enumerates one way in which.
205th, obtain feature detection result；
If the similarity of N number of first regression result meets condition of similarity, described N number of first time is calculated
Sum up the average of fruit, obtain fisrt feature testing result, the fisrt feature testing result is cached.
206th, facial features localization is carried out to the second target image
After feature detection device has been obtained for fisrt feature testing result, feature detection device is to
Two target images carry out facial features localization, obtain the second key point feature.Second target image is
The next frame target image of target image corresponding with the fisrt feature testing result.
207th, regressing calculation is carried out to the second key point feature according to N number of benchmark respectively；
Feature detection device according to N number of benchmark, is returned to the second key point feature respectively
Computing, obtains N number of second regression result.
208th, calculate the similarity of N number of second regression result；
Feature detection device calculates the similarity of N number of second regression result.
209th, judge whether the similarity of N number of second regression result meets condition of similarity；
Feature detection device judges whether the similarity of N number of second regression result meets condition of similarity,
If the similarity of N number of second regression result meets condition of similarity, execution step 210；If the N
The similarity of individual second regression result is unsatisfactory for condition of similarity, then execution step 211.
210th, calculate the average of N number of second regression result；
If the similarity of N number of second regression result meets condition of similarity, described N number of second is calculated
The average of regression result, obtains second feature testing result.
211st, output returns knot with the characteristic similarity highest one second of the fisrt feature testing result
Really.
If the similarity of N number of second regression result is unsatisfactory for condition of similarity, by the fisrt feature
Testing result does Characteristic Contrast respectively with N number of second regression result, exports and examines with the fisrt feature
One the second regression result of characteristic similarity highest of result is surveyed, as second feature testing result.
In embodiments of the present invention, feature detection device judges single previous frame according to the similarity of consecutive frame
The accuracy of face critical point detection result, further increases the accuracy of Face datection, this solution party
Case is adapted to the Accuracy evaluation of the face critical point detection of video flowing.
Feature detection device to performing characteristic detection method in the embodiment of the present invention is illustrated below, please
Refering to Fig. 3, including：
Critical point detection unit 301, for carrying out facial features localization to first object image, obtains first
Key point feature；
Unit 302 is returned, according to N number of benchmark, the first key point feature is entered for respectively
Row regressing calculation, obtains N number of first regression result, and the N is the integer more than 1, the station meter
Spend the benchmark for facial features localization region；
Similarity calculated 303, for calculating the similarity of N number of first regression result；
Average calculation unit 304, if the similarity for N number of first regression result meets similar bar
Part, then calculate the average of N number of first regression result, obtain fisrt feature testing result.
Further,
The critical point detection unit 301 is additionally operable to, and carries out facial features localization to the second target image,
Obtain the second key point feature；
The recurrence unit 302 is additionally operable to, respectively according to N number of benchmark, to second key point
Feature carries out regressing calculation, obtains N number of second regression result；
The similarity calculated 303 is additionally operable to, and calculates the similarity of N number of second regression result；
The average calculation unit 304 is additionally operable to, if the similarity of N number of second regression result meets
Condition of similarity, then calculate the average of N number of second regression result, obtain second feature testing result；
Described device also includes：Characteristic Contrast unit 305, if for N number of second regression result
Similarity is unsatisfactory for condition of similarity, then by the fisrt feature testing result respectively with described N number of second time
Sum up fruit and do Characteristic Contrast, export the characteristic similarity highest one with the fisrt feature testing result
Second regression result, as second feature testing result.
Specifically, the critical point detection unit 301 specifically for：
Face location and size detection are carried out to the first object image；
Based on the face location and the size detection result, normalizing is carried out to the first object image
Change computing；
The first object image after to normalization computing carries out Scale invariant features transform SIFT feature and carries
Take, obtain the first key point feature.
Specifically, the similarity calculated 303 specifically for：
Euclidean distance between calculating in N number of first regression result respectively two-by-two；
Or,
Calculate the Euclidean distance of corresponding first regression result of two neighboring benchmark.
Specifically, the Characteristic Contrast unit 305 specifically for：
Calculate respectively the fisrt feature testing result respectively with N number of second regression result it is European away from
From obtaining N number of Euclidean distance result of calculation, contrast the size of N number of Euclidean distance result of calculation.
The concrete operations flow process of above-mentioned unit may be referred to said method embodiment, and here is omitted.
In several embodiments provided herein, it should be understood that disclosed apparatus and method can
To realize by another way.For example, device embodiment described above is only schematic,
For example, the division of the unit, only a kind of division of logic function, can have in addition when actually realizing
Dividing mode, such as multiple units or component can with reference to or be desirably integrated into another system, or
Some features can be ignored, or not perform.It is another, shown or discussed coupling each other or
Direct-coupling or communication connection can be the INDIRECT COUPLINGs or communication link by some interfaces, device or unit
Connect, can be electrical, mechanical or other forms.
The unit as separating component explanation can be or may not be physically separate, work
For the part that unit shows can be or may not be physical location, you can be local to be located at one,
Or can also be distributed on multiple NEs.Can select according to the actual needs part therein or
Person's whole unit is realizing the purpose of this embodiment scheme.
In addition, each functional unit in each embodiment of the invention can be integrated in a processing unit,
Can also be that unit is individually physically present, it is also possible to which two or more units are integrated in a list
In unit.Above-mentioned integrated unit both can be realized in the form of hardware, it would however also be possible to employ software function list
The form of unit is realized.
If the integrated unit is realized and as independent production marketing using in the form of SFU software functional unit
Or when using, can be stored in a computer read/write memory medium.Based on such understanding, this
Part that the technical scheme of invention is substantially contributed to prior art in other words or the technical scheme
Completely or partially can be embodied in the form of software product, the computer software product is stored in one
In storage medium, use including some instructions so that computer equipment (can be personal computer,
Server, or the network equipment etc.) perform all or part of step of each embodiment methods described of the invention
Suddenly.And aforesaid storage medium includes：USB flash disk, portable hard drive, read only memory (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), magnetic disc or CD
Etc. it is various can be with the medium of store program codes.
The above, the only specific embodiment of the present invention, but protection scope of the present invention is not limited to
In this, any those familiar with the art the invention discloses technical scope in, can be easily
Expect change or replacement, should all be included within the scope of the present invention.Therefore, protection of the invention
Scope described should be defined by scope of the claims.
Claims (10)
1. a kind of characteristic detection method, it is characterised in that include：
Facial features localization is carried out to first object image, the first key point feature is obtained；
Respectively according to N number of benchmark, regressing calculation is carried out to the first key point feature, N is obtained
Individual first regression result, the N are the integer more than 1, and the benchmark is facial features localization area
The benchmark in domain；
Calculate the similarity of N number of first regression result；
If the similarity of N number of first regression result meets condition of similarity, described N number of first is calculated
The average of regression result, obtains fisrt feature testing result.
2. method according to claim 1, it is characterised in that described to obtain fisrt feature detection knot
After fruit, also include：
Facial features localization is carried out to the second target image, the second key point feature is obtained；
Respectively according to N number of benchmark, regressing calculation is carried out to the second key point feature, N is obtained
Individual second regression result；
Calculate the similarity of N number of second regression result；
If the similarity of N number of second regression result meets condition of similarity, described N number of second is calculated
The average of regression result, obtains second feature testing result；
If the similarity of N number of second regression result is unsatisfactory for condition of similarity, by the fisrt feature
Testing result does Characteristic Contrast respectively with N number of second regression result, exports and examines with the fisrt feature
One the second regression result of characteristic similarity highest of result is surveyed, as second feature testing result.
3. method according to claim 1, it is characterised in that described that first object image is carried out
Facial features localization, obtains the first key point feature, including：
Face location and size detection are carried out to the first object image；
Based on the face location and the size detection result, normalizing is carried out to the first object image
Change computing；
The first object image after to normalization computing carries out Scale invariant features transform SIFT feature and carries
Take, obtain the first key point feature.
4. method according to claim 1, it is characterised in that the calculating is described N number of first time
Sum up the similarity of fruit, including：
Euclidean distance between calculating in N number of first regression result respectively two-by-two；
Or,
Calculate the Euclidean distance of corresponding first regression result of two neighboring benchmark.
5. method according to claim 2, it is characterised in that by the fisrt feature testing result
Characteristic Contrast is done with N number of second regression result respectively, including：
Calculate respectively the fisrt feature testing result respectively with N number of second regression result it is European away from
From obtaining N number of Euclidean distance result of calculation, contrast the size of N number of Euclidean distance result of calculation.
6. a kind of feature detection device, it is characterised in that include：
Critical point detection unit, for carrying out facial features localization to first object image, obtains the first pass
Key point feature；
Unit is returned, according to N number of benchmark, the first key point feature is carried out back for respectively
Return computing, obtain N number of first regression result, the N is the integer more than 1, the benchmark is
The benchmark in facial features localization region；
Similarity calculated, for calculating the similarity of N number of first regression result；
Average calculation unit, if the similarity for N number of first regression result meets condition of similarity,
The average of N number of first regression result is then calculated, fisrt feature testing result is obtained.
7. device according to claim 6, it is characterised in that
The critical point detection unit is additionally operable to, and carries out facial features localization to the second target image, obtains
Second key point feature；
The recurrence unit is additionally operable to, respectively according to N number of benchmark, to the second key point feature
Regressing calculation is carried out, N number of second regression result is obtained；
The similarity calculated is additionally operable to, and calculates the similarity of N number of second regression result；
The average calculation unit is additionally operable to, if the similarity of N number of second regression result meets similar
Condition, then calculate the average of N number of second regression result, obtain second feature testing result；
Described device also includes：Characteristic Contrast unit, if for the similar of N number of second regression result
Degree is unsatisfactory for condition of similarity, then the fisrt feature testing result is returned knot with described N number of second respectively
Fruit is cooked Characteristic Contrast, exports the characteristic similarity highest one second with the fisrt feature testing result
Regression result, as second feature testing result.
8. device according to claim 6, it is characterised in that the critical point detection unit is concrete
For：
Face location and size detection are carried out to the first object image；
Based on the face location and the size detection result, normalizing is carried out to the first object image
Change computing；
The first object image after to normalization computing carries out Scale invariant features transform SIFT feature and carries
Take, obtain the first key point feature.
9. device according to claim 6, it is characterised in that the similarity calculated is concrete
For：
Euclidean distance between calculating in N number of first regression result respectively two-by-two；
Or,
Calculate the Euclidean distance of corresponding first regression result of two neighboring benchmark.
10. device according to claim 7, it is characterised in that the Characteristic Contrast unit is concrete
For：
Calculate respectively the fisrt feature testing result respectively with N number of second regression result it is European away from
From obtaining N number of Euclidean distance result of calculation, contrast the size of N number of Euclidean distance result of calculation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510657140.0A CN106570445B (en) | 2015-10-13 | 2015-10-13 | A kind of characteristic detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510657140.0A CN106570445B (en) | 2015-10-13 | 2015-10-13 | A kind of characteristic detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106570445A true CN106570445A (en) | 2017-04-19 |
| CN106570445B CN106570445B (en) | 2019-02-05 |
Family
ID=58508198
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510657140.0A Active CN106570445B (en) | 2015-10-13 | 2015-10-13 | A kind of characteristic detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106570445B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080037836A1 (en) * | 2006-08-09 | 2008-02-14 | Arcsoft, Inc. | Method for driving virtual facial expressions by automatically detecting facial expressions of a face image |
| CN102360421A (en) * | 2011-10-19 | 2012-02-22 | 苏州大学 | Face identification method and system based on video streaming |
| CN102629320A (en) * | 2012-03-27 | 2012-08-08 | 中国科学院自动化研究所 | Ordinal measurement statistical description face recognition method based on feature level |
| CN104517104A (en) * | 2015-01-09 | 2015-04-15 | 苏州科达科技股份有限公司 | Face recognition method and face recognition system based on monitoring scene |
-
2015
- 2015-10-13 CN CN201510657140.0A patent/CN106570445B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080037836A1 (en) * | 2006-08-09 | 2008-02-14 | Arcsoft, Inc. | Method for driving virtual facial expressions by automatically detecting facial expressions of a face image |
| CN102360421A (en) * | 2011-10-19 | 2012-02-22 | 苏州大学 | Face identification method and system based on video streaming |
| CN102629320A (en) * | 2012-03-27 | 2012-08-08 | 中国科学院自动化研究所 | Ordinal measurement statistical description face recognition method based on feature level |
| CN104517104A (en) * | 2015-01-09 | 2015-04-15 | 苏州科达科技股份有限公司 | Face recognition method and face recognition system based on monitoring scene |
Non-Patent Citations (2)
| Title |
|---|
| 姚文韬 等: "基于多尺度SDM模型的人脸对齐算法", 《控制工程》 * |
| 黄靓 等: "基于多尺度特征提取与多元回归分析的人脸识别", 《光学与光电技术》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106570445B (en) | 2019-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108197532A (en) | The method, apparatus and computer installation of recognition of face | |
| Kim et al. | Boundary preserving dense local regions | |
| CN111241989B (en) | Image recognition method and device and electronic equipment | |
| CN108416902B (en) | Real-time object identification method and device based on difference identification | |
| CN110020592A (en) | Object detection model training method, device, computer equipment and storage medium | |
| US20230033052A1 (en) | Method, apparatus, device, and storage medium for training image processing model | |
| CN108549870A (en) | A kind of method and device that article display is differentiated | |
| CN105938559A (en) | Digital image processing using convolutional neural networks | |
| CN107808120A (en) | Glasses localization method, device and storage medium | |
| WO2020134102A1 (en) | Article recognition method and device, vending system, and storage medium | |
| CN109325538A (en) | Object detection method, device and computer readable storage medium | |
| CN107194361A (en) | Two-dimentional pose detection method and device | |
| CN110610575B (en) | Coin identification method and device and cash register | |
| CN111275060A (en) | Recognition model updating processing method and device, electronic equipment and storage medium | |
| US20170213074A1 (en) | Decoy-based matching system for facial recognition | |
| CN106022227A (en) | Gesture identification method and apparatus | |
| CN107918767A (en) | Object detection method, device, electronic equipment and computer-readable medium | |
| CN105740808A (en) | Human face identification method and device | |
| CN110796135A (en) | Target positioning method and device, computer equipment and computer storage medium | |
| CN104281572A (en) | Target matching method and system based on mutual information | |
| Fernández et al. | Robust Real‐Time Traffic Surveillance with Deep Learning | |
| CN109829401A (en) | Traffic sign recognition method and device based on double capture apparatus | |
| Dimiccoli et al. | Exploiting t-junctions for depth segregation in single images | |
| CN109919626B (en) | High-risk bank card identification method and device | |
| CN114255435A (en) | Method and device for detecting abnormality of transport device, electronic apparatus, and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |