High-risk group early warning method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a high-risk group early warning method. The method comprises the following steps: acquiring a monitoring image in a target area; performing micro-expression recognition processing based on the monitoring image to obtain micro-expressions of all people in the target area; in all personnel in the target area, determining the personnel whose micro-expressions accord with high-risk tendency micro-expressions as target personnel; acquiring risk assessment information of the target person in a preset dimension, and performing high risk assessment on the target person based on the risk assessment information to obtain a risk assessment value of the target person; and based on the risk assessment value, determining whether the target person is a high-risk person in the target area. The relevance between the micro-expressions and the high-risk personnel is comprehensively considered, so that the accuracy is higher when early warning is performed on the high-risk personnel, and misjudgment can be avoided as much as possible when evaluation is performed in different dimensions.

Description

Technical Field

[0001] The present invention relates to the field of urban governance technology, and in particular to a high-risk population early warning method, device, electronic equipment and storage medium. Background Art

[0002] In case of hurtful behaviors, such as hurting children or medical staff, traditional security measures often adopt the method of increasing the number of security personnel and having security personnel patrol for safety precautions. This method is inefficient and is prone to untimely prevention. Existing methods often use monitoring technology to collect monitoring images in the area, and use data analysis and machine learning methods to perform safety precaution detection, but the accuracy is low and it is often impossible to issue early warnings. It can only issue an alarm after the hurtful behavior occurs. Therefore, how to provide a high-risk population warning method that can provide early warnings for high-risk populations and still ensure a high accuracy rate has become an urgent problem to be solved. Summary of the invention

[0003] The embodiment of the present invention provides a high-risk population early warning method, which aims to solve the problem of low accuracy of existing high-risk population early warning. The target person is determined by identifying the micro-expressions of all people in the target area, and the risk assessment value of the target person is evaluated in different dimensions. According to the risk assessment value, it is determined whether the target person is a high-risk person in the target area. The correlation between micro-expressions and high-risk persons is comprehensively considered, so that the accuracy rate is higher when issuing high-risk population early warnings, and the above-mentioned evaluation in different dimensions can avoid misjudgment as much as possible.

[0004] In a first aspect, an embodiment of the present invention provides a high-risk population early warning method, the method comprising the following steps:

[0005] Acquire surveillance images within the target area;

[0006] Performing micro-expression recognition processing based on the surveillance image to obtain micro-expressions of all people in the target area;

[0007] Among all persons in the target area, persons whose micro-expressions match those of high-risk micro-expressions are identified as target persons;

[0008] Acquire risk assessment information of the target person in preset dimensions, and perform a high-risk assessment on the target person based on the risk assessment information to obtain a risk assessment value of the target person;

[0009] Based on the risk assessment value, it is determined whether the target person is a high-risk person in the target area.

[0010] Optionally, performing micro-expression recognition processing based on the surveillance image to obtain micro-expressions of all persons in the target area includes:

[0011] Perform face feature point extraction processing based on the monitored image to obtain the face feature points of each person among all the people in the target area;

[0012] Compare the face feature points of each person with a preset micro-expression feature point distribution map of high-risk tendencies;

[0013] Take the person corresponding to the face feature points that match the preset micro-expression feature point distribution map of high-risk tendencies as the target person.

[0014] Optionally, the risk assessment information includes spatio-temporal information and social information. Based on the risk assessment information, perform a high-risk assessment on the target person to obtain the risk assessment value of the target person, including:

[0015] Perform spatio-temporal risk assessment on the target person based on the spatio-temporal information and a preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target person;

[0016] Perform social risk assessment on the target person based on the social information and a preset social coefficient to obtain the social risk value of the target person;

[0017] Determine the risk assessment value of the target person based on the spatio-temporal risk value and the social risk value.

[0018] Optionally, the spatio-temporal information includes historical monitored images. Based on the spatio-temporal information and a preset spatio-temporal coefficient, perform spatio-temporal risk assessment on the target person to obtain the spatio-temporal risk value of the target person, including:

[0019] Based on the historical monitored images, determine at least one of the action trajectory, vehicle-person correlation, and offline shopping record of the target person;

[0020] Perform spatio-temporal risk assessment on the target person based on at least one of the action trajectory, vehicle-person correlation, and offline shopping record of the target person and the preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target person.

[0021] Optionally, the determination of the offline shopping record of the target person based on the historical monitored images includes:

[0022] In the historical monitored images, determine the shopping images of the target person;

[0023] Perform item recognition processing based on the shopping images to determine the purchased items of the target person and the purchase time corresponding to the purchased items;

[0024] Determine the offline shopping records of the target person based on the purchased items and the corresponding purchase times of the purchased items.

[0025] Optionally, the determining the vehicle-person correlation of the target person based on the historical surveillance images includes:

[0026] In the historical surveillance images, determine a target image including the target person and a target vehicle;

[0027] Based on the target image, determine the appearance frequency and appearance locations where the target person and the target vehicle appear simultaneously;

[0028] Based on the appearance frequency and appearance locations, determine the vehicle-person correlation of the target person.

[0029] Optionally, the social information includes social activity data, and the performing a social risk assessment on the target person based on the social information and a preset social coefficient to obtain a social risk value of the target person includes:

[0030] Based on the social activity data, determine the social relationships and interaction patterns of the target person;

[0031] Based on the social relationships, interaction patterns, and the preset social coefficient, perform a social risk assessment on the target person to obtain a social risk value of the target person.

[0032] In a second aspect, an embodiment of the present invention further provides a high-risk population warning device, and the high-risk population warning device includes:

[0033] A first acquisition module, configured to acquire surveillance images within a target area;

[0034] A first recognition module, configured to perform micro-expression recognition processing based on the surveillance images to obtain the micro-expressions of all persons within the target area;

[0035] A first determination module, configured to, among all persons within the target area, determine a person whose micro-expression conforms to a high-risk tendency micro-expression as a target person;

[0036] A risk assessment module, configured to acquire risk assessment information of the target person in a preset dimension, and based on the risk assessment information, perform a high-risk assessment on the target person to obtain a risk assessment value of the target person;

[0037] A second determination module, configured to determine whether the target person is a high-risk person within the target area based on the risk assessment value.

[0038] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the high-risk population warning method provided in an embodiment of the present invention when executing the computer program.

[0039] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the high-risk population early warning method provided in the embodiment of the invention are implemented.

[0040] In an embodiment of the present invention, a surveillance image in a target area is obtained; micro-expression recognition processing is performed based on the surveillance image to obtain the micro-expressions of all persons in the target area; among all persons in the target area, persons whose micro-expressions conform to high-risk tendency micro-expressions are determined as target persons; risk assessment information of the target person in preset dimensions is obtained, and based on the risk assessment information, a high-risk risk assessment is performed on the target person to obtain the risk assessment value of the target person; based on the risk assessment value, it is determined whether the target person is a high-risk person in the target area. The target person is determined by identifying the micro-expressions of all persons in the target area, the risk assessment value of the target person is evaluated in different dimensions, and whether the target person is a high-risk person in the target area is determined according to the risk assessment value. The correlation between micro-expressions and high-risk persons is comprehensively considered, so that the accuracy rate is higher when issuing early warnings for high-risk groups, and the above-mentioned evaluation in different dimensions can avoid misjudgment as much as possible. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 is a flow chart of a high-risk population early warning method provided by an embodiment of the present invention;

[0043] Figure 2 It is a structural schematic diagram of a high-risk population early warning device provided in an embodiment of the present invention;

[0044] Figure 3 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] 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 only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] As Figure 1 shown, Figure 1 is a flowchart of a high-risk population warning method provided by an embodiment of the present invention, including:

[0047] 101. Obtain surveillance images within the target area;

[0048] In the embodiments of the present invention, the above-mentioned high-risk population warning method can be applied to a regional supervision platform. The above-mentioned regional supervision platform can be constructed by a server or a server cluster. The above-mentioned server or server cluster can be any electronic device with functions such as image processing, image recognition, data storage, and data transmission.

[0049] The above-mentioned regional supervision platform can monitor the above-mentioned target area through the above-mentioned surveillance images and give an alarm when high-risk personnel appear in the above-mentioned target area.

[0050] The above-mentioned target area can be any area that needs to be supervised. For example, it can be areas such as hospitals, schools, railway stations, shopping malls, etc. The above-mentioned surveillance images can be an image sequence composed of continuous image frames or composed of discontinuous single-frame image frames.

[0051] An image acquisition device can be deployed within the above-mentioned target area, and the above-mentioned target area is image-acquired by the above-mentioned image acquisition device to obtain the above-mentioned surveillance images. Or it can also be through satellite equipment to image-acquire the above-mentioned target area to obtain the above-mentioned surveillance images.

[0052] 102. Perform micro-expression recognition processing based on the surveillance images to obtain the micro-expressions of all personnel within the target area;

[0053] In the embodiments of the present invention, the above-mentioned micro-expression can be a facial expression that flashes instantaneously and can reveal a person's true feelings and emotions. The above-mentioned micro-expression can last for as short as 1 / 25 seconds. Although a subconscious micro-expression may only last for an instant, it can often more truly reflect a person's mental state. Usually, the above-mentioned micro-expression is caused by emotions and is not controlled by thinking, or is due to habit.

[0054] The above micro-expression recognition process can construct a micro-expression recognition model to be trained through any deep learning algorithm, and train the above micro-expression recognition model to be trained with sample micro-expression images to obtain a trained micro-expression recognition model. Input the above monitoring images into the above trained micro-expression recognition model for micro-expression recognition processing, and the micro-expressions of all personnel in the above target area can be obtained. The above deep learning algorithm can be an optical flow method, Gabor wavelet transform, local binary pattern (LBP), etc. The above sample micro-expression images include corresponding micro-expression labels. The above sample micro-expression images can be surprised micro-expression images, disgusted micro-expression images, angry micro-expression images, fearful micro-expression images, sad micro-expression images, happy micro-expression images, etc. The above micro-expression labels can be surprised micro-expression labels, disgusted micro-expression labels, angry micro-expression labels, fearful micro-expression labels, sad micro-expression labels, happy micro-expression labels, etc. The above micro-expression labels can include micro-expression action labels corresponding to each micro-expression. For example, the micro-expression action labels corresponding to a happy micro-expression image can be a micro-expression action label of slightly squinted eyes and a micro-expression action label of upturned corners of the mouth. The micro-expressions corresponding to a sad micro-expression image can be a micro-expression action label of downturned corners of the mouth and a micro-expression action label of drooping head.

[0055] During the training process, the above micro-expression recognition model to be trained can learn various micro-expressions and corresponding micro-expression actions in the above sample micro-expression images. For example, the micro-expression actions corresponding to a happy micro-expression can be slightly squinted eyes and upturned corners of the mouth, and the micro-expression actions corresponding to a sad micro-expression can be downturned corners of the mouth and drooping head.

[0056] 103. Among all the personnel in the target area, those whose micro-expressions conform to high-risk tendency micro-expressions are determined as target personnel;

[0057] In the embodiment of the present invention, the above high-risk tendency micro-expressions can be wide-open eyes, lowered eyebrows, downturned corners of the mouth, closed mouth, flickering eyes, etc. Among them, wide-open eyes can indicate that the person is concentrating, may be evaluating a target or preparing to take action. Lowered eyebrows can indicate that the person may feel angry or dissatisfied and may act to harm others. Downturned corners of the mouth indicate that the person is feeling frustrated or dissatisfied and is in a negative mood and may act to harm others. A closed mouth indicates that the person may be trying to control their emotions or is thinking about how to take action. Flickering eyes indicate that the person may be considering taking a certain action or is examining the surrounding targets or environment.

[0058] Specifically, template images of microexpressions with high-risk tendencies can be collected, and the microexpressions of all personnel within the target area can be compared with the above template images. Those personnel with successful comparisons are regarded as the above target personnel. The above template images can be collected from channels such as movies, TV dramas, newspapers, and news videos.

[0059] 104. Obtain the risk assessment information of the target personnel in a preset dimension, and based on the risk assessment information, conduct a high-risk assessment on the target personnel to obtain the risk assessment value of the target personnel.

[0060] In the embodiments of the present invention, the above preset dimension can be one dimension or multiple dimensions, and can include dimensions such as time and space dimension, social dimension, etc. It can be understood that the more the number of the above preset dimensions, the more comprehensive the dimensions considered when conducting a high-risk assessment on the target personnel, and the more accurate its risk assessment value.

[0061] Specifically, the risk assessment information of the above target personnel in the time and space dimension can be obtained, a time and space risk assessment is conducted on the target personnel to obtain the time and space risk assessment value, the risk assessment information of the above target personnel in the social dimension is obtained, a social risk assessment is conducted on the target personnel to obtain the social risk assessment value of the above target personnel, and the above time and space risk assessment value and the above social risk assessment value are normalized to obtain the risk assessment value of the above target personnel.

[0062] 105. Based on the risk assessment value, determine whether the target personnel is a high-risk personnel within the target area.

[0063] In the embodiments of the present invention, the risk assessment threshold can be determined according to the type of the target area. The above risk assessment value is compared with the above risk assessment threshold. If the above risk assessment value is greater than or equal to the above risk assessment threshold, then the above target personnel is a high-risk personnel within the above target area. It can be understood that the above high-risk personnel can be highly dangerous personnel for the above target area.

[0064] Specifically, the types of the above-mentioned target areas may include hospital type, school type, square type, station type, etc. The pedestrian flow and personnel composition of different types of target areas are different. For example, if the type of the above-mentioned target area is school type, during weekdays, the pedestrian flow is the largest, and most of the personnel composition are minors. Then, a relatively low risk assessment threshold can be set to avoid missed judgments as much as possible and ensure the safety of this area. If the type of the above-mentioned target area is hospital type, generally, the pedestrian flow in the hospital is larger on rest days, and the personnel composition is relatively complex, including medical staff, patients, accompanying personnel, etc., and the age span is also relatively large. Therefore, a slightly lower risk assessment threshold can be set to avoid missed judgments and ensure the safety of this area. If the type of the above-mentioned target area is square type, generally, during rest time, the pedestrian flow in the square is larger, including square dance teams, walking people, etc., and the age span is smaller compared to hospitals and schools. Therefore, a slightly higher risk assessment threshold can be set to avoid misjudgments and ensure the safety of this area. If the type of the above-mentioned target area is station type, generally, only during holidays, the pedestrian flow is relatively high, and only during holidays, the age span will be relatively large. Therefore, a relatively high risk assessment threshold can be set to avoid misjudgments.

[0065] It can be understood that the risk assessment threshold corresponding to the target area of the above-mentioned station type is the highest, the risk assessment threshold corresponding to the target area of the square type is less than the risk assessment threshold corresponding to the target area of the above-mentioned station type, the risk assessment threshold corresponding to the target area of the hospital type is less than the risk assessment threshold corresponding to the target area of the above-mentioned square type, and the risk assessment threshold corresponding to the target area of the above-mentioned school type is less than the risk assessment threshold corresponding to the target area of the above-mentioned hospital type.

[0066] Or, based on the above-mentioned station type, the risk assessment threshold can be obtained, and then the risk assessment threshold can be dynamically adjusted according to the pedestrian flow in different time periods of different types of target areas. If the pedestrian flow in the target area during the current time period is relatively high, the risk assessment threshold of the target area during the current time period can be correspondingly increased; if the pedestrian flow in the target area during the current time period is relatively low, the risk assessment threshold of the target area during the current time period can be correspondingly decreased.

[0067] In an embodiment of the present invention, a surveillance image in a target area is obtained; micro-expression recognition processing is performed based on the surveillance image to obtain the micro-expressions of all persons in the target area; among all persons in the target area, persons whose micro-expressions conform to high-risk tendency micro-expressions are determined as target persons; risk assessment information of the target person in preset dimensions is obtained, and based on the risk assessment information, a high-risk risk assessment is performed on the target person to obtain the risk assessment value of the target person; based on the risk assessment value, it is determined whether the target person is a high-risk person in the target area. The target person is determined by identifying the micro-expressions of all persons in the target area, the risk assessment value of the target person is evaluated in different dimensions, and whether the target person is a high-risk person in the target area is determined according to the risk assessment value. The correlation between micro-expressions and high-risk persons is comprehensively considered, so that the accuracy rate is higher when issuing early warnings for high-risk groups, and the above-mentioned evaluation in different dimensions can avoid misjudgment as much as possible.

[0068] It is understandable that in the specific implementation of this application, it involves monitoring images, social information, social activity data and other related data. When the embodiments in this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0069] Optionally, in the step of performing micro-expression recognition processing based on surveillance images to obtain the micro-expressions of all people in the target area, facial feature point extraction processing can also be performed based on the surveillance images to obtain the facial feature points of each person in the target area; the facial feature points of each person are compared with a preset high-risk tendency micro-expression feature point distribution map; and the people corresponding to the facial feature points that meet the preset high-risk tendency micro-expression feature point distribution map are taken as target persons.

[0070] In the embodiment of the present invention, a facial feature point extraction model can be constructed by any facial feature point extraction algorithm, and the above-mentioned monitoring image can be input into the above-mentioned facial feature point extraction model to perform the above-mentioned facial feature point extraction process to obtain the facial feature points of each person in the above-mentioned target area. The above-mentioned facial feature point extraction algorithm can be an MTCNN algorithm, a Haar feature algorithm, a cascade classifier algorithm, an MTCNN algorithm, a RetinaFace, an Eigenfaces, a Fisherfaces algorithm, etc.

[0071] Specifically, template images of microexpressions with high-risk tendencies can be collected through channels such as film and television dramas, newspapers, and news videos. The template images of microexpressions with high-risk tendencies are input into the above-mentioned face feature point extraction model for face feature extraction processing to obtain the above-mentioned preset distribution map of high-risk tendency microexpression feature points. The face feature points of each person are compared with the preset distribution map of high-risk tendency microexpression feature points, and the person corresponding to the face feature points that conform to the preset distribution map of high-risk tendency microexpression feature points is used as the above-mentioned target person.

[0072] Optionally, in the step of performing a high-risk assessment on the target person based on the risk assessment information to obtain the risk assessment value of the target person, a spatio-temporal risk assessment can also be performed on the target person based on the spatio-temporal information and the preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target person; a social risk assessment can be performed on the target person based on the social information and the preset social coefficient to obtain the social risk value of the target person; based on the spatio-temporal risk value and the social risk value, the risk assessment value of the target person is determined.

[0073] In the embodiment of the present invention, the above-mentioned risk assessment information includes spatio-temporal information and social information. The preset spatio-temporal coefficient and the preset social coefficient can be set through historical assessment experience. A spatio-temporal risk assessment is performed on the target person through the preset spatio-temporal coefficient and the above-mentioned spatio-temporal information to obtain the spatio-temporal risk value of the target person. A social risk assessment is performed on the above-mentioned target person based on the social information and the preset social coefficient to obtain the social risk value of the target person. Normalization processing is performed based on the above-mentioned spatio-temporal risk value and the above-mentioned social risk value to obtain the risk assessment value of the target person.

[0074] Optionally, in the step of performing a spatio-temporal risk assessment on the target person based on the spatio-temporal information and the preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target person, at least one of the action trajectory, vehicle-person correlation, and offline shopping record of the target person can also be determined based on the historical surveillance images; a spatio-temporal risk assessment is performed on the target person based on at least one of the action trajectory, vehicle-person correlation, and offline shopping record of the target person and the preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target person.

[0075] In the embodiment of the present invention, the above-mentioned spatio-temporal information includes historical surveillance images. The historical surveillance images can include historical surveillance time and historical surveillance location. The appearance location of the above-mentioned target person can be determined according to the above-mentioned historical surveillance location in the map of the target area. Each appearance location is connected in series according to the order of the above-mentioned historical surveillance time to obtain the above-mentioned action trajectory.

[0076] The target vehicle and the target person can be determined from the above historical surveillance images. The association between the person and the vehicle can be determined based on the frequency of simultaneous appearance and the locations where the above target vehicle and target person appear. If the frequency of simultaneous appearance is high, the association between the person and the vehicle is strong; otherwise, it is weak. The closer the locations where the target person and the target vehicle appear simultaneously, the stronger the association between the person and the vehicle; otherwise, it is weak. The above target vehicle can be the vehicle in the image that includes the target person in the above historical surveillance images.

[0077] The shopping image of the above target person can be determined from the above historical surveillance images. The shopping image can include the shopping image acquisition time and the acquisition location. Object recognition processing is performed based on the above shopping image to obtain the object detection frame of the objects in the above shopping image. Person recognition processing is performed based on the above shopping image to obtain the detection frame of the above target person. The items purchased by the above target person are determined according to the relative positions of the detection frame of the above target person and the above object detection frame in the above shopping image. The purchase time of the above purchased items is determined according to the acquisition time of the above shopping image.

[0078] Based on at least one of the action trajectory of the target person, the association between the person and the vehicle, and the offline shopping record, and a preset spatio-temporal coefficient, a spatio-temporal risk assessment is performed on the target person to obtain the above spatio-temporal risk value.

[0079] It should be noted that a basic spatio-temporal risk value can be set for each person. According to the above action trajectory, it can be determined whether the above target person often enters and exits places where idle people gather. If so, the above basic spatio-temporal risk value can be correspondingly increased through the above spatio-temporal risk coefficient; if not, the above basic spatio-temporal risk value can be correspondingly decreased through the above spatio-temporal risk coefficient. Or, according to the above association between the person and the vehicle, it can be determined whether the above target person and the target vehicle often appear in public places. If so, it indicates that the above target person may be scouting the area, and the above basic spatio-temporal risk value can be correspondingly increased through the above spatio-temporal risk coefficient; if not, the above basic spatio-temporal risk value can be correspondingly decreased through the above spatio-temporal risk coefficient. Or, through the above offline shopping record, it can be determined whether the above target person purchases dangerous items such as chemical materials and controlled knives. If so, the above basic spatio-temporal risk value can be correspondingly increased through the above spatio-temporal risk coefficient; if not, the above basic spatio-temporal risk value can be correspondingly decreased through the above spatio-temporal risk coefficient.

[0080] Specifically, the above spatio-temporal risk value is calculated based on the information of comprehensive sub-dimensions. The above spatio-temporal risk coefficient can include multiple sub-spatio-temporal risk coefficients, and can be illustrated by the following spatio-temporal risk calculation formula:

[0081]

[0082] Wherein, the above-mentioned T represents the above-mentioned spatio-temporal risk value, the above-mentioned N1 represents the spatio-temporal information of the first sub-dimension (i.e., it can be the above-mentioned action trajectory), the above-mentioned X1 can represent the sub-spatio-temporal coefficient corresponding to the above-mentioned first sub-dimension, the above-mentioned N2 represents the spatio-temporal information of the second sub-dimension (i.e., it can be the above-mentioned vehicle-person correlation), the above-mentioned X2 can represent the sub-spatio-temporal coefficient corresponding to the above-mentioned second sub-dimension, the above-mentioned X1 can represent the sub-spatio-temporal coefficient corresponding to the above-mentioned first sub-dimension, the above-mentioned N3 represents the spatio-temporal information of the third sub-dimension (i.e., it can be the above-mentioned offline shopping record), the above-mentioned X3 can represent the sub-spatio-temporal coefficient corresponding to the above-mentioned third sub-dimension, the above-mentioned N n represents the spatio-temporal information of the nth sub-dimension, and the above-mentioned X n can represent the sub-spatio-temporal coefficient corresponding to the above-mentioned nth sub-dimension, the above-mentioned A represents the above-mentioned basic spatio-temporal risk value, and the above-mentioned basic spatio-temporal risk value can be a constant, for example, it can be 1 or 100.

[0083] Optionally, in the step of determining the offline shopping record of the target person based on the historical surveillance image, the shopping image of the target person can also be determined in the historical surveillance image; the items purchased by the target person and the purchase time corresponding to the purchased items are determined by performing item recognition processing on the shopping image; the offline shopping record of the target person is determined based on the purchased items and the purchase time corresponding to the purchased items.

[0084] In the embodiment of the present invention, the shopping places passed by the above-mentioned target person can be determined according to the above-mentioned action trajectory, the shopping image of the above-mentioned target person can be determined by calling the historical surveillance image of the above-mentioned shopping place, the item detection frame of the above-mentioned item is obtained by performing item recognition processing on the above-mentioned shopping image, the items purchased by the target person are determined according to the relative position of the above-mentioned item detection frame and the above-mentioned target person in the above-mentioned image, and the purchase time corresponding to the above-mentioned purchased items is determined according to the acquisition time of the historical surveillance image of the above-mentioned shopping place. The purchased items are sorted based on the purchase time corresponding to the purchased items to obtain the offline shopping record table of the above-mentioned target person, and the offline item purchase record of the above-mentioned target person is checked through the offline shopping record table of the above-mentioned target person.

[0085] Optionally, in the step of determining the vehicle-person correlation of the target person based on the historical surveillance image, the target image including the target person and the target vehicle can also be determined in the historical surveillance image; the appearance frequency and appearance location where the target person and the target vehicle appear simultaneously are determined based on the target image; the vehicle-person correlation of the target person is determined based on the appearance frequency and appearance location.

[0086] In an embodiment of the present invention, vehicle recognition processing, license plate recognition processing, and personnel recognition processing can be performed based on the above-mentioned historical surveillance images. In the above-mentioned historical surveillance images, personnel recognition processing is performed to obtain the historical surveillance images of the above-mentioned target personnel and the corresponding personnel recognition detection frames. Based on the historical surveillance images of the above-mentioned target personnel, vehicle recognition processing and license plate recognition processing are performed to obtain vehicle detection frames and license plate character recognition results. According to the relative positions of the above-mentioned personnel recognition detection frames and the above-mentioned vehicle detection frames, the vehicles that are close to the position of the above-mentioned target personnel in the image are determined as candidate target vehicles. According to the license plate characters corresponding to the candidate target vehicles, the same vehicle with the same license plate characters among all candidate target vehicles is determined, and the same vehicle with the same license plate characters is used as the above-mentioned target vehicle. According to the acquisition time and acquisition location of the above-mentioned historical surveillance images, the occurrence frequency and occurrence location where the above-mentioned target personnel and target vehicle appear simultaneously are determined. If the occurrence frequency is higher, the relevance between the person and the vehicle is stronger; otherwise, it is weaker. If the occurrence locations are closer, the relevance between the person and the vehicle is stronger; otherwise, it is weaker.

[0087] Optionally, in the step of performing social risk assessment on the target personnel based on social information and a preset social coefficient to obtain the social risk value of the target personnel, the social relationship and interaction pattern of the target personnel can also be determined based on social activity data; based on the social relationship, interaction pattern, and preset social coefficient, social risk assessment is performed on the target personnel to obtain the social risk value of the target personnel.

[0088] In an embodiment of the present invention, the above-mentioned social information includes social activity data. Specifically, the social relationship and interaction pattern of the target personnel can be obtained by analyzing the social activity data of the target personnel through PageRank and social network analysis methods. The above-mentioned social activity data can include the types of personnel contacted by the above-mentioned target personnel, interaction records with others on public platforms on the Internet, offline interaction records, etc.

[0089] Specifically, a basic social risk value can be set. When the social relationship of the above-mentioned target personnel includes personnel with bad records and the span of the interaction pattern is large (that is, it includes multiple interaction patterns, including but not limited to offline communication, Internet platform communication, etc.), the above-mentioned basic social risk value can be increased accordingly through the above-mentioned preset social risk coefficient; otherwise, the above-mentioned basic social risk value can be decreased accordingly through the above-mentioned preset social risk coefficient.

[0090] As Figure 2 shown, an embodiment of the present invention also provides a high-risk population warning device, including:

[0091] A first acquisition module 201, configured to acquire surveillance images within a target area;

[0092] The first recognition module 202 is configured to perform micro-expression recognition processing based on the monitoring image to obtain the micro-expressions of all persons within the target area;

[0093] The first determination module 203 is configured to determine, among all persons within the target area, the persons whose micro-expressions conform to high-risk tendency micro-expressions as target persons;

[0094] The risk assessment module 204 is configured to obtain the risk assessment information of the target person in a preset dimension, and perform a high-risk assessment on the target person based on the risk assessment information to obtain the risk assessment value of the target person;

[0095] The second determination module 205 is configured to determine whether the target person is a high-risk person within the target area based on the risk assessment value.

[0096] Optionally, the first recognition module 202 includes:

[0097] The feature extraction sub-module is configured to perform face feature point extraction processing based on the monitoring image to obtain the face feature points of each person among all persons within the target area;

[0098] The first comparison sub-module is configured to compare the face feature points of each person with a preset high-risk tendency micro-expression feature point distribution map;

[0099] The second comparison sub-module is configured to use the person corresponding to the face feature points that conform to the preset high-risk tendency micro-expression feature point distribution map as the target person.

[0100] Optionally, the risk assessment module 204 includes:

[0101] The first risk assessment sub-module is configured to perform spatio-temporal risk assessment on the target person based on the spatio-temporal information and a preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target person;

[0102] The second risk assessment sub-module is configured to perform social risk assessment on the target person based on the social information and a preset social coefficient to obtain the social risk value of the target person;

[0103] The first determination sub-module is configured to determine the risk assessment value of the target person based on the spatio-temporal risk value and the social risk value.

[0104] Optionally, the first risk assessment sub-module includes:

[0105] The first determination unit is configured to determine at least one of the action trajectory, vehicle-person correlation, and offline shopping record of the target person based on the historical monitoring image;

[0106] A first risk assessment unit for performing spatio-temporal risk assessment on the target person based on at least one of the action trajectory, vehicle-person correlation, and offline shopping record of the target person and the preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target person.

[0107] Optionally, the first determination unit includes:

[0108] A first determination subunit for determining the shopping image of the target person in the historical surveillance images;

[0109] A second determination subunit for performing item recognition processing based on the shopping image to determine the purchased items of the target person and the purchase time corresponding to the purchased items;

[0110] A third determination subunit for determining the offline shopping record of the target person based on the purchased items and the purchase time corresponding to the purchased items.

[0111] Optionally, the first determination unit further includes:

[0112] A fourth determination subunit for determining a target image including the target person and the target vehicle in the historical surveillance images;

[0113] A fifth determination subunit for determining the appearance frequency and appearance location where the target person and the target vehicle appear simultaneously based on the target image;

[0114] A sixth determination subunit for determining the vehicle-person correlation of the target person based on the appearance frequency and appearance location.

[0115] Optionally, the second risk assessment sub-module includes:

[0116] A second determination unit for determining the social relationship and interaction pattern of the target person based on the social activity data;

[0117] A second risk assessment unit for performing social risk assessment on the target person based on the social relationship, interaction pattern, and the preset social coefficient to obtain the social risk value of the target person.

[0118] As Figure 3 shown, an embodiment of the present invention further provides an electronic device, which is characterized by including a processor, and the above-mentioned processor can execute any one of the above-mentioned high-risk population warning methods.

[0119] Specifically, it includes a processor 301 and a memory 302, as well as a computer program stored in the memory 302 and capable of running on the processor 301 to execute the high-risk population warning method, where:

[0120] The processor 301 runs the calculator program of the high-risk population warning method stored in the memory 302 and executes the following steps:

[0121] Obtain the surveillance images within the target area;

[0122] Perform micro-expression recognition processing based on the surveillance images to obtain the micro-expressions of all personnel within the target area;

[0123] Among all the personnel within the target area, determine the personnel whose micro-expressions conform to the high-risk tendency micro-expressions as the target personnel;

[0124] Obtain the risk assessment information of the target personnel in a preset dimension, and based on the risk assessment information, perform a high-risk assessment on the target personnel to obtain the risk assessment value of the target personnel;

[0125] Based on the risk assessment value, determine whether the target personnel is a high-risk personnel within the target area.

[0126] Optionally, the micro-expression recognition processing based on the surveillance images performed by the processor 301 to obtain the micro-expressions of all personnel within the target area includes:

[0127] Perform face feature point extraction processing based on the surveillance images to obtain the face feature points of each person among all the personnel within the target area;

[0128] Compare the face feature points of each person with the preset high-risk tendency micro-expression feature point distribution map;

[0129] Take the person corresponding to the face feature points that conform to the preset high-risk tendency micro-expression feature point distribution map as the target personnel.

[0130] Optionally, the risk assessment information executed by the processor 301 includes spatio-temporal information and social information. The high-risk assessment of the target personnel based on the risk assessment information to obtain the risk assessment value of the target personnel includes:

[0131] Perform spatio-temporal risk assessment on the target personnel based on the spatio-temporal information and a preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target personnel;

[0132] Perform social risk assessment on the target personnel based on the social information and a preset social coefficient to obtain the social risk value of the target personnel;

[0133] Based on the spatio-temporal risk value and the social risk value, determine the risk assessment value of the target person.

[0134] Optionally, the spatio-temporal information executed by the processor 301 includes historical surveillance images. The spatio-temporal risk assessment of the target person based on the spatio-temporal information and a preset spatio-temporal coefficient to obtain the spatio-temporal risk value of the target person includes:

[0135] Based on the historical surveillance images, determine at least one of the action trajectory, vehicle-person correlation, and offline shopping record of the target person;

[0136] Based on at least one of the action trajectory, vehicle-person correlation, and offline shopping record of the target person and the preset spatio-temporal coefficient, conduct a spatio-temporal risk assessment on the target person to obtain the spatio-temporal risk value of the target person.

[0137] Optionally, the determination of the offline shopping record of the target person by the processor 301 based on the historical surveillance images includes:

[0138] In the historical surveillance images, determine the shopping images of the target person;

[0139] Based on the shopping images, perform item recognition processing to determine the purchased items of the target person and the purchase time corresponding to the purchased items;

[0140] Based on the purchased items and the purchase time corresponding to the purchased items, determine the offline shopping record of the target person.

[0141] Optionally, the determination of the vehicle-person correlation of the target person by the processor 301 based on the historical surveillance images includes:

[0142] In the historical surveillance images, determine the target images including the target person and the target vehicle;

[0143] Based on the target images, determine the appearance frequency and appearance location where the target person and the target vehicle appear simultaneously;

[0144] Based on the appearance frequency and appearance location, determine the vehicle-person correlation of the target person.

[0145] Optionally, the social information executed by the processor 301 includes social activity data. The social risk assessment of the target person based on the social information and a preset social coefficient to obtain the social risk value of the target person includes:

[0146] Based on the social activity data, determine the social relationships and interaction patterns of the target person;

[0147] Based on the social relationship, interaction pattern, and the preset social coefficient, conduct a social risk assessment on the target person to obtain the social risk value of the target person.

[0148] The embodiments of the present invention also provide a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements each process of the high-risk population warning method or the high-risk population warning method for the application end provided by the embodiments of the present invention, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0149] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The above computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above various methods. Among them, the above computer-readable storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0150] The above-disclosed are only the preferred embodiments of the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.

Claims

1. A warning method for high-risk groups, characterized in that, The method includes the following steps: Obtain surveillance images within the target area; Perform micro-expression recognition processing based on the surveillance images to obtain the micro-expressions of all persons within the target area; Among all persons within the target area, determine the persons whose micro-expressions conform to high-risk tendency micro-expressions as target persons; Obtain the risk assessment information of the target persons in a preset dimension, and based on the risk assessment information, perform a high-risk assessment on the target persons to obtain the risk assessment values of the target persons; Based on the risk assessment values, determine whether the target persons are high-risk persons within the target area.

2. The high-risk population warning method according to claim 1, characterized in that The performing micro-expression recognition processing based on the surveillance images to obtain the micro-expressions of all persons within the target area includes: Perform face feature point extraction processing based on the surveillance images to obtain the face feature points of each person among all persons within the target area; Compare the face feature points of each person with a preset distribution map of high-risk tendency micro-expression feature points; Take the persons corresponding to the face feature points that conform to the preset distribution map of high-risk tendency micro-expression feature points as the target persons.

3. The high-risk population warning method according to claim 1, characterized in that, The risk assessment information includes spatio-temporal information and social information. The performing a high-risk assessment on the target persons based on the risk assessment information to obtain the risk assessment values of the target persons includes: Perform spatio-temporal risk assessment on the target persons based on the spatio-temporal information and a preset spatio-temporal coefficient to obtain the spatio-temporal risk values of the target persons; Perform social risk assessment on the target persons based on the social information and a preset social coefficient to obtain the social risk values of the target persons; Based on the spatio-temporal risk values and the social risk values, determine the risk assessment values of the target persons.

4. The high-risk population warning method according to claim 3, wherein The spatio-temporal information includes historical surveillance images. The performing spatio-temporal risk assessment on the target persons based on the spatio-temporal information and a preset spatio-temporal coefficient to obtain the spatio-temporal risk values of the target persons includes: Based on the historical surveillance images, determine at least one of the action trajectories, vehicle-person correlations, and offline shopping records of the target persons; Perform spatio-temporal risk assessment on the target persons based on at least one of the action trajectories, vehicle-person correlations, and offline shopping records of the target persons and the preset spatio-temporal coefficient to obtain the spatio-temporal risk values of the target persons.

5. The high-risk population warning method according to claim 4, characterized in that The determining the offline shopping records of the target persons based on the historical surveillance images includes: In the historical surveillance images, determine the shopping images of the target persons; Perform item recognition processing based on the shopping images to determine the purchased items of the target persons and the purchase times corresponding to the purchased items; Based on the purchased items and the purchase times corresponding to the purchased items, determine the offline shopping records of the target persons.

6. The high-risk population warning method according to claim 4, characterized in that, The determining the vehicle-person correlations of the target persons based on the historical surveillance images includes: In the historical surveillance images, determine target images including the target persons and target vehicles; Based on the target images, determine the appearance frequencies and appearance locations where the target persons and target vehicles appear simultaneously; Based on the occurrence frequency and occurrence location, determine the vehicle-person correlation of the target person.

7. The high-risk population warning method according to claim 3, wherein The social information includes social activity data. The social risk assessment of the target person based on the social information and a preset social coefficient to obtain the social risk value of the target person includes: Based on the social activity data, determine the social relationships and interaction patterns of the target person; Based on the social relationships, interaction patterns, and the preset social coefficient, conduct a social risk assessment on the target person to obtain the social risk value of the target person.

8. An early warning device for high-risk groups, characterized in that, The high-risk population warning device includes: A first acquisition module for acquiring surveillance images within a target area; A first recognition module for performing micro-expression recognition processing based on the surveillance images to obtain the micro-expressions of all persons within the target area; A first determination module for, among all persons within the target area, determining as the target person those persons whose micro-expressions conform to high-risk tendency micro-expressions; A risk assessment module for acquiring the risk assessment information of the target person in a preset dimension and, based on the risk assessment information, conducting a high-risk assessment on the target person to obtain the risk assessment value of the target person; A second determination module for, based on the risk assessment value, determining whether the target person is a high-risk person within the target area.

9. An electronic device, characterized in that, Includes: A memory, a processor, and a computer program stored on the memory and executable on the processor, where when the processor executes the computer program, the steps in the high-risk population warning method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is executed by the processor, the steps in the high-risk population warning method according to any one of claims 1 to 7 are implemented.

Images