Panic behavior recognition method and system based on panic semantic model, and medium

Abstract

A panic behavior recognition method and system based on a panic semantic model, and a medium. The method comprises the following steps: invoking an AI interface to perform real-time recognition on an audio stream to obtain semantic information, and by means of the semantic information and a basic concept, performing hierarchical division, and recognizing key speech segments; on the basis of a locating coordinate matrix, determining the specific hierarchical level and position of each key speech segment; and on the basis of the specific hierarchical level and position, using a description matrix to obtain the weight of the key speech segment, performing consistency matching on the degree of panic in a scenario, and determining whether a panic behavior has occurred, wherein the construction steps of a panic semantic reasoning network model are as follows: selecting a panic scenario as a description object; using a web ontology language to define a basis concept in the selected panic scenario; supplementing knowledge elements in the panic scenario, in order to improve a panic semantic model; and defining a reasoning rule, and constructing the panic semantic reasoning network model. Compared with the prior art, the method improves the accuracy of pedestrian panic behavior recognition.

Description

A panic behavior recognition method and system based on a panic semantic model and a medium TECHNICAL FIELD

[0001] The present application relates to the field of panic behavior recognition, and in particular to a panic behavior recognition method and system based on a panic semantic model and a medium. BACKGROUND

[0002] [According to Rule 26, correct 25.04.2025] Panic behavior of pedestrians is an important factor affecting the stability of crowds. How to recognize panic behavior of pedestrians is of great significance for scientific management and emergency evacuation of crowds in public places. At present, the recognition of panic behavior mainly focuses on the recognition of abnormal postures of pedestrians. There are few related researches on the recognition of panic behavior through a panic semantic reasoning network model in complex crowd scenes.

[0003] There are still some deficiencies in the current research: 1) At present, the recognition of panic behavior mainly focuses on the recognition of abnormal postures of pedestrians. There are few related researches on the recognition of panic scenes through a panic semantic model. 2) Panic scenes have diversity. Different panic scene knowledge elements have different weights in the recognition process of panic scenes, which will hinder the recognition of panic scenes. At present, there are few related researches on eliminating the hindrance to improve the accuracy of panic scene recognition. SUMMARY

[0004] The purpose of the present application is to overcome the defects of the prior art and provide a panic behavior recognition method and system based on a panic semantic model and a medium to improve the accuracy of panic behavior recognition.

[0005] The purpose of the present application can be achieved by the following technical solutions:

[0006] A panic behavior recognition method based on a panic semantic model, comprising the following steps:

[0007] An AI interface is called to recognize audio streams in real time as semantic information. The semantic information and basic concepts are hierarchically divided to recognize key segments in the semantic information;

[0008] According to the positioning coordinate matrix, the specific level and position of each key segment in the panic semantic model are determined;

[0009] According to the specific level and position of the key segment, the weight of the key segment is obtained by using the description matrix of the panic semantic reasoning network model, the consistency of the panic degree in the scene is matched, and it is judged whether panic behavior occurs in the scene;

[0010] The construction steps of the panic semantic reasoning network model are as follows:

[0011] selecting a panic scenario as a description object;

[0012] defining basic concepts under the selected panic scenario using a web ontology language;

[0013] supplementing knowledge elements under the panic scenario to perfect a panic semantic model under the panic scenario;

[0014] defining reasoning rules based on the panic semantic model to construct a panic semantic reasoning network model with panic semantic analysis capability and panic event reasoning capability.

[0015] Further, the semantic information includes multiple types of shouting, cries for help, and conversation contents in the crowd.

[0016] Further, the panic scenario includes multiple types of medical nuisance scenarios, natural disaster scenarios, and crowded scenarios.

[0017] Further, the knowledge elements under the panic scenario include multiple types of key segments, crowd states, and occurrence conditions.

[0018] Further, the hierarchical positioning coordinate matrix is as follows:

[0019] wherein, i = col r -1, col r represents the total length of the key segments, A i represents that only the last hierarchical key segment is different under the condition that the total length of the segments is the same, and q+1 represents the total number of key words in the hierarchical key segment.

[0020] Further, the description matrix of the panic semantic reasoning network model is as follows:

[0021] wherein, represents the position information of the q+1th key segment in the hierarchy, represents the weight occupied by the q+1th key segment in the hierarchy,

[0022] Further, the description matrix includes the key segment information and the weight information.

[0023] Further, when the panic semantic reasoning network model matches the key segment related to the panic and the weighted structure exceeds the threshold value, the matching result of the panic semantic reasoning network model is γ = 1, otherwise, γ = 0.

[0024] According to another aspect of the present application, there is provided a computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, and the computer program, when executed by a processor, implements the panic behavior recognition method based on the panic semantic model.

[0025] According to another aspect of the present application, there is provided a panic behavior recognition system based on a panic semantic model, comprising:

[0026] A key segment recognition module is configured to call an AI interface, recognize audio streams in real time as semantic information, perform hierarchical division through the semantic information and basic concepts, and recognize key segments in the semantic information.

[0027] A hierarchical positioning module is configured to determine specific hierarchies and positions of each key segment in the panic semantic model according to a positioning coordinate matrix.

[0028] A panic behavior judgment module is configured to obtain weights of the key segments by using a description matrix of a panic semantic reasoning network model according to the specific hierarchies and positions of the key segments, perform consistent matching on panic degrees in a scene, and judge whether a panic behavior occurs in the scene.

[0029] The panic semantic reasoning network model is constructed according to the following steps:

[0030] A panic scene is selected, and the panic scene is taken as a description object.

[0031] A network ontology language is used to define basic concepts in the selected panic scene.

[0032] Knowledge elements in the panic scene are supplemented to perfect a panic semantic model in the panic scene.

[0033] Inference rules are defined based on the panic semantic model, and a panic semantic reasoning network model with panic semantic analysis ability and panic event reasoning ability is constructed.

[0034] Compared with the prior art, the present application has the following beneficial effects:

[0035] 1. When a panic behavior occurs, the present application can judge whether a panic event such as a disaster event or a terrorist attack occurs in a scene through real-time voice recognition, and can perform early warning, effectively avoiding the occurrence of panic behaviors, and increasing the accuracy of pedestrian panic behavior recognition.

[0036] 2. The present application performs hierarchical division through key semantic information and basic concepts, greatly reduces the performance requirements of a computer for key segment recognition through hierarchical design structure, and enhances the instantaneity of key segment detection.

[0037] 3. The present application can accurately identify whether panic behavior occurs in the scene by optimizing the weight through statistical methods, and can judge other panic scenes by changing the weight of the key segment, and also provides a new idea for subsequent crowd stability analysis and related research to avoid the spread of panic behavior. BRIEF DESCRIPTION OF DRAWINGS

[0038] Fig. 1 is a flowchart of a panic behavior recognition method based on a panic semantic model according to the present application;

[0039] Fig. 2 is a structural diagram of a panic semantic reasoning network model;

[0040] Fig. 3 is a description matrix derivation logic diagram of the panic semantic reasoning network model. DETAILED DESCRIPTION

[0041] The present application will be described in detail below in conjunction with the drawings and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present application, and gives a detailed implementation and specific operation process, but the protection scope of the present application is not limited to the following embodiments.

[0042] Embodiment 1

[0043] The present embodiment provides a panic behavior recognition method based on a panic semantic model, as shown in Fig. 1, which includes the following steps:

[0044] S1, call the AI interface to identify the audio stream into semantic information in real time, divide the semantic information and basic concepts into hierarchical, and identify the key segment in the semantic information.

[0045] [According to Rule 26 Correction 25.06.2025] This embodiment takes a certain injury event as an example, uses a certain injury video, and verifies the panic semantic reasoning network model through a certain panic event video. The video records a panic event of a certain injury, including crowd chaos, cries for help, panic behavior, etc.

[0046] [According to Rule 26 Correction 25.06.2025] The audio in the certain event video is called to the Baidu AI interface to realize the conversion from voice to text, and the shouting, cries for help, dialogue content and other semantic information in the crowd are captured and identified. Input these semantic information into the panic semantic reasoning network model, from the text obtained by voice recognition, the model identifies the key segment related to the panic scene, which includes "cutting people", "save lives", "run fast", "someone is injured", etc.

[0047] The hierarchical division is made by semantic information and basic concepts. The first field in the longer key semantic information is in a higher level. For example, in "I want to kill" and "kill", the computer searches the semantic information word by word (from the first word). Assuming that "I want to kill" is located, "I" is located in the i-th layer, "want" is searched from the i+1-th layer, and "kill" is searched from the i+2-th layer. However, "kill" is directly located in the i+2-th layer, and the previous layers are filled with 0 as the position code. Of course, due to the different initial positioning information, the influence of "I want to kill" and "kill" on whether the panic behavior occurs in the scene is different. We find that such a design structure can greatly reduce the performance requirements of the computer and enhance the real-time detection. When the computer identifies a potential key segment, if the subsequent field is not matched, it is considered as a non-key segment, such as "I want to eat" is not considered as a key semantic information.

[0048] S2, according to the positioning coordinate matrix, determine the specific level and position of each key segment in the panic semantic model.

[0049] The positioning coordinate matrix is as follows:

[0050] In the formula, i = col r -1, col r A represents the total length of the key segment i q+1 represents the total number of key words in the level where the key segment is located.

[0051] According to the disaster event, the medical dispute event, the stampede event, and the terrorist attack, four panic scenes are taken as examples for analysis, random investigation is carried out, and the total number of participants in the investigation is 300. The groups are divided as follows:

[0052] (1) The elderly group (60 years old and below): 62 people, of which 29 are males and 33 are females;

[0053] (2) The middle-aged group (40 to 60 years old): 98 people, of which 61 are males and 37 are females;

[0054] (3) The youth group (16 to 40 years old): 140 people, of which 65 are males and 75 are females.

[0055] The statistical method is used to optimize the investigation results and the weight. The weight information table is shown in Table 1.

[0056] Table 1: Semantic model key segment weight statistics table

[0057] S3, obtaining the weight of the key sentence according to the description matrix of the panic semantic reasoning network model of the specific level and position of the key sentence, and performing consistent matching on the panic degree in the scene to determine whether panic behavior occurs in the scene.

[0058] The construction steps of the panic semantic reasoning network model are as follows:

[0059] 1) Selecting a panic scene. First, it is necessary to consider whether the panic scene is relatively common, and try to select those scenes that frequently occur and are easy to cause panic, such as medical harassment scenes, natural disaster scenes, and crowded scenes. Secondly, consider whether the panic scene can be identified and prevented in advance. The construction of the reasoning network model for such scenes has more practical application value. Finally, consider whether there is sufficient data support for the scene in order to train and verify.

[0060] 2) Using OWL (Web Ontology Language) to define the basic concepts in the selected panic scene, such as the panic semantic expression being one of the sources of forming panic psychology, and the panic psychology inducing the occurrence of panic behavior.

[0061] 3) Perfecting the panic semantic model in the scene, and supplementing the knowledge elements in the panic scene, including key sentences, crowd state, and occurrence conditions.

[0062] 4) In order to make the model have reasoning ability, define the corresponding reasoning rules, for example, in the process of a crowded stampede event, when identifying key sentences such as "can't breathe", "help", "someone fell down", "someone was trampled to death", etc., the weight of the sentence is used to make a consistent judgment on the panic degree in the scene. When the threshold is exceeded, it is judged that panic behavior has occurred in the scene.

[0063] The specific reasoning network model is shown in FIG. 2.

[0064] The description matrix of the panic semantic reasoning network model is as follows:

[0065] In the formula, represents the position information of the q+1th key sentence in the level, represents the weight of the q+1th key sentence in the level,

[0066] The derivation logic diagram of the description matrix of the panic semantic reasoning network model is shown in FIG. 3.

[0067] The matching result of the panic semantic reasoning network model is as follows: γ∈(0,1)

[0068] In the formula, γ is the matching result of the panic semantic reasoning network model.

[0069] Further, when the panic semantic reasoning network model matches to the key sentence related to panic and the weight structure exceeds the threshold, the matching result of the panic semantic reasoning network model is γ = 1, otherwise, γ = 0.

[0070] In this embodiment, the key sentence such as "cutting people" is identified, and the panic semantic reasoning network model determines the specific level and position of each key sentence in the panic semantic reasoning network model according to the positioning coordinate matrix. Then, according to the weight information provided in the description matrix, the influence of each key sentence on the panic degree of the current scene is calculated, and the key sentence with large weight has greater influence on the judgment of the panic scene. By combining the weight and reasoning rules of multiple key sentences, the current scene is inferred and judged, and the system matches to multiple high-weight key sentences at the same time, reaches a certain threshold, and the matching result of the panic semantic reasoning network model is γ = 1, so it can be determined that the input video exists panic behavior.

[0071] Embodiment 2

[0072] The embodiment provides a computer readable storage medium, and a computer program is stored on the computer readable storage medium. When the computer program is executed by a processor, a panic behavior identification method based on a panic semantic model can be realized.

[0073] The rest is the same as embodiment 1.

[0074] Embodiment 3

[0075] The embodiment provides a panic behavior identification system based on a panic semantic model, which comprises:

[0076] The audio data acquisition module is responsible for real-time acquisition of audio data in the environment, and transmits the acquired audio stream to the key sentence identification module for subsequent processing.

[0077] The key sentence identification module is used to call an AI interface, identify the audio stream into semantic information in real time, divide the levels through the semantic information and basic concepts, and identify the key sentences in the semantic information.

[0078] The hierarchical positioning module is used to determine the specific level and position of each key sentence in the panic semantic model according to the positioning coordinate matrix.

[0079] The panic behavior judgment module is used to obtain the weight of the key sentence by using the description matrix of the panic semantic reasoning network model according to the specific level and position of the key sentence, and to perform consistent matching on the panic degree in the scene to judge whether panic behavior occurs in the scene.

[0080] The data display module displays the panic behavior identification result and related data to the user in a visual manner, so as to facilitate real-time monitoring and subsequent analysis.

[0081] The alarm prompt module: integrate the alarm prompt function in the data display module, when the system detects the panic behavior, can timely remind the relevant personnel through the vision (word) and hearing (sound) way.

[0082] The overall execution process of the system is: the audio data acquisition module captures the environmental audio in real time through the microphone device, the key segment recognition module receives the audio stream, converts it into text semantic information by using the AI interface, and recognizes the key segment, the hierarchical positioning module determines the specific level and position of each key segment in the panic semantic model according to the positioning coordinate matrix, the panic behavior judgment module calculates the key segment weight, performs consistency matching, and judges whether there is a panic behavior; the data display module presents the recognition result to the user in a visual form, provides real-time monitoring and alarm function, when the panic behavior judgment module detects the high-risk panic behavior, the alarm prompt module alarms in time through the sound and vision way.

[0083] Among them, the construction steps of the panic semantic reasoning network model are as follows:

[0084] Select a panic scene, take the panic scene as the description object;

[0085] Use network ontology language to define the basic concepts under the selected panic scene;

[0086] Supplement the knowledge elements under the panic scene to perfect the panic semantic model under the panic scene;

[0087] Define the reasoning rules based on the panic semantic model, and construct the panic semantic reasoning network model with panic semantic analysis ability and panic event reasoning ability.

[0088] The rest is the same as example 1.

Claims

1. A method for recognizing a panic behavior based on a panic semantic model, characterized in that, The method comprises the following steps: calling an AI interface to identify audio streams in real time as semantic information, performing hierarchical division through the semantic information and basic concepts, and identifying key speech segments in the semantic information; determining the specific level and position of each key speech segment in a panic semantic model according to a positioning coordinate matrix; obtaining the weight of the key speech segment according to the specific level and position of the key speech segment by using a description matrix of a panic semantic reasoning network model, performing consistent matching on the panic degree in a scene, and judging whether panic behavior occurs in the scene; wherein the construction steps of the panic semantic reasoning network model are as follows: selecting a panic scene, and taking the panic scene as a description object; defining the basic concepts in the selected panic scene by using a web ontology language; supplementing the knowledge elements in the panic scene to perfect the panic semantic model in the panic scene; defining reasoning rules based on the panic semantic model to construct a panic semantic reasoning network model with panic semantic analysis capability and panic event reasoning capability. 2.The method of claim 1, wherein, The semantic information includes multiple types of shouting, cries for help and dialogue content in a crowd. 3.The method of claim 1, wherein, The panic scene includes multiple types of medical nuisance scenes, natural disaster scenes and crowded scenes. 4.The method of claim 1, wherein, The knowledge elements in the panic scene include multiple types of key speech segments, crowd states and occurrence conditions. 5.The method of claim 1, wherein, The hierarchical positioning coordinate matrix is as follows: where i = col r -1, col r denotes the total length of the key phrase, A i denotes the key phrase whose last level is different from the other key phrases with the same total length of the phrase, and q+1 denotes the total number of the key words in the level where the key phrase is located. 6.The method of claim 1, wherein, The description matrix of the panic semantic reasoning network model is as follows: In the formulae, position information of the (q+1)th key segment in the hierarchy, a weight of a (q+1)th key segment in the hierarchy, 7. The method of claim 6, wherein the method further comprises: The description matrix includes key speech segment information and weight information. 8.The method of claim 1, wherein, When the panic semantic reasoning network model matches a key speech segment related to panic and the weighted structure exceeds a threshold, the matching result of the panic semantic reasoning network model is γ = 1, otherwise, γ = 0.

9. A computer-readable storage medium, characterized in that, The computer program stored on the computer readable storage medium can be executed by the processor to implement the panic behavior identification method based on the panic semantic model according to any one of claims 1 to 8.

10. A panic behavior recognition system based on a panic semantic model, characterized by, It comprises: a key speech segment identification module for calling an AI interface to identify audio streams in real time as semantic information, performing hierarchical division through the semantic information and basic concepts, and identifying key speech segments in the semantic information; a hierarchical positioning module for determining the specific level and position of each key speech segment in a panic semantic model according to a positioning coordinate matrix; a panic behavior judgment module for obtaining the weight of the key speech segment according to the specific level and position of the key speech segment by using a description matrix of a panic semantic reasoning network model, performing consistent matching on the panic degree in a scene, and judging whether panic behavior occurs in the scene; wherein the construction steps of the panic semantic reasoning network model are as follows: selecting a panic scene, and taking the panic scene as a description object; defining the basic concepts in the selected panic scene by using a web ontology language; supplementing the knowledge elements in the panic scene to perfect the panic semantic model in the panic scene; defining reasoning rules based on the panic semantic model to construct a panic semantic reasoning network model with panic semantic analysis capability and panic event reasoning capability.

Images