CN113987244A

CN113987244A - Human body image gathering method and device, computer equipment and storage medium

Info

Publication number: CN113987244A
Application number: CN202111068356.5A
Authority: CN
Inventors: 潘华东; 夏鲁宾; 殷俊
Original assignee: Zhejiang Dahua Technology Co Ltd
Current assignee: Zhejiang Dahua Technology Co Ltd
Priority date: 2021-09-13
Filing date: 2021-09-13
Publication date: 2022-01-28

Abstract

The application relates to a human body image gathering method, a human body image gathering device, computer equipment and a computer readable storage medium, wherein human body characteristics and attribute characteristics in each human body image to be gathered are obtained; integrating the human body characteristics and the attribute characteristics of each human body image of the to-be-gathered file to obtain the clustering characteristics of each human body image of the to-be-gathered file; and clustering the human body images of the plurality of files to be gathered according to the similarity among the clustering characteristics of the human body images of the plurality of files to be gathered to form a human body image file. In the document gathering process of the human body image, the human body characteristics of the human body image to be gathered are combined with the attribute characteristics, the input information amount of the human body image to be gathered is enlarged, the document gathering strategy of the human body image is improved by means of the enlarged input information amount, and therefore the accuracy rate of the document gathering of the human body image is improved.

Description

Human body image gathering method and device, computer equipment and storage medium

Technical Field

The present application relates to the field of image processing technologies, and in particular, to a human body image gathering method and apparatus, a computer device, and a computer-readable storage medium.

Background

In the technical field of security protection, personnel information gathering is an important technology, and is helpful for technical personnel to locate action tracks and common foothold of suspects. Human body image gathering is a key link in personnel information gathering, and the main function is to gather human body images captured by a security camera into files, wherein the human body images in each file belong to the same person, and the images of different files belong to different persons.

In the related technology, a picture to be gathered and characteristic values of each existing representative picture in an existing file are obtained; acquiring a first similarity with the highest similarity between the characteristic value of the picture to be archived and the characteristic values of the existing representative pictures in the existing archive; comparing the first similarity with a preset similarity threshold, and judging a preset similarity region to which the first similarity belongs according to a comparison result; and selecting a corresponding preset document-gathering strategy according to the judgment result to gather the document of the picture to be gathered.

However, the feature information of the picture is easily affected by various environmental factors, and under different environments, the similarity of the feature of the picture is biased, so that the similarity of different people may be higher than that of the same person, thereby causing a problem of a wrong gathering of the file, and having a low accuracy rate of the gathering of the file. Aiming at the problem of low gear-gathering accuracy in the related technology, no effective solution is provided at present.

Disclosure of Invention

In view of the above, it is necessary to provide a human body image gathering method, device, computer device, and computer readable storage medium to solve the problem of low accuracy of gathering in the related art.

In a first aspect, an embodiment of the present application provides a method for gathering human body images, including the following steps:

acquiring human body characteristics and attribute characteristics in each human body image of the document to be gathered, wherein the attribute characteristics represent other attribute information in the human body image of the document to be gathered except the human body information represented by the human body characteristics;

integrating the human body characteristics and the attribute characteristics of each human body image of the to-be-gathered file to obtain the clustering characteristics of each human body image of the to-be-gathered file;

and clustering the human body images of the plurality of files to be gathered according to the similarity among the clustering characteristics of the human body images of the plurality of files to be gathered to form a human body image file.

In some embodiments, the acquiring the human body features and the attribute features in the human body image of each document to be gathered includes:

inputting the human body image of each document to be gathered into a trained feature extraction network to obtain the human body features of the human body image of each document to be gathered; and inputting each human body image of the document to be gathered into the trained attribute recognition network to obtain the attribute characteristics of each human body image of the document to be gathered.

In some embodiments, the integrating, for each human body image of the document to be gathered, the human body feature and the attribute feature therein to obtain a clustering feature of each human body image of the document to be gathered includes:

and aiming at each human body image of the to-be-gathered file, integrating the human body characteristics and the attribute characteristics in the human body image into a trained full-connection network to obtain the clustering characteristics of each human body image of the to-be-gathered file.

In some embodiments, the clustering, according to the similarity between the clustering features of the human body images of the multiple documents to be gathered, the human body images of the multiple documents to be gathered to form the human body image archive includes:

dividing the human body images of the to-be-gathered files into a plurality of image sets with the same place according to the shooting place of each human body image of the to-be-gathered files; clustering a plurality of human body images of the to-be-gathered files in each same image set according to the similarity between the clustering features of the human body images of the to-be-gathered files in the same image set at the same place to obtain a plurality of initial files;

determining the archive attribute of each initial archive in a plurality of initial archives according to the attribute characteristics of the human body images in the initial archives;

clustering a plurality of initial archives with the same archive attribute to form an archive set with the same attribute;

and merging the plurality of initial archives in the same-attribute archive set according to the similarity among the plurality of initial archives to obtain the human body image archive.

In some embodiments, after merging the plurality of initial archives according to the similarity between the plurality of initial archives in the same-attribute archive set to obtain the human body image archive, the method further includes:

and traversing each human body image in the human body image file, if the attribute characteristics of the human body image do not accord with the file attributes of the human body image file in which the human body image is positioned, removing the human body image from the human body image file in which the human body image is positioned, and storing the human body image into a scattered image set.

In some embodiments, after the clustering the human body images of the plurality of documents to be gathered according to the similarity between the clustering features of the human body images of the plurality of documents to be gathered to form a human body image archive, the method further includes:

traversing each human body image in the human body image file, if a space-time logic conflict exists between the human body images in the human body image file, removing one of the two human body images with the space-time logic conflict from the human body image file according to the similarity between the human body image with the space-time conflict and the clustering features of the other human body images in the human body image file, and storing the two human body images in a scattered image set.

In some of these embodiments, the method further comprises:

and under the condition that only one human body image exists in the human body image file, removing the human body image from the human body image file and storing the human body image into a scattered image set.

In some of these embodiments, the method further comprises:

calculating the similarity between the clustering features of each human body image in the scattered image set and each human body image in the human body image archive with the same attribute;

and merging the human body images in the scattered image set into the human body image file under the condition that the average value of the similarity between the clustering features of any one human body image in the scattered image set and all the human body images in the human body image file with the same attribute exceeds a set similarity threshold value and no space-time logic conflict exists between the human body image in the scattered image set and all the human body images in the human body image file with the same attribute.

In some of these embodiments, the method further comprises:

clustering a plurality of human body images which have the same attribute characteristics and do not have space-time logic conflict in the scattered image sets to form a plurality of image sets with the same attribute;

and clustering the human body images in each same attribute image set according to the similarity between the clustering characteristics of the human body images in the same attribute image set to form a scattered archive.

In some of these embodiments, the attribute features include: one or more of a hair style feature, a hat feature, and a garment color feature.

In a second aspect, in this embodiment, there is provided a human body image gathering device, including: the system comprises a first acquisition module, a second acquisition module and a gear-gathering module:

the first acquisition module is used for acquiring human body characteristics and attribute characteristics in each human body image of the document to be gathered, wherein the attribute characteristics represent other attribute information in the human body image of the document to be gathered except the human body information represented by the human body characteristics;

the second obtaining module is used for integrating the human body features and the attribute features of each human body image of the to-be-gathered file to obtain the clustering features of each human body image of the to-be-gathered file;

and the document gathering module is used for clustering a plurality of human body images to be gathered to form a human body image file according to the similarity between the clustering characteristics of the plurality of human body images to be gathered.

In a third aspect, there is provided in this embodiment a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method according to the first aspect when executing the computer program.

In a fourth aspect, in the present embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the method according to the first aspect as described above.

The human body image gathering method, the human body image gathering device, the computer equipment and the computer readable storage medium have the advantages that the human body characteristics and the attribute characteristics in each human body image to be gathered are obtained; integrating the human body characteristics and the attribute characteristics of each human body image of the to-be-gathered file to obtain the clustering characteristics of each human body image of the to-be-gathered file; and clustering the human body images of the plurality of files to be gathered according to the similarity among the clustering characteristics of the human body images of the plurality of files to be gathered to form a human body image file. In the document gathering process of the human body image, the human body characteristics of the human body image to be gathered are combined with the attribute characteristics, the input information amount of the human body image to be gathered is enlarged, the document gathering strategy of the human body image is improved by means of the enlarged input information amount, and therefore the accuracy rate of the document gathering of the human body image is improved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:

fig. 1 is an application scene diagram of a human body image gathering method provided according to an embodiment of the application;

FIG. 2 is a first flowchart of a human body image gathering method according to an embodiment of the present disclosure;

FIG. 3 is a second flowchart of a human body image gathering method according to an embodiment of the present application;

FIG. 4 is a schematic structural diagram of a human body image gathering device according to an embodiment of the present application;

fig. 5 is a schematic structural diagram of a computer device provided according to an embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments provided in the present application without any inventive step are within the scope of protection of the present application.

It is obvious that the drawings in the following description are only examples or embodiments of the present application, and that it is also possible for a person skilled in the art to apply the present application to other similar contexts on the basis of these drawings without inventive effort. Moreover, it should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous implementation-specific decisions must be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which may vary from one implementation to another.

Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the specification. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of ordinary skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments without conflict.

Unless defined otherwise, technical or scientific terms referred to herein shall have the ordinary meaning as understood by those of ordinary skill in the art to which this application belongs. Reference to "a," "an," "the," and similar words throughout this application are not to be construed as limiting in number, and may refer to the singular or the plural. The present application is directed to the use of the terms "including," "comprising," "having," and any variations thereof, which are intended to cover non-exclusive inclusions; for example, a process, method, system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to the listed steps or elements, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. Reference to "connected," "coupled," and the like in this application is not intended to be limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as referred to herein means two or more. "and/or" describes an association relationship of associated objects, meaning that three relationships may exist, for example, "A and/or B" may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship. Reference herein to the terms "first," "second," "third," and the like, are merely to distinguish similar objects and do not denote a particular ordering for the objects.

Fig. 1 is an application scene diagram of a human body image gathering method according to an embodiment of the present application. As shown in fig. 1, both the server 101 and the mobile terminal 102 may perform data transmission via a network. The mobile terminal 102 is configured to collect a human body image of a to-be-gathered document, and transmit the human body image of the to-be-gathered document to the server 101. After the server 101 receives the human body image of the document to be gathered, acquiring human body features and attribute features in each human body image of the document to be gathered; integrating the human body characteristics and the attribute characteristics of each human body image of the to-be-gathered file to obtain the clustering characteristics of each human body image of the to-be-gathered file; and clustering the human body images of the plurality of files to be gathered according to the similarity among the clustering characteristics of the human body images of the plurality of files to be gathered to form a human body image file. The server 101 may be implemented by an independent server or a server cluster composed of a plurality of servers, and the mobile terminal 102 may be an image capturing device such as a camera and a mobile phone.

The embodiment of the application provides a human body image gathering method, which can be used for gathering human body images in the technical field of image processing, and as shown in fig. 2, the method comprises the following steps:

step S210, obtaining the human body characteristics and the attribute characteristics in each human body image of the document to be gathered, wherein the attribute characteristics represent other attribute information except the human body information represented by the human body characteristics in the human body image of the document to be gathered.

The human body image to be gathered can be an initial human body image directly shot by a video camera or a camera, or an initial human body image acquired from a human body image library. In order to obtain a better gathering effect, preprocessing operations such as denoising and enhancement processing can be performed on the initial human body image, so as to obtain a human body image to be gathered.

Human body information in the human body characteristic representation human body image in the application is the representation of human body coarse-grained characteristics in the human body image, the less obvious characteristics can be ignored by the human body characteristics, and the expression form of the human body characteristics is a multi-dimensional vector. The attribute features are the representation of fine-grained features of the human body in the human body image and represent the specific attribute information of the human body in the human body image. Further, the attribute features may represent other attribute information in the human body image than the human body information represented by the human body features. The attribute feature may be one or more of a hair style feature, a hat feature, and a clothing color feature. The hair style characteristic, the hat characteristic and the clothes color characteristic are obvious attribute characteristics capable of distinguishing different human bodies, and the attributes of different human bodies can be represented through one or more of the hair style characteristic, the hat characteristic and the clothes color characteristic.

For example, the color of the hat characterized by the body feature is a dark color, while the hat color characterized by the attribute feature is a specific color black. The representation of the attribute feature may also be a multi-dimensional vector. And the attribute information of the human body image corresponds to the attribute characteristics of the human body image one by one. For example, the attribute information of the human body image is that the hat color is black, but the attribute feature of the human body image is a multi-dimensional vector, and the characteristic means that the hat color is black after analysis. One attribute corresponds to one attribute information and one attribute feature. Generally, there are a plurality of attributes of the human body, such as hair style, hat color, coat color, etc., so there may be a plurality of attribute information and attribute features corresponding to the human body image. Specifically, attribute features of the human body image are extracted, and the attribute features can be formulated according to actual requirements. The human body features and the attribute features in the human body image to be gathered can be extracted by utilizing the existing neural network.

Step S220, integrating the human body characteristics and the attribute characteristics of each human body image of the to-be-gathered file to obtain the clustering characteristics of each human body image of the to-be-gathered file.

And integrating the human body characteristics and the attribute characteristics to generate relevance between elements of the human body characteristics and elements of the attribute characteristics, so as to obtain the clustering characteristics of the human body image of each document to be clustered. After the integration operation, the expression form of the clustering feature is a multi-dimensional vector. Specifically, the human body features and the attribute features of the human body image of each document to be gathered can be spliced into a long vector, and then the long vector is input into the existing neural network for integration to obtain the final clustering features of the human body image of each document to be gathered.

And step S230, clustering the human body images of the plurality of files to be gathered according to the similarity among the clustering characteristics of the human body images of the plurality of files to be gathered to form a human body image file.

Specifically, the clustering feature similarity between every two human body images to be gathered is calculated according to a cosine similarity calculation method, if the similarity of the clustering features between every two human body images in a set of the human body images to be gathered exceeds a first set threshold, all the human body images in the set are gathered together to form a human body image file, and finally all the human body images can form a plurality of personal body image files. The first set threshold may be adjusted according to actual requirements.

In the related art, only the features of the pictures are used for gathering the human body images, however, the feature information of the pictures is easily influenced by various environmental factors, and under different environments, the similarity of the picture features can be deviated, so that the similarity of different people can be higher than that of the same person, and the problem of gathering errors is easily caused. Through the steps S210 to S240, the human body characteristics and the attribute characteristics in each human body image of the document to be gathered are obtained; integrating the human body characteristics and the attribute characteristics of each human body image of the to-be-gathered file to obtain the clustering characteristics of each human body image of the to-be-gathered file; and clustering the human body images of the plurality of files to be gathered according to the similarity among the clustering characteristics of the human body images of the plurality of files to be gathered to form a human body image file. According to the method and the device, in the gathering process of the human body image, the human body feature set attribute feature of the human body image to be gathered is utilized, the input information quantity of the human body image to be gathered is enlarged, the gathering strategy of the human body image is improved by utilizing the enlarged input information quantity, and therefore the accuracy rate of the gathering of the human body image is improved.

In one embodiment, the method for acquiring the human body features and the attribute features in the human body image of each document to be gathered based on the step S210 includes the following steps:

inputting the human body image of each document to be gathered into the trained feature extraction network to obtain the human body features of the human body image of each document to be gathered; and inputting the human body image of each document to be gathered into the trained attribute recognition network to obtain the attribute characteristics of the human body image of each document to be gathered.

Specifically, the feature extraction network and the attribute recognition network are trained in advance, and can be directly used for extracting the human body features and the attribute features of the human body image to be gathered. In order to ensure the accuracy of extracting the human body characteristics and the attribute characteristics, the characteristic extraction network and the attribute identification network can be updated regularly by using historical data.

The extraction of the human body features and the attribute features can be realized by adopting the existing machine learning technology. The trained feature extraction network and attribute recognition network are used for extracting the human body features and attribute features of the human body image to be gathered, so that the accuracy of extracting the human body features and the attribute features can be ensured.

In one embodiment, the step S220 of integrating the human body features and the attribute features of each human body image of the document to be gathered to obtain the clustering features of each human body image of the document to be gathered includes the following steps:

and inputting the human body characteristics and the attribute characteristics into the trained fully-connected network for integrating aiming at the human body image of each document to be gathered to obtain the clustering characteristics of the human body image of each document to be gathered.

Specifically, the fully-connected network is trained in advance, and can be directly used for integrating human body features and attribute features of a human body image to be gathered. In order to ensure the accuracy of the cluster feature acquisition, the fully-connected network can be updated regularly.

The integration of the human body features and the attribute features of the human body image to be gathered can be realized by the existing information integration technology, such as weighted summation or a neural network. In the embodiment, the trained full-connection network is used for integrating the human body features and the attribute features of the human body image to be gathered, so that the accuracy of acquiring the clustering features can be ensured.

As shown in fig. 3, in one embodiment, the clustering the human body images of the multiple documents to be gathered based on the similarity between the clustering features of the human body images of the multiple documents to be gathered in step S230 to form the human body image file includes the following steps:

step S231, dividing the human body images of the plurality of files to be gathered into a plurality of image sets in the same place according to the shooting place of each human body image of each file to be gathered; and clustering the human body images of the multiple documents to be gathered in each same-place image set according to the similarity between the clustering characteristics of the human body images of the documents to be gathered in the same-place image set to obtain multiple initial documents.

Specifically, when capturing a human body image, the existing camera can display the shooting location on the human body image, so that the shooting location of the human body image to be captured can be obtained while the human body image to be captured is obtained. In the case where the human body image to be focused does not show the shooting location, the shooting location of the human body image may also be read by the display content of the human body image to be focused. According to different shooting places, the human body image to be gathered is divided into a plurality of co-location image sets. The method comprises the steps of calculating the similarity of clustering features between every two human body images of files to be gathered in the same image set at the same place, calculating according to a cosine similarity algorithm, if the similarity of the clustering features between every two human body images of all the files to be gathered in one set of the human body images of the files to be gathered exceeds a second set threshold value, clustering all the human body images of the files to be gathered in the set to form an initial file, and finally forming a plurality of initial files by all the human body images of the files to be gathered. The second set threshold may be adjusted according to actual requirements.

Step S232, determining the archive attribute of each initial archive in the plurality of initial archives according to the attribute characteristics of the human body image in the initial archive.

If the attribute characteristics of the same attribute of all the human body images of the files to be gathered in the current initial file are the same, taking the attribute characteristics as the file attribute of the current initial file; if the attribute features of the same attribute of all the human body images to be gathered in the current initial file are not completely the same, the majority of the attribute features are taken as the file attribute of the current initial file.

As another possible implementation manner, the profile attribute of each of the plurality of initial profiles may be determined according to the attribute information of the human body image in the initial profile. The existing camera can display the attribute information on the human body image when capturing the human body image, so that the attribute information of the human body image to be gathered can be acquired while the human body image to be gathered is acquired. In the case that the attribute information is not displayed on the human body image to be gathered, the attribute information of the human body image may also be read through the display content of the human body image to be gathered. If the attribute information of the same attribute of all the human body images of the files to be gathered in the current initial file is the same, taking the attribute information as the file attribute of the current initial file; if the attribute information of the same attribute of all the human body images to be gathered in the current initial file is not completely the same, the majority of the attribute information is taken as the file attribute of the current initial file.

In step S233, a plurality of initial files with the same file attribute are clustered to form a file set with the same attribute.

Specifically, under the condition that the archive attributes of a plurality of initial archives are the same, clustering the plurality of initial archives to form an archive set with the same attribute.

Step S234, in the same attribute file set, combining a plurality of initial files according to the similarity between the plurality of initial files to obtain a human body image file.

And calculating the similarity between the clustering features of the human body images to be clustered in each file among the initial files of the same-attribute file set. For example, the first initial archive and the second initial archive have the same archive attribute, the first initial archive has human images 1, 2, and 3, the second initial archive has human images 11, 12, and 13, then the similarity of clustering features is calculated between the human image 1 and the human images 11, 12, and 13, the similarity of clustering features is calculated between the human image 2 and the human images 11, 12, and 13, the similarity of clustering features is calculated between the human image 3 and the human images 11, 12, and 13, and the similarity of 9 clustering features is obtained, and the first initial archive and the second initial archive are merged when the maximum value of the similarity of the 9 clustering features exceeds a third set threshold value. According to the merging result, a plurality of human body image files can be obtained. The third set threshold value can be adjusted according to actual requirements.

In the steps S231 to S234, the human body images of the same person are effectively reduced into a plurality of human body image files by performing the archive combining operation on the files of different shooting locations.

In one embodiment, after merging the plurality of initial files according to the similarity between the plurality of initial files in the same attribute file set in step S234 to obtain the human body image file, the method for gathering the human body image further includes the following steps:

step S235, each human body image in the human body image file is traversed, if the attribute characteristics of the human body image do not accord with the file attributes of the human body image file where the human body image is located, the human body image is removed from the human body image file where the human body image is located, and the human body image is stored in a scattered image set.

Due to the fact that the attribute information of the human body images corresponds to the attribute characteristics one to one, the human body images can be removed from the human body image files under the condition that the attribute information of one human body image is judged to be not matched with the file attributes of the human body image files where the human body image is located, the attribute information, namely the attribute characteristics, of all the human body images in the same human body image file are guaranteed to be the same, and the accuracy of file aggregation is effectively improved.

In one embodiment, after the human body images of the multiple documents to be gathered are clustered according to the similarity between the clustering features of the human body images of the multiple documents to be gathered to form the human body image archive in step S230, the document gathering method for the human body images further includes the following steps:

traversing each human body image in the human body image file, if the human body images in the human body image file have space-time logic conflict, removing one of the two human body images with space-time logic conflict from the human body image file according to the similarity between the current human body image with space-time conflict and the clustering characteristics of other human body images in the current human body image file, and storing the human body image in a scattered image set.

Specifically, the existing camera can display the snapshot time on the human body image when the human body image is snapshot, so that the snapshot time of the human body image to be gathered can be obtained while the human body image to be gathered is obtained. And under the condition that the human body image of the to-be-gathered file does not display the snapshot time, the snapshot time of the human body image can be read through the storage time of the human body image of the to-be-gathered file. And if the time interval between the two human body images is smaller than the set time threshold and the space distance is larger than the set space threshold, the two human body images are said to have space-time logic conflict. In the same human body image file, if the space-time logic conflict exists between any two human body images, calculating the similarity between the two human body images and the clustering features of other human body images in the human body image file, removing the human body image with the average value of the similarity between the human body image and the clustering features of other human body images in the human body image file, and storing the human body image with a scattered image set. The set time threshold and the set space threshold can be adjusted according to actual requirements.

The human body images with space-time logic conflicts are removed from the current human body image files, so that the accuracy of file aggregation can be further improved.

In one embodiment, the method for gathering the human body image provided by the embodiment of the application further includes the following steps: and under the condition that the human body image file only has one human body image, removing the human body image from the human body image file and storing the human body image into a scattered image set.

Specifically, when only one human body image exists in the current human body image file, the current human body image file does not have the need, so the human body image is removed from the current human body image file and stored into a scattered image set to prepare for subsequent processing.

Further, the method for gathering the human body image provided by the embodiment of the application further comprises the following steps:

and calculating the similarity between the clustering characteristics of each human body image in the scattered image set and each human body image in the human body image file with the same attribute.

The same attribute means that the attribute characteristics of the human body image are the same as the file attribute of the human body image file. And under the condition that the average value of the similarity between the clustering features of any one human body image in the scattered image set and all the human body images in the human body image file with the same attribute exceeds a set similarity threshold value and no space-time logic conflict exists between the human body image in the scattered image set and all the human body images in the human body image file with the same attribute, merging the human body images in the scattered image set into the human body image file.

Specifically, if there are a plurality of human body image files whose file attributes are the same as the attribute features of the human body images in the current scattered image, then calculating the average values of the similarities between the human body images in the current scattered image and the cluster features of the human body images in the plurality of human body image files of the same attribute, comparing the average values with a set similarity threshold, if there are a plurality of average values exceeding the set similarity threshold, sorting the average values from big lane to small lane, according to the sequence, verifying whether space-time logic conflict exists between the human body images in the current scattered images and all the human body images in the corresponding human body image files, and merging the human body images in the current scattered images into the current human body image file under the condition that space-time logic conflict does not exist between the human body images in the current scattered images and all the human body images in the current human body image file. The set similarity threshold can be adjusted according to actual requirements.

The steps improve the utilization rate of the human body images in the scattered image set by adding the processing steps of the human body images in the scattered image set.

clustering a plurality of human body images which have the same attribute characteristics and do not have space-time logic conflict in a scattered image set to form a plurality of image sets with the same attribute;

and clustering the plurality of human body images in each same attribute image set according to the similarity among the clustering characteristics of the plurality of human body images in the same attribute image set to form a scattered archive.

Specifically, the human body images in the scattered image sets with the same attribute characteristics and without space-time collision are clustered to form a plurality of same-attribute image sets, and the similarity between the clustering characteristics of every two human body images in the same-attribute image set is calculated according to a cosine similarity calculation method. If the similarity of the clustering characteristics of every two human body images in a set of human body images exceeds a set threshold value, all the human body images in the set are clustered to form a scattered archive, and finally, the human body images in all the scattered image sets can form a plurality of scattered archives. The set threshold value can be adjusted according to actual requirements.

The steps further improve the utilization rate of the human body images in the scattered image set by adding the processing to the human body images in the scattered image set.

In one embodiment, the attribute features include: one or more of a hair style feature, a hat feature, and a garment color feature.

The hair style characteristic, the hat characteristic and the clothes color characteristic are obvious attribute characteristics capable of distinguishing different human bodies, and the attributes of different human bodies can be represented through one or more of the hair style characteristic, the hat characteristic and the clothes color characteristic.

In this embodiment, a method for gathering a human body image is further provided, where the process includes the following steps:

step S310, acquiring a human body image of a document to be gathered.

Step S311, inputting the human body image to be gathered into the trained feature extraction network to obtain the human body features of the human body image to be gathered; inputting the human body image of the document to be gathered into the trained attribute identification network to obtain attribute characteristics of the human body image of the document to be gathered, wherein the attribute characteristics represent other attribute information in the human body image except the human body information represented by the human body characteristics.

And step S312, inputting the human body characteristics and the attribute characteristics into the trained fully-connected network for integration to obtain the clustering characteristics.

Step 313, dividing the human body image to be gathered into a plurality of image sets according to the shooting place of the human body image to be gathered; the shooting places of the images in the same image set are the same.

And step S314, clustering each image set based on the similarity among the clustering characteristics to obtain a plurality of initial archives.

Step S315, determining the file attribute of the initial file according to the attribute characteristics of the human body image of the file to be gathered in the initial file.

Step S316, clustering the initial files according to the file attributes of the initial files to form a file set with the same attributes; and merging the initial files according to the similarity between the initial files in the same attribute file set to obtain a plurality of first files.

Step S317, traversing the human body image in each first archive, if there is a situation that the attribute characteristics of the human body image do not match the archive attributes of the first archive in which the human body image is located, removing the human body image from the first archive in which the human body image is located, and storing the human body image into a scattered image set.

Step S318, traversing the human body images in each first file, and if the space-time logic conflict exists between any two human body images in the first file, based on the similarity between the clustering characteristics, removing one human body image of the two human body images with the space-time logic conflict from the first file, and storing the two human body images in a scattered image set.

And step S319, under the condition that the first file only has one human body image, removing the human body image from the first file and storing the human body image into a scattered image set.

Step S320, traverse the scattered images in the scattered image set, obtain a first archive having the same attribute characteristics as those of each scattered image, and calculate the similarity between the scattered images and the clustering characteristics of the human body images in the first archive having the same attributes.

Step S321, merging the scattered images into the first file under the condition that the average value of the similarity between the clustering features of the scattered images and the human body images in the first file with the same attribute exceeds a set similarity threshold value and no space-time logic conflict exists between the scattered images and all the human body images in the first file with the same attribute.

Step S322, the scattered images with the same attribute characteristics and without space-time conflict are combined together, and the combined scattered image set is clustered based on the similarity between the clustering characteristics to form a plurality of second archives.

Fig. 4 is a schematic diagram of a human body image gathering device according to an embodiment of the present invention, and as shown in fig. 4, there is provided a human body image gathering device 40, which includes a first obtaining module 41, a second obtaining module 42 and a gathering module 43, wherein:

a first obtaining module 41, configured to obtain a human body feature and an attribute feature in each human body image of the to-be-gathered file, where the attribute feature represents other attribute information in the human body image of the to-be-gathered file except for human body information represented by the human body feature;

the second obtaining module 42 is configured to integrate, for each human body image of the to-be-gathered document, the human body features and the attribute features thereof to obtain a clustering feature of each human body image of the to-be-gathered document;

and the document gathering module 43 is configured to cluster the plurality of human body images to be gathered to form a human body image file according to the similarity between the clustering features of the plurality of human body images to be gathered.

The human body image gathering device 40 acquires the human body characteristics and attribute characteristics in each human body image to be gathered; integrating the human body characteristics and the attribute characteristics of each human body image of the to-be-gathered file to obtain the clustering characteristics of each human body image of the to-be-gathered file; and clustering the human body images of the plurality of files to be gathered according to the similarity among the clustering characteristics of the human body images of the plurality of files to be gathered to form a human body image file. In the document gathering process of the human body image, the human body characteristics of the human body image to be gathered are combined with the attribute characteristics, the input information amount of the human body image to be gathered is enlarged, the document gathering strategy of the human body image is improved by means of the enlarged input information amount, and therefore the accuracy rate of the document gathering of the human body image is improved.

In one embodiment, the first obtaining module 41 is further configured to input the human body image of each to-be-gathered file into the trained feature extraction network, so as to obtain the human body feature of the human body image of each to-be-gathered file; and inputting the human body image of each document to be gathered into the trained attribute recognition network to obtain the attribute characteristics of the human body image of each document to be gathered.

In one embodiment, the second obtaining module 42 is further configured to, for each human body image of the to-be-gathered file, input human body features and attribute features thereof into the trained fully-connected network for integration, so as to obtain a clustering feature of each human body image of the to-be-gathered file.

In one embodiment, the document gathering module 43 is further configured to divide the human body images to be gathered into a plurality of co-location image sets according to the shooting location of each human body image to be gathered; clustering the human body images of the multiple files to be gathered in each same-place image set according to the similarity between the clustering characteristics of the human body images of the files to be gathered in the same-place image set to obtain multiple initial files;

determining the file attribute of each initial file in a plurality of initial files according to the attribute characteristics of the human body image in the initial file;

clustering a plurality of initial files with the same file attribute to form a file set with the same attribute;

and merging the plurality of initial archives according to the similarity among the plurality of initial archives in the same attribute archives set to obtain the human body image archives.

In one embodiment, the human body image gathering device 40 further includes a first removing module, configured to merge the plurality of initial files according to similarities between the plurality of initial files in the same attribute file set to obtain a human body image file, traverse each human body image in the human body image file, and if there is a case that an attribute feature of the human body image does not match an attribute of the human body image file in which the human body image is located, remove the human body image from the human body image file in which the human body image is located, and store the human body image in a scattered image set.

In one embodiment, the human body image gathering device 40 further includes a second removing module, configured to cluster the multiple human body images to be gathered according to similarity between the clustering features of the multiple human body images to be gathered to form a human body image archive, traverse each human body image in the human body image archive, and if a spatiotemporal logic conflict exists between the human body images in the human body image archive, remove one of the two human body images with the spatiotemporal logic conflict from the human body image archive according to similarity between the current human body image with the spatiotemporal conflict and clustering features of other human body images in the current human body image archive, and store the one of the two human body images with the spatiotemporal logic conflict to a scattered image set.

In one embodiment, the human body image gathering device 40 further includes a third removing module, configured to remove the human body image from the human body image file and store the human body image in the scattered image set when there is only one human body image in the human body image file.

In one embodiment, the human body image clustering device 40 further includes a merging module, configured to calculate a similarity between each human body image in the scattered image set and a clustering feature of each human body image in the human body image archive with the same attribute;

and under the condition that the average value of the similarity between the clustering features of any one human body image in the scattered image set and all the human body images in the human body image file with the same attribute exceeds a set similarity threshold value and no space-time logic conflict exists between the human body image in the scattered image set and all the human body images in the human body image file with the same attribute, merging the human body images in the scattered image set into the human body image file.

In one embodiment, the human body image clustering device 40 further includes a scattered archive module, configured to cluster a plurality of human body images in a scattered image set, which have the same attribute characteristics and do not have spatio-temporal logic conflicts, to form a plurality of same-attribute image sets;

The above modules may be functional modules or program modules, and may be implemented by software or hardware. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 5. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The memory provides an environment for the operating system and the computer programs to run in the non-volatile storage medium. The database of the computer device is used for storing a preset configuration information set. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to realize the human body image gathering method.

In one embodiment, a computer device is provided, which may be a terminal. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The memory provides an environment for the operating system and the computer programs to run in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of documenting a human image. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.

Those skilled in the art will appreciate that the architecture shown in fig. 5 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:

In one embodiment, the processor, when executing the computer program, further performs the steps of:

dividing the human body images of the multiple documents to be gathered into a plurality of image sets in the same place according to the shooting place of the human body image of each document to be gathered; clustering the human body images of the multiple files to be gathered in each same-place image set according to the similarity between the clustering characteristics of the human body images of the files to be gathered in the same-place image set to obtain multiple initial files;

In one embodiment, after the initial files are merged according to the similarity between the initial files in the same attribute file set to obtain the human body image file, the processor executes the computer program to further implement the following steps:

and traversing each human body image in the human body image file, if the attribute characteristics of the human body image do not accord with the file attributes of the human body image file where the human body image is located, removing the human body image from the human body image file where the human body image is located, and storing the human body image into a scattered image set.

In one embodiment, after clustering the human body images of the multiple documents to be gathered according to the similarity between the clustering features of the human body images of the multiple documents to be gathered to form the human body image archive, the processor executes the computer program to further implement the following steps:

and under the condition that the human body image file only has one human body image, removing the human body image from the human body image file and storing the human body image into a scattered image set.

calculating the similarity between the clustering characteristics of each human body image in the scattered image set and each human body image in the human body image file with the same attribute;

The storage medium acquires the human body characteristics and the attribute characteristics in each human body image of the document to be gathered; integrating the human body characteristics and the attribute characteristics of each human body image of the to-be-gathered file to obtain the clustering characteristics of each human body image of the to-be-gathered file; and clustering the human body images of the plurality of files to be gathered according to the similarity among the clustering characteristics of the human body images of the plurality of files to be gathered to form a human body image file. In the document gathering process of the human body image, the human body characteristics of the human body image to be gathered are combined with the attribute characteristics, the input information amount of the human body image to be gathered is enlarged, the document gathering strategy of the human body image is improved by means of the enlarged input information amount, and therefore the accuracy rate of the document gathering of the human body image is improved.

It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to be limiting. All other embodiments, which can be derived by a person skilled in the art from the examples provided herein without any inventive step, shall fall within the scope of protection of the present application.

It is obvious that the drawings are only examples or embodiments of the present application, and it is obvious to those skilled in the art that the present application can be applied to other similar cases according to the drawings without creative efforts. Moreover, it should be appreciated that in the development of any such actual implementation, as in any engineering or design project, numerous implementation-specific decisions must be made to achieve the developers' specific goals, such as compliance with system-related and business-related constraints, which may vary from one implementation to another.

The term "embodiment" is used herein to mean that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is to be expressly or implicitly understood by one of ordinary skill in the art that the embodiments described in this application may be combined with other embodiments without conflict.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the patent protection. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A method for gathering files of human body images is characterized by comprising the following steps:

2. The method for gathering the human body image according to claim 1, wherein the obtaining of the human body features and the attribute features in each human body image to be gathered comprises:

3. The method for gathering the human body images according to claim 1, wherein the integrating the human body features and the attribute features of each human body image to be gathered to obtain the clustering features of each human body image to be gathered comprises:

4. The human body image gathering method according to claim 1, wherein the clustering the human body images of the plurality of documents to be gathered according to the similarity between the clustering features of the human body images of the plurality of documents to be gathered to form the human body image file comprises:

5. The method for gathering human body images according to claim 4, wherein after the initial archives are merged according to similarities between the initial archives in the homonymy archives set to obtain the human body image archives, the method further comprises:

6. The human body image gathering method as claimed in claim 1, wherein after clustering the human body images of the plurality of documents to be gathered according to the similarity between the clustering features of the human body images of the plurality of documents to be gathered to form a human body image archive, the method further comprises:

7. The method for gathering the human body image according to claim 1, wherein the method further comprises:

8. The method for gathering the human body image according to any one of claims 5 to 7, wherein the method further comprises:

9. The method for gathering the human body image according to any one of claims 5 to 7, wherein the method further comprises:

10. The method for gathering the human body image according to claim 1, wherein the attribute features comprise: one or more of a hair style feature, a hat feature, and a garment color feature.

11. A gathering device for human body images, the device comprising: the system comprises a first acquisition module, a second acquisition module and a gear-gathering module:

12. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 10 are implemented by the processor when executing the computer program.

13. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 10.