CN110413815A - Portrait clusters cleaning method and device - Google Patents
Portrait clusters cleaning method and device Download PDFInfo
- Publication number
- CN110413815A CN110413815A CN201910684614.9A CN201910684614A CN110413815A CN 110413815 A CN110413815 A CN 110413815A CN 201910684614 A CN201910684614 A CN 201910684614A CN 110413815 A CN110413815 A CN 110413815A
- Authority
- CN
- China
- Prior art keywords
- portrait
- archives
- picture
- veritification
- similarity
- 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
- 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
- G06F16/53—Querying
-
- 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
- G06F16/53—Querying
- G06F16/532—Query formulation, e.g. graphical querying
-
- 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
- G06F16/55—Clustering; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Processing Or Creating Images (AREA)
- Image Analysis (AREA)
Abstract
This disclosure relates to a kind of portrait cluster cleaning method and device, wherein it includes: obtaining step that portrait, which clusters cleaning method, obtains portrait picture, and clustered according to the similarity of portrait picture, forms portrait archives;Veritify step, it is veritified based on the archive feature value of portrait archives to object is veritified, judges the veritification similarity for veritifying object and portrait archives, and veritify the size relation of threshold value, wherein, veritifying object includes portrait picture and/or other portrait archives in addition to portrait archives;Cleaning step, according to veritification as a result, cleaning portrait archives.By the method, the cleaning to portrait archives is realized, there are the incidence reductions that miscellany picture, archives division, candid photograph image filing rate are low inside the portrait archives after making cleaning.
Description
Technical field
This disclosure relates to facial image processing technical field, it is specifically related to a kind of portrait cluster cleaning method and device.
Background technique
With the deep development of China smart city, safe city, bright as snow Engineering Strategy, covering is gradually built at present
The face identification system of public domain, critical position.The identifying system can the passerby to process carry out candid photograph identification, and combine
Early warning of deploying to ensure effective monitoring and control of illegal activities is realized in public security fugitive personnel's portrait library, emphasis personnel's portrait library.To the magnanimity portrait picture that history is captured, people is utilized
Face identifying system analyzes the characteristic value and structured attributes of every portrait picture, such as gender, age bracket, can be realized to scheme
Search the preliminary portrait retrieval application such as figure, attribute retrieval.And portrait clustering technique is combined, it can will belong to the difference of the same person
The image clustering that time, different location are captured to the same person as in archives, to realize the portrait picture that will be captured, finally
The people being mapped in reality.
But the portrait archives that the face identification system established at present is generated by clustering are often faced with that there are miscellanies
Picture, captures the low problem of image filing rate at archives division.
Summary of the invention
In order to overcome problems of the prior art, the disclosure provides a kind of portrait cluster cleaning method and device.
In a first aspect, the embodiment of the present disclosure provides a kind of portrait cluster cleaning method comprising obtaining step obtains portrait
Picture, and clustered according to the similarity of portrait picture, form portrait archives;Veritify step, the archives based on portrait archives
Characteristic value is veritified to object is veritified, and judges the veritification similarity for veritifying object and portrait archives, and veritify the size of threshold value
Relationship, wherein veritifying object includes portrait picture and/or other portrait archives in addition to portrait archives;Cleaning step, according to
It veritifies as a result, cleaning portrait archives.
In one example, after obtaining step, portrait clusters cleaning method further include: files step in real time, obtains in real time
Current portrait picture obtains similar in real time according to the archive feature value of the picture feature value of current portrait picture and portrait archives
Current portrait picture is incorporated to portrait archives if similarity is greater than or equal to real time threshold in real time by degree.
In one example, cleaning step includes that if veritification object is portrait picture, and portrait picture is not archived in portrait shelves
Portrait picture then when veritifying similarity more than or equal to threshold value is veritified, is incorporated to portrait archives by case;If veritifying object to behave
As picture, and portrait picture has been archived in portrait archives, then, when veritify similarity be less than veritify threshold value when, by portrait picture from
It is rejected in portrait archives;If veritifying object is other portrait archives in addition to portrait archives, when veritify similarity be greater than or
When equal to veritifying threshold value, two portrait archives are merged.
In one example, after the cleaning step, portrait clusters cleaning method further include: step is updated, according to portrait archives
In include portrait picture, update portrait archives archive feature value.
In one example, step is veritified further include: the veritification period for veritifying step is determined, according to the veritification period to veritification object
It is veritified.
In one example, veritifying the period includes that picture veritifies period and archives veritification period, and picture veritifies the period to veritify core
Test the period that object is portrait picture;The archives veritification period is to veritify object as other portrait shelves in addition to portrait archives
The period of case;The picture veritification period is shorter than archives and veritifies the period.
In one example, portrait picture does not file including filing portrait picture and portrait picture, wherein filing portrait picture is
The portrait picture of portrait archives has been formed or has been incorporated to according to similarity;It is not formed or simultaneously according to similarity for not filing portrait picture
Enter the portrait picture of portrait archives;Veritifying threshold value includes that the first veritification threshold value and second veritify threshold value, wherein first veritifies threshold value
For the veritification threshold value of the veritification similarity for judging to file portrait picture and portrait archives;Second veritifies threshold value as judging
The veritification threshold value of the veritification similarity of portrait picture and portrait archives is not filed;First, which veritifies threshold value, veritifies threshold value less than second.
In one example, real time threshold includes the first real time threshold and the second real time threshold, and the first real time threshold is less than the second real time threshold;
Filing step in real time further include: obtain the acquisition device ID and the previous portrait being incorporated in portrait archives of current portrait picture
The acquisition device ID of picture;If the acquisition device ID of current portrait picture, with the previous portrait picture being incorporated in portrait archives
Acquisition device ID it is identical, if then in real time similarity be greater than or equal to the first real time threshold, current portrait picture is incorporated to people
As archives;If the acquisition device ID of current portrait picture, the acquisition device with the previous portrait picture being incorporated in portrait archives
ID is different, if then similarity is greater than or equal to the second real time threshold in real time, current portrait picture is incorporated to portrait archives.
In one example, real time threshold includes the first real time threshold and the second real time threshold, and the first real time threshold is less than second
Real time threshold;Filing step in real time further include: obtain the acquisition time of current portrait picture and the acquisition dress of current portrait picture
ID is set, the acquisition device ID and the previous portrait being incorporated in portrait archives with the previous portrait picture being incorporated in portrait archives
The acquisition time of picture;The acquisition device ID of current portrait picture, with adopting for the previous portrait picture being incorporated in portrait archives
Acquisition means ID is identical, if the acquisition time of current portrait picture, the acquisition with the previous portrait picture being incorporated in portrait archives
The time interval of time is less than or equal to time threshold, if then similarity is greater than or equal to the first real time threshold in real time, will work as
Forefathers are incorporated to portrait archives as picture;The acquisition device ID of current portrait picture, with the previous portrait being incorporated in portrait archives
The acquisition device ID of picture is identical, if the acquisition time of current portrait picture, with the previous portrait figure being incorporated in portrait archives
The time interval of the acquisition time of piece is greater than time threshold, will if then similarity is greater than or equal to the second real time threshold in real time
Current portrait picture is incorporated to portrait archives;If the acquisition device ID of current portrait picture, is incorporated in portrait archives with previous
The acquisition device ID of portrait picture is different, if then similarity is greater than or equal to the second real time threshold in real time, by current portrait figure
Piece is incorporated to portrait archives.
Second aspect, the embodiment of the present disclosure provide a kind of portrait cluster cleaning device, which, which has, realizes above-mentioned first
The function for the portrait cluster cleaning method that aspect is related to.The function can also be executed by hardware realization by hardware
Corresponding software realization.The hardware or software include one or more modules corresponding with above-mentioned function.
In one example, portrait cluster cleaning device includes obtaining module, for obtaining portrait picture, and according to portrait picture
Similarity clustered, formed portrait archives;Module is veritified, for the archive feature value based on portrait archives to veritification object
It is veritified, judges the veritification similarity for veritifying object and portrait archives, and veritify the size relation of threshold value, wherein veritification pair
As including portrait picture and/or other portrait archives in addition to portrait archives;Cleaning module is used for according to veritification as a result, clear
Portrait archives are washed, threshold value is veritified if veritifying similarity and being greater than or equal to, veritifies object and be incorporated to portrait archives, if veritifying similarity
Less than threshold value is veritified, then veritifies object and rejected from portrait archives.
The third aspect, the embodiment of the present disclosure provide a kind of electronic equipment, wherein electronic equipment includes: memory, for depositing
Storage instruction;And processor, the portrait of the instruction execution first aspect for calling memory to store cluster cleaning method.
Fourth aspect, the embodiment of the present disclosure provide a kind of computer readable storage medium, wherein computer-readable storage medium
Matter is stored with computer executable instructions, and computer executable instructions when executed by the processor, execute the portrait of first aspect
Cluster cleaning method.
The disclosure provides a kind of portrait cluster cleaning method and device, wherein portrait clusters cleaning method by veritifying step
Suddenly, i.e., the archive feature value based on established portrait archives is veritified to object is veritified.If portrait archives and veritification object
Veritification similarity be greater than or equal to veritify threshold value, then by veritify object be incorporated to portrait archives;If above-mentioned veritification similarity is less than
Threshold value is veritified, then will veritify object and rejected from portrait archives.By the method, the cleaning to portrait archives is realized, is made clear
There are the incidence reductions that miscellany picture, archives division, candid photograph image filing rate are low inside portrait archives after washing.
Detailed description of the invention
The following detailed description is read with reference to the accompanying drawings, above-mentioned and other purposes, the feature of disclosure embodiment
It will become prone to understand with advantage.In the accompanying drawings, embodiment of the present disclosure is shown by way of example rather than limitation,
Wherein:
Fig. 1 shows the portrait cluster cleaning method schematic diagram of embodiment of the present disclosure offer;
Fig. 2 shows another portraits that the embodiment of the present disclosure provides to cluster cleaning method schematic diagram;
Fig. 3 shows another portrait cluster cleaning method schematic diagram of embodiment of the present disclosure offer;
Fig. 4 shows a kind of portrait cluster cleaning device schematic diagram of embodiment of the present disclosure offer;
Fig. 5 shows a kind of electronic equipment schematic diagram of embodiment of the present disclosure offer.
Specific embodiment
The principle and spirit of the disclosure are described below with reference to several illustrative embodiments.It should be appreciated that providing this
A little embodiments are used for the purpose of making those skilled in the art can better understand that realizing the disclosure in turn, and be not with any
Mode limits the scope of the present disclosure.
Although being noted that the statements such as " first " used herein, " second " to describe implementation of the disclosure mode not
Same module, step and data etc., still the statement such as " first ", " second " is merely in different modules, step and data etc.
Between distinguish, and be not offered as specific sequence or significance level.In fact, the statements such as " first ", " second " are complete
It may be used interchangeably.
Fig. 1 is the schematic diagram that a kind of portrait that the embodiment of the present disclosure provides clusters cleaning method.As shown in Figure 1, portrait is poly-
Class cleaning method 100 includes obtaining step S101, veritifies step S102 and cleaning step S103.
As a kind of possible embodiment, portrait picture can be obtained by obtaining step S101, and according to getting
The similarity of portrait picture is clustered, to form portrait archives.At this point it is possible to which the portrait archives of formation are regarded as initial
Portrait archives, according to the later period file in real time as a result, with portrait picture constantly being incorporated to, the portrait archives of formation also can be therewith
It updates, variation.In practical applications, portrait picture can also be returned when the portrait picture of acquisition reaches certain amount
Shelves, cluster, to obtain portrait archives.
In application process, fusion calculation can be carried out according to the characteristic value for the portrait picture for being incorporated to portrait archives, with
The archive feature of corresponding portrait archives out, and the archive feature value of portrait archives is obtained according to archive feature.It is counted by fusion
The archive feature of obtained portrait archives more preferable can must integrate the clearest portrait picture of various pieces, to obtain quality
The archive feature of higher portrait archives, so that the archive feature value of the portrait archives arrived is more accurate.
For example, by similarity cluster calculation, if the portrait picture P1 and portrait picture P2 that obtain belong to a certain portrait
Portrait archives, for ease of description, now enable in the portrait archives only have portrait picture P1 and portrait picture P2.Wherein, portrait
Picture P1 has tetra- key position information of A, B, C, D, and portrait picture P2 has tetra- key position information of B, C, D, E, then should
The synthesis key position of portrait archives then include A, B, C, D, E this five key position (such as A, B, C, D, E be respectively eyes,
Nose, mouth, eyebrow, lip, face mask, P1 is because be the information that side face figure can not obtain key position E), and calculate people
As the readability of picture P1 and portrait picture P2 key position B, C, D having jointly, take in two portrait pictures, a certain phase
With the highest key position of key position clarity as the key position in the synthesis key position for calculating the portrait archives, then
The archive feature value of portrait archives is obtained by the synthesis key position of portrait archives.For example, if B, C two of portrait picture P1 are crucial
The clarity at position higher than portrait picture P2 two key position of B, C clarity, the D key position of portrait picture P1 it is clear
The clarity of D key position of the degree lower than portrait picture P2, then the synthesis key position of the portrait archives is A (P1), B
(P1)、C(P1)、D(P2)、E(P2)。
As a kind of deformation, in the application, the portrait figure of a picture optimal quality can also be chosen in portrait archives
Piece is as master map, and using the characteristic value of the master map as the archive feature value of portrait archives.It, can be with during choosing master map
Comprehensive test is carried out from the shooting angle of portrait picture, light conditions and picture clarity etc..
It is deformed as another kind, can also be according to whole portrait pictures in portrait archives, that is, it has been archived in portrait
All portrait pictures of archives generate the archive feature value of portrait archives.For example, it may be by taking each filed people
As the average value of the characteristic value of picture, as the archive feature value of the portrait archives.
In veritifying step S102, established portrait archives are based on, the archive feature value of the portrait archives, and root are obtained
Carry out veritification analysis to object is veritified according to archive feature value, i.e., according to the characteristic value for the characteristic value and portrait archives for veritifying object,
Veritification similarity between the two is obtained, and judges whether the veritification similarity is greater than or equal to veritification threshold value.Wherein, threshold is veritified
The size of value can according to the actual situation different and different.
Cleaning step S103, the veritification obtained according to veritification step S102 is as a result, clean portrait archives.Namely
It according to the veritification similarity for veritifying object and portrait archives, and veritifies the size relation of threshold value: being greater than or wait if veritifying similarity
In veritifying threshold value, then the veritification object is incorporated to the portrait archives;If veritify similarity be less than veritify threshold value, will veritify object from
It is rejected in portrait archives.
The portrait cluster cleaning method 100 that the disclosure provides is by veritifying step, i.e., based on established portrait archives
Archive feature value is veritified to object is veritified.If portrait archives and the veritification similarity for veritifying object, which are greater than or equal to, veritifies threshold
Value will then veritify object and be incorporated to portrait archives;If above-mentioned veritification similarity, which is less than, veritifies threshold value, object will be veritified from portrait shelves
It is rejected in case.By the method, the cleaning to portrait archives is realized, there are miscellany figures inside the portrait archives after making cleaning
The low incidence reduction of image filing rate is captured in piece, archives division.
Since the portrait archives of formation can be under the jurisdiction of the addition of the portrait picture of the portrait archives with the later period, and to the people
As archives are updated, correspondingly, the archive feature of portrait archives and archive feature value can also update accordingly, therefore, can deposit
In the portrait picture for the filing having been incorporated into the past, or the portrait picture that do not file, there are errors with the cluster result of portrait archives
The case where occur.For example, a portrait picture A if it exists, by being calculated, the similarity with portrait archives M is 79%, is
Convenient for explanation, now enabling similarity threshold is 80%.Since portrait picture A does not reach similar to the similarity of portrait archives M
Threshold value is spent, then, portrait picture A will not be incorporated into portrait archives M.Now there are another portrait picture B, the portrait figures again
Piece B and the similarity of portrait picture A are very high, it can are interpreted as, portrait picture B and portrait picture A can be considered same
Portrait.By being calculated, the similarity of the portrait picture B and portrait archives M are 85%, since similarity has been more than similarity
Threshold value, then, portrait picture B will be incorporated into portrait archives M.Therefore, portrait archives M will include portrait picture B, accordingly
, being incorporated to based on portrait picture B is also updated by the archive feature of portrait archives M.Due to portrait picture A and portrait figure
The similarity of piece B is very high, in addition the update of the archive feature of portrait archives M, and archive feature value also accordingly updates, at this point, if
Again veritify portrait picture A and portrait archives M similarity, in very maximum probability, it will occur portrait picture A with it is updated
The occurrence of similarity of portrait archives M is more than similarity threshold.
Similarly, a certain portrait picture for having been incorporated into portrait archives also will with the continuous renewal of portrait archives if it exists
It will appear the occurrence of similarity of the portrait picture and updated portrait archives is lower than similarity threshold.
In order to reduce the influence of portrait archives accuracy of the such case to formation, following embodiment can also be implemented.
As a kind of possible embodiment, if veritification object is portrait picture, and the portrait picture is based on and portrait shelves
The difference of the similarity of case is not archived in the portrait archives, then, in cleaning step S103, when the portrait that do not file
Picture is greater than or equal to the veritification similarity of the portrait archives and veritifies threshold value, then the portrait picture that do not file this is incorporated to this
Portrait archives, and it is based on this, update the portrait archives.
If veritification object is portrait picture, and the portrait picture has been archived in this based on the similarity with portrait archives
Portrait archives, then, the veritification similarity in cleaning step S103, when the filed portrait picture, with the portrait archives
It is rejected from the portrait archives less than threshold value is veritified then by the filed portrait picture, and is based on this, update the portrait shelves
Case.
If veritifying object is other portrait archives except first portrait archives, that is, mentioned above, it is counted as
Except " initial portrait archives ", other established portrait archives, then, in cleaning step S103, when two portrait shelves
The veritification similarity of case, which is greater than or equal to, veritifies threshold value, then merges amount portrait archives, and be based on this, update the portrait shelves
Case.By veritifying the veritification similarity between portrait archives and portrait archives again, can effectively reduce be under the jurisdiction of it is same
The portrait picture of portrait captured leads to a portion portrait due to objective factor, such as light, scene, angle reason
The characteristic value of picture and another part portrait picture differs greatly, by algorithm cluster to two or more portrait archives
In.That is, the occurrence of clustering cleaning method 100 by portrait, reducing the division of portrait archives.
As a kind of deformation, in veritifying step S102, veritifying object can be portrait picture, which can be
According to cluster calculation as a result, the portrait picture for the filing being already incorporated into portrait archives, is also possible to not be incorporated into portrait
The portrait picture that do not file in archives.In practical applications, can whithin a period of time, the portrait archives that will have been formed,
With all portrait pictures during this period of time obtained, again including the portrait picture filed and the portrait picture that do not file
It is veritified, to reduce the low situation of portrait image archive rate that the portrait archives that cluster is formed miscellany picture occur or obtain
Occur.
Correspondingly, in practical applications, in veritifying step S102, veritify object can also include simultaneously portrait picture and
In addition to above-mentioned portrait archives, other established portrait archives.In a kind of possible embodiment, in addition to including above-mentioned obtain
It takes step S101, veritify other than step S102 and cleaning step S103, after cleaning step S103, portrait clusters cleaning method
100 further include updating step S104.
As shown in Fig. 2, including in the portrait archives according to obtained from by cleaning step S103 in updating step S104
Portrait picture the archive feature value of the portrait archives is updated by fusion calculation method.It, can be with by updating step S104
The archive feature value for obtaining the portrait archives after over cleaning, enables the archive feature value to be formed more to represent portrait archives
Truth, but also a series of judgements that the later period is made based on the archive feature value, for example, being based on established portrait shelves
The archive feature value of case is veritified to object is veritified, can be more accurate.
In a kind of possible embodiment, veritifying step S102 further includes further determining that the veritification for veritifying step S102
Period, and veritified according to the period is veritified to object is veritified, to guarantee the accuracy of the portrait archives in the veritification period.Its
In, the length for veritifying the period can have more actual conditions and determine.
In practical applications, in a certain phase tail for veritifying the period, the portrait archives formed within the veritification period are obtained, and
By fusion calculation, the corresponding archive feature value of the portrait archives is obtained.Then according to the archive feature value to veritify object into
Row is veritified, and is judged the veritification similarity and is veritified the size relation of threshold value.
Further, veritifying the period includes that picture veritifies period and archives veritification period, wherein the picture veritification period is to use
The period that object is portrait picture is veritified to veritify;The archives veritification period is to veritify object for veritifying as in addition to portrait archives
Other portrait archives period.Wherein, the picture veritification period can be shorter than the archives veritification period.Since archives veritification can be
It being carried out on the basis of picture is veritified, the veritification for archives can be and further clean established portrait archives, because
This, the time in period is veritified by extending archives, on the basis of the accuracy for the portrait archives for guaranteeing to be formed, can effectively be subtracted
The calculation amount of few portrait cluster cleaning method 100.
It is veritified in the period in picture, by veritifying step S102, the picture for treating veritification is veritified, and is tied according to veritifying
Fruit cleans portrait archives, so that the portrait archives formed are more nearly the corresponding portrait of portrait archives.
For example, the portrait archives that formation is updated in the period will be veritified in the picture if it is one day that picture, which veritifies the period, with
All portrait pictures for obtaining in the period are veritified in the picture, portrait picture including filing and the portrait picture progress that do not file
Secondary veritification cluster calculation one by one.If the portrait picture of part filing and the veritification similarity of portrait archives, which are less than, veritifies threshold
Value, then remove the portrait picture that the part is filed from portrait archives;If portrait picture and portrait archives that part is not filed
Veritification similarity be greater than or equal to veritify threshold value, then the portrait picture that do not file of the part is incorporated to portrait archives.In turn
It ensure that the archives purity of the portrait archives, and improve the filing rate of the portrait picture of acquisition.
As a kind of possible embodiment, veritified in the period in archives, it can also be by veritifying step S102, to complete
After veritifying the veritification and cleaning that object is portrait picture, and obtained portrait archives, with other in addition to the portrait archives
Portrait archives carry out archives veritification.That is, veritified in the period in archives, it can be by the shelves of all portrait archives of formation
Pattern characteristics value compares, if the veritification similarity between certain two portrait archives, which is greater than or equal to, veritifies threshold value, then it is assumed that
The two corresponding portraits of portrait archives are same person's picture, therefore the two portrait archives can be classified as the same person as shelves
Case, and then the occurrence of archives divide can be effectively reduced.
As a kind of possible embodiment, when veritifying object is portrait picture, since the portrait picture veritified includes
According to similarity calculation, it can be incorporated to the portrait picture of portrait archives, filing portrait picture can be referred to as here;It is veritified
Portrait picture further includes the portrait picture that portrait archives can not be incorporated to according to similarity calculation, can be referred to as not return here
Shelves portrait picture.In veritifying step S102, in order to improve veritification accuracy, portrait picture and portrait archives are not filed in judgement
Veritification similarity when, can properly increase veritify threshold value size.Therefore, in practical applications, threshold value point will can be veritified
Threshold value and second is veritified for first and veritifies threshold value, wherein the first veritification threshold value is for judging to file portrait picture and portrait shelves
The veritification similarity of case;Second veritification threshold value is the veritification similarity for judging not file portrait picture Yu portrait archives
's.In practical applications, the second veritification threshold value can be enabled to be greater than first and veritify threshold value, meanwhile, the second veritification threshold value is also greater than core
Test threshold value, do not file the permit standard that portrait picture is incorporated to portrait archives by improving, come improve formation portrait archives standard
Exactness.
It is emphasized that filing portrait picture is initially archived in portrait archives in addition to including being clustered based on similarity
Portrait picture other than, further include that established people can be equally archived in real-time archiving process based on similarity calculation
As the portrait picture of archives;Correspondingly, not filing portrait picture in addition to including clustering based on similarity, in initial archiving process
In, it cannot be archived in other than the portrait picture in portrait archives, further include being based on similarity calculation, in real-time archiving process,
The portrait picture that cannot be equally archived in portrait archives.
It should also be noted that, since the first veritification threshold value is the veritification for judging to file portrait picture and portrait archives
Similarity, therefore, usually from the point of view of, first veritify threshold value can with veritify threshold size it is consistent.It in practical applications, can be with
Difference according to the actual situation, to veritify threshold value for reference, to be adjusted correspondingly to the first veritification threshold value.
As a kind of possible embodiment, in addition to including above-mentioned obtaining step S101, veritifying step S102, cleaning step
Except S103 and update step S104, after obtaining step S101, portrait cluster cleaning method 100 further includes real-time filing step
Rapid S105.As shown in figure 3, in filing step S105 in real time current portrait picture can be obtained in real time, and according to current portrait
The picture feature value of picture and the archive feature value of portrait archives, obtain real-time similarity, if similarity is greater than or equal in real time
Current portrait picture is then incorporated to portrait archives by real time threshold.In practical applications, the size of real time threshold can be according to reality
It needs, to veritify threshold value as reference, is adjusted correspondingly, it is, in principle, that real time threshold, is similarly and veritifies the one of threshold value
Kind.
Further, real time threshold includes the first real time threshold and the second real time threshold, wherein the first real time threshold is less than
Second real time threshold, also, the first real time threshold, lower than threshold value is veritified, the second real time threshold, which is greater than, veritifies threshold value.
In real-time archiving process, it is also necessary to obtain the acquisition device ID of current portrait picture and previous be incorporated to portrait
The acquisition device ID of portrait picture in archives.If the acquisition device ID of current portrait picture, is incorporated to portrait archives with previous
In portrait picture acquisition device ID it is identical, if then in real time similarity be greater than or equal to the first real time threshold, forefathers will be worked as
As picture is incorporated to portrait archives.It should be noted that " the previous portrait picture being incorporated in portrait archives " refers to, and work as forefathers
As picture is compared, filing portrait picture in portrait archives is incorporated to be newest.
Due to current portrait picture and the previous portrait picture being incorporated in portrait archives be by same acquisition device, and
It is collected within the time closed on, then it is reasonable that the two portrait pictures, which are likely to be, is under the jurisdiction of same portrait shelves
Case.It is therefore possible to use threshold value of the numerical values recited lower than veritification threshold value, that is, the first real time threshold, to current portrait picture
Similarity analysis is carried out with portrait archives.If current portrait picture is the portrait picture for being under the jurisdiction of portrait archives, lower core
Test threshold value can the low portrait picture of guarantee section similarity normally filed;If current portrait picture is not to be under the jurisdiction of the portrait
The portrait picture of archives, then since the characteristic value of portrait picture and portrait archives itself differs greatly, it will not be because of core
It is lower to test threshold value, and current portrait picture is caused to be filed by mistake.
Correspondingly, in real-time archiving process, if the acquisition device ID of current portrait picture, is incorporated to portrait shelves with previous
The acquisition device ID of portrait picture in case is different, will be current if then similarity is greater than or equal to the second real time threshold in real time
Portrait picture is incorporated to portrait archives.It is by not with the previous portrait picture being incorporated in portrait archives due to current portrait picture
It is collected with acquisition device, it is therefore possible to use numerical values recited is higher than the threshold value for veritifying threshold value, that is, the second real-time threshold
Value carries out similarity analysis to current portrait picture and portrait archives.Higher veritification threshold value, that is, the second real time threshold energy
Enough accuracy for guaranteeing to file obtained portrait archives in real time, to reduce miscellany picture, that is, are not belonging to the portrait archives
Portrait picture, but situation about being present in the portrait archives occurs.
As a kind of possible embodiment, in real-time archiving process, need to obtain the acquisition time of current portrait picture
With the acquisition device ID of current portrait picture, with the acquisition device ID of the previous portrait picture being incorporated in portrait archives and previous
The acquisition time of a portrait picture being incorporated in portrait archives.
The acquisition device ID, the acquisition device ID with the previous portrait picture being incorporated in portrait archives of current portrait picture
It is identical, if the acquisition time of current portrait picture, the time with the acquisition time of the previous portrait picture being incorporated in portrait archives
Interval is less than or equal to time threshold, if then similarity is greater than or equal to the first real time threshold in real time, by current portrait picture
It is incorporated to portrait archives.
By the way that time threshold is arranged, that is to say, that if the acquisition time of current portrait picture, is incorporated to portrait shelves with previous
The acquisition time of portrait picture in case, time interval between the two is shorter, is less than or equal to time threshold, then has reason to recognize
For, there are biggish probability, current portrait picture and the previous portrait picture being incorporated in the portrait archives be under the jurisdiction of it is same
The portrait picture of portrait archives.It is therefore possible to use numerical values recited is lower than the threshold value for veritifying threshold value, that is, the first real-time threshold
Value carries out similarity analysis to current portrait picture and portrait archives.
It is alternatively possible, the acquisition device ID of current portrait picture, with the previous portrait picture being incorporated in portrait archives
Acquisition device ID it is identical, if the acquisition time of current portrait picture, with adopting for the previous portrait picture being incorporated in portrait archives
The time interval for collecting the time is greater than time threshold, if then similarity is greater than or equal to the second real time threshold in real time, will work as forefathers
As picture is incorporated to portrait archives.That is, if the acquisition time of current portrait picture, is incorporated in portrait archives with previous
The acquisition time of portrait picture, time interval between the two is too long, and is greater than time threshold, then it is reasonable that, works as forefathers
As picture and the previous portrait picture being incorporated in portrait archives are not necessarily under the jurisdiction of the portrait picture of same portrait archives.
It is therefore possible to use numerical values recited is higher than the threshold value for veritifying threshold value, that is, the second real time threshold, to current portrait picture and people
As archives carry out similarity analysis.To be further ensured that the accuracy for the portrait archives to be formed.
Correspondingly, if the acquisition device ID of current portrait picture, with the previous portrait picture being incorporated in portrait archives
Acquisition device ID is different, if then similarity is greater than or equal to the second real time threshold in real time, current portrait picture is incorporated to portrait
Archives.It with the previous portrait picture being incorporated in portrait archives is acquired by different acquisition device due to current portrait picture
Arrive, therefore the threshold value for veritifying threshold value can be higher than using numerical values recited, that is, the second real time threshold, to current portrait picture and
Portrait archives carry out similarity analysis.Higher veritification threshold value, that is, the second real time threshold can guarantee that filing obtains in real time
Portrait archives accuracy, to reduce miscellany picture situation present in the portrait archives.
Based on identical inventive concept, the embodiment of the present disclosure also provides a kind of portrait cluster cleaning device 200.Such as Fig. 4 institute
Show, portrait cluster cleaning device 200 includes obtaining module 201, veritifies module 202 and cleaning module 203.
Module 201 is obtained, is clustered for obtaining portrait picture, and according to the similarity of portrait picture, portrait is formed
Archives;Module 202 is veritified, is veritified for the archive feature value based on portrait archives to object is veritified, judges to veritify object
With the veritification similarity of portrait archives, and the size relation of veritification threshold value, wherein veritify object and include portrait picture and/or remove
Other portrait archives except portrait archives;Cleaning module 203 is used for according to veritification as a result, portrait archives are cleaned, if veritifying phase
It is greater than or equal to like degree and veritifies threshold value, then veritify object and be incorporated to portrait archives, veritifies threshold value if veritifying similarity and being less than, veritify
Object is rejected from the portrait archives.In one example, portrait cluster cleaning device 200 further includes that update module 204 is used for
According to the portrait picture for including in portrait archives, the archive feature value of portrait archives is updated.
In one example, it veritifies step 202 to be also used to, determines the veritification period for veritifying step, according to the veritification period to veritification
Object is veritified.
In one example, veritifying the period includes that picture veritifies period and archives veritification period, and picture veritifies the period to veritify core
Test the period that object is portrait picture;It is that veritify object be other people in addition to the portrait archives that archives, which veritify the period,
As the period of archives;The picture veritification period is shorter than archives and veritifies the period.
In one example, portrait picture does not file including filing portrait picture and portrait picture, wherein filing portrait picture is
The portrait picture of portrait archives has been formed or has been incorporated to according to similarity;It is not formed or simultaneously according to similarity for not filing portrait picture
Enter the portrait picture of portrait archives;Veritifying threshold value includes that the first veritification threshold value and second veritify threshold value, wherein first veritifies threshold value
For judging to file the veritification threshold value of the veritification similarity of portrait picture and portrait archives;Second veritification threshold value is not returned for judging
The veritification threshold value of the veritification similarity of shelves portrait picture and portrait archives;First, which veritifies threshold value, veritifies threshold value less than second.
In one example, portrait cluster cleaning device 200 further includes real-time profiling module 205, works as forefathers for obtaining in real time
Picture picture obtains real-time similarity according to the archive feature value of the picture feature value of current portrait picture and portrait archives, if real
When similarity be greater than or equal to real time threshold, then current portrait picture is incorporated to portrait archives.
In one example, real time threshold includes the first real time threshold and the second real time threshold, and the second real time threshold is greater than first
Real time threshold;Real-time profiling module 205 is also used to: being obtained the acquisition device ID of current portrait picture and previous is incorporated to portrait
The acquisition device ID of portrait picture in archives;If the acquisition device ID of current portrait picture, is incorporated to portrait archives with previous
In portrait picture acquisition device ID it is identical, if then in real time similarity be greater than or equal to the first real time threshold, forefathers will be worked as
As picture is incorporated to portrait archives;If the acquisition device ID of current portrait picture, with the previous portrait figure being incorporated in portrait archives
The acquisition device ID of piece is different, if then similarity is greater than or equal to the second real time threshold in real time, current portrait picture is incorporated to
Portrait archives.
In one example, real time threshold includes the first real time threshold and the second real time threshold, and the second real time threshold is greater than first
Real time threshold;Real-time profiling module 205 is also used to: obtaining the acquisition of the acquisition time and current portrait picture of current portrait picture
Device ID, acquisition device ID and the previous people being incorporated in portrait archives with the previous portrait picture being incorporated in portrait archives
As the acquisition time of picture;The acquisition device ID of current portrait picture, with adopting for the previous portrait picture being incorporated in portrait archives
Acquisition means ID is identical, if the acquisition time of current portrait picture, when acquisition with the previous portrait picture being incorporated in portrait archives
Between time interval be less than or equal to time threshold, if then in real time similarity be greater than or equal to the first real time threshold, will be current
Portrait picture is incorporated to portrait archives;The acquisition device ID of current portrait picture, with the previous portrait picture being incorporated in portrait archives
Acquisition device ID it is identical, if the acquisition time of current portrait picture, with adopting for the previous portrait picture being incorporated in portrait archives
The time interval for collecting the time is greater than time threshold, if then similarity is greater than or equal to the second real time threshold in real time, will work as forefathers
As picture is incorporated to portrait archives;If the acquisition device ID of current portrait picture, with the previous portrait figure being incorporated in portrait archives
The acquisition device ID of piece is different, if then similarity is greater than or equal to the second real time threshold in real time, current portrait picture is incorporated to
Portrait archives.
Fig. 5 shows a kind of electronic equipment 30 that an embodiment of the disclosure provides.As shown in figure 5, the disclosure
The a kind of electronic equipment 30 that one embodiment provides, wherein the electronic equipment 30 includes memory 310, processor 320, defeated
Enter/export (Input/Output, I/O) interface 330.Wherein, memory 310, for storing instruction.Processor 320, for adjusting
Cleaning method is clustered with the instruction execution disclosure portrait that memory 310 stores.Wherein, processor 320 respectively with memory
310, I/O interface 330 connects, such as can be attached by bindiny mechanism's (not shown) of bus system and/or other forms.
Memory 310 can be used for storing program and data, the program including the cluster cleaning of portrait involved in the embodiment of the present disclosure, processing
Device 320 by operation be stored in the program of memory 310 thereby executing the various function application and data of electronic equipment 30 at
Reason.
Processor 320 can use digital signal processor (Digital Signal in the embodiment of the present disclosure
Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), programmable patrol
At least one of volume array (Programmable Logic Array, PLA) example, in hardware realizes, the processor 320
It can be central processing unit (Central Processing Unit, CPU) or there is data-handling capacity and/or instruction
The combination of one or more of the processing unit of other forms of executive capability.
Memory 310 in the embodiment of the present disclosure may include one or more computer program products, the computer
Program product may include various forms of computer readable storage mediums, such as volatile memory and/or non-volatile deposit
Reservoir.The volatile memory for example may include random access memory (Random Access Memory, RAM) and/
Or cache memory (cache) etc..The nonvolatile memory for example may include read-only memory (Read-Only
Memory, ROM), flash memory (Flash Memory), hard disk (Hard Disk Drive, HDD) or solid state hard disk
(Solid-State Drive, SSD) etc..
In the embodiment of the present disclosure, I/O interface 330 can be used for receiving input instruction (such as number or character information, and
Generate key signals input related with the user setting of electronic equipment 30 and function control etc.), it can also be output to the outside various
Information (for example, image or sound etc.).In the embodiment of the present disclosure I/O interface 330 may include physical keyboard, function button (such as
Volume control button, switch key etc.), mouse, operating stick, trace ball, microphone, one in loudspeaker and touch panel etc.
It is a or multiple.
In some embodiments, present disclose provides a kind of computer readable storage medium, the computer-readable storages
Media storage has computer executable instructions, and computer executable instructions when executed by the processor, execute described above appoint
Where method.
Although description operation in a particular order in the accompanying drawings should not be construed as requiring specific shown in
Sequence or serial order operate to execute these operations, or shown in requirement execution whole to obtain desired result.In
In specific environment, multitask and parallel processing be may be advantageous.
Disclosed method and device can be completed using standard programming technology, using rule-based logic or its
His logic realizes various method and steps.It should also be noted that herein and the terms used in the claims " device "
" module " is intended to include using the realization of a line or multirow software code and/or hardware realization and/or for receiving input
Equipment.
One or more combined individually or with other equipment can be used in any step, operation or program described herein
A hardware or software module are executed or are realized.In one embodiment, software module use includes comprising computer program
The computer program product of the computer-readable medium of code is realized, can be executed by computer processor any for executing
Or whole described step, operation or programs.
For the purpose of example and description, the preceding description of disclosure implementation is had been presented for.Preceding description is not poor
The disclosure is restricted to exact form disclosed by also not the really wanting of act property, according to the above instruction there is likely to be various modifications and
Modification, or various changes and modifications may be obtained from the practice of the disclosure.Select and describe these embodiments and be in order to
Illustrate the principle and its practical application of the disclosure, so that those skilled in the art can be to be suitable for the special-purpose conceived
Come in a variety of embodiments with various modifications and using the disclosure.
Claims (12)
1. a kind of portrait clusters cleaning method, wherein the described method includes:
Obtaining step obtains portrait picture, and is clustered according to the similarity of the portrait picture, forms portrait archives;
Step is veritified, the archive feature value based on the portrait archives is veritified to object is veritified, and judges the veritification object
With the veritification similarity of the portrait archives, and the size relation of veritification threshold value, wherein the veritification object includes portrait picture
And/or other portrait archives in addition to the portrait archives;
Cleaning step, according to the veritification as a result, cleaning the portrait archives.
2. according to the method described in claim 1, wherein, after the obtaining step, the method also includes:
In real time filing step, obtain current portrait picture in real time, according to the picture feature value of the current portrait picture with it is described
The archive feature value of portrait archives, obtains real-time similarity, if the real-time similarity is greater than or equal to real time threshold,
The current portrait picture is incorporated to the portrait archives.
3. method according to claim 1 or 2, wherein the cleaning step includes:
If the veritification object is portrait picture, and the portrait picture is not archived in the portrait archives, then, when the veritification
When similarity is greater than or equal to the veritification threshold value, the portrait picture is incorporated to the portrait archives;
If the veritification object is portrait picture, and the portrait picture has been archived in the portrait archives, then, when the veritification
When similarity is less than the veritification threshold value, the portrait picture is rejected from the portrait archives;
If the veritification object be in addition to the portrait archives described in other portrait archives, when the veritification similarity
When more than or equal to the veritification threshold value, two portrait archives are merged.
4. method according to claim 1 or 2, wherein after the cleaning step, the method also includes:
It updates step and the archive feature of the portrait archives is updated according to the portrait picture for including in the portrait archives
Value.
5. method according to claim 1 or 2, wherein the veritification step further include:
It determines the veritification period for veritifying step, the veritification is carried out to the veritification object according to the veritification period.
6. according to the method described in claim 5, wherein,
The veritification period includes that picture veritifies period and archives veritification period, and the picture veritifies the period to veritify the veritification
Object is the period of the portrait picture;It is that veritify the veritifications object be described to remove the portrait shelves that the archives, which veritify the period,
The period of other portrait archives except case;
The picture veritification period is shorter than the archives and veritifies the period.
7. method described in any one of -6 claims according to claim 1, wherein
The portrait picture is including filing portrait picture and does not file portrait picture, wherein according to the filing portrait picture
Similarity has formed or has been incorporated to the portrait picture of the portrait archives;The portrait picture of not filing is not formed according to similarity
Or it is incorporated to the portrait picture of the portrait archives;
The veritification threshold value includes that the first veritification threshold value and second veritify threshold value, wherein described first veritifies threshold value as sentencing
The veritification threshold value for the veritification similarity for filing portrait picture and the portrait archives of breaking;Described second veritifies threshold
Value is the veritification threshold value for judging the veritification similarity for not filing portrait picture and the portrait archives;Institute
It states the first veritification threshold value and is less than the second veritification threshold value.
8. according to the method described in claim 2, wherein,
The real time threshold includes the first real time threshold and the second real time threshold, and it is real that first real time threshold is less than described second
When threshold value;
The real-time filing step further include: obtain the current portrait picture acquisition device ID and it is previous be incorporated to it is described
The acquisition device ID of the portrait picture in portrait archives;
If the acquisition device ID of the current portrait picture, with the previous portrait figure being incorporated in the portrait archives
The acquisition device ID of piece is identical, will be described current if then the real-time similarity is greater than or equal to first real time threshold
Portrait picture is incorporated to the portrait archives;
If the acquisition device ID of the current portrait picture, with the previous portrait figure being incorporated in the portrait archives
The acquisition device ID of piece is different, will be described current if then the real-time similarity is greater than or equal to second real time threshold
Portrait picture is incorporated to the portrait archives.
9. according to the method described in claim 2, wherein,
The real time threshold includes the first real time threshold and the second real time threshold, and it is real that first real time threshold is less than described second
When threshold value;
The real-time filing step further include: obtain the acquisition time and the current portrait picture of the current portrait picture
Acquisition device ID, and the acquisition device ID of the previous portrait picture being incorporated in the portrait archives and previous is incorporated to institute
State the acquisition time of the portrait picture in portrait archives;
The acquisition device ID of the current portrait picture, with the previous portrait picture being incorporated in the portrait archives
Acquisition device ID it is identical, if the acquisition time of the current portrait picture, previous be incorporated in the portrait archives with described
The portrait picture acquisition time time interval be less than or equal to time threshold, if then the real-time similarity be greater than or
Equal to first real time threshold, then the current portrait picture is incorporated to the portrait archives;
The acquisition device ID of the current portrait picture, with the previous portrait picture being incorporated in the portrait archives
Acquisition device ID it is identical, if the acquisition time of the current portrait picture, previous be incorporated in the portrait archives with described
The portrait picture acquisition time time interval be greater than time threshold, if then the real-time similarity be greater than or equal to institute
The second real time threshold is stated, then the current portrait picture is incorporated to the portrait archives;
If the acquisition device ID of the current portrait picture, with the previous portrait figure being incorporated in the portrait archives
The acquisition device ID of piece is different, will be described current if then the real-time similarity is greater than or equal to second real time threshold
Portrait picture is incorporated to the portrait archives.
10. a kind of portrait clusters cleaning device, wherein described device includes:
Module is obtained, is clustered for obtaining portrait picture, and according to the similarity of the portrait picture, portrait shelves are formed
Case;
Module is veritified, is veritified for the archive feature value based on the portrait archives to object is veritified, judges the veritification
The veritification similarity of object and the portrait archives, and veritify the size relation of threshold value, wherein the veritification object includes portrait
Picture and/or other portrait archives in addition to the portrait archives;
Cleaning module, for being veritified according to described as a result, the portrait archives are cleaned, if the veritification similarity is greater than or equal to
The veritification threshold value, then the veritification object is incorporated to the portrait archives, if the veritification similarity is less than the veritification threshold value,
Then the veritification object is rejected from the portrait archives.
11. a kind of electronic equipment, wherein the electronic equipment includes:
Memory, for storing instruction;And
Processor, portrait cluster described in any one of instruction execution claim 1-9 for calling memory storage are clear
Washing method.
12. a kind of computer readable storage medium, wherein
The computer-readable recording medium storage has computer executable instructions, and the computer executable instructions are by handling
When device executes, perform claim requires portrait described in any one of 1-9 to cluster cleaning method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910684614.9A CN110413815B (en) | 2019-07-26 | 2019-07-26 | Portrait clustering cleaning method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910684614.9A CN110413815B (en) | 2019-07-26 | 2019-07-26 | Portrait clustering cleaning method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110413815A true CN110413815A (en) | 2019-11-05 |
CN110413815B CN110413815B (en) | 2022-05-24 |
Family
ID=68363513
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910684614.9A Active CN110413815B (en) | 2019-07-26 | 2019-07-26 | Portrait clustering cleaning method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110413815B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111242040A (en) * | 2020-01-15 | 2020-06-05 | 佳都新太科技股份有限公司 | Dynamic face clustering method, device, equipment and storage medium |
CN111427873A (en) * | 2020-03-12 | 2020-07-17 | ***科技(杭州)有限公司 | Data cleaning method and system |
CN111522974A (en) * | 2020-04-26 | 2020-08-11 | 深圳惟远智能技术有限公司 | Real-time filing method and device |
CN112818867A (en) * | 2021-02-02 | 2021-05-18 | 浙江大华技术股份有限公司 | Portrait clustering method, equipment and storage medium |
CN114639143A (en) * | 2022-03-07 | 2022-06-17 | 北京百度网讯科技有限公司 | Portrait filing method, equipment and storage medium based on artificial intelligence |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105404897A (en) * | 2015-11-23 | 2016-03-16 | 昆明理工大学 | Method for cleaning solar photosphere bright spot in astronomical image |
US10133942B2 (en) * | 2016-07-05 | 2018-11-20 | Nauto Global Limited | System and method for automatic driver identification |
CN109145138A (en) * | 2018-09-10 | 2019-01-04 | 北京点网聚科技有限公司 | A kind of cover choosing method, device, electronic equipment and storage medium |
CN109241310A (en) * | 2018-07-25 | 2019-01-18 | 南京甄视智能科技有限公司 | The data duplicate removal method and system of face image database |
CN109710780A (en) * | 2018-12-28 | 2019-05-03 | 上海依图网络科技有限公司 | A kind of archiving method and device |
CN109800318A (en) * | 2018-12-28 | 2019-05-24 | 上海依图网络科技有限公司 | A kind of archiving method and device |
CN112307239A (en) * | 2020-10-29 | 2021-02-02 | 泰康保险集团股份有限公司 | Image retrieval method, device, medium and equipment |
-
2019
- 2019-07-26 CN CN201910684614.9A patent/CN110413815B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105404897A (en) * | 2015-11-23 | 2016-03-16 | 昆明理工大学 | Method for cleaning solar photosphere bright spot in astronomical image |
US10133942B2 (en) * | 2016-07-05 | 2018-11-20 | Nauto Global Limited | System and method for automatic driver identification |
CN109241310A (en) * | 2018-07-25 | 2019-01-18 | 南京甄视智能科技有限公司 | The data duplicate removal method and system of face image database |
CN109145138A (en) * | 2018-09-10 | 2019-01-04 | 北京点网聚科技有限公司 | A kind of cover choosing method, device, electronic equipment and storage medium |
CN109710780A (en) * | 2018-12-28 | 2019-05-03 | 上海依图网络科技有限公司 | A kind of archiving method and device |
CN109800318A (en) * | 2018-12-28 | 2019-05-24 | 上海依图网络科技有限公司 | A kind of archiving method and device |
CN112307239A (en) * | 2020-10-29 | 2021-02-02 | 泰康保险集团股份有限公司 | Image retrieval method, device, medium and equipment |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111242040A (en) * | 2020-01-15 | 2020-06-05 | 佳都新太科技股份有限公司 | Dynamic face clustering method, device, equipment and storage medium |
CN111242040B (en) * | 2020-01-15 | 2022-08-02 | 佳都科技集团股份有限公司 | Dynamic face clustering method, device, equipment and storage medium |
CN111427873A (en) * | 2020-03-12 | 2020-07-17 | ***科技(杭州)有限公司 | Data cleaning method and system |
CN111427873B (en) * | 2020-03-12 | 2023-03-14 | ***科技(杭州)有限公司 | Data cleaning method and system |
CN111522974A (en) * | 2020-04-26 | 2020-08-11 | 深圳惟远智能技术有限公司 | Real-time filing method and device |
CN112818867A (en) * | 2021-02-02 | 2021-05-18 | 浙江大华技术股份有限公司 | Portrait clustering method, equipment and storage medium |
CN112818867B (en) * | 2021-02-02 | 2024-05-31 | 浙江大华技术股份有限公司 | Portrait clustering method, equipment and storage medium |
CN114639143A (en) * | 2022-03-07 | 2022-06-17 | 北京百度网讯科技有限公司 | Portrait filing method, equipment and storage medium based on artificial intelligence |
CN114639143B (en) * | 2022-03-07 | 2024-04-16 | 北京百度网讯科技有限公司 | Portrait archiving method, device and storage medium based on artificial intelligence |
Also Published As
Publication number | Publication date |
---|---|
CN110413815B (en) | 2022-05-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110413815A (en) | Portrait clusters cleaning method and device | |
CN111090398B (en) | Garbage recycling method, device and equipment for solid state disk and readable storage medium | |
CN103890724B (en) | Information processing apparatus, method for controlling information processing apparatus, host device, and performance evaluation method used for external storage device | |
JP5934439B2 (en) | Event-based simulation of continuous-time spiking neural networks | |
CN105894464B (en) | A kind of medium filtering image processing method and device | |
US11734103B2 (en) | Behavior-driven die management on solid-state drives | |
CN108733579A (en) | The method and solid-state disk of the window of opportunity prompt of consistency operation in solid-state disk are provided | |
CN108701259A (en) | The system and method that the data handled using binary classifier carry out create-rule | |
US11550686B2 (en) | Adaptable online breakpoint detection over I/O trace time series via deep neural network autoencoders re-parameterization | |
CN106527966A (en) | Garbage collection in ssd drives | |
CN109978006B (en) | Face image clustering method and device | |
CN108734092A (en) | Personage's authentication device | |
WO2020259543A1 (en) | Hot data prediction method based on joint optimization of echo state network | |
Zhao et al. | Neuronal population reconstruction from ultra-scale optical microscopy images via progressive learning | |
KR102267487B1 (en) | Method and apparatus for operating independent classification model using metadata | |
CN113986561A (en) | Artificial intelligence task processing method and device, electronic equipment and readable storage medium | |
CN111639517A (en) | Face image screening method and device | |
CN115953430B (en) | Video single-target tracking method and system based on improved batch sample loss function | |
CN110874601B (en) | Method for identifying running state of equipment, state identification model training method and device | |
CN110321971A (en) | Image data cleaning method, image processing method and device | |
CN110060287A (en) | Facial image nose shaping methods and device | |
CN111091022A (en) | Machine vision efficiency evaluation method and system | |
CN110929706B (en) | Video frequency selecting method, device and storage medium | |
CN107846327A (en) | A kind of processing method and processing device of network management performance data | |
CN110909210A (en) | Video screening method and device 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 | ||
TR01 | Transfer of patent right |
Effective date of registration: 20230804 Address after: No. 1268, 1f, building 12, neijian Middle Road, Xisanqi building materials City, Haidian District, Beijing 100096 Patentee after: BEIJING KUANGSHI TECHNOLOGY Co.,Ltd. Address before: 100190 A block 2, South Road, Haidian District Academy of Sciences, Beijing 313 Patentee before: BEIJING KUANGSHI TECHNOLOGY Co.,Ltd. Patentee before: Qingdao Guangshi Technology Co.,Ltd. |
|
TR01 | Transfer of patent right |