Abnormal behavior warning method, device and application

Abstract

This application proposes an abnormal behavior warning method, device and application, which uses a progressive approach to integrate artificial intelligence analysis methods and social hazard assessment methods to monitor and track the perpetrators. First, a preset image feature model is used to compare and analyze the behavior images, and then a preset behavior factor feature model is used to compare and analyze the behavior factors. In some cases, the four-dimensional social hazard assessment method is finally used to evaluate the behavior of the perpetrators. It can timely and accurately warn of possible abnormal behaviors, reduce social conflicts and play a risk prevention role.

Description

Technical Field

[0001] The present application relates to the field of data analysis, and in particular to an abnormal behavior early warning method, device and application. Background Art

[0002] Socially harmful behaviors often occur due to the extreme behavior of people who are hostile to society. How to timely control and warn of the dangerous behavior of such people to avoid the escalation of the situation is a major challenge in social governance.

[0003] The means currently adopted by the existing technology for conflict and dispute investigation system is: each community (village) reports conflict and dispute incidents and implements hierarchical dynamic management. If the conflict and dispute incidents cannot be mediated within the community, such conflict and dispute incidents will be reported layer by layer to the street or relevant management department to achieve hierarchical management of conflict and dispute incidents. However, this method is relatively inefficient in investigating conflict and dispute issues, and the risk assessment of conflict and dispute incidents depends on the subjective consciousness of the assessor. There is a problem of being unable to timely and accurately warn of conflict and dispute incidents that may evolve into socially harmful behaviors, and it is also impossible to track and warn the parties involved in conflict and dispute incidents that may evolve into socially harmful behaviors. There is a problem of one-size-fits-all or neglect of management.

[0004] Of course, there are currently recognition algorithms that can identify the behavior of the perpetrators, but the current recognition algorithms can only identify standard behaviors, while the abnormal behaviors of the perpetrators are often changeable, which makes the current recognition algorithms prone to misjudgment or inaccurate recognition. Summary of the Invention

[0005] The embodiments of the present application provide a method, device, and application for warning of abnormal behavior, which analyzes conflict and dispute events to determine the perpetrators who may have abnormal behavior, and makes a scientific and accurate risk assessment of the perpetrators based on comprehensive video images and perpetrator factors, so as to timely warn of abnormal behavior.

[0006] In a first aspect, an embodiment of the present application provides a method for early warning of abnormal behavior, the method comprising:

[0007] Selecting a person who behaves abnormally, and capturing at least one behavioral image and at least one behavioral factor of the person by at least one camera assembly, wherein the behavioral factor is used to characterize a characteristic dimension of abnormal behavior;

[0008] Inputting all behavioral images of each camera component into a preset image feature model to obtain at least one single image convolution value, combining and analyzing all the single image convolution values ​​to obtain an image analysis probability value, and if the image analysis probability value is greater than a set image threshold, determining that the person has exhibited abnormal behavior corresponding to the preset image feature model;

[0009] If the image analysis probability value is less than the image setting threshold, all behavioral factors of each camera component are input into the preset behavioral factor feature model to obtain at least one single behavioral convolution value, and all the single behavioral convolution values ​​are combined and analyzed to obtain a behavioral analysis probability value. If the behavioral analysis probability value is greater than the behavioral setting threshold, it is determined that the actor has abnormal behavior corresponding to the preset behavioral factor feature model.

[0010] In a second aspect, an embodiment of the present application provides an abnormal behavior warning device, comprising:

[0011] An analysis data acquisition unit is used to select a person and collect at least one behavior image and at least one behavior factor of the person through at least one camera component, wherein the behavior factor is used to characterize the characteristic dimension of abnormal behavior;

[0012] A behavior image analysis unit is configured to input all behavior images of each camera assembly into a preset image feature model to obtain at least one single image convolution value, and to combine and analyze all the single image convolution values ​​to obtain an image analysis probability value. If the image analysis probability value is greater than a set image threshold, it is determined that the person has exhibited abnormal behavior corresponding to the preset image feature model.

[0013] The behavior factor analysis unit is used to input all the behavior factors of each camera component into the preset behavior factor characteristic model to obtain at least one single behavior convolution value when the image analysis probability value is less than the image setting threshold, and to combine and analyze all the single behavior convolution values ​​to obtain a behavior analysis probability value. If the behavior analysis probability value is greater than the behavior setting threshold, it is determined that the actor has abnormal behavior corresponding to the preset behavior factor characteristic model.

[0014] In a third aspect, an embodiment of the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the abnormal behavior warning method.

[0015] In a fourth aspect, an embodiment of the present application provides a readable storage medium, in which a computer program is stored. The computer program includes a program code for controlling a process to execute a process, and the process includes the abnormal behavior warning method.

[0016] The main contributions and innovations of the embodiments of this application are as follows:

[0017] The embodiments of this application use a progressive approach to integrate artificial intelligence analysis methods and social harm assessment methods to monitor and track perpetrators. First, behavioral images are compared and analyzed using a preset image feature model, and then behavioral factors are compared and analyzed using a preset behavioral factor feature model. In some cases, if accurate abnormal behavior cannot be obtained through the previous methods, the four-dimensional social harm assessment method is finally used to conduct a behavioral assessment of the perpetrator. This can provide timely and accurate warnings of possible abnormal behavior, reduce social conflicts, and play a risk prevention role.

[0018] The details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 is a flow chart of an abnormal behavior warning method according to an embodiment of the present application;

[0021] Figure 2 It is a logical flow diagram of the abnormal behavior warning method according to an embodiment of the present application.

[0022] Figure 3 is a structural block diagram of a data storage device according to an embodiment of the present application;

[0023] Figure 4 Schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The implementations described in the following exemplary embodiments are not intended to represent all implementations consistent with one or more embodiments of this specification. Rather, they are merely examples of apparatuses and methods consistent with certain aspects of one or more embodiments of this specification, as detailed in the appended claims.

[0025] It should be noted that in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the method may include more or fewer steps than those described in this specification. In addition, a single step described in this specification may be broken down into multiple steps for description in other embodiments, and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0026] Example 1

[0027] The embodiment of the present application provides an abnormal behavior early warning method, which adopts a step-by-step deepening idea, integrates artificial intelligence analysis technology and risk assessment methods, and makes timely early warnings for abnormal behaviors of actors layer by layer. Specifically, refer to Figure 1 , the method comprising:

[0028] Selecting a person who behaves abnormally, and capturing at least one behavioral image and at least one behavioral factor of the person by at least one camera assembly, wherein the behavioral factor is used to characterize a characteristic dimension of abnormal behavior;

[0029] Inputting all behavioral images of each camera component into a preset image feature model to obtain at least one single image convolution value, combining and analyzing all the single image convolution values ​​to obtain an image analysis probability value, and if the image analysis probability value is greater than a set image threshold, determining that the person has exhibited abnormal behavior corresponding to the preset image feature model;

[0030] If the image analysis probability value is less than the image setting threshold, all behavioral factors of each camera component are input into the preset behavioral factor feature model to obtain at least one single behavioral convolution value, and all the single behavioral convolution values ​​are combined and analyzed to obtain a behavioral analysis probability value. If the behavioral analysis probability value is greater than the behavioral setting threshold, it is determined that the actor has abnormal behavior corresponding to the preset behavioral factor feature model.

[0031] It is worth mentioning that this solution uses a step-by-step approach to evaluate the behavior of the perpetrator, which is linked together to improve the accuracy of the assessment of socially harmful behaviors. First, a preset image feature model and the behavioral image of the perpetrator are compared and analyzed. If the result of the first analysis does not meet the specified requirements, the preset behavioral factor feature model and the behavioral factor of the perpetrator are compared and analyzed again. If the result of the second analysis still does not meet the specified requirements, the social harm assessment method is used to analyze the perpetrator. This solution takes into account the computational complexity and recognition efficiency of the recognition algorithm. The behavioral factors are not identified every time, and this solution can identify the behavioral images before identifying the behavioral factors. This can reduce the algorithm accuracy requirements for the behavioral factor recognition algorithm. In addition, in some cases, the social harm assessment method provided by this solution evaluates the behavior of the perpetrator from the perspective of influencing factors in four dimensions, and then promptly and scientifically issues early warnings for the perpetrator's abnormal behavior.

[0032] The abnormal behavior warning method provided by this solution can provide warnings of abnormal behavior against specific perpetrators. In embodiments of this solution, the perpetrator can be manually selected or identified through a social conflict and dispute investigation system. When this solution's abnormal behavior warning method is applied to conflict and dispute risk management scenarios, the perpetrator can be selected as a party involved in the conflict and dispute. For example, the perpetrator can be a party involved in an illegal demolition incident or a party involved in a neighborhood dispute.

[0033] After the actor's identity information is determined, the actor's behavioral image and behavioral factors can be captured using a camera assembly. In an embodiment of this solution, the actor's behavioral image and behavioral factors can be collected using a camera assembly placed on the street. The collected behavioral image and behavioral factors are uploaded and stored using network technology, and then compared and analyzed using a preset image feature model and / or a preset behavioral factor feature model.

[0034] In some embodiments, in "capturing at least one behavioral image and at least one behavioral factor of the actor through at least one camera component", the behavioral images and behavioral factors acquired by the camera component are screened to obtain behavioral images containing the actor and behavioral factors corresponding to the actor.

[0035] In addition, it is worth mentioning that the behavioral factors and behavioral images can be collected simultaneously, and there is no need to collect the behavioral factors again.

[0036] In “inputting all behavioral images of each camera component into a preset image feature model to obtain at least one single image convolution value”, all the behavioral images are encoded to obtain behavioral image encoding, and all the behavioral image encodings of each camera component are input into a preset image feature model to obtain at least one single image convolution value, wherein each single image convolution value corresponds to each camera component.

[0037] The single image convolution value obtained in this solution represents the convolution value between all behavioral images from the same camera assembly. Accordingly, in the step of "inputting all behavioral image codes from each camera assembly into a preset image feature model to obtain at least one single image convolution value," all behavioral image codes from the same camera assembly are convolved using a convolutional neural network to obtain a single image convolution value.

[0038] This solution pre-stores multiple categories of trained pre-set image feature models, each corresponding to a specific category of abnormal behavior. For example, these pre-set image feature models include models for holding knives, jumping from buildings, jumping into rivers, and suicide.

[0039] In the step of "combining and analyzing all single image convolution values ​​to obtain image analysis probability values", all the single image convolution values ​​are subjected to a convolution operation again to obtain a combined image convolution value, and the combined image convolution value is converted into the image analysis probability value.

[0040] In one embodiment of the present solution, a preset SIGMOID function is used to convert the combined image convolution value to obtain an image analysis probability value in the range of [0, 1], which represents the probability of the abnormal behavior category of the actor.

[0041] If the image analysis probability value is greater than the image set threshold, it is determined that the person has engaged in abnormal behavior corresponding to the preset image feature model. The person's behavior image can then be stored in a database corresponding to the abnormal behavior category to provide data samples for subsequent model training.

[0042] Furthermore, this plan can also send early warning information of the perpetrators of abnormal behaviors that endanger society and their behaviors to the corresponding business departments in a timely manner to remind the business departments to take corresponding feedback measures, and then the business departments will take emergency measures against the perpetrators and focus on tracking and managing them to achieve "one person, one file" management.

[0043] If the image analysis probability value is less than the image set threshold, it is impossible to confirm whether the actor has abnormal behavior through the behavioral image. However, this does not mean that the actor does not have abnormal behavior. Therefore, more detailed behavioral factors need to be evaluated.

[0044] In this scheme, in "inputting all behavioral factors of each camera component into a preset behavioral factor feature model to obtain at least one single behavioral convolution value", all the behavioral factors are encoded to obtain behavioral factor codes, and all the behavioral factor codes of each camera component are input into a preset behavioral factor feature model to obtain at least one single behavioral convolution value, wherein each single behavioral convolution value corresponds to each camera component.

[0045] This solution pre-stores multiple categories of trained preset behavior factor feature models, where each preset behavior factor feature model corresponds to a category of abnormal behavior, and the preset behavior factor feature models and preset image feature models of this solution are in a corresponding relationship. It is worth mentioning that the behavior factors of this solution are used to characterize the feature dimensions of abnormal behavior. For example, if the abnormal behavior is "hand-held knife", the behavior factors are: long knife, short knife, kitchen knife, long stick, machete and axe; if the abnormal behavior is "jumping from a building", the behavior factors are rooftop and high-rise building; if the abnormal behavior is "jumping into a river", the behavior factors are bridge and river; if the abnormal behavior is "suicide", the behavior factors are dynamite, explosive pack, gas tank and lighter.

[0046] In “inputting all behavioral factors of each camera component into a preset behavioral factor characteristic model to obtain at least one single behavioral convolution value”, all the behavioral factors are encoded to obtain behavioral factor codes, and all the behavioral factor codes of each camera component are input into a preset behavioral factor characteristic model to obtain at least one single behavioral convolution value, wherein each single behavioral convolution value corresponds to each camera component.

[0047] The single behavioral convolution value obtained in this solution represents the convolution value between all behavioral factors of the same camera component. Accordingly, in the step of "inputting all behavioral factor codes of each camera component into a preset behavioral factor feature model to obtain at least one single behavioral convolution value," all behavioral factor codes of the same camera component are convolved using a convolutional neural network to obtain a single behavioral convolution value.

[0048] In the step of "combining and analyzing all single behavior convolution values ​​to obtain a behavior analysis probability value", all the single behavior convolution values ​​are subjected to a convolution operation again to obtain a combined behavior convolution value, and the combined behavior convolution value is converted into the behavior analysis probability value.

[0049] In one embodiment of the present solution, the combined behavior convolution value is converted using a preset SIGMOID function to obtain a behavior analysis probability value in the range [0, 1], which represents the probability of the behavior abnormality category of the actor.

[0050] If the behavior analysis probability value is greater than the behavior setting threshold, the person is determined to have engaged in abnormal behavior corresponding to the preset behavior factor feature model. The person's behavior factors and behavior images can then be stored in a database corresponding to abnormal behavior categories to provide data samples for subsequent model training.

[0051] Similarly, this plan can also form early warning information about people who engage in abnormal behaviors that endanger society and their behaviors, and send it to the corresponding business departments in a timely manner to remind them to take corresponding feedback measures. The business departments can then take emergency measures against the people and focus on tracking and managing them, realizing "one person, one file" management.

[0052] If the behavior analysis probability value is less than the behavior setting threshold, it is impossible to confirm whether the actor has abnormal behavior through the behavior factor. However, this does not mean that the actor does not have abnormal behavior. Therefore, it is necessary to conduct a social harm assessment.

[0053] That is to say, if the behavior analysis probability value is less than the behavior setting threshold, the basic information of the actor, the event information corresponding to the actor, the image analysis probability value and the behavior analysis probability value are obtained and analyzed to obtain the event occurrence possibility score, the actor's appeal score, the accident consequence score and the social impact score. The social hazard assessment score is obtained by summing the event occurrence possibility score, the actor's appeal score, the accident consequence score and the social impact score.

[0054] In this solution, the event likelihood score is positively correlated with the probability of abnormal behavior occurring, and the event likelihood score is positively correlated with the image analysis probability value and the behavior analysis probability value. In other words, the larger the image analysis probability value and / or the behavior analysis probability value, the higher the event likelihood score. This solution categorizes the event likelihood score into five levels: 0, 1, 3, 6, and 10.

[0055] The actor's appeal score is positively correlated with the intensity and length of the actor's appeal. The actor's appeal score is related to the actor's basic information. This information includes the number of times the actor has filed complaints about conflicts and disputes. A higher number of complaints corresponds to a higher actor's appeal score. This solution categorizes the actor's appeal score into five levels: 0, 1, 3, 6, and 10.

[0056] The accident consequence score is directly correlated with the harm caused by the abnormal behavior. It is related to the event information, which includes the type and severity of the conflict or dispute. The more complex the conflict or dispute type and the higher the severity, the higher the corresponding accident consequence score. This solution categorizes the accident consequence score into six levels: 0, 1, 3, 6, 8, and 10.

[0057] The social impact score is positively correlated with the social impact of abnormal behavior. The social impact score is related to the event information, which includes the type of conflict and dispute and the severity of the event. The more complex the conflict and dispute type and the higher the severity of the event, the higher the corresponding social impact score.

[0058] In one embodiment of this solution, the index table of the social harm assessment method is shown in Table 1 below:

[0059] Table 1 Social Hazard Assessment Method Indicators

[0060]

[0061] In addition, after obtaining the social harm assessment score, this solution can also perform hierarchical processing on the social harm assessment score based on a custom level. The hierarchical processing table of this solution is shown in Table 2 below:

[0062] Table 2 Stage Processing Table

[0063]

[0064] For example, this plan issues an early warning based on the abnormal behavior of person A who takes a knife to retaliate against society after repeatedly complaining about forced demolition incident B but the complaint remains unresolved.

[0065] First, the basic information of the actor A is collected through the actor basic information collection module, including the basic information of the actor A, the event information of the forced demolition event B, etc. The information collection and process are not the focus of this plan and will not be described in detail.

[0066] The basic information collection module for the perpetrator is connected to the database interface unit via network technology, and the database interface unit is connected to the external interface unit via the WebService interface, providing technical services for the data exchange and sharing, and realizing data exchange and integrated sharing between the basic information collection module for the perpetrator and the external system. Specifically, it is mainly connected with the complaint, comprehensive management and other unit systems through network technology, and the basic information of the perpetrator and the event information entered through the basic information collection module for the perpetrator are automatically compared to prevent the perpetrator from providing false addresses, false events and other information.

[0067] Camera components are used to capture behavioral image data of perpetrator A. Image data information from each camera component is automatically acquired through automatic acquisition. The intelligent processing module then filters and stores the image data by running a built-in behavioral processing program. Five behavioral image data sets are generated for each camera component after filtering. Specifically, the first camera component V1 captures behavioral images of perpetrator A walking on the road holding a long knife, which are coded as code 11, code 12, ..., code 15; the second camera component V2 captures behavioral images of perpetrator A walking on the road holding a long knife, which are coded as code 21, code 22, ..., code 25; the third camera component V3 captures behavioral images of perpetrator A walking on the road holding a long knife, which are coded as code 31, code 32, ..., code 35, and so on. By comparing the image codes code11, code12, ..., code15 of all behaviors captured by the first camera component V1 with the preset image feature model corresponding to holding a knife, and performing a convolution operation using a convolutional neural network CNN (Cable News Network), a single image convolution ycode1 between the images of the camera component can be obtained; by comparing the image codes code21, code22, ..., code25 of all behaviors captured by the second camera component V2 with the preset image feature model of holding a knife, and performing a convolution operation using a convolutional neural network CNN (Cable News Network), a single image convolution value ycode2 between the images of the camera component can be obtained; by comparing the image codes code31, code32, ..., code35 of all behaviors captured by the third camera component V3 with the preset image feature model of holding a knife, and performing a convolution operation using a convolutional neural network CNN (Cable News Network), a single image convolution ycode3 between the images of the camera component can be obtained.

[0068] Perform a combined analysis on the discrete analysis values ​​of each camera component, and perform a convolution operation on the single image convolution values ​​ycode1, ycode2, and ycode3 obtained by each camera component V1, V2, and V3 to obtain the combined image convolution value yCNN between all cameras. Use a pre-set sigmoid function to calculate the probability of abnormal image behavior categories. The obtained probability result is mapped to the range [0, 1] and a probability value is obtained to obtain the image analysis probability value. The combined analysis is as follows:

[0069] P=sigmoid(yCNN); where p is the image analysis probability value.

[0070] If the image analysis probability value is higher than the image analysis threshold, the perpetrator is judged to have engaged in abnormal behavior by holding a knife. If the image analysis probability value is lower than the image analysis threshold, a behavioral factor assessment is performed: the number of behavioral factor data filtered for each camera component is set to 5. Specifically, the first camera component V1 captures the behavioral factor codes code11', code12', ... code15' of perpetrator A holding a knife; the second camera component V2 captures the behavioral factor codes code21', code22', ... code25' of perpetrator A holding a knife; the third camera component V3 captures the behavioral shadow codes code31', code32', ... code35' of perpetrator A holding a knife, and so on. All the behavior factor codes code11', code12', ... code15' collected by the first camera component V1 are compared with the preset behavior factor feature model corresponding to holding a knife, and a convolution operation is performed using a convolutional neural network CNN (Cable News Network) to obtain a single behavior convolution ycode1' between each image of the camera component; all the behavior factor codes code21', code22', ... code25' collected by the second camera component V2 are compared with the preset behavior factor feature model of holding a knife, and a convolution operation is performed using a convolutional neural network CNN (Cable News Network) to obtain a single behavior convolution value ycode2' between each image of the camera component; all the behavior factor codes code31', code32', ... code35' collected by the third camera component V3 are compared with the preset behavior factor model of holding a knife, and a convolution operation is performed using a convolutional neural network CNN (Cable News Network) to obtain a single behavior convolution ycode3' between each image of the camera component.

[0071] Perform a combined analysis on the discrete analysis values ​​of each camera component. Convolve the single behavior convolution values ​​ycode1', ycode2', and ycode3' obtained by each camera component V1, V2, and V3 again to obtain the combined behavior convolution value yCNN' across all cameras. Calculate the probability of abnormal behavior categories using a pre-set sigmoid function. The resulting probability is mapped to the range [0, 1] to obtain a probability value, and the probability value of the comfort analysis is obtained. The combined analysis is as follows:

[0072] P=sigmoid(yCNN'); where p is the probability value of behavior analysis.

[0073] The behavior analysis probability value is compared with the behavior analysis threshold. If the behavior analysis probability value is less than the behavior analysis threshold, it cannot be determined whether the perpetrator has abnormal behavior. In this case, this solution conducts a risk assessment based on the perpetrator's basic information, the corresponding event information, the image analysis probability value, and the behavior analysis probability value, using the indicators shown in Table 2 for this assessment.

[0074] Example 2

[0075] Based on the same concept, Figure 3 , this application also proposes an abnormal behavior early warning device, including:

[0076] An analysis data acquisition unit 301 is configured to select a person and capture at least one behavioral image and at least one behavioral factor of the person through at least one camera assembly, wherein the behavioral factor is used to characterize a characteristic dimension of abnormal behavior;

[0077] The behavior image analysis unit 302 is configured to input all behavior images of each camera assembly into a preset image feature model to obtain at least one single image convolution value, and to combine and analyze all the single image convolution values ​​to obtain an image analysis probability value. If the image analysis probability value is greater than a preset image threshold, it is determined that the person has engaged in abnormal behavior corresponding to the preset image feature model.

[0078] The behavior factor analysis unit 303 is used to input all the behavior factors of each camera component into the preset behavior factor characteristic model to obtain at least one single behavior convolution value when the image analysis probability value is less than the image setting threshold, and to combine and analyze all the single behavior convolution values ​​to obtain a behavior analysis probability value. If the behavior analysis probability value is greater than the behavior setting threshold, it is determined that the actor has abnormal behavior corresponding to the preset behavior factor characteristic model.

[0079] If there is any overlap between the technical features in this embodiment and the technical features mentioned in Example 1, please refer to the introduction of Example 1.

[0080] Example 3

[0081] This embodiment also provides an electronic device, referring to Figure 4 , including a memory 404 and a processor 402, wherein the memory 404 stores a computer program, and the processor 402 is configured to run the computer program to execute the steps in any of the above-mentioned abnormal behavior warning method embodiments.

[0082] Specifically, the processor 402 may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0083] Memory 404 may include a large-capacity memory 404 for data or instructions. By way of example, and not limitation, memory 404 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 404 may include removable or non-removable (or fixed) media. Where appropriate, memory 404 may be internal or external to the data processing device. In certain embodiments, memory 404 is non-volatile memory. In certain embodiments, memory 404 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM) or a flash memory (FLASH), or a combination of two or more of these. In appropriate circumstances, the RAM may be a static random access memory (SRAM) or a dynamic random access memory (DRAM), wherein the DRAM may be a fast page mode dynamic random access memory 404 (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0084] The memory 404 may be used to store or cache various data files that need to be processed and / or used for communication, as well as computer program instructions of the possible abnormal behavior early warning method executed by the processor 402 .

[0085] The processor 402 implements any one of the abnormal behavior warning methods in the above embodiments by reading and executing computer program instructions stored in the memory 404 .

[0086] Optionally, the electronic device may further include a transmission device 406 and an input / output device 408 , wherein the transmission device 406 is connected to the processor 402 , and the input / output device 408 is connected to the processor 402 .

[0087] Transmission device 406 can be used to receive or transmit data via a network. Specific examples of such networks may include wired or wireless networks provided by the electronic device's communications provider. In one embodiment, the transmission device includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, transmission device 406 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0088] The input and output device 408 is used to input or output information. In this embodiment, the input information may be behavior images, behavior factors, etc., and the output information may be behavior analysis probability values, image analysis probability values, analysis results, etc.

[0089] Optionally, in this embodiment, the processor 402 may be configured to execute the following steps through a computer program:

[0090] S101, selecting an actor, and capturing at least one behavioral image and at least one behavioral factor of the actor using at least one camera assembly, wherein the behavioral factor is used to characterize characteristic dimensions of abnormal behavior;

[0091] S102: Input all behavioral images of each camera assembly into a preset image feature model to obtain at least one single image convolution value, combine and analyze all the single image convolution values ​​to obtain an image analysis probability value, and if the image analysis probability value is greater than a set image threshold, determine that the person has engaged in abnormal behavior corresponding to the preset image feature model;

[0092] S103. If the image analysis probability value is less than the image setting threshold, all behavioral factors of each camera component are input into the preset behavioral factor feature model to obtain at least one single behavioral convolution value, and all the single behavioral convolution values ​​are combined and analyzed to obtain a behavioral analysis probability value. If the behavioral analysis probability value is greater than the behavioral setting threshold, it is determined that the actor has abnormal behavior corresponding to the preset behavioral factor feature model.

[0093] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be repeated here.

[0094] In general, various embodiments may be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects of the invention may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device, but the invention is not limited thereto. Although various aspects of the invention may be shown and described as block diagrams, flow charts, or using some other graphical representation, it should be understood that, as non-limiting examples, the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or a controller or other computing device, or some combination thereof.

[0095] Embodiments of the present invention can be implemented by computer software, which is executable by the data processor of the mobile device, such as in the processor entity, or is implemented by hardware, or is implemented by a combination of software and hardware. Computer software or programs (also referred to as program products) including software routines, applets and / or macros can be stored in any device-readable data storage medium, and they include program instructions for performing specific tasks. The computer program product can include one or more computer executable components configured to perform the embodiment when the program is running. One or more computer executable components can be at least one software code or a part thereof. In addition, at this point, it should be noted that any box of the logic flow in the figure can represent a program step, or interconnected logical circuits, boxes and functions, or a combination of program steps and logical circuits, boxes and functions. The software can be stored in physical media such as memory chips or storage blocks implemented in the processor, magnetic media such as hard disks or floppy disks, and optical media such as, for example, DVDs and their data variants, CDs. Physical media is non-transient media.

[0096] Those skilled in the art should understand that the technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0097] The above embodiments merely illustrate several embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for early warning of abnormal behavior, characterized in that: The following steps are involved: Selecting a person who behaves abnormally, and capturing at least one behavioral image and at least one behavioral factor of the person by at least one camera assembly, wherein the behavioral factor is used to characterize a characteristic dimension of abnormal behavior; Inputting all behavioral images of each camera component into a preset image feature model to obtain at least one single image convolution value, combining and analyzing all the single image convolution values ​​to obtain an image analysis probability value, and if the image analysis probability value is greater than a set image threshold, determining that the person has exhibited abnormal behavior corresponding to the preset image feature model; If the image analysis probability value is less than the image setting threshold, all behavioral factors of each camera component are input into the preset behavioral factor feature model to obtain at least one single behavioral convolution value, and all the single behavioral convolution values ​​are combined and analyzed to obtain a behavioral analysis probability value. If the behavioral analysis probability value is greater than the behavioral setting threshold, it is determined that the actor has abnormal behavior corresponding to the preset behavioral factor feature model. If the behavioral analysis probability value is less than the behavioral setting threshold, the actor's basic information, the event information corresponding to the actor, the image analysis probability value and the behavioral analysis probability value are obtained and analyzed to obtain an event probability score, an actor's appeal score, an accident consequence score and a social impact score. The event probability score, the actor's appeal score, the accident consequence score and the social impact score are summed to obtain a social hazard assessment score. After obtaining the social hazard assessment score, the social hazard assessment score can also be graded based on a custom level.

2. The abnormal behavior early warning method according to claim 1, characterized in that: In "inputting all behavioral images of each camera component into a preset image feature model to obtain at least one single image convolution value", all the behavioral images are encoded to obtain behavioral image encoding, and all the behavioral image encodings of each camera component are input into a preset image feature model to obtain at least one single image convolution value, wherein each single image convolution value corresponds to each camera component.

3. The abnormal behavior early warning method according to claim 1, characterized in that: In the step of "combining and analyzing all single image convolution values ​​to obtain image analysis probability values", all the single image convolution values ​​are subjected to a convolution operation again to obtain a combined image convolution value, and the combined image convolution value is converted into the image analysis probability value.

4. The abnormal behavior early warning method according to claim 1, characterized in that: In "inputting all behavioral factors of each camera component into a preset behavioral factor characteristic model to obtain at least one single behavioral convolution value", all the behavioral factors are encoded to obtain behavioral factor codes, and all the behavioral factor codes of each camera component are input into a preset behavioral factor characteristic model to obtain at least one single behavioral convolution value, wherein each single behavioral convolution value corresponds to each camera component.

5. The abnormal behavior early warning method according to claim 1, characterized in that: In the step of "combining and analyzing all single behavior convolution values ​​to obtain a behavior analysis probability value", all the single behavior convolution values ​​are subjected to a convolution operation again to obtain a combined behavior convolution value, and the combined behavior convolution value is converted into the behavior analysis probability value.

6. The abnormal behavior early warning method according to claim 1, characterized in that: Multiple categories of trained preset image feature models are pre-stored, and each of the preset image feature models corresponds to a category of abnormal behavior.

7. The abnormal behavior early warning method according to claim 1, characterized in that: Pre-stored are multiple categories of trained preset behavior factor feature models, and the preset behavior factor feature models and the preset image feature models are set accordingly.

8. An abnormal behavior warning device, characterized in that: include: An analysis data acquisition unit is used to select a person and collect at least one behavior image and at least one behavior factor of the person through at least one camera component, wherein the behavior factor is used to characterize the characteristic dimension of abnormal behavior; A behavior image analysis unit is configured to input all behavior images of each camera assembly into a preset image feature model to obtain at least one single image convolution value, and to combine and analyze all the single image convolution values ​​to obtain an image analysis probability value. If the image analysis probability value is greater than a set image threshold, it is determined that the person has exhibited abnormal behavior corresponding to the preset image feature model. The behavior factor analysis unit is used to input all the behavior factors of each camera component into the preset behavior factor characteristic model to obtain at least one single behavior convolution value when the image analysis probability value is less than the image setting threshold, and combine and analyze all the single behavior convolution values ​​to obtain the behavior analysis probability value; if the behavior analysis probability value is greater than the behavior setting threshold, it is determined that the actor has abnormal behavior corresponding to the preset behavior factor characteristic model; if the behavior analysis probability value is less than the behavior setting threshold, the actor's basic information, the event information corresponding to the actor, the image analysis probability value and the behavior analysis probability value are obtained and analyzed to obtain the event possibility score, the actor's appeal score, the accident consequence score and the social impact score; the event possibility score, the actor's appeal score, the accident consequence score and the social impact score are summed to obtain the social hazard assessment score; after obtaining the social hazard assessment score, the social hazard assessment score can also be graded based on a custom level.

9. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to run the computer program to execute the abnormal behavior early warning method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, wherein the computer program includes a program code for controlling a process to execute a process, wherein the process includes the abnormal behavior early warning method according to any one of claims 1 to 7.

Images