CN117011774A - Cluster behavior understanding system based on scene knowledge element - Google Patents
Cluster behavior understanding system based on scene knowledge element Download PDFInfo
- Publication number
- CN117011774A CN117011774A CN202310422133.7A CN202310422133A CN117011774A CN 117011774 A CN117011774 A CN 117011774A CN 202310422133 A CN202310422133 A CN 202310422133A CN 117011774 A CN117011774 A CN 117011774A
- Authority
- CN
- China
- Prior art keywords
- behavior
- knowledge
- cluster
- data
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000005065 mining Methods 0.000 claims abstract description 21
- 230000006399 behavior Effects 0.000 claims description 140
- 238000012544 monitoring process Methods 0.000 claims description 33
- 238000004422 calculation algorithm Methods 0.000 claims description 28
- 230000002159 abnormal effect Effects 0.000 claims description 17
- 230000004927 fusion Effects 0.000 claims description 14
- 238000004458 analytical method Methods 0.000 claims description 13
- 238000011160 research Methods 0.000 claims description 8
- 239000000284 extract Substances 0.000 claims description 6
- 238000010801 machine learning Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 3
- 238000011065 in-situ storage Methods 0.000 claims description 3
- 238000011835 investigation Methods 0.000 claims description 3
- 238000000034 method Methods 0.000 abstract description 13
- 230000008569 process Effects 0.000 abstract description 4
- 230000003068 static effect Effects 0.000 abstract description 4
- 238000005516 engineering process Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 230000009471 action Effects 0.000 description 5
- 238000000605 extraction Methods 0.000 description 5
- 230000005856 abnormality Effects 0.000 description 4
- 238000010276 construction Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 206010000117 Abnormal behaviour Diseases 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005381 potential energy Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000002269 spontaneous effect Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2216/00—Indexing scheme relating to additional aspects of information retrieval not explicitly covered by G06F16/00 and subgroups
- G06F2216/03—Data mining
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the fields of computer vision, public safety and the like, in particular to a cluster behavior understanding system based on scene knowledge elements. Aiming at the situation that scene definition is narrow in the existing cluster behavior understanding and knowledge outside a video scene is not considered, the invention provides a cluster behavior understanding method integrating data and knowledge, the cluster behavior understanding based on scene knowledge elements is realized through processes of video data acquisition and feature mining, scene modeling integrating data and knowledge and the like, and the scheme not only considers dynamic personnel information in a video image, but also considers static background information in the video image and knowledge information outside the video image. The invention can more comprehensively and truly describe the scene of the cluster behavior, and improve the speed and the accuracy of the cluster behavior identification.
Description
Technical Field
The invention relates to the fields of computer vision, public safety, behavior analysis, artificial intelligence, intelligent monitoring and the like, in particular to a cluster behavior understanding system based on scene knowledge elements.
Background
Clusters are crowds with a certain cohesive force, and the behavior of the clusters is neither simple crowd gathering behavior nor the sum of scattered individual behaviors. Cluster behavior refers to collective behavior that affects social public safety that occurs in a relatively spontaneous, unorganized, and unstable situation because of some general impact and encouragement. Common cluster behaviors are aggregation, rotation, evacuation, etc. The places with larger crowd flow, such as government buildings, traffic intersections, squares, markets, airports and the like, and higher crowd density are easy to generate clustering behaviors. The method has the advantages that the method can bring more casualties to the untimely prevention, inaccurate identification and improper treatment of cluster behaviors, and causes larger social influence and even disorder.
Understanding of the cluster behavior means not only identifying whether the cluster behavior is or is not, but also studying whether the cluster behavior is abnormal or is not. The traditional video monitoring system can only complete shooting, storage and playback, cannot realize an automatic early warning function, and if cluster behaviors are required to be recognized in time, abnormal phenomena are researched and judged, real-time monitoring of video images is required to be completed manually, and the probability of misjudgment and missed judgment can be improved due to the superposition of fatigue. The cluster behaviors are automatically processed by using related technologies such as digital image processing, computer vision, machine learning and the like, namely intelligent elements are integrated into a video image processing system, so that the robot brain is assisted to quickly capture and recognize the cluster behaviors.
Currently, many students analyze clustered behavior or abnormal behavior from a surveillance video perspective. One class of students analyzes the clustered behavior based on modeling of individual objects. Based on convolutional neural network to capture the dynamic characteristics of individuals, fully connected conditional random fields are used to analyze the interactions among people in group activities, and the potential functions of the interactions are beneficial to realizing group behavior classification. Group activity recognition depends on a few participants on several key frames, a semantic relation graph is constructed to represent the relation between people, the semantic relation graph is perfected according to two Markov decision processes, the action characteristics of important people are strengthened, and the action characteristics of irrelevant people are restrained. Another class of students analyzes the overall characteristics of the event based on overall modeling, and from a system perspective, the crowd behavior. A method for detecting and locating abnormal behavior in crowd scene analysis using a social strength model. Based on detection of potential energy and kinetic energy, a crowd distribution index is defined for detecting the gathering condition of pedestrians, then the kinetic energy is calculated through the optical flow and the crowd distribution index, and abnormal behaviors are analyzed based on a threshold value.
The existing behavior recognition technology is mainly aimed at some individual actions in a simple scene, the behaviors of the crowd events are complex, and the actual occurrence scene is complex, so that scene modeling is necessary to conduct scene modeling on the crowd actions, and whether abnormal crowd actions occur is inferred by analyzing information in the scene. The existing cluster behavior understanding method is essentially video image understanding, and is mostly based on analysis of monitoring images only, and scene definition is narrow. The real scene not only contains video data shot by the monitoring camera, but also knowledge outside the monitoring range, such as the experience of police or the related rules of government attendance, and the knowledge outside the monitoring is also important to scene analysis and event identification.
Disclosure of Invention
The invention provides a cluster behavior understanding system based on scene knowledge elements, which aims to solve the problem that the existing cluster behavior understanding method is used for understanding video images basically, is mainly based on analysis of monitoring images, is limited in scene definition and causes inaccurate cluster behavior judgment.
The invention is realized by adopting the following technical scheme: the system sequentially comprises an information acquisition and feature mining layer, a data and knowledge fusion modeling layer and a scene reproduction and behavior understanding layer; the information acquisition and feature mining layer comprises a camera video data module for acquiring monitoring data/video images and a behavior priori knowledge module for acquiring monitoring external knowledge and rule information; the system also comprises a video characteristic data set obtained by characteristic mining of the monitoring data/video images and a behavior characteristic knowledge set obtained by characteristic mining of the monitoring external knowledge and the rule information;
the data and knowledge fusion modeling layer comprises a behavior feature data set and a background feature data set which are extracted from a video feature data set through an algorithm, and a behavior record library, manager knowledge and a research rule which are extracted from the behavior feature knowledge set through the algorithm; the behavior characteristic data set and the background characteristic data set jointly form a data element, and the behavior record library, the manager knowledge and the research rule jointly form a knowledge element; constructing an information element model by fusing the data element and the knowledge element;
the scene reproduction and behavior understanding layer subdivides the information elements into crowd information elements and background information elements, wherein the crowd information elements comprise crowd quantity changes and behavior changes; the background information element comprises an inner video background information element and an outer video background information element, wherein the inner video background information element and the crowd information element are sources of data when the information element model is built, and the outer video background information element is a source of knowledge when the information element model is built; through the classification representation, the cluster site is reproduced as comprehensively and truly as possible, and the prior knowledge and the existing rules are combined to comprehensively analyze the cluster behaviors acquired by the video, so as to judge whether the cluster behaviors are abnormal.
In the understanding system, an information acquisition and feature mining layer is a scene understanding basis, and feature mining is mainly carried out on monitoring data acquired by a camera and manually acquired knowledge and rules by using a related algorithm. The data and knowledge fusion modeling layer is a key of scene understanding, realizes cluster scene information element construction based on multi-source information by fusing data and knowledge, and provides information support for subsequent cluster behavior research and judgment. The scene reproduction and behavior understanding layer mainly represents scene elements based on scene information elements, reproduces a cluster scene as comprehensively and truly as possible, comprehensively analyzes cluster behaviors by combining priori knowledge and existing rules, and judges whether the cluster behaviors are abnormal or not.
The main purpose of feature mining is to extract relevant features in video images, namely analyzing the upper video stream by technical means, and extracting relevant attributes of people or objects, such as: profile, trajectory, skeleton, direction of movement, speed, etc. In intelligent understanding and recognition of the cluster behaviors, more remarkable image features of people or objects, such as running of people, banner and the like, often exist, and the features provide important judgment basis for whether the machine recognition is the cluster behaviors or not and whether the cluster behaviors are abnormal or not. In addition, according to different scenes, the relevant attributes of the normal cluster behaviors are combed, a cluster behavior priori knowledge system is established, and the characteristic differences of the normal cluster behaviors and the abnormal cluster behaviors are clear. The key technology of the process is an image recognition algorithm, and the image information is processed, analyzed and understood mainly through computer technology. The traditional image recognition algorithm mainly utilizes characteristics such as texture, shape and the like to be combined with a machine learning classification algorithm for recognition. With the rapid development of electronic information technology, the computing capability is greatly improved, intelligent feature recognition is realized by an algorithm based on deep learning, no manual design of features is needed, and most of the deep learning algorithms are far more than the traditional algorithm in feature recognition accuracy.
The situation of public places is complex and changeable, data information in the monitoring video is only a part of the situation, static scene priori knowledge and social security related knowledge are also included, the construction of scenes from a single angle of the monitoring video is not sound, the data and the knowledge are necessary to be fused, a scene model is constructed by utilizing multi-source heterogeneous information, and the accurate and efficient recognition of cluster behaviors is realized. And (3) fusing data-knowledge to construct a information element model, carrying out scene modeling based on the information element model, and utilizing the information element to represent targets and attributes and relations thereof in the scene so as to realize comprehensive and real description of the scene. The scene information element comprehensively describes the real scene of the cluster behavior from three angles of the information element, the information element attribute and the association relation among the attributes, and provides information support for comprehensive research and judgment of the subsequent cluster behavior.
Further, the data and knowledge fusion modeling layer firstly judges whether the data/video image is a cluster behavior in advance through monitoring the data/video image, if the data/video image is not the cluster behavior, the analysis and judgment of the subsequent cluster behavior are terminated, and unnecessary computer resources and time consumption are reduced; if the cluster behavior belongs to the cluster behavior, other features are continuously mined, and data elements are obtained and fused with knowledge elements to form information elements.
The data and knowledge fusion modeling layer judges whether the video data is in a cluster behavior or not in advance through a person number abnormality judging module; the abnormal number judgment module comprises a crowd density chart generated by the video image, the number of people is estimated through the crowd density chart, meanwhile, an experience knowledge base and a behavior record base are combined to judge whether the crowd behavior is the group behavior, if the crowd behavior is not the group behavior, the subsequent analysis and judgment of the group behavior are terminated; if the cluster behavior is the cluster behavior, other features are started to be mined; the experience knowledge base is the knowledge of a manager, namely a database formed by summarizing the generalized cluster behavior characteristics through the ways of in-situ investigation, case base arrangement and interviewing police officers, and meanwhile, a new rule can be mined through a machine learning algorithm to update the experience knowledge base; the behavior record library refers to the time, place and scale information of the recorded cluster behaviors stored in government authorities.
The invention has the beneficial effects that: the cluster behavior has the characteristics of large space-time scale, heterogeneous information multiple sources, complex target characteristics and the like. Most of the existing cluster behavior understanding methods are based on analysis of monitoring images, and scene definition is narrow. The scene in reality not only contains the data information acquired by the monitoring camera, but also knowledge information outside the monitoring range such as police experience and related rules. The scene modeling is carried out by fusing the data and the knowledge, the scene can be more comprehensively and accurately represented, the cluster behavior understanding system is built based on the scene knowledge elements constructed by the multi-source information, the speed and the precision of cluster behavior recognition are improved, and the intelligent level of the public safety prevention and control system is improved.
Drawings
FIG. 1 builds a model framework diagram based on a clustered behavior understanding system of scene information elements.
Fig. 2 is a schematic diagram of a key feature extraction module.
FIG. 3 is a schematic diagram of a knowledge data fusion modeling method.
Fig. 4 is a schematic diagram of a module for judging abnormality of the number of people.
Detailed Description
The invention provides an overall framework of a cluster behavior understanding system based on scene knowledge elements, which is shown in fig. 1 and comprises three layers, namely an information acquisition and feature mining layer, a data and knowledge fusion modeling layer and a scene reproduction and behavior understanding layer. Wherein the solid rectangle represents data, the long-dashed rectangle represents knowledge, the short-dashed circle represents information, the solid line represents knowledge driving, the short-dashed line represents information driving, and the circle represents information element composed of data source and knowledge element. The information acquisition and feature mining layer is a scene understanding basis, and mainly utilizes a related algorithm to perform feature mining on monitoring data acquired by a camera and manually acquired knowledge and rules. The data and knowledge fusion modeling layer is a key of scene understanding, realizes cluster scene information element construction based on multi-source information by fusing data and knowledge, and provides information support for subsequent cluster behavior research and judgment. The scene reproduction and behavior understanding layer mainly represents scene elements based on scene information elements, reproduces a cluster scene as comprehensively and truly as possible, comprehensively analyzes cluster behaviors by combining priori knowledge and existing rules, and judges whether the cluster behaviors are abnormal or not.
The main purpose of feature mining is to extract relevant features in video images, namely analyzing the upper video stream by technical means, and extracting relevant attributes of people or objects, such as: profile, trajectory, skeleton, direction of movement, speed, etc. In intelligent understanding and recognition of the cluster behaviors, more remarkable image features of people or objects, such as running of people, banner and the like, often exist, and the features provide important judgment basis for whether the machine recognition is the cluster behaviors or not and whether the cluster behaviors are abnormal or not. In addition, according to different scenes, the relevant attributes of the normal cluster behaviors are combed, a cluster behavior priori knowledge system is established, and the characteristic differences of the normal cluster behaviors and the abnormal cluster behaviors are clear. The key technology of the process is an image recognition algorithm, and the image information is processed, analyzed and understood mainly through computer technology. The traditional image recognition algorithm mainly utilizes characteristics such as texture, shape and the like to be combined with a machine learning classification algorithm for recognition. With the rapid development of electronic information technology, the computing capability is greatly improved, intelligent feature recognition is realized by an algorithm based on deep learning, no manual design of features is needed, and most of the deep learning algorithms are far more than the traditional algorithm in feature recognition accuracy. Fig. 2 is a schematic diagram showing the corresponding feature information extracted from the video image by a plurality of known algorithms in the information acquisition and feature mining layer, and the plurality of algorithms and the extracted corresponding feature information form a key feature extraction module. The information acquisition and feature mining layer extracts information of potential threat personnel and important attention personnel from the monitoring data/video images through a face recognition algorithm, extracts information of whether a banner exists or not and whether weapons exist or not from the monitoring data/video images through a target detection algorithm, and extracts information of cluster population, direction consistency, cluster center and distribution uniformity from the monitoring data/video images through a group recognition algorithm.
The situation of public places is complex and changeable, data information in the monitoring video is only a part of the situation, static scene priori knowledge and social security related knowledge are also included, the construction of scenes from a single angle of the monitoring video is not sound, the data and the knowledge are necessary to be fused, a scene model is constructed by utilizing multi-source heterogeneous information, and the accurate and efficient recognition of cluster behaviors is realized. And (3) fusing data-knowledge to construct a information element model, carrying out scene modeling based on the information element model, and utilizing the information element to represent targets and attributes and relations thereof in the scene so as to realize comprehensive and real description of the scene. The scene information element comprehensively describes the real scene of the cluster behavior from three angles of the information element, the information element attribute and the association relation among the attributes, and provides information support for comprehensive research and judgment of the subsequent cluster behavior.
The scene information elements built by fusing data and knowledge can be subdivided into crowd information elements and background information elements. Crowd information elements are the subject of video analysis and are also the main subject of feature extraction. Crowd information elements are mainly focused on the quantity change and behavior change of crowd. The background information element refers to an element that is static in a video in a certain time and space range, and the background is an element that needs to be determined first in video image analysis. Whether the background is found or not is accurate, and the found background is one of key factors of video analysis accuracy. Considering the easy places and the influence range of the cluster behaviors, the concept and the attribute extraction of the background information element are mainly carried out at key positions such as party administrative offices, traffic intersections, airports, stations, squares, markets and the like. The background information element can be subdivided into a video inner background information element and a video outer background information element, wherein the video inner background information element and the crowd information element are sources of data during scene modeling, and the video outer background information element is sources of knowledge during scene modeling and comprises experiences of related staff, related policy regulations of government outbound and the like. Taking the example of a background information element in the video, the background information element can be subdivided into different object information elements, such as: market information elements, station information elements, etc., which are composed of element information elements, for example: building information elements, road information elements, sign information elements, and the like.
FIG. 3 is a schematic diagram of a knowledge data fusion modeling method, wherein a data and knowledge fusion modeling layer firstly judges whether a cluster behavior is preset by monitoring data/video images, and if the cluster behavior is not preset, the analysis and judgment of the subsequent cluster behavior are terminated; if the scene belongs to the cluster behavior, other features are continuously mined, a one-to-one correspondence relationship is established between the scene real world and the video virtual world through scene mapping, namely digitization, so that data elements are obtained, and the data elements and the knowledge elements are fused to form information elements.
Based on the acquired data and knowledge, a people number abnormality judging module can be constructed to judge in advance, mainly, the monitoring video which is directly acquired is primarily screened through a group activity record library, an experience knowledge base, a crowd counting algorithm and the like, and whether the cluster behavior is judged is finished, and a specific flow is shown in fig. 4. Because the people are the main bodies of the cluster behaviors, if no people exist or the number of people is very small in the monitoring video, advanced attribute mining and extraction are not needed, and unnecessary computer resources and time consumption are reduced as much as possible.
The cluster behavior record library refers to a statistical table containing large mass activity information. Most governments will dictate that large mass campaigns must be reported in advance to government-related departments at the site of the campaign. Therefore, government authorities can store information such as time, place, scale and the like of the recorded cluster behaviors, and the data is associated with the position information of the cameras, so that the screening of scenes of abnormal cluster behaviors can be assisted. The experience knowledge base is a database formed by summarizing the generalized cluster behavior characteristics through various approaches such as in-situ investigation, case library arrangement, interview police officers and the like. Meanwhile, a new rule can be mined through a machine learning algorithm, and an experience knowledge base can be updated. The experience knowledge base has the same function as the cluster behavior record base, performs preliminary screening on video images, and performs feature mining and scene modeling on video fragments with abnormal cluster behaviors.
After the people number abnormality judging module performs pre-screening, if the cluster behavior is confirmed to occur, the cluster behavior which is abnormal or not needs to be further screened, so that further scene modeling needs to be performed, and the cluster scene is analyzed more deeply. The scene information meta-model constructed by fusing the data and the priori knowledge can more accurately judge whether the cluster behaviors in the video are abnormal or not.
Claims (4)
1. A cluster behavior understanding system based on scene knowledge elements is characterized in that: the system sequentially comprises an information acquisition and feature mining layer, a data and knowledge fusion modeling layer and a scene reproduction and behavior understanding layer; the information acquisition and feature mining layer comprises a camera video data module for acquiring monitoring data/video images and a behavior priori knowledge module for acquiring monitoring external knowledge and rule information; the system also comprises a video characteristic data set obtained by characteristic mining of the monitoring data/video images and a behavior characteristic knowledge set obtained by characteristic mining of the monitoring external knowledge and the rule information;
the data and knowledge fusion modeling layer comprises a behavior feature data set and a background feature data set which are extracted from a video feature data set through an algorithm, and a behavior record library, manager knowledge and a research rule which are extracted from the behavior feature knowledge set through the algorithm; the behavior characteristic data set and the background characteristic data set jointly form a data element, and the behavior record library, the manager knowledge and the research rule jointly form a knowledge element; constructing an information element model by fusing the data element and the knowledge element;
the scene reproduction and behavior understanding layer subdivides the information elements into crowd information elements and background information elements, wherein the crowd information elements comprise crowd quantity changes and behavior changes; the background information element comprises an inner video background information element and an outer video background information element, wherein the inner video background information element and the crowd information element are sources of data when the information element model is built, and the outer video background information element is a source of knowledge when the information element model is built; through the classification representation, the cluster site is reproduced as comprehensively and truly as possible, and the prior knowledge and the existing rules are combined to comprehensively analyze the cluster behaviors acquired by the video, so as to judge whether the cluster behaviors are abnormal.
2. The system for understanding the cluster behavior based on the scene knowledge elements according to claim 1, wherein the information acquisition and feature mining layer extracts information of potential threat personnel and important attention personnel from the monitored data/video images through a face recognition algorithm, extracts information of existence of banners and weapons from the monitored data/video images through a target detection algorithm, and extracts information of cluster population, direction consistency, cluster center and distribution uniformity from the monitored data/video images through a group recognition algorithm.
3. The system for understanding the cluster behavior based on the scene knowledge elements according to claim 1 or 2, wherein the data and knowledge fusion modeling layer firstly judges whether the cluster behavior is the cluster behavior in advance by monitoring the data/video image, and if the cluster behavior is not the cluster behavior, the analysis and judgment of the subsequent cluster behavior are terminated; if the cluster behavior belongs to the cluster behavior, other features are continuously mined, and data elements are obtained and fused with knowledge elements to form information elements.
4. The system for understanding the cluster behavior based on the scene knowledge elements as claimed in claim 3, wherein the data and knowledge fusion modeling layer judges whether the video data is the cluster behavior in advance by constructing a population anomaly judgment module; the abnormal number judgment module comprises a crowd density chart generated by video images, the number of people is estimated through the crowd density chart, meanwhile, an experience knowledge base and a behavior record base are combined to judge whether the crowd is a trunking behavior, if the crowd is not the trunking behavior, the subsequent analysis and judgment of the trunking behavior are stopped, and if the crowd is the trunking behavior, other characteristics are started to be mined; the experience knowledge base is the knowledge of a manager, namely a database formed by summarizing the generalized cluster behavior characteristics through the ways of in-situ investigation, case base arrangement and interviewing police officers, and meanwhile, a new rule can be mined through a machine learning algorithm to update the experience knowledge base; the behavior record library refers to the time, place and scale information of the recorded cluster behaviors stored in government authorities.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310422133.7A CN117011774A (en) | 2023-04-19 | 2023-04-19 | Cluster behavior understanding system based on scene knowledge element |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310422133.7A CN117011774A (en) | 2023-04-19 | 2023-04-19 | Cluster behavior understanding system based on scene knowledge element |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117011774A true CN117011774A (en) | 2023-11-07 |
Family
ID=88564296
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310422133.7A Pending CN117011774A (en) | 2023-04-19 | 2023-04-19 | Cluster behavior understanding system based on scene knowledge element |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117011774A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117576634A (en) * | 2024-01-16 | 2024-02-20 | 浙江大华技术股份有限公司 | Anomaly analysis method, device and storage medium based on density detection |
-
2023
- 2023-04-19 CN CN202310422133.7A patent/CN117011774A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117576634A (en) * | 2024-01-16 | 2024-02-20 | 浙江大华技术股份有限公司 | Anomaly analysis method, device and storage medium based on density detection |
| CN117576634B (en) * | 2024-01-16 | 2024-05-28 | 浙江大华技术股份有限公司 | Anomaly analysis method, device and storage medium based on density detection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106997629B (en) | Access control method, apparatus and system | |
| Dong et al. | Face detection in security monitoring based on artificial intelligence video retrieval technology | |
| CN109271554B (en) | Intelligent video identification system and application thereof | |
| Saxena et al. | Crowd behavior recognition for video surveillance | |
| CN110428522A (en) | A kind of intelligent safety and defence system of wisdom new city | |
| Swathi et al. | Crowd behavior analysis: A survey | |
| CN109858365A (en) | A kind of special population Assembling Behavior analysis method, device and electronic equipment | |
| CN107659754B (en) | Effective concentration method for monitoring video under condition of tree leaf disturbance | |
| CN111353338B (en) | Energy efficiency improvement method based on business hall video monitoring | |
| CN109101944A (en) | A kind of real-time video monitoring algorithm identifying rubbish of jettisoninging into river | |
| CN101859436A (en) | Large-amplitude regular movement background intelligent analysis and control system | |
| CN112329691A (en) | Monitoring video analysis method and device, electronic equipment and storage medium | |
| Ullah et al. | Gaussian mixtures for anomaly detection in crowded scenes | |
| CN110245603A (en) | A kind of group abnormality behavior real-time detection method | |
| CN117011774A (en) | Cluster behavior understanding system based on scene knowledge element | |
| CN111191637A (en) | Crowd concentration detection and presentation method based on unmanned aerial vehicle video acquisition | |
| Patil et al. | Suspicious movement detection and tracking based on color histogram | |
| CN112766119A (en) | Method for accurately identifying strangers and constructing community security based on multi-dimensional face analysis | |
| CN111353343A (en) | Business hall service standard quality inspection method based on video monitoring | |
| Wang et al. | Traffic camera anomaly detection | |
| Khan et al. | Comparative study of various crowd detection and classification methods for safety control system | |
| CN104077571A (en) | Method for detecting abnormal behavior of throng by adopting single-class serialization model | |
| Feris et al. | Case study: IBM smart surveillance system | |
| CN111160150A (en) | Video monitoring crowd behavior identification method based on depth residual error neural network convolution | |
| CN115294519A (en) | Abnormal event detection and early warning method based on lightweight network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |