Target Personnel Analysis Method, Device, Computer Equipment and Storage Medium

Abstract

The present invention discloses a method, device, computer device and storage medium for analyzing target personnel. The method includes obtaining the social relationship network of historical target personnel; obtaining portrait captured pictures; for each captured object included in the portrait captured pictures, determining a picture analysis score corresponding to the captured object according to the portrait captured pictures and the pictures of related personnel; determining an information analysis score corresponding to the captured object according to the captured object information of the captured object, and determining an abnormal analysis score corresponding to the captured object according to the picture analysis score and the information analysis score; obtaining a preset analysis threshold, and performing abnormal behavior analysis on the captured object according to the abnormal analysis score and the preset analysis threshold to determine the abnormal analysis result of the captured object. The present invention improves the comprehensiveness and accuracy of abnormal behavior analysis, and further improves the effectiveness of mining target personnel.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular, to a method, device, computer device, and storage medium for analyzing target personnel. Background Art

[0002] In the prior art, the excavation of personnel or teams is achieved through manual methods, but there are problems of low excavation efficiency and low accuracy. Summary of the Invention

[0003] Embodiments of the present invention provide a method, device, computer device, and storage medium for analyzing target personnel, so as to solve the problems of low efficiency and low accuracy in excavating target personnel or target groups in the prior art.

[0004] A method for analyzing target personnel includes:

[0005] Obtaining the social relationship network of historical target personnel; the social relationship network includes at least one related person, and the related person has a related person picture;

[0006] Obtaining a portrait capture picture; the portrait capture picture includes at least one captured object other than the historical target personnel; the captured object has a captured object information;

[0007] For each captured object included in the portrait capture picture, determining a picture analysis score corresponding to the captured object according to the portrait capture picture and the related person picture;

[0008] Determining an information analysis score corresponding to the captured object according to the captured object information of the captured object, and determining an abnormal analysis score corresponding to the captured object according to the picture analysis score and the information analysis score;

[0009] Obtaining a preset analysis threshold, and performing abnormal behavior analysis on the captured object according to the abnormal analysis score and the preset analysis threshold to determine the abnormal analysis result of each captured object.

[0010] A target personnel analysis device includes:

[0011] A relationship network acquisition module, configured to obtain the social relationship network of historical target personnel; the social relationship network includes at least one related person, and the related person has a related person picture;

[0012] A capture picture acquisition module, configured to obtain a portrait capture picture; the portrait capture picture includes at least one captured object other than the historical target personnel; the captured object has a captured object information;

[0013] The picture analysis score determination module is used to determine the picture analysis score corresponding to each photographed object included in the portrait photographed picture according to the portrait photographed picture and the related personnel picture;

[0014] The abnormal analysis score determination module is used to determine the information analysis score corresponding to the photographed object according to the photographed object information of the photographed object, and determine the abnormal analysis score corresponding to the photographed object according to the picture analysis score and the information analysis score;

[0015] The abnormal analysis module is used to obtain a preset analysis threshold, and perform abnormal behavior analysis on the photographed object according to the abnormal analysis score and the preset analysis threshold to determine the abnormal analysis result of the photographed object.

[0016] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned target personnel analysis method is implemented.

[0017] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned target personnel analysis method is implemented.

[0018] In the above-mentioned target personnel analysis method, device, computer device and storage medium, the method determines the picture analysis score of the photographed object by using the related personnel pictures in the social relationship network of the historical target personnel and the obtained portrait photographed picture. Introducing the social relationship network can make the picture analysis score of the photographed object more comprehensive, thereby improving the accuracy of abnormal behavior analysis of the photographed object. Further, the abnormal analysis result of the photographed object is also assisted and judged by the photographed object information of the photographed object, that is, when performing abnormal behavior analysis, the basic situation of the photographed object (such as information such as age in the photographed object information) is combined, thereby improving the comprehensiveness and accuracy of abnormal behavior analysis, and further improving the effectiveness of target personnel mining. Brief Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments of the present invention. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0020] Figure 1 It is an application environment schematic diagram of the target personnel analysis method in an embodiment of the present invention;

[0021] Figure 2 It is a flowchart of a target person analysis method in an embodiment of the present invention;

[0022] Figure 3 It is a flowchart of step S30 in the target person analysis method in an embodiment of the present invention;

[0023] Figure 4 It is a principle block diagram of a target person analysis device in an embodiment of the present invention;

[0024] Figure 5 It is a schematic diagram of a computer device in an embodiment of the present invention. Detailed implementation manners

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] The target person analysis method provided by the embodiments of the present invention can be applied in an application environment as shown in Figure 1 Specifically, the target person analysis method is applied in a target person analysis system, and the target person analysis system includes a client and a server as shown in Figure 1 The target person analysis method can be implemented through the client or the server in the target person analysis system. The client and the server communicate through a network, aiming to solve the problems of low efficiency and low accuracy in mining target persons or target groups in the prior art. Among them, the client, also known as the user side, refers to a program that provides local services corresponding to the server. The client can be installed on but not limited to various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers.

[0027] In one embodiment, as shown in Figure 2 A target person analysis method is provided. Taking the server in Figure 1 as an example for illustration, the method includes the following steps:

[0028] S10: Obtain the social relationship network of historical target persons; the social relationship network includes at least one related person, and one related person has one related person picture.

[0029] Understandably, a historical target person can be marked after being determined as a target person by the relevant department. The social relationship network is constructed by various relationship persons who have a direct relationship (such as relatives, friends, colleagues, etc.) or an indirect relationship (such as a friend of a relative, a relative of a colleague, etc.) with the historical target person. A relationship person picture is a picture with the facial features of the relationship person.

[0030] S20: Obtain a portrait-taking picture; the portrait-taking picture includes at least one subject other than the historical target person; the subject has a subject information.

[0031] Understandably, the portrait-taking picture can be a picture taken by a camera or other shooting device in different areas (for example, within the community where the historical target person lives, areas including inappropriate locations, etc.). The portrait-taking picture includes the historical target person and other subjects other than the historical target person, or only includes other subjects other than the historical target person; generally, the historical target person will travel with other members of the target group, so the historical target person and other subjects traveling with the historical target person may be captured in the portrait-taking picture; but it is also possible to capture only other subjects other than the historical target person in the community including inappropriate locations. Further, the time of the portrait-taking picture is not limited. For example, taking the time point when the historical target person is confirmed to have abnormal behavior as the dividing time point, the portrait-taking picture can be collected before the dividing time point or after the dividing time point.

[0032] Further, the subject information includes the basic subject information of the subject and voice data information. Among them, the basic subject information can be the age, gender, whether having a job, job type, marital status, etc. of the subject; the voice data information refers to the call data information of the subject, such as the call record between the subject and the historical target person.

[0033] S30: For each subject included in the portrait-taking picture, determine the picture analysis score corresponding to the subject according to the portrait-taking picture and the relationship person picture.

[0034] Specifically, after obtaining the social relationship network and the portrait-taking picture of the historical target person, the portrait-taking picture and the relationship person picture can be compared in terms of features, so as to determine whether there is a subject corresponding to the relationship person in the portrait-taking picture. If there is a subject corresponding to the relationship person in the portrait-taking picture, the picture analysis score can be determined according to the hierarchical weight of the relationship level to which the relationship person corresponding to the subject belongs and the number of times the subject appears in all portrait-taking pictures.

[0035] S40: Determine the information analysis score corresponding to the subject according to the subject information of the subject, and determine the abnormal analysis score corresponding to the subject according to the picture analysis score and the information analysis score.

[0036] Understandably, it is pointed out in the above description that the subject information includes the basic subject information and the voice data information of the subject. Therefore, different scores can be assigned to, for example, age, gender, whether having a job, job type, marital status, and different scores can be assigned to the voice data information. Furthermore, according to the analysis results of the basic subject information and the voice data information, for example, analyzing the age group to which the subject belongs (each age group is assigned a different score), the information analysis score can be determined based on the analysis results of the basic subject information and the voice data information. Further, the sum of the picture analysis score and the information analysis score is determined as the abnormal analysis score corresponding to each subject.

[0037] S50: Obtain a preset analysis threshold, and perform abnormal behavior analysis on the subject according to the abnormal analysis score and the preset analysis threshold to determine the abnormal analysis result of the subject.

[0038] Exemplarily, the preset analysis threshold can be set to 85, 90, etc. Specifically, after determining the abnormal analysis score corresponding to each subject according to the picture analysis score and the information analysis score, obtain the preset analysis threshold, and compare the abnormal analysis score with the preset analysis threshold. If the abnormal analysis score is greater than or equal to the preset analysis threshold, it is determined that the abnormal analysis result of the subject indicates that the subject may be the target person; if the abnormal analysis score is less than the preset analysis threshold, it is determined that the abnormal analysis result of the subject indicates that the subject is not the target person. Furthermore, after determining that the subject is the target person, the subject can be monitored to further determine whether there is any abnormal behavior, thereby improving the efficiency of abnormal behavior analysis.

[0039] In this embodiment, by determining the picture analysis score of the subject according to the pictures of the related persons in the social relationship network of the historical target person and the obtained portrait pictures, introducing the social relationship network can make the picture analysis score of the subject more comprehensive, thereby improving the accuracy of abnormal behavior analysis of the subject. Further, the abnormal analysis result of the subject is also assisted in judgment through the subject information of the subject, that is, the basic situation of the subject (such as information such as age in the subject information) is combined when performing abnormal behavior analysis, thereby improving the comprehensiveness and accuracy of abnormal behavior analysis, and further improving the effectiveness of mining the target person.

[0040] In one embodiment, before step S10, that is, before obtaining the social relationship network of the historical target person, it further includes:

[0041] Identify all related personnel in the preset personnel attribute database who have an associated relationship with the historical target personnel.

[0042] Understandably, the personnel attribute database can be the social platform used by the historical target personnel, or the personnel database of relevant departments. Related personnel refer to those who have a direct associated relationship (such as the wife of the historical target personnel) or an indirect associated relationship (such as the friend of the wife of the historical target personnel) with the historical target personnel.

[0043] Respectively determine the relevant levels between the historical target personnel and each related personnel.

[0044] According to the relevant levels, determine the relationship levels corresponding to the historical target personnel and each related personnel, and set each related personnel in the relationship level corresponding to the related personnel.

[0045] Understandably, the relevant level represents the degree of relevance of the relationship between the related personnel and the historical target personnel. This relevant level can be a direct relationship level, an indirect relationship level, etc. A related personnel belongs to a relationship level, that is, at least one related personnel is included in a relationship level. If no related personnel is included in a certain relationship level, then this relationship level will be deleted.

[0046] Furthermore, the related personnel with a direct relationship level refer to those who have a direct associated relationship with the historical target personnel. Exemplarily, such as the relatives, friends, colleagues, etc. of the historical target personnel. Then, set the related personnel who have a direct relationship with the historical target personnel, that is, the related personnel with a direct relationship level, in the first relationship level.

[0047] Furthermore, the first indirect relationship level is included in the indirect relationship level. The first indirect relationship level refers to those who have a direct relationship with the related personnel in the first relationship level. Exemplarily, such as the parents of the spouse of the historical target personnel, the friends of the friends of the historical target personnel, etc. Then, set the related personnel who have a direct relationship with the related personnel in the first relationship level, that is, the related level between them and the related personnel in the first relationship level is a direct relationship level, and the related level with the historical target personnel is the first indirect relationship level, in the second relationship level.

[0048] Further, the indirect relationship levels may also include, for example, the second indirect relationship level (such as other relationship personnel having a direct association relationship with the relationship personnel at the second relationship level, which can be set at the third relationship level), the third indirect relationship level, etc. It can be understood that the relationship personnel at each indirect relationship level have a direct association level with the relationship personnel at the previous indirect relationship level (such as between the second indirect relationship level and the first indirect relationship level), and the greater the indirect relationship level (such as the second indirect relationship level being greater than the first indirect relationship level), the smaller its association relationship with the historical target personnel.

[0049] After all relationship personnel are set at the corresponding relationship levels of the relationship personnel, it is determined that the social relationship network of the historical target personnel is completed.

[0050] Specifically, after determining the relationship levels corresponding to each relationship personnel according to the relevant levels and setting each relationship personnel in its corresponding relationship level, if all relationship personnel are set after their corresponding relationship levels, it is determined that the social relationship network of the historical target personnel is completed.

[0051] In this embodiment, by crawling the relationship personnel associated with the historical target personnel from the preset personnel attribute library, and determining the corresponding relationship levels according to the relevant levels between each relationship personnel and the historical target personnel, and then setting each relationship personnel in its corresponding relationship level, the social relationship network of the historical target personnel is constructed. In this way, it can be determined whether there are personnel associated with the historical target personnel among the personnel traveling with the historical target personnel through the social relationship network, providing a data basis for calculating the picture analysis score of the shooting object in the subsequent steps, thereby improving the accuracy of target personnel analysis.

[0052] In one embodiment, the portrait shooting pictures include multiple pictures taken at different locations and multiple tracking shooting pictures; the social relationship network includes at least one relationship level; one relationship level includes at least one relationship personnel;

[0053] Such as Figure 3 As shown, in step S30, that is, determining the picture analysis score corresponding to the shooting object according to the portrait shooting pictures and the relationship personnel pictures, includes:

[0054] S301: Determine the total number of times the shooting object appears in all the pictures taken at different locations, and record this total number as the location appearance times; one shooting object corresponds to one location appearance times.

[0055] Understandably, there are multiple relationship levels in the social relationship network. For example, the relationship personnel directly related to the historical target person can be set in the first relationship level, and other relationship personnel directly related to the relationship personnel in the first relationship level can be set in the second relationship level, etc. Exemplarily, the wife of the historical target person can be set in the first relationship level, and the colleague of the wife of the historical target person can be set in the second relationship level.

[0056] Understandably, in the above description, it is pointed out that the portrait capture pictures can be pictures captured by cameras or other capture devices in different areas. Therefore, the pictures captured at this location are pictures captured by cameras or other capture devices within the area where the suspected inappropriate location confirmed by the relevant department is located. The pictures captured at this location may include the historical target person and / or other capture objects.

[0057] Further, after obtaining the pictures captured at the location, human face recognition can be performed on the pictures captured at the location, and then image cropping can be performed on the pictures captured at the location, so that one picture captured at the location is cropped into multiple portrait pictures each containing only one capture object, and the portrait pictures containing the same capture object are grouped into a picture portrait group. Then, determine the number of portrait pictures included in each picture portrait group. This number is the total number of times the capture object corresponding to this picture portrait group appears in all the pictures captured at the location, and record this total number as the location appearance times. Among them, one capture object corresponds to one location appearance times, and this location appearance times also represents the number of times the capture object appears in the inappropriate location.

[0058] S302: Obtain the preset location analysis weight, and determine the location analysis score corresponding to the capture object included in the pictures captured at the location according to the location analysis weight and the location appearance times.

[0059] Understandably, the preset location analysis weight can be set in advance. For example, the preset location analysis weight can also be determined according to the number of areas where different inappropriate locations are located. The preset location analysis weight is the proportion of the location analysis score in the abnormal analysis score of the capture object. Exemplarily, the preset location analysis weight can be set to 2, 3, etc.

[0060] Specifically, after determining the total number of times the capture object appears in all the pictures captured at the location and recording this total number as the location appearance times, obtain the preset location analysis weight, and take the product of the location analysis weight and the location appearance times of the capture object as the location analysis score of this capture object.

[0061] S303: Compare the features of the tracking capture pictures with the pictures of the relationship personnel to obtain the image comparison result of the capture object included in the tracking capture pictures.

[0062] Understandably, the feature comparison in this embodiment is to determine whether there are features in the tracked and photographed image that are the same as the facial features of the related person in the related person's image, that is, to compare a related person's image with the tracked and photographed image one by one, and then it can be determined whether the feature similarity between the facial features in the tracked and photographed image and the facial features in the related person's image meets the preset conditions. The preset conditions can be whether the feature similarity between the facial features in the tracked and photographed image and the facial features in the related person's image is greater than the preset similarity threshold. The preset similarity threshold can be set according to the specific scenario. Exemplarily, the preset similarity threshold can be set to 90%, 95%, etc.

[0063] Furthermore, the image comparison result represents the comparison results of all photographed objects in the tracked and photographed image. When the comparison result of a photographed object represents a successful comparison, it means that the feature similarity between the photographed object and one of the related persons is greater than or equal to the preset similarity threshold, that is, the photographed object and the related person with a successful comparison are the same person; when the comparison result of a photographed object represents a failed comparison, it means that the feature similarity between the photographed object and all related persons is less than the preset similarity threshold, that is, the photographed object does not exist in the social relationship network of the historical target person.

[0064] S304: According to the relationship level and the image comparison result, determine the tracking analysis score of the photographed object included in the tracked and photographed image.

[0065] Specifically, after performing feature comparison between the tracked and photographed image and the related person's image to obtain the image comparison result between the location photographed image and the related person's image, if the comparison between the photographed object and the related person is successful, the tracking analysis score of the photographed object with a successful comparison can be determined according to the weight of the relationship level to which the related person with a successful comparison with the photographed object belongs, and the total number of times the photographed object appears in all tracked and photographed images; if the comparison between the photographed object and the related person fails, the tracking analysis score of the photographed object with a failed comparison is directly determined according to the total number of times the photographed object with a failed comparison appears in all tracked and photographed images.

[0066] S305: According to the location analysis score and the tracking analysis score, determine the image analysis score corresponding to the photographed object.

[0067] Specifically, after determining the tracking analysis score of the photographed object included in the tracking shot picture according to the relationship level and the image comparison result, and determining the location analysis score corresponding to the photographed object in the location shot picture according to the location analysis weight and the number of times the location appears, the sum of the location analysis score and the tracking analysis score corresponding to the same photographed object can be recorded as the picture analysis score of the photographed object; further, when a photographed object only includes the location analysis score, the location analysis score can be directly used as the picture analysis score of the photographed object; when a photographed object only includes the tracking analysis score, the tracking analysis score can be directly used as the picture analysis score of the photographed object. For example, the wife of a historical target person, although there are behaviors such as traveling together and being in the same car between his wife and the historical target person, that is, the wife of the historical target person is photographed in the tracking shot picture, but his wife does not appear in the area belonging to the non-appropriate location, that is, the wife of the historical target person is not photographed in the location shot picture, then the wife of the historical target person only includes the tracking analysis score.

[0068] In this embodiment, the location analysis score is determined according to the total number of times the photographed object appears in the location shot picture, the tracking analysis score is determined according to the image comparison result between the photographed object and the related person and the relationship level, and then the picture analysis score of the photographed object is determined according to the location analysis score and the tracking analysis score. In this way, the situation of the photographed object traveling together and being in the same car with the historical target person (that is, the above-mentioned tracking analysis score) and the situation of the photographed object appearing in the area belonging to the non-appropriate location (that is, the above-mentioned location analysis score) can be combined to perform abnormal behavior analysis based on pictures on the photographed object, thereby improving the comprehensiveness of the abnormal behavior analysis and further improving the accuracy of the abnormal behavior analysis.

[0069] In one embodiment, in step S303, that is, comparing the features of the tracking shot picture with the picture of the related person to obtain the image comparison result of the photographed object included in the tracking shot picture, including:

[0070] Performing face recognition on the tracking shot picture to obtain at least one intercepted face picture; the intercepted face picture refers to a face picture that only includes one photographed object intercepted from the tracking shot picture.

[0071] Associating and recording the intercepted face pictures containing the same photographed object as a photographed face group.

[0072] Understandably, portrait recognition is a method for recognizing different subjects in a tracking and shooting picture, and then different facial features in the tracking and shooting picture can be classified into the portrait picture group of the corresponding subject. Among them, one subject corresponds to one portrait picture group. In one portrait picture group, there are portrait pictures containing this subject intercepted from the tracking and shooting picture, that is, intercepted portrait pictures; there are at least one intercepted portrait pictures in one portrait picture group. For example, when this subject exists in different tracking and shooting pictures, portrait pictures of this subject can be intercepted from different tracking and shooting pictures. In this way, after performing portrait recognition on the tracking and shooting picture and obtaining the portrait picture group, the efficiency of feature comparison between the tracking and shooting picture and the related person picture in subsequent steps can be improved.

[0073] For each portrait picture group, determine the picture similarity between the related person picture and the intercepted portrait pictures in the portrait picture group, and determine the image comparison result of the subject included in the tracking and shooting picture according to the picture similarity.

[0074] Specifically, after performing portrait recognition on the tracking and shooting picture to obtain at least one intercepted portrait picture and associating and recording the intercepted portrait pictures containing the same subject as one portrait picture group, the related person picture can be compared with the intercepted portrait pictures in each portrait picture group, and then the picture similarity between the related person picture and the intercepted portrait pictures in each portrait picture group can be determined. Furthermore, the image comparison result of the subject included in the tracking and shooting picture can be determined according to the picture similarity.

[0075] Furthermore, after determining the picture similarity between the related person picture and the intercepted portrait pictures in each portrait picture group, the picture similarity can be compared with a preset similarity threshold (the preset similarity threshold can be set to 80%, 90%, etc.). If the picture similarity is greater than or equal to the preset similarity threshold, it is determined that the subject corresponding to the portrait picture group to which the portrait picture corresponding to this picture similarity belongs and the related person corresponding to the related person picture compared with the portrait picture corresponding to this picture similarity are the same person. Furthermore, it is determined that the image comparison result of this subject indicates a successful comparison; if the picture similarity between all intercepted portrait pictures in one portrait picture group and all related person pictures is less than the preset similarity threshold, it is determined that the subject corresponding to this portrait picture group does not match all related persons. Furthermore, it is determined that the image comparison result of this subject indicates a failed comparison.

[0076] In this embodiment, after obtaining the captured portrait group by performing portrait recognition on the tracked and captured pictures, the picture of the related person can be compared with the intercepted portrait picture in the captured portrait group, rather than comparing the picture of the related person with any randomly tracked and captured picture. Since the facial features of the same captured object are basically the same, comparing the picture of the related person with the intercepted portrait picture of the same captured object can better notice the differences between the facial features, improving the efficiency and accuracy of feature comparison.

[0077] In one embodiment, in step S304, that is, according to the relationship level and the image comparison result, determining the tracking analysis score of the captured object included in the tracked and captured picture includes:

[0078] Determine the total number of times the captured object appears in all the tracked and captured pictures, and record this total number as the total tracking appearance times; one captured object corresponds to one total tracking appearance times.

[0079] It can be understood that the captured objects included in different tracked and captured pictures may be different. Therefore, the total number of times the captured object appears in all the tracked and captured pictures can be used as the number of times the captured object has behaviors such as walking or riding in the same vehicle as the historical target person, and then this total number is recorded as the total tracking appearance times. One captured object corresponds to one total tracking appearance times.

[0080] Determine the successfully compared objects and the failed compared objects from all the captured objects according to the image comparison result; the successfully compared objects refer to the captured objects that match the related person; the failed compared objects refer to the captured objects that do not match all the related persons.

[0081] It can be understood that the above description indicates that the image comparison result represents the comparison results of all the captured objects in the tracked and captured picture. When the comparison result of a captured object represents a successful comparison, it means that the captured object and the successfully compared related person are the same person, that is, it means that the captured object matches the related person. Thus, the captured object that matches the related person can be determined as the successfully compared object; when the comparison result of a captured object represents a failed comparison, it means that the captured object does not exist in the social relationship network of the historical target person, that is, the captured object does not match all the related persons. Thus, the captured object that does not match all the related persons can be determined as the failed compared object.

[0082] Obtain the level weight corresponding to the relationship level to which the related person matching the successfully compared object belongs, and determine the tracking analysis score corresponding to the successfully compared object according to the level weight and the total tracking appearance times.

[0083] Understandably, as pointed out in the above description, if the image comparison result indicates a successful comparison, then the photographed object and the related person matching it are the same person. Further, after determining the successfully compared objects and the failed compared objects from all photographed objects according to the image comparison result, the hierarchical weight corresponding to the relationship level to which the related person matching the successfully compared object belongs can be obtained, and based on the hierarchical weight and the total number of tracking occurrences, the tracking analysis score corresponding to the successfully compared object can be determined. For example, assuming that each occurrence of a photographed object in a tracked photographed image is counted as one point, the product of the hierarchical weight and the total number of tracking occurrences can be determined as the tracking analysis score. Assuming that each occurrence of a photographed object in a tracked photographed image is counted as two points, the product of the hierarchical weight, the total number of tracking occurrences, and this score can be determined as the tracking analysis score.

[0084] Furthermore, the hierarchical weights can be pre-assigned to each relationship level, and the hierarchical weight of the relationship level closer to the historical target person is greater. For example, the hierarchical weight of the first relationship level is greater than that of the second relationship level. For example, the hierarchical weight of the first relationship level can be set to 10, and the hierarchical weight of the second relationship level can be set to 8, etc.

[0085] Determine the tracking analysis score corresponding to the failed compared object according to the total number of tracking occurrences.

[0086] Specifically, after determining the successfully compared objects and the failed compared objects from all photographed objects according to the image comparison result, the tracking analysis score corresponding to each failed compared object can be directly determined based on the total number of tracking occurrences corresponding to the failed compared object and the score for each occurrence of a photographed object in a tracked photographed image.

[0087] In this embodiment, when the image comparison result of a photographed object indicates a successful comparison, the tracking analysis score of the successfully compared photographed object is determined by the hierarchical weight of the relationship level to which the related person matching the successfully compared photographed object belongs and the total number of tracking occurrences. When the image comparison result of a photographed object indicates a failed comparison, the tracking analysis score can be determined according to the total number of tracking occurrences of the failed compared photographed object. In this way, a higher score weight can be assigned to the photographed objects having an associated relationship with the historical target person, making the tracking analysis scores of each photographed object more in line with the actual situation, thereby improving the accuracy of subsequent steps for analyzing abnormal behaviors of photographed objects.

[0088] In one embodiment, in step S40, that is, determining the information analysis score corresponding to the photographed object according to the photographed object information of the photographed object, includes:

[0089] Obtain the basic object information and voice data information corresponding to the photographed object from the photographed object information.

[0090] Understandably, the photographed object information includes the basic object information of the photographed object and voice data information. Among them, the basic object information can be the age, gender, whether having a job, job type, marital status, etc. of the photographed object; the voice data information refers to the call data information of the photographed object, such as the call record between the photographed object and historical target personnel.

[0091] Obtain a preset abnormal word database; the preset abnormal word database includes at least one abnormal word.

[0092] Understandably, the preset abnormal word database can be obtained by pre-collecting words for some abnormal behaviors in advance, such as words for illegal fund transfer behaviors, etc., that is, abnormal words.

[0093] Match the voice data information with each abnormal word to obtain a word matching result, and determine the voice analysis score corresponding to the photographed object according to the word matching result.

[0094] Specifically, after obtaining the preset abnormal word database, perform word segmentation processing on the voice data information to obtain each voice word in the voice data information, and match each voice word with the abnormal word to obtain a word matching result. Among them, the word matching result includes a successful matching result and a failed matching result; the successful matching result indicates that the voice word and the abnormal word are successfully matched; the failed matching result indicates that the voice word fails to match all abnormal words. Further, the voice analysis score can be determined according to the voice words representing successful matching in the voice data information. For example, if there are five voice words representing successful matching in a voice data information, and it is set that the score corresponding to a voice word representing successful matching is one point, then the voice analysis score corresponding to this voice data information is five points.

[0095] Determine the basic analysis score corresponding to the photographed object according to the basic object information, and determine the information analysis score corresponding to the photographed object according to the basic analysis score and the voice analysis score.

[0096] Specifically, after determining the voice analysis score corresponding to each photographed object according to the word matching result, determine the basic analysis score corresponding to each photographed object according to the basic object information, and determine the information analysis score corresponding to each photographed object by summing the basic analysis score and the voice analysis score.

[0097] Further, the basic object information includes the age, gender, whether having a job, job type, marital status, etc. of the photographed object. First, in this embodiment, it is defined that the gender analysis score of a photographed object with male gender is higher than that of a photographed object with female gender. For example, it is set that the gender analysis score of a photographed object with male gender is five points, then the gender analysis score of a photographed object with female gender is three points.

[0098] Secondly, in this embodiment, the shooting objects are divided into different age groups. For example, the ages are divided into 0 to 7 years old, 8 to 16 years old, 17 to 24 years old, etc., and a certain age analysis score is assigned to each age group. For example, the age analysis score assigned to the age group of 0 to 7 is 0, the age analysis score assigned to the age group of 8 to 16 is 10, the age analysis score assigned to the age group of 17 to 24 is 20, etc. In addition, when the age group is 0 to 7 or over 60, abnormal behavior analysis may not be performed on the shooting objects in these two age groups.

[0099] Further, in this embodiment, it is defined that if the shooting object has a job, the job analysis score of the shooting object can be determined to be -5, and the job analysis score can be adjusted according to the job type. For example, when the job type is a public institution, the job analysis score can be -10. That is, when the shooting object has a job, the job analysis score is a negative score; when the shooting object does not have a job, the job analysis score can be set to zero.

[0100] Further, in this embodiment, it is defined that if the shooting object is in a married state, the marital status analysis score of the shooting object can be determined to be -5; if the shooting object is in an unmarried state, the marital status analysis score of the shooting object can be determined to be zero.

[0101] Further, the basic analysis score of each shooting object is the sum of the gender analysis score, age analysis score, job analysis score, and marital status analysis score.

[0102] In this embodiment, based on the basic object information and voice data information of the shooting object, the information analysis score corresponding to each shooting object is determined, so that the information analysis score can improve the analysis of abnormal behavior of the shooting object, thereby improving the accuracy of abnormal behavior analysis.

[0103] In one embodiment, the voice data information is matched with each abnormal word to obtain a word matching result, including:

[0104] Perform speech recognition on the voice data information to obtain a speech recognition text.

[0105] It can be understood that speech recognition is a method of converting voice data into text data. Speech recognition can be performed on the voice data information through, for example, a hidden Markov model, a neural network model, etc., so as to convert the voice data information into text information, that is, the speech recognition text.

[0106] Perform word segmentation on the speech recognition text to obtain the voice data words in the speech recognition text.

[0107] Understandably, word segmentation processing can be performed by methods such as Jieba word segmentation and neural network model word segmentation. The words in the speech data are the words in the speech data information.

[0108] Perform word vector conversion on the words in the speech data and the abnormal words to obtain the speech word vectors corresponding to the words in the speech data and the abnormal word vectors corresponding to the abnormal words.

[0109] Understandably, word vector conversion is a method of converting text words into word vectors. For example, word vector conversion can be performed through models based on Word2Vec, CBOW models, etc. Specifically, after performing word segmentation processing on the speech recognition text to obtain the words in the speech data of the speech recognition text, perform word vector conversion on all the words in the speech data to obtain the speech word vectors corresponding to each word in the speech data. At the same time, perform word vector conversion on all the abnormal words to obtain the abnormal word vectors corresponding to each abnormal word. That is, one word in the speech data corresponds to one speech word vector; one abnormal word corresponds to one abnormal word vector.

[0110] Determine the word vector distance between the speech word vector and the abnormal word vector, and determine the word matching result according to the word vector distance and the preset vector distance threshold.

[0111] Optionally, the word vector distance can be determined by calculation methods such as norm distance and Euclidean distance. The preset vector distance threshold can be selected according to the specific application scenario. If the requirement for intention recognition is high in the scenario, the preset vector distance threshold can be set to 0.05, 0.1, etc.

[0112] Specifically, after performing word vector conversion on the words in the speech data and the abnormal words to obtain the speech word vectors corresponding one by one to each word in the speech data and the abnormal word vectors corresponding one by one to each abnormal word, determine the word vector distance between the speech word vector and the abnormal word vector, and compare the word vector distance with the preset vector distance threshold. If the word vector distance is less than or equal to the preset vector distance threshold, determine that the word matching result is a matching success result; if the word vector distance is greater than the preset vector distance threshold, determine that the word matching result is a matching failure result.

[0113] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0114] In one embodiment, a target person analysis device is provided, and the target person analysis device corresponds one by one to the target person analysis method in the above embodiment. As Figure 4As shown in the figure, the target person analysis device includes a relationship network acquisition module 10, a captured image acquisition module 20, an image analysis score determination module 30, an anomaly analysis score determination module 40, and an anomaly analysis module 50. The detailed description of each functional module is as follows:

[0115] The relationship network acquisition module 10 is used to acquire the social relationship network of historical target persons; at least one related person is included in the social relationship network, and the related person has a related person image;

[0116] The captured image acquisition module 20 is used to acquire portrait captured images; at least one captured object other than the historical target person is included in the portrait captured images; the captured object has a captured object information;

[0117] The image analysis score determination module 30 is used to determine, for each captured object included in the portrait captured images, an image analysis score corresponding to the captured object according to the portrait captured images and the related person images;

[0118] The anomaly analysis score determination module 40 is used to determine an information analysis score corresponding to the captured object according to the captured object information of the captured object, and determine an anomaly analysis score corresponding to the captured object according to the image analysis score and the information analysis score;

[0119] The anomaly analysis module 50 is used to obtain a preset analysis threshold, and perform anomaly behavior analysis on the captured object according to the anomaly analysis score and the preset analysis threshold to determine an anomaly analysis result of the captured object.

[0120] Preferably, the target person analysis device further includes:

[0121] The related person determination module is used to determine all related persons having an associated relationship with the historical target person in a preset personnel attribute library;

[0122] The correlation level determination module is used to determine the correlation level between the historical target person and each of the related persons respectively;

[0123] The relationship level determination module is used to determine the relationship level corresponding to the historical target person and each of the related persons according to the correlation level, and set each of the related persons in the relationship level corresponding to the related person;

[0124] The social relationship network construction module is used to determine that the construction of the social relationship network of the historical target person is completed after all the related persons are set in the relationship level corresponding to the related person.

[0125] Preferably, the image analysis score determination module 30 includes:

[0126] A location appearance times determination unit, configured to determine the total number of times the photographed object appears in the pictures taken at all the locations, and record the total number of times as the location appearance times; one photographed object corresponds to one location appearance times;

[0127] A location analysis score determination unit, configured to obtain a preset location analysis weight, and determine a location analysis score corresponding to the photographed object included in the picture taken at the location according to the location analysis weight and the location appearance times;

[0128] A feature comparison unit, configured to perform feature comparison between the tracked photographed picture and the picture of the related person, and obtain an image comparison result of the photographed object included in the tracked photographed picture;

[0129] A tracking analysis score determination unit, configured to determine a tracking analysis score of the photographed object included in the tracked photographed picture according to the relationship level and the image comparison result;

[0130] A picture analysis score determination unit, configured to determine a picture analysis score corresponding to the photographed object according to the location analysis score and the tracking analysis score.

[0131] Preferably, the feature comparison unit includes:

[0132] A portrait recognition subunit, configured to perform portrait recognition on the tracked photographed picture to obtain at least one intercepted portrait picture; the intercepted portrait picture refers to a portrait picture that only includes one photographed object intercepted from the tracked photographed picture;

[0133] A picture classification subunit, configured to associatively record the intercepted portrait pictures including the same photographed object as a photographed portrait group;

[0134] An image comparison subunit, configured to, for each of the photographed portrait groups, determine the picture similarity between the picture of the related person and the intercepted portrait pictures in the photographed portrait group, and determine an image comparison result of the photographed object included in the tracked photographed picture according to the picture similarity.

[0135] Preferably, the tracking analysis score determination unit includes:

[0136] A tracking appearance total times determination subunit, configured to determine the total number of times the photographed object appears in all the tracked photographed pictures, and record the total number of times as the tracking appearance total times; one photographed object corresponds to one tracking appearance total times;

[0137] An object recording subunit, configured to determine a successfully matched object and an unsuccessfully matched object from all photographed objects according to the image comparison result; the successfully matched object refers to a photographed object that matches the related person; the unsuccessfully matched object refers to a photographed object that does not match any of the related persons.

[0138] A first score determination subunit, configured to obtain a hierarchical weight corresponding to the relationship level to which the related person matching the successfully matched object belongs, and determine a tracking analysis score corresponding to the successfully matched object according to the hierarchical weight and the total number of tracking occurrences.

[0139] A second score determination subunit, configured to determine a tracking analysis score corresponding to the unsuccessfully matched object according to the total number of tracking occurrences.

[0140] Preferably, the abnormal analysis score determination module 40 includes:

[0141] A data information acquisition unit, configured to acquire basic object information and voice data information corresponding to the photographed object from the photographed object information.

[0142] An abnormal word acquisition unit, configured to acquire a preset abnormal word database; the preset abnormal word database includes at least one abnormal word.

[0143] A word matching unit, configured to match the voice data information with each of the abnormal words to obtain a word matching result, and determine a voice analysis score corresponding to the photographed object according to the word matching result.

[0144] An information analysis score determination unit, configured to determine a basic analysis score corresponding to the photographed object according to the basic object information, and determine an information analysis score corresponding to the photographed object according to the basic analysis score and the voice analysis score.

[0145] Preferably, the word matching unit includes:

[0146] A voice recognition unit, configured to perform voice recognition on the voice data information to obtain a voice recognition text.

[0147] A word segmentation processing unit, configured to perform word segmentation processing on the voice recognition text to obtain voice data words in the voice recognition text.

[0148] A word vector conversion unit, configured to perform word vector conversion on the voice data words and the abnormal words to obtain a voice word vector corresponding to the voice data words and an abnormal word vector corresponding to the abnormal words.

[0149] A vector distance determination unit is configured to determine a word vector distance between the speech word vector and the abnormal word vector, and determine the word matching result according to the word vector distance and a preset vector distance threshold.

[0150] For the specific limitations of the target person analysis device, reference may be made to the limitations of the target person analysis method in the foregoing text, which will not be elaborated here. Each module in the above target person analysis device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so as to facilitate the processor to call and execute the operations corresponding to each of the above modules.

[0151] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as Figure 5 shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the data used in the target person analysis in the above embodiments. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a target person analysis method.

[0152] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the target person analysis method in the above embodiments.

[0153] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, it implements the target person analysis method in the above embodiments.

[0154] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned target personnel analysis method can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned various methods. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0155] Those skilled in the art can clearly understand that for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0156] The above-mentioned embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for analyzing target personnel, characterized in that, it includes: Obtain the social relationship network of historical target personnel; The social relationship network includes at least one related person, and the related person has a picture of the related person; Obtain portrait shooting pictures; the portrait shooting pictures include at least one shooting object other than the historical target personnel; the shooting object has a shooting object information; For each shooting object included in the portrait shooting pictures, determine the picture analysis score corresponding to the shooting object according to the portrait shooting pictures and the pictures of the related persons; Determine the information analysis score corresponding to the shooting object according to the shooting object information of the shooting object, and determine the abnormal analysis score corresponding to the shooting object according to the picture analysis score and the information analysis score; Obtain a preset analysis threshold, and perform abnormal behavior analysis on the shooting object according to the abnormal analysis score and the preset analysis threshold to determine the abnormal analysis result of the shooting object; The portrait shooting pictures include multiple location shooting pictures and multiple tracking shooting pictures; the social relationship network includes at least one relationship level; One relationship level includes at least one of the related persons; The determining the picture analysis score corresponding to the shooting object according to the portrait shooting pictures and the pictures of the related persons includes: Determine the total number of times the shooting object appears in all the location shooting pictures, and record this total number as the location appearance times; one shooting object corresponds to one location appearance times; Obtain a preset location analysis weight, and determine the location analysis score corresponding to the shooting object included in the location shooting pictures according to the location analysis weight and the location appearance times; Perform feature comparison between the tracking shooting pictures and the pictures of the related persons to obtain the image comparison result of the shooting object included in the tracking shooting pictures; Determine the tracking analysis score of the shooting object included in the tracking shooting pictures according to the relationship level and the image comparison result; Determine the picture analysis score corresponding to the shooting object according to the location analysis score and the tracking analysis score; The determining the information analysis score corresponding to the shooting object according to the shooting object information of the shooting object includes: Obtain the basic object information and voice data information corresponding to the shooting object from the shooting object information; Obtain a preset abnormal word database; the preset abnormal word database includes at least one abnormal word; Match the voice data information with each abnormal word to obtain a word matching result, and determine the voice analysis score corresponding to the shooting object according to the word matching result; Determine the basic analysis score corresponding to the shooting object according to the basic object information, and determine the information analysis score corresponding to the shooting object according to the basic analysis score and the voice analysis score.

2. The method for analyzing target personnel according to claim 1, characterized in that, before obtaining the social relationship network of historical target personnel, it includes: Determine all related personnel in the preset personnel attribute library who have an associated relationship with the historical target personnel; Respectively determine the relevant levels between the historical target personnel and each of the related personnel; According to the relevant levels, determine the relationship levels corresponding to the historical target personnel and each of the related personnel, and set each of the related personnel in the relationship level corresponding to the related personnel; After all the related personnel are set in the relationship level corresponding to the related personnel, it is determined that the social relationship network of the historical target personnel is completed.

3. The target personnel analysis method according to claim 1, characterized in that The feature comparison of the tracking captured image and the related personnel image to obtain the image comparison result between the location captured image and the related personnel image includes: Perform face recognition on the tracking captured image to obtain at least one intercepted face image; the intercepted face image refers to a face image intercepted from the tracking captured image that only contains one captured object; Associate and record the intercepted face images containing the same captured object as a captured face group; For each of the captured face groups, determine the image similarity between the related personnel image and the intercepted face images in the captured face group, and determine the image comparison result of the captured object included in the tracking captured image according to the image similarity.

4. The target personnel analysis method according to claim 1, characterized in that The determination of the tracking analysis score of the captured object included in the tracking captured image according to the relationship level and the image comparison result includes: Determine the total number of times the captured object appears in all the tracking captured images, and record this total number as the total tracking appearance times; one captured object corresponds to one total tracking appearance times; Determine the comparison success objects and comparison failure objects from all the captured objects according to the image comparison result; the comparison success objects refer to the captured objects that match the related personnel; the comparison failure objects refer to the captured objects that do not match all the related personnel; Obtain the level weight corresponding to the relationship level to which the related personnel matching the comparison success object belongs, and determine the tracking analysis score corresponding to the comparison success object according to the level weight and the total tracking appearance times; Determine the tracking analysis score corresponding to the comparison failure object according to the total tracking appearance times.

5. The target personnel analysis method according to claim 1, characterized in that The matching of the voice data information with each of the abnormal words to obtain the word matching result includes: Perform voice recognition on the voice data information to obtain a voice recognition text; Perform word segmentation on the voice recognition text to obtain the voice data words in the voice recognition text; Perform word vector conversion on the voice data words and the abnormal words to obtain the voice word vectors corresponding to the voice data words and the abnormal word vectors corresponding to the abnormal words; Determine the word vector distance between the speech word vector and the abnormal word vector, and determine the word matching result according to the word vector distance and a preset vector distance threshold.

6. A target person analysis device, characterized in that the target person analysis device is used to execute the target person analysis method according to any one of claims 1 to 5, and the target person analysis device includes: a relationship network acquisition module, configured to acquire the social relationship network of historical target persons; at least one related person is included in the social relationship network, and the related person has a picture of the related person; a captured picture acquisition module, configured to acquire a portrait captured picture; at least one captured object other than the historical target person is included in the portrait captured picture; the captured object has a captured object information; a picture analysis score determination module, configured to determine, for each captured object included in the portrait captured picture, a picture analysis score corresponding to the captured object according to the portrait captured picture and the picture of the related person; an abnormal analysis score determination module, configured to determine an information analysis score corresponding to the captured object according to the captured object information of the captured object, and determine an abnormal analysis score corresponding to the captured object according to the picture analysis score and the information analysis score; an abnormal analysis module, configured to obtain a preset analysis threshold, and perform abnormal behavior analysis on the captured object according to the abnormal analysis score and the preset analysis threshold to determine the abnormal analysis result of the captured object.

7. A computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that when the processor executes the computer program, it implements the target person analysis method according to any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that when the computer program is executed by a processor, it implements the target person analysis method according to any one of claims 1 to 5.

Images