CN109800672A - A kind of archiving method and device - Google Patents
A kind of archiving method and device Download PDFInfo
- Publication number
- CN109800672A CN109800672A CN201811626344.8A CN201811626344A CN109800672A CN 109800672 A CN109800672 A CN 109800672A CN 201811626344 A CN201811626344 A CN 201811626344A CN 109800672 A CN109800672 A CN 109800672A
- Authority
- CN
- China
- Prior art keywords
- archives
- filing
- face
- class center
- facial image
- 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.)
- Withdrawn
Links
Landscapes
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of archiving method and device, the methods are as follows: obtains multiple facial images in preset period of time；The online archives are for filing the facial image that monitoring device obtains in real time；Multiple described facial images are clustered, pre- filing archives are generated；Wherein, the similarity in any facial image and other pre- filing archives in any pre- filing archives between any facial image is less than first threshold；The similarity of pre- the filing archives and the first face archives is determined according to the similarity between all kinds of centers of the first face archives in all kinds of centers and offline profile of the pre- filing archives at least one pre- filing archives；If the similarity is greater than second threshold, the facial image in the pre- filing archives is included into the first face archives；The first face archives are filed any face archives in the offline profile.
Description
Technical field
The present invention relates to monitoring technology field more particularly to a kind of archiving methods and device.
Background technique
With the fast development of informationization technology, monitoring technology has obtained answering extensively in various industries, various places
With providing many conveniences for the routine work and life of people.But there is also many to ask for monitoring application model single simultaneously
Topic, the analysis of image content for monitoring network are also needed by artificial viewing in real time, are had access to manually.
With the expansion of monitoring network size, video data magnanimity increases, and the manpower that image supervised and had access to occupancy is very big,
It is more and more difficult that useful information or information are obtained from the monitor video of magnanimity.
In addition, currently to the application of the monitor video only reproduction and playback to monitoring scene mostly, it is difficult to meet modern
The demand of public security prevention and control application.
Summary of the invention
The present invention provides a kind of archiving method and device, the archive information of the facial image to improve monitoring device acquisition
Accuracy, help to obtain useful information or information from the monitor video of magnanimity, to adapt to the needs of modern security protection.
Specific technical solution provided in an embodiment of the present invention is as follows:
The embodiment of the present invention provides a kind of archiving method, comprising:
From multiple facial images obtained in online archives in preset period of time；The online archives are used for monitoring device reality
When the facial image that obtains filed；
Multiple described facial images are clustered, pre- filing archives are generated；Wherein, any in any pre- filing archives
It opens similarities in facial image and other pre- filing archives between any facial image and is less than first threshold；
For at least one pre- filing archives, according to first in all kinds of centers and offline profile of the pre- filing archives
Similarity between all kinds of centers of face archives determines the similarity of pre- the filing archives and the first face archives；If
The similarity is greater than second threshold, then the facial image in the pre- filing archives is included into the first face archives；Institute
Stating the first face archives is filed any face archives in the offline profile；The class center is according to the people in archives
What face image determined.
A kind of possible implementation, all kinds of centers according to the pre- filing archives and first in offline profile
Similarity between all kinds of centers of face archives determines the similarity of pre- the filing archives and the first face archives, packet
It includes:
Determine it is described it is pre- filing archives each class center and the first face archives each class center between class between
Similarity；
According between the weight at each class center of the pre- filing archives, each class center of the first face archives
Weight and all kinds of similarities determine the similarity of pre- the filing archives and the first face archives.
A kind of possible implementation, the similarity C of pre- filing archives and the first face archives is according to following
Formula determines:
Wherein, Ai is the weight at i-th of class center of pre- filing archives；N≥i≥1；Bj is the jth of the first face archives
The weight at a class center；M≥j≥1；Ci, j are pre- i-th of class center for filing archives and the first face archives
Similarity between j-th of class center；N, M is the positive integer greater than 1.
The weight of a kind of possible implementation, the class center determines in the following manner, comprising:
The attribute information at the class center is inputted into supervised learning model, determines the weight at the class center；The attribute
Information includes at least one of and combination: picture quality, the temporal information degree of correlation or the spatial information degree of correlation.
A kind of possible implementation, the method also includes:
If it is determined that described in the similarity of all people's face archives is respectively less than in the pre- filing archives and the offline profile
Second threshold then establishes the second face archives for the pre- filing archives, and using the class center of the pre- filing archives as institute
State the class center of the second face archives.
The embodiment of the present invention provides a kind of filing apparatus, comprising:
Acquiring unit, for from online archives obtain preset period of time in multiple facial images；The online archives are used
File in the facial image that monitoring device obtains in real time；
Processing unit generates pre- filing archives for clustering to multiple described facial images；Wherein, any pre- to return
Similarity in any facial image and other pre- filing archives in shelves archives between any facial image is less than first
Threshold value；For at least one pre- filing archives, according to all kinds of centers of the pre- filing archives with it is the first in offline profile
Similarity between all kinds of centers of face archives determines the similarity of pre- the filing archives and the first face archives；If institute
Similarity is stated greater than second threshold, then the facial image in the pre- filing archives is included into the first face archives；It is described
First face archives are filed any face archives in the offline profile；The class center is according to the face in archives
What image determined.
A kind of possible implementation, the processing unit are specifically used for:
Determine it is described it is pre- filing archives each class center and the first face archives each class center between class between
Similarity；According between the weight at each class center of the pre- filing archives, each class center of the first face archives
Weight and all kinds of similarities determine the similarity of pre- the filing archives and the first face archives.
A kind of possible implementation, the similarity C of pre- filing archives and the first face archives is according to following
Formula determines:
Wherein, Ai is the weight at i-th of class center of pre- filing archives；N≥i≥1；Bj is the jth of the first face archives
The weight at a class center；M≥j≥1；Ci, j are pre- i-th of class center for filing archives and the first face archives
Similarity between j-th of class center；N, M is the positive integer greater than 1.
The weight of a kind of possible implementation, the class center determines in the following manner, comprising:
The attribute information at the class center is inputted into supervised learning model, determines the weight at the class center；The attribute
Information includes at least one of and combination: picture quality, the temporal information degree of correlation or the spatial information degree of correlation.
A kind of possible implementation, the processing unit are also used to:
If it is determined that described in the similarity of all people's face archives is respectively less than in the pre- filing archives and the offline profile
Second threshold then establishes the second face archives for the pre- filing archives, and using the class center of the pre- filing archives as institute
State the class center of the second face archives.
The embodiment of the present invention provides a kind of archival device, including at least one processing unit and at least one storage list
Member, wherein the storage unit is stored with computer program, when described program is executed by the processing unit, so that described
Processing unit executes the step of either method provided in an embodiment of the present invention.
The embodiment of the present invention provides a kind of computer-readable medium, is stored with the computer journey that can be executed by archival device
Sequence, when described program is run in archival device, so that the archival device executes the embodiment of the present invention and provides either method
The step of.
Based on the above-mentioned technical proposal, by the way that multiple class centers are arranged, and weight is set for each class center, it can be effective
The accuracy for improving the archive information for the facial image that monitoring device obtains, facilitates from the monitor video of magnanimity to obtain useful
Information or information, to adapt to the needs of modern security protection.
Detailed description of the invention
Fig. 1 is monitoring system configuration diagram in the embodiment of the present invention；
Fig. 2 is archiving method flow diagram in the embodiment of the present invention；
Fig. 3 is archiving method flow diagram in the embodiment of the present invention；
Fig. 4 is archiving method flow diagram in the embodiment of the present invention；
Fig. 5 is the structural schematic diagram of filing apparatus in the embodiment of the present invention；
Fig. 6 is the structural schematic diagram of archival device in the embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with attached drawing to the present invention make into
It is described in detail to one step, it is clear that described embodiments are only a part of the embodiments of the present invention, rather than whole implementation
Example.Based on the embodiments of the present invention, obtained by those of ordinary skill in the art without making creative efforts
All other embodiment, shall fall within the protection scope of the present invention.
Archiving method in the embodiment of the present application can be applied to security protection, using the facial image of monitoring device acquisition as base
Plinth establishes facial image database, includes multiple face archives in facial image database, and an archives correspond to a people, wherein face shelves
The archiving method of case can be divided into online elite archiving and offline filing according to archiving method.Online elite archiving is for monitoring device
The method that the facial image acquired in real time is filed, offline filing for periodically to the facial image acquired in set period of time into
The method of row filing.For convenience of description, online archives are known as during the face archives online elite archiving in the embodiment of the present invention, this
Face archives in inventive embodiments are known as offline profile in offline archiving process.Offline filing can be used for updating returns online
Shelves.Include real name archives and non-real name archives in online archives, also include real name archives and non-real name archives in offline profile,
In, it include personal identification information in real name archives, for example include identity document in real name archives.
When alert occurs, police can with the face archives in the picture search facial image database of suspect, from
The corresponding face archives of middle acquisition suspect, analyze suspect according to the facial image in the corresponding face archives of suspect later
Activity trajectory, frequent activities the information such as regional, most emerging position, consequently facilitating police arrest suspect.
It include that monitoring is set in the system architecture if Fig. 1 illustrates the applicable system architecture of the embodiment of the present application
Standby 101-111, server 102.Monitoring device 101-111 acquires video flowing in real time, then by the video stream of acquisition to clothes
It is engaged in device 102, includes filing apparatus in server 102, server 102 obtains facial image to be archived from video flowing, then
Facial image to be archived is included into corresponding face archives.Monitoring device 101-111 is connected by wireless network and server 102
It connects, monitoring device is the electronic equipment, such as camera, video camera, video recorder etc. for having acquisition image function.Server 102
It is the server cluster or cloud computing center of a server or several servers composition.
Based on system architecture shown in FIG. 1, the embodiment of the present application provides a kind of process of archiving method, the stream of this method
Journey can be executed by filing apparatus, and filing apparatus can be server 102 shown in FIG. 1, as shown in Figure 2, comprising the following steps:
Step 201: obtaining the first facial image.
Wherein, the first facial image is online facial image to be archived；Online facial image is the prison in operating conditions
The facial image that control equipment obtains in real time.
Step 202: if it is determined that the class center of the first face archives in online archives and the phase of first facial image
It is greater than or equal to first threshold like degree, then first facial image is included into the first face archives.
It wherein, include at least one online archives in facial image database, online archives are for real-time to online facial image
Filing, online archives for filing online facial image in real time；First face archives are any face shelves in online archives
Case；The class center of first face archives is determined according to the facial image in the first face archives.
By above-mentioned archiving method, the archive information for the facial image that monitoring device obtains in real time can be determined in real time, is had
It conducive to the supervision for fast implementing image and has access to, helps to obtain useful information or information from the monitor video of magnanimity,
To adapt to the needs of modern security protection.
In step 202, the class center of the first face archives can be determined according to various ways, as shown in figure 3, of the invention
Embodiment provides a kind of archiving method, comprising:
Step 301: obtaining the first face archives；
Wherein, the first face archives are any face archives in facial image database；Appointing in any face archives
Similarity in one facial image and other face archives between any facial image is less than first threshold；
Step 302: according to an at least facial image for the first face archives, determining the first face archives
Class center；Whether the facial image that the class center of the first face archives is used to determine that monitoring device obtains is included into described first
Face archives.
It is illustrated below according to application scenarios.
The first scene, the first face archives are the face archives created according to the first facial image, in archives only at this time
There are a facial images, can be using the first facial image as the class center of the first face archives.
Similarly, however, it is determined that the class center of any face archives in the online archives and the phase of first facial image
It is both less than the first threshold like degree, then establishes the second face archives in online archives, and first facial image is made
For the class center of the second face archives.
Second of scene includes at least 2 facial images in the first face archives, to improve the efficiency of online elite archiving and returning
Shelves accuracy, a kind of possible implementation, can using the highest facial image of picture quality in the first face archives as
The timing node at the class center of the first face archives, the specific class center for updating the first face archives can be according to specific needs
It determines.
It should be noted that picture quality can be according to the pixel of facial image, clarity, the shooting angle of facial image
Degree, the comprehensive determination of effect of feature extraction etc., it is not limited here.
A kind of possible implementation, can be when setting period arrival, for any face in the online archives
Archives update the class center of the face archives.
A kind of possible implementation can update the class center of face archives, specifically after updating facial image
, may include:
If it is determined that the picture quality of first facial image is greater than the image matter at the class center of the first face archives
Amount, then be updated to first facial image for the class center of the first face archives.
In a kind of possible embodiment, the class center of online archives is one.By such center and the first face figure
Similarity between the class of picture, the similarity being determined as between online archives and the first facial image.
In specific implementation, when determining similarity between class center and the class of the first facial image, it can determine in class first
Then the heart calculates first eigenvector to the first eigenvector of application and the corresponding second feature vector of the first facial image
The distance between second feature vector, according to the distance between first eigenvector and second feature vector determine class center with
Similarity between the class of first facial image.
In alternatively possible embodiment, the class centers of online archives can be it is multiple, determine the first facial image
The similarity between the class at all kinds of centers of online archives, then according to the weight at all kinds of centers of online archives and all kinds of centers
Class between similarity, determine the similarity of online archives and the first facial image.
In specific implementation, multiple class centers of online archives can be determined using following methods, comprising:
Step 1: presetting the quantity at the class center of online archives；
Step 2: the online highest facial image of archival image quality is determined as first kind center；
Step 3: when receiving the first facial image for being included into online archives, however, it is determined that second facial image
Picture quality is greater than third threshold value, it is determined that the similarity at first kind center and the first facial image；
Step 4: if it is determined that the first facial image is determined as the second class center when the similarity is less than second threshold；
Step 5: the first facial image for being included into online archives for next executes above-mentioned steps, until online archives
The data at class center reach pre-set quantity.
Due to previously according to facial image in online archives picture quality and characteristics of image selected from each facial image
Take class center, therefore when filing to the first facial image, only need to by the class center of the first facial image and online archives into
Row compares, determining with the matched online archives of the first facial image, without will in the first facial image and online archives it is every
It opens facial image to be compared, to improve filing efficiency.
When determining similarity between the first facial image and the class at multiple class centers, the first face figure can be first determined respectively
As and similarity between the class at each class center, then according to the weight at each class center by the first facial image and each
Similarity is merged between the class at class center, determines the similarity of online archives and the first facial image, the weight at class center can
To be allocated according to picture quality, can also be allocated according to characteristics of image.
In the specific implementation process, all kinds of centers according to the first face archives and first facial image
Class between similarity, determine the similarity of the first face archives Yu first facial image, comprising:
Step 1: determining respectively similar between first facial image and the class at all kinds of centers of the first face archives
Degree；
Step 2: according to similar between the weight at all kinds of centers of the first face archives and the class at all kinds of centers
Degree determines the similarity for stating the first face archives Yu first facial image.
A kind of possible implementation, the similarity C of first facial image and the first face archives according to
Lower formula determines:
Wherein, Bj is the weight at j-th of class center of the first face archives；M≥j≥1；M is the positive integer greater than 1；Cj
Similarity between class between first facial image and j-th of class center of the first face archives.
A kind of possible implementation, what the weight at all kinds of centers was determined according to following methods, comprising:
For each class center in the first face archives, the attribute information at the class center is inputted into supervised learning
Model determines the weight at the class center；The attribute information includes at least one of and combination: picture quality, time letter
Cease the degree of correlation or the spatial information degree of correlation.
In step 202, specific embodiment may include following two implementation:
A kind of possible implementation, it is described if it is determined that the class center of the first face archives in online archives and described the
The similarity of one facial image is greater than or equal to first threshold, then first facial image is included into the first face archives
In, comprising:
Step 1: all face archives in the traversal online archives, the class center of all face archives and the is determined
The similarity of one facial image；
Step 2: if it is determined that the class center of the first face archives of at least one of described online archives with it is described the first
The similarity of face image is greater than or equal to the first threshold, then chooses similarity at least one described first face archives
Highest first face archives；
Step 3: first facial image is included into the highest first face archives of the similarity.
It for example, include the class center 1 of face archives 1, the class center 2 of face archives 2, face archives in online archives
3 class center 3 determines the similarity 1 of the first facial image Yu class center 1, determines that the first facial image is similar to class center 2
Degree 2, determines the similarity 3 of the first facial image Yu class center 3；If it is determined that similarity 1 and similarity 2 are both greater than first threshold,
And similarity 1 is greater than similarity 2, then the first facial image is included into face archives 1.
A kind of possible implementation, it is described if it is determined that the class center of the first face archives in online archives and described the
The similarity of one facial image is greater than or equal to first threshold, then first facial image is included into the first face archives
In, comprising:
Choose any face archives in the online archives, however, it is determined that go out first face in the online archives
The similarity of the class center of archives and first facial image is greater than or equal to the first threshold, then by first face
Image is included into the first face archives.
It for example, include the class center 1 of face archives 1, the class center 2 of face archives 2, face archives in online archives
3 class center 3, however, it is determined that the similarity 1 at the first facial image and class center 1 is less than first threshold, it is determined that the first face figure
As the similarity 2 with class center 2；If it is determined that the similarity 2 at the first facial image and class center 2 is less than first threshold, it is determined that
The similarity 3 of first facial image and class center 3；If it is determined that similarity 3 is greater than first threshold, then the first facial image is included into
Face archives 3.
It should be noted that sequence of first facial image compared with face archives can be random selection, it can also root
It is determined according to the priority of face archives, it is not limited here.The priority of face archives can be according to the sky of the first facial image
Between information and/or temporal information generate.Facial image can carry spatial information and temporal information, wherein spatial information is to grab
Clap mark and the position of the monitoring device of facial image.Temporal information is the time that monitoring device captures facial image.Example
Property, the spatial information and temporal information of facial image A is as follows: spatial information: camera identification are as follows: CAMERA_1 takes the photograph
As head position: Pudong New Area, Shanghai.Temporal information: 2018-10-20 10:07:21.
Specifically, multiple screening strategies can be determined according to spatial information, successively determine the first face according to screening strategy
The online archives of each priority of image, wherein the corresponding spatial dimension of screening strategy is smaller, and determining priority is higher.Often
First facial image is compared by the online archives for determining a priority with the online archives of the priority, when this is preferential
When there are online archives matched with the first facial image in the online archives of grade, the first facial image is included into matched online
Archives.
Since the life range of many people is relatively fixed, a people may often be captured by the camera of certain areas,
Therefore when the first facial image to real-time grasp shoot is filed, first believed using the spatial information of the first facial image as priori
Breath, online archives are filtered out from facial image database, are then compared the first facial image with the online archives filtered out
After file, the efficiency of online elite archiving can be effectively improved.
Illustratively, the spatial information for setting the first facial image includes: camera identification as CAMERA_1, camera position
It is set to Pudong New Area, Shanghai.The sequence according to spatial dimension from small to large is set, determines the following screening of the first facial image
Strategy:
Screening strategy 1: the online of the 1st priority is filtered out from facial image database according to " camera identification CAMERA_1 "
Archives.
Screening strategy 2: the online archives of the 2nd priority are filtered out from facial image database according to " Pudong New District ".
Screening strategy 3: the online archives of the 3rd priority are filtered out from facial image database according to " Shanghai City ".
Screening strategy 4: online archives of all online archives as the 4th priority in facial image database.
Further, set according to screening strategy 1 filtered out from facial image database the online archives of the 1st priority as
First facial image and online archives 1 are then compared by line archives 1 and online archives 2, determine the first facial image and online
When archives 1 match, the first facial image is included into online archives 1.
When online archives matched with the first facial image being not present in the online archives for determining K priority, from people
The online archives of K+1 priority are obtained in face image library.Determining in the online archives of K+1 priority whether there is and first
The matched online archives of facial image, the online archives up to finding matched online archives or each priority, which compare, to be finished.
Illustratively, set according to screening strategy 1 filtered out from facial image database the online archives of the 1st priority as
First facial image is compared with online archives 1 and online archives 2 respectively, determines first by line archives 1 and online archives 2
Facial image is mismatched with online archives 1 and online archives 2, then the 2nd is filtered out from facial image database according to screening strategy 2
The online archives of priority.The online archives of the 2nd priority are set as online archives 1, online archives 2, online archives 3, online
Archives 4.First facial image is compared with online archives 3 and online archives 4 respectively, the first facial image is and online
Archives 1 and online archives 2 compare, therefore can no longer compare at this time.Exist when in online archives 3 and online archives 4
When one online archives is matched with the first facial image, the first facial image is included into matched online archives, when online archives 3
When mismatching with the first facial image with online archives 4, it is preferential that the 3rd is filtered out from facial image database according to screening strategy 3
The online archives of grade, and so on, until being found in facial image and online archives matched in the first facial image or the
One facial image and the online archives comparison of the 4th priority finish.
Exist in the specific online archives for determining K priority using following methods matched online with the first facial image
Archives:
For any online archives in the online archives of K priority, according to the class center of online archives with it is the first
Similarity between the class of face image determines the similarity of online archives Yu first facial image.
If similarity is greater than or equal to first threshold, it is determined that exist and the first face in the online archives of N priority
The online archives of images match.
Specifically, for the online archives of different priorities, the first threshold of setting is not also identical, wherein priority is got over
Height, corresponding first threshold are smaller.Illustratively, the corresponding first threshold of the 1st priority is set as T0, and the 2nd priority is corresponding
First threshold be T1, the corresponding first threshold of the 3rd priority be T2, the corresponding first threshold of the 4th priority be T3, then T0 <
T1<T2<T3。
For the accuracy for further increasing the first face archive, in conjunction with Fig. 2, as shown in figure 4, the embodiment of the present invention is also
A kind of archiving method is provided, comprising:
Step 401: from multiple facial images obtained in online archives in preset period of time；
Wherein, the online archives are for filing the facial image that monitoring device obtains in real time；
Wherein, preset period of time can be divided with one day for granularity；For example, preset period of time can be one day, i.e., at the T days
0 point, obtain the T days monitoring device candid photograph face images.
It should be noted that online archives can be cancelled to described more to avoid influence of the online archives to offline profile
Open the archive operation that facial image carries out.
A kind of possible implementation can also include: that revocation is updated in the preset period of time before step 301
Online archives；The online archives of the update are the online archives that multiple described facial images are carried out with online elite archiving；It is described more
Opening facial image is the facial image that monitoring device obtains in real time in the preset period of time.
A kind of possible implementation can also include: to restore offline before preset period of time before step 401
Archives.
Step 402: multiple described facial images being clustered, pre- filing archives are generated；
Wherein, any facial image in any facial image and other pre- filing archives in any pre- filing archives
Between similarity be less than first threshold；
Step 403: being directed at least one pre- filing archives, however, it is determined that the class center of the pre- filing archives and offline profile
In the first face archives class center between similarity be greater than second threshold, then by it is described it is pre- filing archives in facial image
It is included into the first face archives；
Wherein, the first face archives are filed any face archives in the offline profile；Offline profile is described
The archives filed before preset period of time；Class center is determined according to the facial image in archives.
It is clustered by multiple facial images to preset period of time, and is filed again, determine the class of offline profile
Center, relative to the archiving method of online archives, effectively increases the accuracy of the filing of archives after filing again.
In addition, after step 403, the offline profile after filing can directly be covered to the online archives of preset period of time.
In conjunction with the example in step 401, in the T+1 days progress online elite archivings, the initial archives of online archives were to file at the T days
Offline profile.It carries out can effectively improve the accuracy of filing for online elite archiving relative to being directly used in line archives, and
The accuracy at class center is improved, and then improves the accuracy of online elite archiving.
In step 402, the class center of filing archives can be determined according to various ways in advance.
For example, the first scene, however, it is determined that own in the class center of the pre- filing archives and the offline profile
Face archives class center between similarity be respectively less than the second threshold, then establish the second face for the pre- filing archives
Archives, and using the class center of the pre- filing archives as the class center of the second face archives.
Second of scene, it is pre- to file including at least 2 facial images in archives, for the efficiency and filing for improving online elite archiving
Accuracy, a kind of possible implementation, can will in pre- filing archives the highest facial image of picture quality as returning in advance
The class center of shelves archives, the timing node at the specific class center for updating pre- filing archives can for offline profile filing when
Between, it also can according to need determination.It should be noted that picture quality can be according to the pixel of facial image, clarity, face
The shooting angle of image, the comprehensive determination of effect of feature extraction etc., it is not limited here.
In a kind of possible embodiment, the class center of offline profile is one.By the class center of offline profile and in advance
Similarity between the class at the class center of filing archives, the similarity being determined as between offline profile and pre- filing archives.
In specific implementation, between the class center for determining offline profile and the class at the class center of pre- filing archives when similarity,
Can determine first class center to the corresponding second feature in class center of the first eigenvector of application and pre- filing archives to
Then amount calculates the distance between first eigenvector and second feature vector, according to first eigenvector and second feature to
The distance between amount determines similarity between the class center of offline profile and the class at the class center of pre- filing archives.
Under the scene, after step 403, can also include:
If it is determined that the picture quality at the class center of the pre- filing archives is greater than the class center of the first face archives
The class center of the first face archives is then updated to the class center of the pre- filing archives by picture quality.
In alternatively possible embodiment, the class center of offline profile can be it is multiple, determine each of offline profile
Similarity between class center and the class at the class center of pre- filing archives, then according to the weight at all kinds of centers of online archives and respectively
Similarity between the class at class center determines the similarity of online archives and the first facial image.
In the specific implementation process, the method for determining similarity between the class at multiple class centers may include:
Step 1: using highest first facial image of picture quality in the first face archives as first face
The first kind center of archives；
Step 2: if it is determined that in the first face archives the second facial image picture quality be greater than third threshold value, and
Similarity between second facial image and first facial image is less than the 4th threshold value, then by the second face figure
As the second class center as the first face archives；4th threshold value is greater than the second threshold.
For example, multiple class centers of offline profile can be determined using following methods, comprising:
Step 1: presetting the quantity at the class center of offline profile；
Step 2: the highest facial image of offline profile picture quality is determined as first kind center；
Step 3: after determining that pre- filing archives are included into the first face archives, however, it is determined that the second people in pre- filing archives
The picture quality of face image is greater than third threshold value, it is determined that similar between first kind center and the class at all kinds of centers of pre- filing archives
Degree；
Step 4: if it is determined that the second facial image is determined as in the second class when such similarity is less than second threshold
The heart；
Step 5: above-mentioned steps are executed for next pre- filing archives for being included into offline profile, until online archives
The data at class center reach pre-set quantity.
Due to previously according to facial image in offline profile picture quality and characteristics of image selected from each facial image
Class center is taken, therefore when filing to pre- filing archives, it need to will only file in the class center of archives and the class of offline profile in advance
The heart is compared, determination and the pre- filing matched offline profile of archives, without all face figures that will file in advance in archives
As being compared with every facial image in offline profile, to improve filing efficiency；In addition, by being arranged in multiple classes
The heart effectively increases the accuracy rate of filing, controls archives ratio, so that the same person will not generate excessive archives, to improve
The robustness of filing.
Between the class at multiple class centers at the class center and offline profile for determining pre- filing archives when similarity, can first it divide
Similarity between each class center of archives and the class at each class center of the first face archives Que Ding not filed in advance, then basis
The weight at each class center carries out similarity between all kinds of centers of pre- filing archives and the class at all kinds of centers of offline profile
Fusion determines the similarity at each class center of offline profile and pre- filing archives, and the weight at class center can be according to image matter
Amount is allocated, and can also be allocated according to characteristics of image.
In the specific implementation process, all kinds of centers according to the first face archives and the pre- filing archives
Similarity between the class at each class center determines the similarity of the first face archives and the pre- filing archives, comprising:
Step 1: determining each class center of the pre- filing archives and all kinds of centers of the first face archives respectively
Class between similarity；
Step 2: according to the weight at all kinds of centers of the first face archives, each class center of pre- filing archives
Similarity between weight and the class at all kinds of centers determines in each class for stating the first face archives and the pre- filing archives
The similarity of the heart.
Each class center of a kind of possible implementation, the pre- filing archives is similar to the first face archives
Degree C is determined according to the following formula:
Wherein, Ai is the weight at i-th of class center of pre- filing archives；N≥i≥1；Bj is the jth of the first face archives
The weight at a class center；M≥j≥1；Ci, j are pre- i-th of class center for filing archives and the first face archives
Similarity between j-th of class center.N, M are the positive integer greater than 1.
A kind of possible implementation, what the weight at all kinds of centers was determined according to following methods, comprising:
The attribute information at the class center is inputted into supervised learning model, determines the weight at the class center；The attribute
Information includes at least one of and combination: picture quality, the temporal information degree of correlation or the spatial information degree of correlation.
For example, the mass fraction of image can be divided into high, medium and low three classes；Space time information can be divided into strong correlation (
Same camera is on the same day), weak related (same camera or on the same day), uncorrelated (neither same camera, nor together
One day).The weight at class center can be determined according to the mass fraction and the space time information degree of correlation of image.
Based on the same technical idea, the embodiment of the invention provides a kind of filing apparatus, as shown in figure 5, the device 500
Include:
Acquiring unit 501 is used for acquiring unit, for obtaining the first face archives；The first face archives are face
Any face archives in image library；Wherein, appoint in any facial image in any face archives and other face archives
Similarity between one facial image is less than first threshold；
Processing unit 502 determines described the first for at least facial image according to the first face archives
The class center of face archives；Whether the facial image that the class center of the first face archives is used to determine that monitoring device obtains is included into
The first face archives.
A kind of possible implementation, the processing unit 502, is specifically used for:
By highest first facial image of picture quality in the first face archives, it is determined as the first face archives
First kind center.
A kind of possible implementation, the class center of the first face archives include multiple class centers, and the processing is single
Member 502 is also used to:
If it is determined that the similarity between the second facial image and first facial image in the first face archives
Less than second threshold, then using second facial image as the second class center of the first face archives.
A kind of possible implementation, the processing unit 502, is also used to:
Determine that the picture quality of second facial image is greater than third threshold value.
A kind of possible implementation, the processing unit 502, is also used to:
The attribute information at the class center is inputted into supervised learning model, determines the weight at the class center；The attribute
Information includes at least one of and combination: picture quality, the temporal information degree of correlation or the spatial information degree of correlation.
Based on the same technical idea, the embodiment of the present application provides a kind of archival device, as shown in fig. 6, including at least
One processor 601, and the memory 602 connecting at least one processor do not limit processor in the embodiment of the present application
Specific connection medium between 601 and memory 602 is connected as in Fig. 5 by bus between processor 601 and memory 602
Example.Bus can be divided into address bus, data/address bus, control bus etc..
In the embodiment of the present application, memory 602 is stored with the instruction that can be executed by least one processor 601, at least
The instruction that one processor 601 is stored by executing memory 602, can execute step included in aforementioned archiving method.
Wherein, processor 601 is the control centre of archival device, can use various interfaces and connection archival device
Various pieces, by running or executing the instruction being stored in memory 602 and calling is stored in number in memory 602
According to realize filing.Optionally, processor 601 may include one or more processing units, and processor 601 can integrate application
Processor and modem processor, wherein the main processing operation system of application processor, user interface and application program etc.,
Modem processor mainly handles wireless communication.It is understood that above-mentioned modem processor can not also be integrated into
In processor 1801.In some embodiments, processor 601 and memory 602 can be realized on the same chip, in some realities
It applies in example, they can also be realized respectively on independent chip.
Processor 601 can be general processor, such as central processing unit (CPU), digital signal processor, dedicated integrated
Circuit (Application Specific Integrated Circuit, ASIC), field programmable gate array or other can
Perhaps transistor logic, discrete hardware components may be implemented or execute the application implementation for programmed logic device, discrete gate
Each method, step and logic diagram disclosed in example.General processor can be microprocessor or any conventional processor
Deng.The step of method in conjunction with disclosed in the embodiment of the present application, can be embodied directly in hardware processor and execute completion, Huo Zheyong
Hardware and software module combination in processor execute completion.
Memory 602 is used as a kind of non-volatile computer readable storage medium storing program for executing, can be used for storing non-volatile software journey
Sequence, non-volatile computer executable program and module.Memory 602 may include the storage medium of at least one type,
It such as may include flash memory, hard disk, multimedia card, card-type memory, random access storage device (Random Access
Memory, RAM), static random-access memory (StaticRandom Access Memory, SRAM), may be programmed read-only deposit
Reservoir (Programmable Read OnlyMemory, PROM), read-only memory (Read Only Memory, ROM), electrification
Erasable Programmable Read Only Memory EPROM (Electrically Erasable Programmable Read-Only Memory,
EEPROM), magnetic storage, disk, CD etc..Memory 602 can be used for carrying or storing have instruction or data
The desired program code of structure type and can by any other medium of computer access, but not limited to this.The application is real
Applying the memory 602 in example can also be circuit or other devices that arbitrarily can be realized store function, for storing program
Instruction and/or data.
Based on the same inventive concept, the embodiment of the present application provides a kind of computer-readable medium, and being stored with can be by returning
The computer program that shelves equipment executes, when described program is run in archival device, so that the archival device executes filing
The step of method.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method or computer program product.
Therefore, complete hardware embodiment, complete software embodiment or embodiment combining software and hardware aspects can be used in the present invention
Form.It is deposited moreover, the present invention can be used to can be used in the computer that one or more wherein includes computer usable program code
The shape for the computer program product implemented on storage media (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.)
Formula.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications may be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (10)
1. a kind of archiving method characterized by comprising
From multiple facial images obtained in online archives in preset period of time；The online archives for obtaining monitoring device in real time
The facial image taken is filed；
Multiple described facial images are clustered, pre- filing archives are generated；Wherein, any people in any pre- filing archives
Similarity in face image and other pre- filing archives between any facial image is less than first threshold；
For at least one pre- filing archives, according to the first face in all kinds of centers and offline profile of the pre- filing archives
Similarity between all kinds of centers of archives determines the similarity of pre- the filing archives and the first face archives；If described
Similarity is greater than second threshold, then the facial image in the pre- filing archives is included into the first face archives；Described
One face archives are filed any face archives in the offline profile；The class center is according to the face figure in archives
As determining.
2. the method as described in claim 1, which is characterized in that described according to the pre- all kinds of centers for filing archives and offline
Similarity between all kinds of centers of the first face archives in archives determines the pre- filing archives and the first face archives
Similarity, comprising:
Determine that each class center of the pre- filing archives is similar between the class between each class center of the first face archives
Degree；
According to the weight between the weight at each class center of the pre- filing archives, each class center of the first face archives
And all kinds of similarities, determine the similarity of pre- the filing archives and the first face archives.
3. method according to claim 2, which is characterized in that the pre- filing archives are similar to the first face archives
Degree C is determined according to the following formula:
Wherein, Ai is the weight at i-th of class center of pre- filing archives；N≥i≥1；Bj is j-th of class of the first face archives
The weight at center；M≥j≥1；Ci, j are j-th of pre- i-th of the class center and the first face archives for filing archives
Similarity between class center；N, M is the positive integer greater than 1.
4. the weight of method as described in any one of claims 1 to 3, the class center determines in the following manner, comprising:
The attribute information at the class center is inputted into supervised learning model, determines the weight at the class center；The attribute information
Including at least one of and combination: picture quality, the temporal information degree of correlation or the spatial information degree of correlation.
5. the method as described in claim 1, which is characterized in that the method also includes:
If it is determined that the similarity of all people's face archives is respectively less than described second in the pre- filing archives and the offline profile
Threshold value then establishes the second face archives for the pre- filing archives, and using the class center of the pre- filing archives as described the
The class center of two face archives.
6. a kind of filing apparatus characterized by comprising
Acquiring unit, for from online archives obtain preset period of time in multiple facial images；The online archives for pair
The facial image that monitoring device obtains in real time is filed；
Processing unit generates pre- filing archives for clustering to multiple described facial images；Wherein, any pre- filing shelves
Similarity in any facial image and other pre- filing archives in case between any facial image is less than first threshold；
For at least one pre- filing archives, according to the first face archives in all kinds of centers and offline profile of the pre- filing archives
All kinds of centers between similarity, determine it is described it is pre- filing archives and the first face archives similarity；If described similar
Degree is greater than second threshold, then the facial image in the pre- filing archives is included into the first face archives；It is described the first
Face archives are filed any face archives in the offline profile；The class center is true according to the facial image in archives
Fixed.
7. device as claimed in claim 6, which is characterized in that the processing unit is specifically used for:
Determine that each class center of the pre- filing archives is similar between the class between each class center of the first face archives
Degree；According to the weight between the weight at each class center of the pre- filing archives, each class center of the first face archives
And all kinds of similarities, determine the similarity of pre- the filing archives and the first face archives.
8. device as claimed in claim 7, which is characterized in that the pre- filing archives are similar to the first face archives
Degree C is determined according to the following formula:
Wherein, Ai is the weight at i-th of class center of pre- filing archives；N≥i≥1；Bj is j-th of class of the first face archives
The weight at center；M≥j≥1；Ci, j are j-th of pre- i-th of the class center and the first face archives for filing archives
Similarity between class center；N, M is the positive integer greater than 1.
9. a kind of archival device, which is characterized in that including at least one processing unit and at least one storage unit, wherein
The storage unit is stored with computer program, when described program is executed by the processing unit, so that the processing unit
Perform claim requires the step of 1~5 any claim the method.
10. a kind of computer-readable medium, which is characterized in that it is stored with the computer program that can be executed by archival device, when
Described program in archival device when running, so that the archival device perform claim requires the step of 1~5 any the method
Suddenly.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811626344.8A CN109800672A (en) | 2018-12-28 | 2018-12-28 | A kind of archiving method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811626344.8A CN109800672A (en) | 2018-12-28 | 2018-12-28 | A kind of archiving method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109800672A true CN109800672A (en) | 2019-05-24 |
Family
ID=66558090
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811626344.8A Withdrawn CN109800672A (en) | 2018-12-28 | 2018-12-28 | A kind of archiving method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109800672A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507424A (en) * | 2020-04-27 | 2020-08-07 | 上海依图网络科技有限公司 | Data processing method and device |
| CN112686141A (en) * | 2020-12-29 | 2021-04-20 | 杭州海康威视数字技术股份有限公司 | Personnel filing method and device and electronic equipment |
-
2018
- 2018-12-28 CN CN201811626344.8A patent/CN109800672A/en not_active Withdrawn
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507424A (en) * | 2020-04-27 | 2020-08-07 | 上海依图网络科技有限公司 | Data processing method and device |
| CN111507424B (en) * | 2020-04-27 | 2023-10-27 | 上海依图网络科技有限公司 | Data processing method and device |
| CN112686141A (en) * | 2020-12-29 | 2021-04-20 | 杭州海康威视数字技术股份有限公司 | Personnel filing method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109784217A (en) | A kind of monitoring method and device | |
| CN109800668A (en) | A kind of archiving method and device | |
| CN109740004A (en) | A kind of archiving method and device | |
| CN109783672A (en) | A kind of archiving method and device | |
| CN109800322A (en) | A kind of monitoring method and device | |
| CN109800329A (en) | A kind of monitoring method and device | |
| CN109815829A (en) | A kind of method and device of determining passerby track | |
| CN109783685A (en) | A kind of querying method and device | |
| AU2017250159B2 (en) | Video recording method, server, system, and storage medium | |
| CN109800674A (en) | A kind of archiving method and device | |
| CN109726674A (en) | A kind of face identification method and device | |
| CN104050449B (en) | A kind of face identification method and device | |
| CN110675433A (en) | Video processing method and device, electronic equipment and storage medium | |
| CN109800673A (en) | A kind of archiving method and device | |
| CN109740003B (en) | Filing method and device | |
| CN109784220A (en) | A kind of method and device of determining passerby track | |
| CN109800664A (en) | A kind of method and device of determining passerby track | |
| CN109783663A (en) | A kind of archiving method and device | |
| CN109800672A (en) | A kind of archiving method and device | |
| CN109308704A (en) | Background elimination method, device, computer equipment and storage medium | |
| CN109800675A (en) | A kind of method and device of the identification image of determining face object | |
| CN109063628A (en) | Face identification method, device, computer equipment and storage medium | |
| CN109872541A (en) | A kind of information of vehicles analysis method and device | |
| CN109783664A (en) | A kind of archiving method and device | |
| CN109558861A (en) | A kind of method and device of determining passerby track |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200507 Address after: Room 01 and 02, 1st floor, No. 180 Yizhou Road, Xuhui District, Shanghai Applicant after: SHANGHAI YITU NETWORK SCIENCE & TECHNOLOGY Co.,Ltd. Address before: Room 01 and 02, 1st floor, No. 180 Yizhou Road, Xuhui District, Shanghai Applicant before: SHANGHAI YITU NETWORK SCIENCE & TECHNOLOGY Co.,Ltd. Applicant before: SHENZHEN YITU INFORMATION TECHNOLOGY Co.,Ltd. |
|
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190524 |