CN106845532B - A kind of screening sample method - Google Patents
A kind of screening sample method Download PDFInfo
- Publication number
- CN106845532B CN106845532B CN201611270693.1A CN201611270693A CN106845532B CN 106845532 B CN106845532 B CN 106845532B CN 201611270693 A CN201611270693 A CN 201611270693A CN 106845532 B CN106845532 B CN 106845532B
- Authority
- CN
- China
- Prior art keywords
- probability
- picture
- files
- personage
- pictures
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Library & Information Science (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention is suitable for image contract correlation technique improvement areas, provides a kind of screening sample method, the screening sample method includes the following steps：A, the average value of the similar value of picture in two files to be compared is calculated；B, the probability under the average value is found out according to obtained average value；C, whether it is same personage according to picture in two files of probabilistic determination；It is when probability is bigger, then bigger for same personage's probability, it is when probability is smaller, then smaller for same personage's probability.By being the probability of same personage between calculating Photo folder, picture is illustrated according to the sequence of probability from big to small on screening implement interface and is screened for user, it to be all pictures of same personage that the locking of user quickly can be allowed, which to have greatly with current image,.This method can greatly accelerate the speed and efficiency of screening, while can find duplicate data all in training sample as far as possible.
Description
Technical field
The invention belongs to image contract correlation technique improvement areas more particularly to a kind of screening sample method and system.
Background technology
Face is the important information of people, is to discriminate between the important evidence of different people, therefore face alignment is compared with fingerprint, iris
Etc. technologies are more natural, more direct alignments.
Face alignment is by the face of image or video input by extracting specific face characteristic information, and in database
Registered face characteristic information compares, and obtains matched face extremely similarity, be confirmed whether be with face in database
It is same.
Face alignment all has very important effect in many instances, such as the video multimedia message in cell phone multimedia message, people
Machine interface, permission control, intelligent monitoring system etc..Accuracy, precision and the robustness problem of comparison are always what industry was concerned about
Main problem.
Art methods are to go to judge whether between different files be same personage successively by artificial mode.It should
Method efficiency is low, and time-consuming, and not accurate enough.
Invention content
The purpose of the present invention is to provide a kind of screening sample methods, it is intended to solve prior art efficiency is low, time-consuming, no
Enough accurate technical problems.
The invention is realized in this way a kind of screening sample method, the screening sample method includes the following steps：
A, the average value of the similar value of picture in two files to be compared is calculated；
B, the probability under the average value is found out according to obtained average value；
C, whether it is same personage according to picture in two files of probabilistic determination；Then it is same when probability is bigger
Personage's probability is bigger, when probability is smaller, then smaller for same personage's probability.
The present invention further technical solution be：It is further comprising the steps of in the step A：
A1, the similar value for calculating all pictures in two files successively；
A2, simultaneously average value is found out in turn according to obtained all similar value summations.
The present invention further technical solution be：The average value of picture is s in two files in the step A2avg, public
Formula is：M, n are to have picture number respectively in file A and B,For picture a in AiScheme in B
Piece bjSimilar value.
The present invention further technical solution be：It is P that picture, which is the probability of same people, in two files in the step BAB,
Its formula isWherein, savgFor the average value of picture in two files, PscFor
Two pictures are the probability of same personage, 1-PscFor the probability that two pictures are different personages, Pf(s) it is different for two pictures
Personage and the probability that similarity is s, Pt(s) it is two pictures are identical personage and similarity is s probability.
Another object of the present invention is to provide screening sample method, the screening sample method includes the following steps：
A, the similar value of all pictures in two files is calculated successively；
B, it is summed according to obtained all similar values；
C, being acquired according to similar value and calculate average value；
D, judge whether picture is same people in two files according to obtained average value, when average value is higher, then
Picture in two files is that the probability of same personage is bigger, and when average value is lower, then the picture in two files is
The probability of same personage is smaller.
The present invention further technical solution be：Similar value in the step a is sAB, formula isWherein, m, n are the number of picture in two files,For picture a in AiWith picture b in BjPhase
Like value.
Another object of the present invention is to provide a kind of screening sample method, the screening sample method includes following step
Suddenly：
(1), the probability that picture in two files to be compared is same personage is calculated successively；
(2), obtained all probability value additions are found out into average value；
(3), judge whether picture is same personage in two files according to the average probability value acquired, when average probability value
It is bigger, then in two files picture be same person probability it is bigger, when average probability be worth it is smaller, then scheme in two files
Piece is that the probability of same person is smaller.
The present invention further technical solution be：According to the data after Local Artificial sampling screening in the step (1)
Statistics is P with the probability that two pictures of calculating are same personagesc, formula is：N is in pictures
Picture number, S be identical personage's picture logarithm.
The present invention further technical solution be：Two pictures are identical personage in the step (1) and similarity is s's
Probability is Pt(s), formula isThe probability that two pictures are different personages and similarity is s is Pf(s), public
Formula isN is the picture number in pictures, and S is identical personage's picture logarithm, and TS is same person
Object and similarity are the picture logarithm of s, FS be not be same personage and picture logarithm that similarity is s.
The beneficial effects of the invention are as follows：By calculate Photo folder between be same personage probability, by picture according to
The sequence of probability from big to small is illustrated on screening implement interface screens for user, and the locking of user quickly can be allowed to scheme with current
Piece, which has greatly, to be all pictures of same personage.This method can greatly accelerate the speed and efficiency of screening, while can
Duplicate data all in training sample is found as far as possible.
Description of the drawings
Fig. 1 is the flow chart one of screening sample method provided in an embodiment of the present invention.
Fig. 2 is the flowchart 2 of screening sample method provided in an embodiment of the present invention.
Fig. 3 is the flow chart 3 of screening sample method provided in an embodiment of the present invention.
Specific implementation mode
Fig. 1 shows the flow chart of screening sample method provided by the invention, and details are as follows：
Step S11 calculates two average values with the similar value of picture in comparison document folder；It is summed by similar value flat
Averaging method and probabilistic method are combined, and can first calculate the average value of picture similar value in two files to be compared；Wherein ask
Average value calculates the similar value of all pictures in two files successively first；For the file that any two is to be compared,
The similar value of picture and all pictures in another file in a file can be calculated successively；Secondly, according to obtained
All similar value summations simultaneously and then find out average value；Then all similar values for calculating gained are summed, and then finds out average value；
Wherein, the average value of picture is s in two filesavg, formula is：M, n are in file A and B
There is picture number respectively,For picture a in AiWith picture b in BjSimilar value.
Step S12 finds out the probability under the average value according to obtained average value；The average value is found out according to the average value
Under probability be PAB, wherein it is P that picture, which is the probability of same people, in two filesAB, formula isWherein, savgFor the average value of picture in two files, PscFor two pictures
For the probability of same personage, 1-PscFor the probability that two pictures are different personages, Pf(s) it is different personages and phase for two pictures
Like the probability that degree is s, Pt(s) it is two pictures are identical personage and similarity is s probability；Pf(s), Pt(s) it is according to early period
Two functions that a large amount of experimental data comes out, s is some similarity.
Whether step S13 is same personage according to picture in two files of probabilistic determination；When probability is bigger, then for
Same personage's probability is bigger, when probability is smaller, then smaller for same personage's probability.The probability that can be acquired according to front
To indicate that picture is the possibility of same personage between both of these documents are pressed from both sides.When the probability is bigger, then show that both of these documents are pressed from both sides
Between picture be same personage possibility it is bigger.It is on the contrary then when the probability is smaller, then show picture between both of these documents folder
Possibility for same personage is smaller.
As shown in Fig. 2, another object of the present invention is to provide a kind of flow chart of screening sample method, it is described in detail such as
Under：
Step S21 calculates the similar value of all pictures in two files successively；For the file that any two is to be compared
Folder can calculate the similar value of picture and all pictures in another file in a file successively；Wherein similar value is sAB,
Its formula isWherein, m, n are the number of picture in two files,For picture a in AiScheme in B
Piece bjSimilar value.Assuming that have m in file A and B to be compared respectively, n pictures,For picture a in AiWith picture b in Bj's
The similar value of similar value, AB files is：
Step S22 sums according to obtained all similar values；All similar values for calculating gained are summed,
In, at thisMolecule is summed for similar value in formula, and the value of the formula is exactly average value, that is, the phase of AB files
Like value.
Step S23, acquired according to similar value and calculate average value；Utilize the sum got required by similar value, Jin Erqiu
Go out average value.
Step S24 judges whether picture is same people in two files, when average value is higher according to obtained average value
When, then the picture in two files is that the probability of same personage is bigger, when average value is lower, then the figure in two files
Piece is that the probability of same personage is smaller.Picture is same between can be used for weighing two files using the average value asked
The possibility size of personage then shows that picture is the possibility of same personage between both of these documents are pressed from both sides when average value is higher
Bigger, when average value is lower, then the picture in two files is that the probability of same personage is smaller.
As shown in figure 3, another object of the present invention is to provide a kind of flow chart of screening sample method, it is described in detail such as
Under：
Step S31 calculates two with the probability that picture in comparison document folder is same personage successively；Any two is waited for
Comparison document presss from both sides, and can be that the probability of same personage is with picture between two files to weigh picture between two files
The possibility of same personage.It is same people to obtain two pictures in a prodigious pictures according to Local Artificial sampling statistics
The probability of object.According to many experiments statistics can obtain, probability when arbitrary two pictures are identical personage and similarity is s with
And the probability that arbitrary two pictures are different personages and similarity is s.Arbitrary two figures can be found out according to two above probability
Piece is the probability of same personage, can calculate in file to be compared picture and institute in another file in a file successively
It is the probability of same personage to have picture, and the data statistics after being screened according to Local Artificial sampling is same with two pictures of calculating
The probability of one personage is Psc, formula is：N is the picture number in pictures, and S is same task
Picture logarithm.It is P in the probability that two pictures are identical personage and similarity is st(s), formula isTwo
The probability that picture is different personages and similarity is s is Pf(s), formula isN is pictures
In picture number, S be same task picture logarithm, the picture logarithm that TS is identical personage and similarity is s, FS be not be same
One personage and the picture logarithm that similarity is s.
According to the data statistics and calculating after Local Artificial sampling screening, can obtain, in a prodigious pictures
In, two pictures are that the probability of same personage is Psc, two pictures are that the probability of different personages is for 1-Psc。
Assuming that experiment picture is concentrated and shares N pictures, wherein identical personage's picture logarithm is S, can learn：
(2) it can be obtained according to many experiments, the probability that two pictures are identical personage and similarity is s is Pt(s), two
The probability that pictures are different personages and similarity is s is Pf(s)。
Assuming that experiment picture, which is concentrated, shares N pictures, statistics obtains identical personage's picture logarithm S, is identical personage and phase
It is TS like the picture logarithm that degree is s, is not same personage and picture logarithm that similarity is s is FS, then can calculates：
It calculates：
(1) picture a in A files is calculatediWith picture b in B filesjSimilarity, be denoted as.According to aiAnd bjPhase
Like degree, calculate aiAnd bjFor the probability of same personage：
(2) assume there is m, n pictures in A and B files respectively, A and B files are the probability of same personage：
I.e.
Obtained all probability value additions are found out average value by step S32；It is all general by what is obtained in step S31
Rate value finds out its average value after being added.
Step S33 judges whether picture is same personage in two files according to the average probability value acquired, when averagely general
Rate value is bigger, then picture is that the probability of same person is bigger in two files, when average probability is worth smaller, then two files
Middle picture is that the probability of same person is smaller.Using the average probability found out in step S32, scheme to weigh in both of these documents folder
Piece is the possibility of same personage, wherein the average probability found out is bigger, then picture is same between showing both of these documents folder
The possibility of personage is bigger.Conversely, the average probability wherein found out is smaller, then picture is same between showing both of these documents folder
The possibility of personage is smaller.
When meeting linear relationship between the alignment score s and P (s) of two pictures, it is apparent from one formula of scheme (9) and scheme
Three formula (7) are of equal value, but the experimental results showed that between s and P (s) and be unsatisfactory for linear relationship, it is in this case, public
The result of calculation of formula (9) is inaccurate, and formula (7) can accurately more calculate the probability that two files are same personage.Institute
It is more more acurrate than the result of calculation of scheme one reliable with scheme three.
The average value s of the two file picture similarities of scheme 2avgIndicate two file pictures be same personage can
Energy property, scheme one use savgCorresponding probabilityIndicate that two file pictures are the possibility of same personage.Assuming that A files and
The average value of B, C file similarity is respectively sAB, sAC, corresponding probability is respectivelyIt clearly can function P (s)
For increasing function, even sAB≥sACThenSo it is known that the measurement mode of scheme two and scheme one is of equal value,
Its weighing result is also consistent.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement etc., should all be included in the protection scope of the present invention made by within refreshing and principle.
Claims (9)
1. a kind of screening sample method, which is characterized in that the screening sample method includes the following steps：
A, the average value of the similar value of picture in two files to be compared is calculated；
B, the probability under the average value is found out according to obtained average value；
C, whether it is same personage according to picture in two files of probabilistic determination；Then it is the same personage when probability is bigger
Probability is bigger, when probability is smaller, then smaller for same personage's probability.
2. screening sample method according to claim 1, which is characterized in that further comprising the steps of in the step A：
A1, the similar value for calculating all pictures in two files successively；
A2, simultaneously average value is found out in turn according to obtained all similar value summations.
3. screening sample method according to claim 2, which is characterized in that picture in two files in the step A2
Average value is savg, formula is：M, n are to have picture number, s respectively in file A and BaibjFor
Picture a in AiWith picture b in BjSimilar value.
4. screening sample method according to claim 3, which is characterized in that picture is in two files in the step B
The probability of same people is PAB, formula isWherein, savgFor in two files
The average value of picture, PscFor the probability that two pictures are same personage, 1-PscFor the probability that two pictures are different personages, Pf
(s) it is the probability that two pictures are different personages and similarity is s, Pt(s) it is that two pictures are identical personage and similarity is s
Probability.
5. a kind of screening sample method, which is characterized in that the screening sample method includes the following steps：
A, the similar value of all pictures in two files is calculated successively；
B, it is summed according to obtained all similar values；
C, being acquired according to similar value and calculate average value；
D, judge whether picture is same people in two files according to obtained average value, when average value is higher, then two
Picture in file is that the probability of same personage is bigger, and when average value is lower, then the picture in two files is same
The probability of personage is smaller.
6. screening sample method according to claim 5, which is characterized in that the similar value in the step a is sAB, public
Formula isWherein, m, n are the number of picture in two files, saibjFor picture a in AiWith picture in B
bjSimilar value.
7. a kind of screening sample method, which is characterized in that the screening sample method includes the following steps：
(1), the probability that picture in two files to be compared is same personage is calculated successively；
(2), obtained all probability value additions are found out into average value；
(3), judge whether picture is same personage in two files according to the average probability value acquired, when average probability value is got over
Greatly, then picture is that the probability of same person is bigger in two files, when average probability is worth smaller, then picture in two files
It is smaller for the probability of same person.
8. screening sample method according to claim 7, which is characterized in that taken out according to Local Artificial in the step (1)
Data statistics after sample screening with to calculate two pictures be the probability of same personage is Psc, formula is：N is the picture number in pictures, and S is identical personage's picture logarithm.
9. screening sample method according to claim 8, which is characterized in that two pictures are identical in the step (1)
The probability that personage and similarity are s is Pt(s), formula isTwo pictures are different personages and similarity is s
Probability is Pf(s), formula isN is the picture number in pictures, and S is identical figure map
Piece logarithm, the picture logarithm that TS is identical personage and similarity is s, FS be not be same personage and picture pair that similarity is s
Number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611270693.1A CN106845532B (en) | 2016-12-30 | 2016-12-30 | A kind of screening sample method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611270693.1A CN106845532B (en) | 2016-12-30 | 2016-12-30 | A kind of screening sample method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106845532A CN106845532A (en) | 2017-06-13 |
| CN106845532B true CN106845532B (en) | 2018-07-20 |
Family
ID=59116986
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611270693.1A Active CN106845532B (en) | 2016-12-30 | 2016-12-30 | A kind of screening sample method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106845532B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241311A (en) * | 2018-08-09 | 2019-01-18 | 上海常仁信息科技有限公司 | A kind of view field's automatic screening system and method based on robot |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6272249B1 (en) * | 1997-03-21 | 2001-08-07 | Sharp Kabushiki Kaisha | Image processing device |
| CN101546424A (en) * | 2008-03-24 | 2009-09-30 | 富士通株式会社 | Method and device for processing image and watermark detection system |
| CN101888469A (en) * | 2009-05-13 | 2010-11-17 | 富士通株式会社 | Image processing method and image processing device |
| CN102509303A (en) * | 2011-11-22 | 2012-06-20 | 鲁东大学 | Binarization image registration method based on improved structural similarity |
-
2016
- 2016-12-30 CN CN201611270693.1A patent/CN106845532B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6272249B1 (en) * | 1997-03-21 | 2001-08-07 | Sharp Kabushiki Kaisha | Image processing device |
| CN101546424A (en) * | 2008-03-24 | 2009-09-30 | 富士通株式会社 | Method and device for processing image and watermark detection system |
| CN101888469A (en) * | 2009-05-13 | 2010-11-17 | 富士通株式会社 | Image processing method and image processing device |
| CN102509303A (en) * | 2011-11-22 | 2012-06-20 | 鲁东大学 | Binarization image registration method based on improved structural similarity |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106845532A (en) | 2017-06-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11210503B2 (en) | Systems and methods for facial representation | |
| Ma et al. | Detect rumors using time series of social context information on microblogging websites | |
| CN106202041B (en) | A kind of method and apparatus of entity alignment problem in solution knowledge mapping | |
| CN107341220B (en) | Multi-source data fusion method and device | |
| CN106295482B (en) | A kind of update method and device of face database | |
| US20160117347A1 (en) | Method and system of using image recognition and geolocation signal analysis in the construction of a social media user identity graph | |
| US10311288B1 (en) | Determining identity of a person in a digital image | |
| CN103942705A (en) | Advertisement classified match pushing method and system based on human face recognition | |
| JP2010130695A (en) | Method and system for aligning sequence of images | |
| CN112818867B (en) | Portrait clustering method, equipment and storage medium | |
| CN104615715A (en) | Social network event analyzing method and system based on geographic positions | |
| CN109064294A (en) | A kind of time of fusion factor, the drug recommended method of text feature and correlation | |
| CN110458091A (en) | Recognition of face 1 based on position screening is than N algorithm optimization method | |
| WO2022142903A1 (en) | Identity recognition method and apparatus, electronic device, and related product | |
| CN113254696B (en) | Cover image acquisition method and device | |
| CN112395582A (en) | Job entry management method, system, computer device and storage medium | |
| CN106845532B (en) | A kind of screening sample method | |
| CN102915449B (en) | A kind of photo classification method | |
| CN102799616A (en) | Outlier point detection method in large-scale social network | |
| CN104778388A (en) | Method and system for identifying same user under two different platforms | |
| CN111652260B (en) | Face clustering sample number selection method and system | |
| Joshi et al. | Data-efficient contrastive language-image pretraining: Prioritizing data quality over quantity | |
| CN106980989B (en) | Merchant recommendation method based on user behavior characteristic analysis | |
| CN106897726A (en) | The finding method and device of Missing Persons | |
| CN112084956A (en) | Special face crowd screening system based on small sample learning prototype network |
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 |