CN109670470B - Pedestrian relationship identification method, device and system and electronic equipment - Google Patents
Pedestrian relationship identification method, device and system and electronic equipment Download PDFInfo
- Publication number
- CN109670470B CN109670470B CN201811616722.4A CN201811616722A CN109670470B CN 109670470 B CN109670470 B CN 109670470B CN 201811616722 A CN201811616722 A CN 201811616722A CN 109670470 B CN109670470 B CN 109670470B
- Authority
- CN
- China
- Prior art keywords
- relationship
- group
- target
- target group
- score
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a pedestrian relationship identification method, a device, a system and electronic equipment, which relate to the technical field of image processing and comprise the steps of analyzing the direction information of pedestrians in a target range in real time according to a group identification algorithm to obtain a target group; extracting the face features of all group members in a target group; performing pairwise cross verification on the group members according to the face characteristics to obtain a membership score between each two group members; determining target group members according to the membership score and a preset threshold value, and generating a membership network according to the target group members and the membership score; and connecting the target group members in the relationship network through edges, wherein the edges take the relationship score as weight. The invention can identify the more complex relationship chain which possibly appears in the pedestrian group, and has stronger practicability.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a method, an apparatus, a system, and an electronic device for identifying a pedestrian relationship.
Background
With the popularization of cameras, image processing techniques are also gradually applied to various scenes for identifying pedestrian relationships, such as shopping malls, streets, squares and the like. The pedestrian relationship is identified in advance, so that merchants can be helped to identify members or frequent customers in advance, and accurate marketing and other services can be provided for the customers; other systems can be assisted in the public safety field to identify the single falling masses and suspicious groups, thereby maintaining public safety and public safety. The human face has a certain inheritance, and a method for judging whether the human face belongs to the parent-child relationship or not according to two human face pictures is provided. However, for pedestrians with complex relationships, these methods for identifying relationships are difficult to apply, and cannot identify relationships of people.
Disclosure of Invention
In view of the above, the present invention provides a method, an apparatus, a system and an electronic device for identifying a pedestrian relationship, so as to identify a complex relationship of a pedestrian.
In a first aspect, an embodiment of the present invention provides a pedestrian relationship identification method, where the method includes: analyzing the direction information of the pedestrians in the target range in real time according to a group identification algorithm to obtain a target group; extracting the face features of all group members in a target group; performing pairwise cross verification on the group members according to the face characteristics to obtain a membership score between each two group members; determining target group members according to the membership score and a preset threshold value, and generating a membership network according to the target group members and the membership score; and connecting the target group members in the relationship network through edges, wherein the edges take the relationship score as weight.
With reference to the first aspect, an embodiment of the present invention provides a first possible implementation manner of the first aspect, where the method further includes: respectively calculating the confidence scores of the families for the plurality of the relationship networks through a graph theory algorithm; and sequencing the confidence scores of the families, and respectively generating the confidence degrees of the plurality of the family relationship networks according to the sequencing result.
With reference to the first aspect and the first possible implementation manner thereof, an embodiment of the present invention provides a second possible implementation manner of the first aspect, where the step of calculating the family confidence scores for the multiple affinity networks through a graph theory algorithm includes: calculating the degree of each target group member; the degree is the times that the target group member has cross relation with other target group members in the relativity relation network; calculating the average degree of the relativity network according to the following formula:
wherein the content of the first and second substances,<k>deg (f) is the average degreei) Is a member f of the target groupiThe degree of (c) is M is the number of target group members; the average weight of the affinity net is calculated according to the following formula:
where K is the average weight, KijIs a weight, P is an edgeThe number of (2); calculating a family confidence score according to the following formula: ScoreF ═<k>K, wherein ScoreF is the confidence score of the family, i and j are the numbers of any two target population members, and i is more than or equal to 1<j≤M。
With reference to the first aspect, an embodiment of the present invention provides a third possible implementation manner of the first aspect, where the step of extracting the face features of each group member in the target group includes: extracting the face features of each group member in the target group according to a target algorithm, wherein the target algorithm at least comprises the following steps: an ASM algorithm, an SDM algorithm, or a neural network algorithm.
With reference to the first aspect, an embodiment of the present invention provides a fourth possible implementation manner of the first aspect, where the step of performing pairwise cross validation on the group members according to the face features to obtain a membership score between each two group members includes: and performing pairwise cross validation on the group members through an NRML algorithm to obtain the membership score between every two group members.
With reference to the first aspect, an embodiment of the present invention provides a fifth possible implementation manner of the first aspect, where the step of determining a target group member according to the affinity score and a preset threshold, and generating an affinity network according to the target group member and the affinity score includes: judging whether the membership score is larger than a preset threshold value or not; if so, determining the group member corresponding to the membership score as the target group member; and generating a relationship network according to the target group member and the relationship score.
In a second aspect, an embodiment of the present invention further provides a pedestrian relationship identification apparatus, including: the group determination module is used for analyzing the direction information of the pedestrians in the target range in real time according to a group identification algorithm to obtain a target group; the characteristic extraction module is used for extracting the face characteristics of all group members in the target group; the scoring module is used for respectively carrying out pairwise cross verification on the group members according to the face characteristics to obtain the membership score between each two group members; the identification module is used for determining target group members according to the membership score and a preset threshold value and generating a membership network according to the target group members and the membership score; and connecting the target group members in the relationship network through edges, wherein the edges take the relationship score as weight.
With reference to the second aspect, an embodiment of the present invention provides a first possible implementation manner of the second aspect, where the apparatus further includes: a confidence module to: respectively calculating the confidence scores of the families for the plurality of the relationship networks through a graph theory algorithm; and sequencing the confidence scores of the families, and respectively generating the confidence degrees of the plurality of the family relationship networks according to the sequencing result.
In a third aspect, an embodiment of the present invention further provides a pedestrian relationship identification system, including: the system comprises a family library, an image acquisition module, a group library, a group identification module, a face detection module and a relative identification module; the image acquisition module is used for acquiring pedestrian information and sending the pedestrian information to the group library; the group library is used for receiving and storing pedestrian information; the group library is in communication connection with the group identification module; the group identification module is used for analyzing the pedestrian information in the target range in real time according to a group identification algorithm to obtain a target group; the face detection module is used for extracting the face features of all group members in the target group; the face detection module is in communication connection with the group identification module; the relative identification module is used for respectively carrying out pairwise cross verification on the group members according to the face characteristics to obtain a relative relationship score between each two group members; the relative identification module is also used for determining a target group member according to the relative score and a preset threshold value and generating a relative network according to the target group member and the relative score; connecting target group members in the relativity network through edges, wherein the edges take the relativity score as weight; the family library is in communication connection with the relative identification module and is used for storing the information generated by the relative identification module.
In a fourth aspect, an embodiment of the present invention further provides an electronic device, including: a memory for storing a computer program; a processor for executing a computer program stored in the memory to cause the apparatus to perform the pedestrian relationship identification method of any one of the first aspects.
The embodiment of the invention has the following beneficial effects: the embodiment of the invention provides a pedestrian relationship identification method, a device, a system and electronic equipment, which analyze the orientation information of pedestrians in a target range in real time through a group identification algorithm so as to obtain a target group in real time, extract the face characteristics of each group member in the target group, perform pairwise cross verification on the group members by taking the face characteristics as the basis, thereby obtaining the membership score between each two group members, and then determine the target group members according to the membership score and a preset threshold value, thereby generating a membership network by combining the membership score. The embodiment of the invention can identify the more complex relationship chain which possibly appears in the pedestrian group, and has stronger practicability.
Additional features and advantages of the disclosure will be set forth in the description which follows, or in part may be learned by the practice of the above-described techniques of the disclosure, or may be learned by practice of the disclosure.
In order to make the aforementioned objects, features and advantages of the present disclosure more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a pedestrian relationship identification method according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of a pedestrian relationship identification method according to an embodiment of the present invention;
fig. 3 is an implementation manner of a face key point model in the pedestrian relationship identification method according to the embodiment of the present invention;
fig. 4 is a schematic diagram of a relationship network in the pedestrian relationship identification method according to the embodiment of the present invention;
fig. 5 is a schematic block diagram of a pedestrian relationship identification system according to an embodiment of the present invention;
fig. 6 is a block diagram of a pedestrian relationship identification apparatus according to an embodiment of the present invention;
fig. 7 is a block diagram schematically illustrating a structure of an electronic device according to an embodiment of the present invention.
Icon:
61-a population determination module; 62-a feature extraction module; 63-a scoring module; 64-an identification module; 71-a memory; 72-processor.
Detailed Description
To make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
At present, the existing algorithm of other parent-child relationship is difficult to be used in the scene of pedestrian relationship identification, and the following limitations exist specifically: 1. the pedestrian scene is complicated, is difficult to screen out the target pedestrian group, in same camera field of vision, can have a plurality of pedestrian's people's face simultaneously, is difficult to judge which pedestrian belongs to the target pedestrian group that needs to discern. On the other hand, since the pedestrian is in motion, it is difficult to completely photograph each face in the target pedestrian population in one camera view. 2. The relativity is complex, most of the current algorithms for judging the relativity are aimed at 1 to 1 judgment, such as judging the relationship between father and son, father and daughter, mother and son and mother and daughter, but a complex relativity chain, such as aunt, grandma, epigen and the like, may appear in a pedestrian population, and an identification method for the complex relativity chain is lacked.
Based on this, the pedestrian relationship identification method, the device, the system and the electronic device provided by the embodiment of the invention can identify the pedestrian group by tracking the pedestrian track and the intimacy, then cross identification verification is carried out on each member in the group by adopting the intimacy identification algorithm, the intimacy relationship among multiple persons is identified, the confidence score of the family is further calculated by utilizing the graph theory algorithm, and the family with the intimacy relationship can be identified more accurately.
For the convenience of understanding the embodiment, a detailed description will be first given of a pedestrian relationship identification method disclosed in the embodiment of the present invention.
Example 1
An embodiment 1 of the present invention provides a method for identifying a pedestrian relationship, which is described with reference to a flowchart of the method for identifying a pedestrian relationship shown in fig. 1, and the method includes the following steps:
and S102, analyzing the direction information of the pedestrians in the target range in real time according to a group recognition algorithm to obtain a target group.
The group identification algorithm can judge the group affiliation of the pedestrian by tracking the change of the intimacy degree between the pedestrian and the group, so as to identify the group. And identifying the group affiliation of each pedestrian through a certain group identification algorithm by using the direction information of each pedestrian. The target population may be any one of a plurality of populations. The target range may be considered to be set as required.
And step S104, extracting the face features of all group members in the target group.
Referring to an implementation of the face key point model shown in fig. 3, eyebrow, eye, mouth, nose and face contour feature key points of a face can be detected by using a face key point detection algorithm.
And step S106, performing pairwise cross validation on the group members according to the face characteristics to obtain the membership score between each two group members.
The high or low of the affinity score is used to describe the magnitude of the likelihood of affinity between two population members. And (3) taking the face characteristics as the input of a relationship recognition algorithm, and carrying out pairwise calculation on all group members to obtain a relationship score.
Step S108, determining target group members according to the membership score and a preset threshold value, and generating a membership network according to the target group members and the membership score; and connecting the target group members in the relationship network through edges, wherein the edges take the relationship score as weight.
The preset threshold may be set or adjusted empirically. Whether the group corresponding to the membership score is a target group member or not can be determined by comparing the membership score with a preset threshold value. The members of the target group are all related directly or indirectly. The affinity network is used for describing affinity between target group members, the target group members can be used as nodes in the affinity network, and the affinity score can be used as the weight of an edge connecting two nodes in the affinity network.
The embodiment of the invention provides a pedestrian relationship identification method, which comprises the steps of analyzing the direction information of pedestrians in a target range in real time through a group identification algorithm so as to obtain a target group in real time, extracting the face characteristics of all group members in the target group, carrying out pairwise cross verification on the group members by taking the face characteristics as the basis so as to obtain the membership score between every two group members, and determining the target group members according to the membership score and a preset threshold value so as to generate a membership network by combining the membership score. The embodiment of the invention can identify the more complex relationship chain which possibly appears in the pedestrian group, and has stronger practicability.
In view of facilitating evaluation of the recognition result, the method further comprises the steps of: respectively calculating the confidence scores of the families for the plurality of the relationship networks through a graph theory algorithm; and sequencing the confidence scores of the families, and respectively generating the confidence degrees of the plurality of the family relationship networks according to the sequencing result.
The use of the graph theory algorithm is based on a naive idea: the more intricate and complicated the estimated relationships between individuals, the higher the confidence that the individuals as a whole are in a family. And evaluating the complexity of the relationship among the relationships in the relationship network by using the family confidence score, and sequencing the family confidence score to obtain the credibility of the relationship network.
In order to facilitate the analysis by using the family confidence score, the step of respectively calculating the family confidence score for a plurality of relationship networks by a graph theory algorithm comprises the following steps:
(1) calculating the degree of each target group member; the degree is the number of times that the target group member has a cross relationship with other target group members in the relationship network.
Referring to the schematic diagram of the relationship network in the pedestrian relationship identification method shown in fig. 4, the degree is the number of edges from adjacent nodes to a certain node, the target group member is taken as a node, and the calculation degree is the number of edges from the node to adjacent target group members, that is, the number of times that the target group member has a cross relationship with other target group members in the relationship network.
(2) Calculating the average degree of the relativity network according to the following formula:
wherein the content of the first and second substances,<k>deg (f) is the average degreei) Is a member f of the target groupiAnd M is the number of target group members.
As can be seen from the graph theory analysis, < k > is ≧ 1, and the more complex the cross-relationship between members, the greater < k >.
(3) The average weight of the affinity net is calculated according to the following formula:
where K is the average weight, KijP is the number of edges.
(4) Calculating a family confidence score according to the following formula: ScoreF is a family confidence score, i and j are numbers of any two target population members, and i is more than or equal to 1 and less than or equal to j and less than or equal to M.
And taking the product of the average degree and the average weight as the family confidence score ScoreF of the target group member. The information of the family and the members of the family can be recorded in a family library, and the ranking in the family library is based on ScoreF, and when ScoreF is higher, the identification result is more credible.
In order to extract the face features more accurately or efficiently, the step of extracting the face features of each group member in the target group comprises the following steps: extracting the face features of each group member in the target group according to a target algorithm, wherein the target algorithm at least comprises the following steps: an ASM algorithm, an SDM algorithm, or a neural network algorithm.
The ASM (Active Shape Model) algorithm is divided into two steps of training and searching. During training, the position constraint of each characteristic point is established, and the local characteristic of each specific point is constructed. And during searching, iteratively matching. An SDM (Supervised Descent Method) algorithm is a supervision and Descent Method, and belongs to a Method for solving the problem of nonlinear minimization. The neural network algorithm has the characteristics of large-scale parallel processing, distributed information storage, good self-organizing and self-learning capabilities and the like. The specific algorithm used can be determined according to actual needs.
In order to more accurately describe the relationship between the group members, the step of performing pairwise cross validation on the group members according to the human face characteristics to obtain the relationship score between every two group members comprises the following steps: and (3) performing pairwise cross validation on the group members through an NRML (neighbor likelihood Learning) algorithm to obtain the membership score between every two group members.
Referring to a schematic diagram of a relationship network in the pedestrian relationship identification method shown in fig. 4, one implementation manner of the relationship identification algorithm is as follows:
recording the population as G, recording the total number of population members in the population as N, and recording the population members as gi, namely G ═ gi |1 ≦ i ≦ N }; calculating the relative relationship score of any two group members gi and gj in the group by using NRML algorithm to obtain kijWherein 1 is less than or equal to i<j≤N。
In order to obtain a more practical relative network, determining target group members according to the relative score and a preset threshold value, and generating the relative network according to the target group members and the relative score, wherein the method comprises the following steps: judging whether the membership score is larger than a preset threshold value or not; if so, determining the group member corresponding to the membership score as the target group member; and generating a relationship network according to the target group member and the relationship score.
For example, using a preset threshold kT, a low-value filtering is performed on the group member affinity scores, that is: when k isij<kT, then the population members gi and gj are identified as not relatives; will be when kij>When kT is carried out, the corresponding M group members have the relationship of relatives, the M group members are determined as target group members, the M group members form a family F, and the family F can be recorded as the individual FiIf the logarithm of the relationship is P, i.e., the number of edges, F ═ FiI is more than or equal to 1 and less than or equal to M, and the I form a undirected network graph with weight, namely the relativity network.
Referring to the flow diagram of the pedestrian relationship identification method shown in fig. 2, the method, the device, the system and the electronic device for identifying the pedestrian relationship provided by the embodiment of the invention effectively eliminate the contingency of face similarity between strangers, and break through the application range that only 1-to-1 relationship can be identified in the traditional algorithm through a cross-validation method. In addition, the reliability of overall judgment of the plurality of cross relatives is improved through the analysis and calculation of the relative network graph. The method specifically comprises the steps of identifying a pedestrian group by tracking the track and the intimacy of pedestrians, and preliminarily screening the possibility that strangers with similar growth phases are misjudged as relatives; then, cross identification verification is carried out on each member in the group by adopting a relative identification algorithm, and possible relative can be screened out more reliably according to the face characteristics; and finally, a graph theory algorithm is utilized to endow the families with the cross relativity relationship with higher confidence scores, so that the confidence level of the family identification result is further improved.
Example 2
Embodiment 2 of the present invention provides a pedestrian relationship recognition apparatus, referring to a structural block diagram of the pedestrian relationship recognition apparatus shown in fig. 6, the apparatus including:
the group determination module 61 is used for analyzing the direction information of the pedestrians in the target range in real time according to a group identification algorithm to obtain a target group; a feature extraction module 62, configured to extract face features of each group member in the target group; the scoring module 63 is configured to perform pairwise cross validation on the group members according to the face features to obtain a membership score between each two group members; the identification module 64 is used for determining the target group members according to the membership score and a preset threshold value, and generating a membership network according to the target group members and the membership score; and connecting the target group members in the relationship network through edges, wherein the edges take the relationship score as weight.
The device also includes: a confidence module to: respectively calculating the confidence scores of the families for the plurality of the relationship networks through a graph theory algorithm; sequencing the confidence scores of the families and respectively generating the confidence degrees of the plurality of the family relationship networks according to the sequencing result
The device provided by the embodiment of the present invention has the same implementation principle and technical effect as the method embodiments, and for the sake of brief description, reference may be made to the corresponding contents in the method embodiments without reference to the device embodiments.
Example 3
An embodiment 3 of the present invention provides a pedestrian relationship recognition system, which refers to a schematic block diagram of a structure of a pedestrian relationship recognition system shown in fig. 5, and includes:
the system comprises a family library, an image acquisition module, a group library, a group identification module, a face detection module and a relative identification module; the image acquisition module is used for acquiring pedestrian information and sending the pedestrian information to the group library; the group library is used for receiving and storing pedestrian information; the group library is in communication connection with the group identification module; the group identification module is used for analyzing the pedestrian information in the target range in real time according to a group identification algorithm to obtain a target group; the face detection module is used for extracting the face features of all group members in the target group; the face detection module is in communication connection with the group identification module; the relative identification module is used for respectively carrying out pairwise cross verification on the group members according to the face characteristics to obtain a relative relationship score between each two group members; the relative identification module is also used for determining a target group member according to the relative score and a preset threshold value and generating a relative network according to the target group member and the relative score; connecting target group members in the relativity network through edges, wherein the edges take the relativity score as weight; the family library is in communication connection with the relative identification module and is used for storing the information generated by the relative identification module.
The family library comprises the family confidence score of each family, the ID numbers of family members, the face characteristics and the relatives. The family is the above-mentioned relativity network. The image acquisition module is used for controlling the camera to record the video. The group library comprises group ID numbers, group orientation information, group members and orientation information of the group members, and is used for marking and recording different groups, and the group identification module analyzes the group attribution of each pedestrian at the moment by utilizing the orientation information and the historical group attribution information of each pedestrian. The face detection module is used for identifying and extracting key feature points of the face. The relative identification module is used for analyzing the members in the group pairwise, and finally identifying the family members in the group pairwise through graph theory analysis.
It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the system described above may refer to the corresponding process in the foregoing method embodiment, and is not described herein again.
Example 4
Embodiment 4 of the present invention provides an electronic device, referring to a block diagram schematically illustrating a structure of the electronic device shown in fig. 7, where the electronic device includes:
a memory 71 for storing a computer program; a processor 72 for executing the computer program stored in the memory to cause the apparatus to perform the pedestrian relationship identification method as in any one of embodiment 1.
The electronic device provided by the embodiment of the invention has the same technical characteristics as the pedestrian relationship identification method provided by the embodiment, so that the same technical problems can be solved, and the same technical effects can be achieved.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A pedestrian relationship identification method, comprising:
analyzing the direction information of the pedestrians in the target range in real time according to a group identification algorithm to obtain a target group;
extracting the face features of all group members in the target group;
performing pairwise cross validation on the group members according to the face features to obtain a relationship score between each two group members;
determining target group members according to the affiliation scores and a preset threshold value, and generating an affiliation network according to the target group members and the affiliation scores; and connecting the target group members in the relationship network through edges, wherein the edges take the relationship scores as weights.
2. The pedestrian relationship identification method according to claim 1, characterized by further comprising:
respectively calculating family confidence scores for the plurality of the relationship networks through a graph theory algorithm;
and sequencing the plurality of family confidence scores, and respectively generating the credibility of the plurality of the relatives according to the sequencing result.
3. The pedestrian relationship identification method according to claim 2, wherein the step of calculating the family confidence scores for the plurality of the relationship networks, respectively, by a graph theory algorithm includes:
calculating the degree of each target group member; the degree is the number of times that the target group member has a cross relationship with other target group members in the relationship network;
calculating the average degree of the relativity network according to the following formula:
wherein the content of the first and second substances,<k>deg (f) is the average degreei) Is a member f of the target groupiThe degree of (c) is M is the number of target group members;
calculating the average weight of the relationship net according to the following formula:
where K is the average weight, KijIs weight, P is number of sides;
calculating the family confidence score according to the following formula:
ScoreF＝<k>*K
wherein ScoreF is the confidence score of the family, i and j are the numbers of any two target group members, i is more than or equal to 1 and j is more than or equal to M.
4. The pedestrian relationship recognition method according to claim 1, wherein the step of extracting the face features of each group member in the target group includes:
extracting the face features of each group member in the target group according to a target algorithm, wherein the target algorithm at least comprises one of the following steps: an ASM algorithm, an SDM algorithm, or a neural network algorithm.
5. The pedestrian relationship identification method according to claim 1, wherein the step of performing pairwise cross validation on the group members according to the face features to obtain a membership score between each two group members comprises:
and performing pairwise cross validation on the group members through an NRML algorithm to obtain the membership score between every two group members.
6. The pedestrian relationship identification method according to claim 1, wherein the step of determining a target group member according to the relationship score and a preset threshold value, and generating a relationship net according to the target group member and the relationship score comprises:
judging whether the membership score is larger than a preset threshold value or not;
if so, determining the group member corresponding to the membership score as a target group member;
and generating a relationship network according to the target group members and the relationship scores.
7. A pedestrian relationship recognition apparatus, characterized by comprising:
the group determination module is used for analyzing the direction information of the pedestrians in the target range in real time according to a group identification algorithm to obtain a target group;
the characteristic extraction module is used for extracting the face characteristics of all group members in the target group;
the scoring module is used for respectively carrying out pairwise cross validation on the group members according to the face features to obtain the membership score between each two group members;
the identification module is used for determining a target group member according to the affiliation score and a preset threshold value and generating an affiliation network according to the target group member and the affiliation score; and connecting the target group members in the relationship network through edges, wherein the edges take the relationship scores as weights.
8. The pedestrian relationship recognition apparatus of claim 7, further comprising a confidence module to:
respectively calculating family confidence scores for the plurality of the relationship networks through a graph theory algorithm;
and sequencing the plurality of family confidence scores, and respectively generating the credibility of the plurality of the relatives according to the sequencing result.
9. A pedestrian relationship recognition system, comprising: the system comprises a family library, an image acquisition module, a group library, a group identification module, a face detection module and a relative identification module;
the image acquisition module is used for acquiring pedestrian information and sending the pedestrian information to the group library;
the group library is used for receiving and storing the pedestrian information; the group library is in communication connection with the group identification module;
the group identification module is used for analyzing pedestrian information in a target range in real time according to a group identification algorithm to obtain a target group;
the face detection module is used for extracting the face features of all group members in the target group; the face detection module is in communication connection with the group identification module;
the relative identification module is used for respectively carrying out pairwise cross verification on the group members according to the face features to obtain a relative relationship score between each two group members;
the relative identification module is further used for determining a target group member according to the relative score and a preset threshold value, and generating a relative network according to the target group member and the relative score; connecting the target group members in the relationship net through edges, wherein the edges take the relationship scores as weights;
the family library is in communication connection with the relative identification module and is used for storing the information generated by the relative identification module.
10. An electronic device, comprising:
a memory for storing a computer program;
a processor for executing the computer program stored in the memory to cause the apparatus to perform the pedestrian relationship identification method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811616722.4A CN109670470B (en) | 2018-12-27 | 2018-12-27 | Pedestrian relationship identification method, device and system and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811616722.4A CN109670470B (en) | 2018-12-27 | 2018-12-27 | Pedestrian relationship identification method, device and system and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109670470A CN109670470A (en) | 2019-04-23 |
| CN109670470B true CN109670470B (en) | 2020-12-01 |
Family
ID=66146860
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811616722.4A Active CN109670470B (en) | 2018-12-27 | 2018-12-27 | Pedestrian relationship identification method, device and system and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109670470B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112825083B (en) * | 2019-11-20 | 2024-06-07 | 深圳云天励飞技术有限公司 | Method, device and equipment for constructing group relation network and readable storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2016091383A (en) * | 2014-11-06 | 2016-05-23 | 富士通株式会社 | Portable terminal apparatus, screen control method and screen control program |
| CN106776781B (en) * | 2016-11-11 | 2018-08-24 | 深圳云天励飞技术有限公司 | A kind of human relation network analysis method and device |
| CN107992795B (en) * | 2017-10-27 | 2021-08-31 | 江西高创保安服务技术有限公司 | Group partner based on population information base and real name call record and its head and target identification method |
| CN108596735A (en) * | 2018-04-28 | 2018-09-28 | 北京旷视科技有限公司 | Information-pushing method, apparatus and system |
-
2018
- 2018-12-27 CN CN201811616722.4A patent/CN109670470B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN109670470A (en) | 2019-04-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108182394B (en) | Convolutional neural network training method, face recognition method and face recognition device | |
| Fu et al. | Fast crowd density estimation with convolutional neural networks | |
| CN108921051B (en) | Pedestrian attribute identification network and technology based on cyclic neural network attention model | |
| Ding et al. | Violence detection in video by using 3D convolutional neural networks | |
| US20200012923A1 (en) | Computer device for training a deep neural network | |
| JP6598746B2 (en) | Apparatus, program, and method for tracking an object in consideration of an image area of another object | |
| CN108229335A (en) | It is associated with face identification method and device, electronic equipment, storage medium, program | |
| CN109977895B (en) | Wild animal video target detection method based on multi-feature map fusion | |
| CN111222500A (en) | Label extraction method and device | |
| CN111738054B (en) | Behavior anomaly detection method based on space-time self-encoder network and space-time CNN | |
| CN110674680B (en) | Living body identification method, living body identification device and storage medium | |
| CN113033523B (en) | Method and system for constructing falling judgment model and falling judgment method and system | |
| CN112926522B (en) | Behavior recognition method based on skeleton gesture and space-time diagram convolution network | |
| CN111177469A (en) | Face retrieval method and face retrieval device | |
| Harrou et al. | Malicious attacks detection in crowded areas using deep learning-based approach | |
| CN111967433A (en) | Action identification method based on self-supervision learning network | |
| CN113205002A (en) | Low-definition face recognition method, device, equipment and medium for unlimited video monitoring | |
| CN107633527B (en) | Target tracking method and device based on full convolution neural network | |
| CN113076963B (en) | Image recognition method and device and computer readable storage medium | |
| CN109670470B (en) | Pedestrian relationship identification method, device and system and electronic equipment | |
| CN117765348A (en) | Target detection model deployment method, target detection method and electronic equipment | |
| Mucha et al. | Depth and thermal images in face detection-a detailed comparison between image modalities | |
| CN114120076B (en) | Cross-view video gait recognition method based on gait motion estimation | |
| CN111753601A (en) | Image processing method and device and storage medium | |
| CN112749686B (en) | Image detection method, image detection device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |