A method, system, device and medium for analyzing the behavior of a group of people

Abstract

The present invention provides a method, system, device and medium for analyzing the collective behavior of people, including: obtaining face images of target objects in a specified place, and performing feature analysis on the face images to obtain face features corresponding to the face images that meet the set face metrics; wherein, the face metrics include: face angle and / or face quality score; clustering the face features to obtain face profiles corresponding to the same target object; cleaning the face profiles of each target object, and selecting the cleaned corresponding face profiles according to different target areas in the specified place for statistical analysis of the collective behavior of people to obtain the statistical results of the collective behavior of people corresponding to the target areas, wherein the cleaning process includes: cleaning the face profiles according to the gender distribution corresponding to the face features in the face profiles, the number of face features that cannot be merged, and / or the increment of face features within a set time; the present invention can effectively improve the accuracy of statistical analysis of the collective behavior of people.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a method, system, device and medium for analyzing the behavior of a group of people. Background Art

[0002] With the increasing development of artificial intelligence and the optimization and iteration of model algorithms such as deep learning and neural networks, applications based on pedestrians and faces have received increasing attention, and the demand analysis of pedestrians is particularly important. However, the existing technologies are difficult to accurately analyze the needs of new people. How to accurately analyze the big data of pedestrian group behavior by using resources such as surveillance images of the target place and pedestrian information records is a major problem that needs to be solved urgently at present. Taking a shopping mall as an example, how to improve the accuracy of analyzing the behavior of a group of people in the shopping mall and improve the shopping experience of users is the most important work for shopping mall operators.

[0003] The traditional methods for analyzing the behavior of a group of people mainly include two methods: detecting the mobile phone mac address and the number of wifi connections. For the methods of mac address detection and wifi connection number, first, most mobile phones can block wifi detection. Second, with the development of 5G, the willingness of people to connect to wifi has decreased significantly and almost no one will connect to wifi anymore. Therefore, the behavior of the group of people detected by these two methods is also less than the real data. None of the above methods can perceive the attributes of the behavior of a group of people, such as age distribution, gender, etc. Therefore, the accuracy of analyzing the behavior of a group of people is not high. Summary of the Invention

[0004] In view of the above problems existing in the prior art, the present invention proposes a method, system, device and medium for analyzing the behavior of a group of people, mainly solving the problem of low accuracy in counting the behavior of a group of people by traditional methods.

[0005] In order to achieve the above object and other objects, the technical solutions adopted by the present invention are as follows.

[0006] A method for analyzing the behavior of a group of people includes:

[0007] Obtaining a face image of a target object in a specified place, and performing feature analysis on the face image to obtain a face feature corresponding to the face image that meets the set face index; wherein, the face index includes: face angle and / or face quality score;

[0008] Clustering the face features to obtain a face profile corresponding to the same target object;

[0009] Clean the face files of each target object, and select the corresponding cleaned face files according to different target areas in the specified venue for statistical analysis of the group behavior of people, so as to obtain the statistical results of the group behavior of people in the corresponding target area. Among them, the cleaning process includes: cleaning the face files according to the gender distribution corresponding to the face features in the face files, the number of face features that cannot be merged, and / or the increment of face features within a set time.

[0010] Optionally, after obtaining the statistical results of the group behavior of people in the corresponding target area, it further includes:

[0011] According to the statistical results of the group behavior of people in different target areas, obtain the depth distribution and residence time distribution of each target object in the specified venue, where the depth represents the number of target areas visited by the target object;

[0012] Adjust the business layout of the specified venue according to the depth distribution and residence time distribution.

[0013] Optionally, cleaning the face files according to the gender distribution corresponding to the face features in the face files includes:

[0014] According to the face image acquisition time nodes corresponding to the face features in the face files, determine one or more capture trajectories included in the face files, and identify the gender of the target object corresponding to the face features included in each capture trajectory according to the corresponding face features;

[0015] Select the number of face features with the highest gender ratio as the denominator, calculate the gender ratio of the target object in each capture trajectory, and determine whether the gender ratio exceeds the set threshold range. If it exceeds, delete the corresponding face file.

[0016] Optionally, cleaning the face files according to the number of face features that cannot be merged includes:

[0017] Merge the face files obtained by clustering according to the face feature similarity. If two or more face files corresponding to the same target object still cannot be merged after exceeding the set time limit, and the number of face features included in the unmergeable face files is less than the set quantity threshold, delete the corresponding face files.

[0018] Optionally, cleaning the face files according to the increment of face features within a set time includes:

[0019] Judge the increment of face features in the corresponding face files within the set time threshold. If the increment exceeds the set increment threshold, delete the corresponding face files.

[0020] Optionally, after cleaning the face files, it further includes:

[0021] Obtain the point information of the acquisition devices corresponding to the face images of each face file, where the point information includes: a unique identifier, geographical location information, and tag information;

[0022] Filter the point information of the acquisition devices according to the proportion of invalid face images obtained by the acquisition devices, and turn off the acquisition devices that exceed the set proportion threshold; or, turn off the acquisition devices corresponding to the face files that cannot be merged, where the invalid face images are determined according to the face metrics.

[0023] Optionally, after obtaining the statistical results of the group behavior of the characters in the target area, it further includes:

[0024] When two face files need to be merged, if the statistical results of the group behavior of the characters have been completed based on the face files to be merged, delete the statistical results, and after completing the merging of the face files, re-perform the statistical analysis of the group behavior of the characters in the corresponding target area to obtain the corrected results of the group behavior of the characters.

[0025] A system for analyzing the group behavior of characters, including:

[0026] A feature acquisition module, configured to acquire face images of target objects in a specified place, and perform feature analysis on the face images to obtain face features corresponding to the face images that meet the set face metrics; where the face metrics include: face angle and / or face quality score;

[0027] An archive acquisition module, configured to cluster the face features to obtain face files corresponding to the same target object;

[0028] A module for statistical analysis of the group behavior of characters, configured to perform cleaning processing on the face files of each target object, and perform statistical analysis of the group behavior of characters on the corresponding face files that have been cleaned according to different target areas in the specified place to obtain the statistical results of the group behavior of characters in the corresponding target area, where the cleaning processing includes: performing cleaning processing on the face files according to the gender distribution corresponding to the face features in the face files, the number of face features that cannot be merged, and / or the increment of face features within a set time.

[0029] A device for analyzing the group behavior of characters, including:

[0030] One or more processors; and

[0031] One or more machine-readable media storing instructions, which when executed by the one or more processors, cause the device to execute the method for analyzing the group behavior of characters.

[0032] A machine-readable medium stores instructions that, when executed by one or more processors, cause a device to perform the method for analyzing the behavior of a group of people described above.

[0033] As described above, a method, system, device, and medium for analyzing the behavior of a group of people according to the present invention have the following beneficial effects.

[0034] By screening face images based on face metrics and cleaning face profiles, the accuracy and reliability of the statistical results of the behavior of a group of people can be ensured; the target area is delimited according to requirements, and the statistical analysis results corresponding to the requirements can be quickly obtained based on the statistical results of the behavior of a group of people, and the scene structure or layout of a designated place such as a shopping mall can be adjusted in a timely manner to quickly respond to customer needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a schematic flowchart of the method for analyzing the behavior of a group of people in an embodiment of the present invention.

[0036] Figure 2 It is a module diagram of the system for analyzing the behavior of a group of people in an embodiment of the present invention.

[0037] Figure 3 It is a schematic structural diagram of the device in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] The following specifically illustrates the embodiments of the present invention through specific examples. Those skilled in the art can easily understand the other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0039] It should be noted that the drawings provided in the following embodiments only illustrate the basic concept of the present invention schematically. Therefore, only the components related to the present invention are shown in the drawings, rather than being drawn according to the number, shape, and size of the components in actual implementation. The types, quantities, and proportions of the components in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0040] Please refer to Figure 1 , the present invention provides a method for analyzing the behavior of a group of people, including the following steps.

[0041] Step S01: Obtain face images of target objects in a designated place, and perform feature analysis on the face images to obtain face features corresponding to the face images that meet the set face metrics; wherein, the face metrics include: face angle and / or face quality score;

[0042] In step S02, cluster the face features to obtain a face profile corresponding to the same target object;

[0043] In step S03, clean the face profiles of each target object, and select the corresponding cleaned face profiles according to different target areas in the specified venue for statistical analysis of the group behavior of people to obtain the statistical results of the group behavior of people corresponding to the target areas. Among them, the cleaning process includes: cleaning the face profiles according to the gender distribution corresponding to the face features in the face profiles, the number of face features that cannot be merged, and / or the increment of face features within a set time.

[0044] In one embodiment, in step S01, the specified venue may include venues with large crowds such as shopping malls and supermarkets. The specific venue can be set according to actual application requirements and is not limited here. The face image information can be obtained through a collection device such as a surveillance camera deployed in the specified venue. The captured face image information obtained through the surveillance camera may include the actual face image, device ID, capture time, etc. The face image is identified and obtained from the captured scene image through a face detection algorithm. The face detection algorithm can adopt a neural network model, such as target detection models like MTCNN and YOLO series. The captured scene image may include one or more faces, and the face detection algorithm needs to identify all possible face images in the captured scene image. The specific face recognition process is not elaborated here.

[0045] After obtaining the face image from the captured scene, perform face feature extraction on the face image. Specifically, the face parsing service can be used to extract the face feature vector and structured information of the face image to be clustered. Among them, the face parsing service is generally one or more multi-task neural network parsing models. The multi-task neural network parsing model means that one model can simultaneously identify some or all of the structured information and face feature information of the face image. Exemplarily, if the face parsing model is a multi-task neural network model, then the face features can be parsed through this model and all structured information values can be regressively predicted. The face structured information may include face pitch angle, horizontal angle, face quality score, gender, age, whether wearing a mask, etc. Exemplarily, such as face pitch angle: 0, horizontal angle: 10, face quality score: 80, gender: 0, age: 30, whether wearing a mask: 1, etc.

[0046] Screen the faces to be clustered based on the above-mentioned face structured information, eliminate low-quality faces, and obtain the actual faces to be clustered. Exemplarily, due to factors such as the relative capture orientation and illumination of the capture camera, the faces to be clustered with too large pitch angles, too large horizontal angles, and low face quality scores are usually of very low recognizability, and will result in poor corresponding face feature quality, which has a large and continuous negative impact on the final clustering effect. These can all be regarded as dirty data and filtered out without participating in face clustering, which is beneficial to improving clustering efficiency and accuracy. The faces to be clustered can be screened by setting face metrics. Exemplarily, it can be set that when the pitch angle of the face > 40 or the horizontal angle > 50 or the face quality data < 30, then the face image is directly put into the waste film library and not clustered. The pitch angle and horizontal angle can be obtained through face feature recognition, and the face quality calculation can be weighted according to features such as the clarity of the face image and whether the mouth and nose are blocked. The specific calculation process is not elaborated here.

[0047] After completing the screening of the face features of the target object, go to step S02.

[0048] In one embodiment, face archives and the class feature centers corresponding to the face archives can be obtained through clustering algorithms such as K-means clustering and probability density clustering. The specific clustering process is prior art and will not be elaborated here. It is also possible to compare the similarity between the feature vectors of the actual faces to be clustered and the class feature centers of the existing face archives. If the maximum similarity of a certain face to the existing face archives meets the set similarity threshold, then the face is classified into the face archive corresponding to the maximum similarity, and the clustering of the face is completed, and the age, gender, and class feature center of the corresponding face archive are updated. If the maximum similarity of a certain face to the existing face archives does not meet the set similarity threshold, a new face archive is created, and the face is classified into the newly created face archive.

[0049] Each face archive corresponds to a target object. After completing the clustering of different target objects to obtain the corresponding face archives, due to differences in the angles or resolutions of face images collected by different capture cameras, the face images of the same target object can be clustered into multiple face archives. Therefore, the clustered face archives can be merged. Specifically, first, it is judged whether the target objects corresponding to the two face archives are the same. If they are the same, then the similarity between the two face archives is further calculated. If the similarity reaches the set threshold, the corresponding face features are merged into the face archive with more face features.

[0050] After obtaining the face archives through step S02, go to step S03.

[0051] In one embodiment, the face archives can be cleaned. Based on the statistical information of the capture trajectories of the face archives, the invalid archives are cleaned. The specific cleaning calculation rules include cleaning the face archives according to the gender distribution corresponding to the face features in the face archives, the number of face features that cannot be merged, and / or the increment of face features within a set time.

[0052] In one embodiment, cleaning the face archives according to the gender distribution corresponding to the face features in the face archives includes:

[0053] According to the face image acquisition time nodes corresponding to each face feature in the face archive, one or more capture trajectories included in the face archive are determined, and the gender of the target object corresponding to the face feature included in each of the capture trajectories is identified according to the corresponding face feature.

[0054] Select the number of face features with the highest gender ratio as the denominator, calculate the gender ratio of the target object in each capture trajectory, and determine whether the gender ratio exceeds the set threshold range. If it exceeds, the corresponding face archive is deleted.

[0055] Specifically, since there may be a situation where a person is captured multiple times by a camera in a short period of time, it is first necessary to remove duplicates from the capture trajectories of the face archives. Then, the male-female gender ratios of each capture trajectory are statistically calculated. When the system identifies an imbalance in the male-female gender ratio, the corresponding face archive is deleted. For example, if the total number of captures of a file is N, and the total number of captures N is scaled to R according to the condition that only 1 photo is taken by the same camera within 1 minute for statistical calculation. According to the previous steps, the male-female gender characteristics of each capture photo have been determined. Calculate the total number of males and females in the set R. The total number of identified males is R1, and the total number of identified females is R2. Compare the sizes of R1 and R2. Take the larger number as the denominator and the smaller number as the numerator. In this example, assume that R1 < R2. Then, when R1 / R2 > 3 / 7, the file is deleted.

[0056] In one embodiment, cleaning the face archives according to the number of face features that cannot be merged includes:

[0057] Merge the face archives obtained through clustering according to the face feature similarity. If two or more face archives corresponding to the same target object still cannot be merged after exceeding the set time limit, and the number of face features included in the unmergeable face archives is less than the set quantity threshold, the corresponding face archives are deleted.

[0058] Specifically, due to the existence of some special captures, such as special capture angles, etc., the file of a certain person may not be merged all the time. However, the real face of such a file actually exists, so this type of file belongs to an invalid file. For example, if the total number of captures of a file is N (N < 3) and the total number of such files has been less than 3 for 15 consecutive days, then this type of file will be deleted.

[0059] In one embodiment, cleaning the face files according to the increment of face features within a set time includes:

[0060] Judging the increment of face features in the corresponding face files within the set time threshold. If the increment exceeds the set increment threshold, then delete the corresponding face files.

[0061] Specifically, as the number of trajectories included in the face files increases, the class center features of the files will be updated. When a file has a clustering error, with the generation of incorrect classification, the eigenvalue of the new features will tend to be averaged, that is, it will no longer have a feature effect, which will cause the captured photos of different people to be clustered into this file. Similarly, as the number of trajectories of different people increases, the features will become more and more averaged, forming a negative cycle, and ultimately resulting in a large number of captures being incorrectly merged, affecting the statistical effect. Therefore, it is necessary to delete the files with too rapid growth of trajectories. Exemplarily, assuming that after the system runs stably for 30 days, it is screened that the maximum growth amount of trajectories of a face file within 1 day is A. When the system identifies that the trajectory growth amount of a certain file within one day is 2A, then this file will be deleted.

[0062] In one embodiment, after cleaning the face files, it further includes:

[0063] Obtaining the point position information of the acquisition device corresponding to each face file, where the point position information includes: unique identifier, geographical location information, and label information;

[0064] Screening the point position information of the acquisition device according to the proportion of invalid face images obtained by the acquisition device, and closing the acquisition device that exceeds the set proportion threshold; or, closing the acquisition device corresponding to the face file that cannot be merged, where the invalid face images are determined according to the face metrics.

[0065] Specifically, obtaining all the camera point position information and classifying the cameras according to actual business needs. The point position information of the capture cameras includes but is not limited to the unique identifier of the capture camera, longitude and latitude coordinate information, geographical location information, label information, etc. The label information can be the affiliated business district, affiliated street, affiliated community, affiliated administrative division, etc. Exemplarily, generally, the point position information of the capture cameras is obtained from the capture device management database, and the unique identifier of the capture camera is generally the device ID of the capture camera.

[0066] Screen high-value cameras and turn off low-value cameras. In actual installation and deployment, it is often difficult to know whether the camera position is suitable and whether the portrait capture is good before the system runs. Therefore, it is necessary to screen the camera positions through automatic system analysis. There are mainly two screening methods: Method 1, screen according to the proportion of invalid captures. The system automatically counts the invalid pictures obtained through face index screening, calculates the proportion of invalid capture pictures at this position to the total capture pictures at this position. When this proportion is greater than 40%, it means that this position is not suitable for capture, and the system automatically turns off the face collection at this position; Method 2, screen according to the proportion of face archives that cannot be merged. The system automatically counts the face archives that cannot be merged and only contain a small number of face features, calculates which camera position the capture of this archive comes from, calculates the ratio of the capture pictures of this position archive to the valid capture pictures. When this ratio is greater than 10%, the capture at this position is turned off.

[0067] In one embodiment, when two face archives need to be merged, if the statistical result of the group behavior of the person has been completed based on the face archives to be merged, the statistical result is deleted, and after the face archives are merged, the statistical of the group behavior of the person is re-performed for the corresponding target area to obtain the corrected result of the group behavior of the person.

[0068] Specifically, taking the statistical of the group behavior of the person in the area as an example, assume that there are cameras a1, a2, a3... a10, which form the camera set A, and the business label is the shopping mall A in the labeled area; a1, a2, a3 form the set A1, and the business label represents the food court area of the shopping mall A. Calculate the group behavior volume of the shopping mall A on June 30th. Screen the face archives after cleaning and processing with the condition that there is a capture under the camera set A on June 30th, so as to calculate the group behavior of the shopping mall A on June 30th. Similarly, calculate the group behavior volume of the food court of the shopping mall A on June 30th. Screen the face archives after cleaning and processing with the condition that there is a capture under the camera set A1 on June 30th, so as to calculate the group behavior of the food court of the shopping mall A on June 30th.

[0069] Since for the rapid use of the business, in the actual operation process, it is necessary to first convert the statistical situation of the archives into the statistical of the group behavior of the person to facilitate the use for the business. However, since the face archives are constantly being merged, and the statistical of the group behavior of the person is usually archived and counted by day, when the face archives change, it is necessary to correct the group behavior of the person. For example, a1 is merged into the A archive, and in all the previous statistical of the group behavior of the person, the statistical of a1 needs to be deleted and the statistical of A needs to be added.

[0070] In one embodiment, according to the statistical results of the behavior of the population in different regions, the depth distribution and residence time distribution of each target object in a specified venue are obtained, where the depth represents the number of target regions traversed by the target object;

[0071] Adjust the business layout of the specified venue according to the depth distribution and residence time distribution. Analysis of the depth of the behavior of the population. Analysis of the time spent wandering in a day for the behavior of the population, and analysis of the depth of wandering, i.e., how many regions have been visited.

[0072] Specifically, taking a shopping mall as an example, the residence time of the target object can be expressed as the wandering time of the target object in a certain area. First, calculate the wandering time of all files, then screen the valid files (i.e., the face files obtained after cleaning), and finally calculate the average value of the wandering time, and add other dimensions such as age and gender to conduct a complete analysis of the wandering time. Exemplarily, the number of files with capture records within one day is N. For each file, calculate the wandering time t0 of each file using the formula of subtracting the capture time t1 of the first track from the capture time t2 of the last track. Determine that the number of files with t0 > 10 minutes is N2, then the average wandering is equal to the sum of those with a wandering time greater than 10 minutes divided by N2. At the same time, the files can be grouped according to the wandering time of 30 minutes, 30 - 60 minutes, 60 - 120 minutes, and more than 120 minutes. Combining the gender and age characteristics of the files themselves, the wandering time of the population behavior of different age groups and genders can be calculated.

[0073] For the above-mentioned analysis of the wandering depth, first, it is necessary to calculate the wandering depth of all files, then screen the valid files, and finally calculate the average value of the wandering depth. By adding other dimensions such as age and gender, a complete analysis of the wandering depth can be carried out. Exemplarily, the number of files with capture records within one day is N. Each file is screened according to the camera sets with different tags. For example, the camera set in Area A with cameras a1 - a10 and the camera set in Area B with cameras b1 - b10. If there is any one capture of a file in a1 - a10, it is considered that the file has been to Area A, and the wandering depth is 1. By analogy, the wandering depth of each file is calculated as N. After removing the files with N = 0, the average wandering depth can be calculated, which is equal to the sum of the wandering depths greater than 0 divided by the number of files with wandering depths greater than 0. At the same time, the files can be grouped according to the depth levels of 1 - 5 / 6 - 10 / 15 - 20 / above 20. Combining the gender and age characteristics of the files themselves, the wandering depths of the behavior of different age groups and genders can be calculated. Step 8, further, the mall operator can conduct corresponding analysis based on the results of the depth analysis. For example, if it is statistically found that most people go to the same area, then this area is the main gathering point of the crowd. Deploying corresponding billboards in this area will achieve the best advertising effect. Another example is that if it is identified that the wandering depth of the entire mall is insufficient and most people only go to a few specific areas, it indicates that the brand association of the mall is not good and no scale effect has been formed. The mall should make corresponding adjustments in aspects such as investment promotion.

[0074] Please refer to Figure 2 , this embodiment provides a system for analyzing the behavior of a group of people, which is used to execute the method for analyzing the behavior of a group of people described in the foregoing method embodiment. Since the technical principle of the system embodiment is similar to that of the foregoing method embodiment, the same technical details will not be repeated.

[0075] In one embodiment, the system for analyzing the behavior of a group of people includes: a feature acquisition module 10, which is used to acquire the face images of target objects in a specified place and perform feature analysis on the face images to obtain the face features corresponding to the face images that meet the set face indexes; wherein, the face indexes include: face angle and / or face quality score; a file acquisition module 11, which is used to cluster the face features to obtain face files corresponding to the same target object; a module 12 for statistically analyzing the behavior of a group of people, which is used to perform cleaning processing on the face files of each target object, and select the corresponding face files that have been cleaned according to different target areas in the specified place for statistically analyzing the behavior of a group of people to obtain the statistical results of the behavior of a group of people corresponding to the target areas. Among them, the cleaning processing includes: performing cleaning processing on the face files according to the gender distribution corresponding to the face features in the face files, the number of face features that cannot be merged, and / or the increment of face features within a set time.

[0076] The feature acquisition module 10 is used to assist in performing step S01 introduced in the foregoing method embodiment; the file acquisition module 11 is used to perform step S02 introduced in the foregoing method embodiment; the population behavior statistics module 12 is used to perform step S03 introduced in the foregoing method embodiment.

[0077] An embodiment of the present application also provides a population behavior analysis device, which may include: one or more processors; and one or more machine-readable media storing instructions thereon, which when executed by the one or more processors, cause the device to execute Figure 1 the method described above. In practical applications, the device may be used as a terminal device or a server. Examples of terminal devices may include: smart phones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, in-vehicle computers, desktop computers, set-top boxes, smart televisions, wearable devices, etc. The embodiments of the present application do not limit the specific devices.

[0078] An embodiment of the present application also provides a machine-readable medium, in which one or more modules (programs) are stored. When the one or more modules are applied to a device, the device can be caused to execute the Figure 1 instructions included in the population behavior analysis method in the embodiments of the present application. The machine-readable medium may be any available medium that a computer can store or a data storage device such as a server or a data center integrating one or more available media. The available medium may be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0079] Referring to Figure 3 , this embodiment provides a device 80, which may be a desktop computer, a portable computer, a smart phone, or other devices. Specifically, the device 80 at least includes: a memory 82 and a processor 83 connected through a bus 81. The memory 82 is used to store a computer program, and the processor 83 is used to execute the computer program stored in the memory 82 to perform all or part of the steps in the foregoing method embodiment.

[0080] The aforementioned system bus can be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The system bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience in representation, only a thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to implement communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include Random Access Memory (RAM), and may also include non-volatile memory, such as at least one disk memory.

[0081] The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0082] In summary, a method, system, device, and medium for analyzing the behavior of a group of people according to the present invention have more data sources, more accurate statistics, and no other conditional restrictions compared with traditional in-depth analysis methods for the behavior of a group of people. In addition, by combining information such as the gender and age of faces, the statistical dimensions are more diverse, which can effectively ensure the accuracy of the pedestrian flow statistics. Therefore, the present invention effectively overcomes various disadvantages in the prior art and has high industrial utilization value.

[0083] The above embodiments are only illustrative of the principles and effects of the present invention, and are not used to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A method for analyzing the behavior of a group of people, characterized in that, Including: Obtain a face image of a target object within a specified venue, and perform feature analysis on the face image to obtain the face features corresponding to the face image that meets the set face metrics; wherein, the face metrics include: face angle and / or face quality score; Cluster the face features to obtain a face profile corresponding to the same target object; Perform cleaning processing on the face profiles of each target object, and select the cleaned corresponding face profiles according to different target areas within the specified venue for statistical analysis of the group behavior of people, to obtain the statistical results of the group behavior of people corresponding to the target area, wherein, the cleaning processing includes: cleaning the face profiles according to the gender distribution corresponding to the face features in the face profiles, the number of face features that cannot be merged, and / or the increment of face features within a set time; cleaning the face profiles according to the gender distribution corresponding to the face features in the face profiles includes: determining one or more capture trajectories included in the face profile according to the time nodes of face image acquisition corresponding to each face feature in the face profile, and identifying the gender of the target object corresponding to the face features included in each capture trajectory according to the corresponding face features; selecting the number of face features with the highest gender ratio as the denominator, calculating the gender ratio of the target object in each capture trajectory, and determining whether the gender ratio exceeds the set threshold range, if it exceeds, delete the corresponding face profile.

2. The method for analyzing the behavior of a group of people according to claim 1, wherein After obtaining the statistical results of the group behavior of people corresponding to the target area, it further includes: According to the statistical results of the group behavior of people in different target areas, obtain the depth distribution and residence time distribution of each target object within the specified venue, wherein, the depth represents the number of target areas traversed by the target object; Adjust the business layout of the specified venue according to the depth distribution and residence time distribution.

3. The method for analyzing the behavior of a group of people according to claim 1, characterized in that, Cleaning the face profiles according to the number of face features that cannot be merged includes: Merge the face profiles obtained by clustering according to the similarity of face features. If two or more face profiles corresponding to the same target object still cannot be merged after exceeding the set time limit, and the number of face features included in the unmergeable face profiles is less than the set quantity threshold, delete the corresponding face profiles.

4. The method for analyzing the behavior of a group of people according to claim 1, characterized in that Cleaning the face profiles according to the increment of face features within a set time includes: Judge the increment of face features in the face profile corresponding within the set time threshold. If the increment exceeds the set increment threshold, delete the corresponding face profile.

5. The method for analyzing the behavior of a group of people according to claim 1, characterized in that, After performing cleaning processing on the face profiles, it further includes: Obtain the point information of the acquisition device corresponding to each face image of the face profile, wherein, the point information includes: unique identifier, geographical location information, and label information; Screen the point information of the acquisition device according to the proportion of invalid face images obtained by the acquisition device, and turn off the acquisition device that exceeds the set proportion threshold; or, turn off the acquisition device corresponding to the unmergeable face profile, wherein, the invalid face images are determined according to the face metrics.

6. The method for analyzing the behavior of a group of people according to claim 1, wherein After obtaining the statistical results of the group behavior of people corresponding to the target area, it further includes: When two face profiles need to be merged, if the statistical results of the group behavior of the person have been completed based on the face profiles to be merged, the statistical results shall be deleted. After the face profiles are merged, the statistical analysis of the group behavior of the person shall be re-performed for the corresponding target area to obtain the corrected results of the group behavior of the person.

7. A system for analyzing the behavior of a group of people, characterized in that, Including: A feature acquisition module, configured to acquire a face image of a target object in a specified place, and perform feature analysis on the face image to obtain a face feature corresponding to the face image that meets the set face index; wherein, the face index includes: face angle and / or face quality score; A profile acquisition module, configured to cluster the face features to obtain face profiles corresponding to the same target object; A group behavior statistics module for cleaning the face profiles of each target object, and selecting the corresponding cleaned face profiles according to different target areas in the specified place to perform statistics on the group behavior of the person to obtain the statistical results of the group behavior of the person corresponding to the target area. The cleaning process includes: cleaning the face profiles according to the gender distribution corresponding to the face features in the face profiles, the number of face features that cannot be merged, and / or the increment of face features within a set time; cleaning the face profiles according to the gender distribution corresponding to the face features in the face profiles includes: determining one or more capture tracks included in the face profile according to the face image acquisition time node corresponding to each face feature in the face profile, and identifying the gender of the target object corresponding to the face feature included in each capture track according to the corresponding face feature; selecting the number of face features with the highest gender ratio as the denominator, calculating the gender ratio of the target object in each capture track, and determining whether the gender ratio exceeds the set threshold range. If it exceeds, the corresponding face profile shall be deleted.

8. A device for analyzing the collective behavior of people, characterized in that, Including: One or more processors; And One or more machine-readable media storing instructions that, when executed by the one or more processors, cause the device to perform the method according to any one of claims 1-6.

9. A machine-readable medium, characterized in that, Storing instructions thereon that, when executed by one or more processors, cause the device to perform the method according to any one of claims 1-6.

Images