CN-122024401-A - Multi-agent large-model security event cognition early-warning system

CN122024401ACN 122024401 ACN122024401 ACN 122024401ACN-122024401-A

Abstract

The invention relates to the technical field of intersection of artificial intelligence and security technology, and discloses a multi-agent large-model security event cognition early warning system which comprises a multi-agent cluster, a large-model core processing module, a data interaction module and an early warning output module, wherein the multi-agent cluster and the large-model core processing module are in bidirectional communication through the data interaction module, the multi-agent cluster is used for collecting multi-source security data and performing preliminary preprocessing, the multi-agent cluster comprises a data perception agent, a data transmission agent and a data filtering agent, and the large-model core processing module comprises an event cognition engine, a risk assessment engine and a model optimization unit. According to the invention, through the cooperative division of multiple intelligent agents and the depth cognition capability of a large model, the system realizes the full-link efficient processing of security data from acquisition to early warning, improves the accuracy of security event identification and the real-time performance of early warning, and provides reliable technical support for complex scene security work.

Inventors

LIAO JUN
LIU DAN
LI BINGSHENG
Shi Mengtao
TIAN XIN

Assignees

重庆市忠信保安服务有限公司

Dates

Publication Date: 20260512
Application Date: 20251225

Claims (10)

1. The multi-agent large-model security event cognition early warning system is characterized by comprising a multi-agent cluster, a large-model core processing module, a data interaction module and an early warning output module; the multi-agent cluster and the large model core processing module are in bidirectional communication through the data interaction module, and the output end of the large model core processing module is connected with the early warning output module; The multi-agent cluster is used for collecting multi-source security data and performing preliminary pretreatment, the large model core processing module is used for performing event cognition and risk assessment on the preprocessed security data, and the early warning output module is used for generating and outputting early warning information according to an assessment result; The multi-agent cluster comprises a data sensing agent, a data transmission agent and a data filtering agent; the large model core processing module comprises an event cognition engine, a risk assessment engine and a model optimization unit; the early warning output module comprises a grading early warning sub-module and a multi-channel pushing sub-module.
2. The multi-agent large model security event cognitive pre-warning system according to claim 1, wherein the data perception agent is configured to collect video data, audio data, environmental sensing data and personnel identity data, the data transmission agent is configured to transmit the collected data by using an encryption transmission protocol, and the data filtering agent is configured to reject redundant data and abnormal interference data.
3. The multi-agent large model security event cognitive early warning system according to claim 1, wherein the event cognitive engine is configured to identify security event types and key features based on a pre-trained large model, the risk assessment engine is configured to calculate a risk level by combining historical event data and real-time scene parameters, and the model optimization unit is configured to update model parameters according to early warning feedback results.
4. The multi-agent large model security event cognitive early warning system according to claim 3, wherein the event cognitive engine comprises a feature extraction sub-module and a semantic understanding sub-module, the feature extraction sub-module is configured to extract visual features, audio features and text features of multi-source data by adopting a deep learning algorithm, and the semantic understanding sub-module is configured to analyze abnormal behavior descriptions and dangerous article identification semantic information.
5. The multi-agent large-model security event cognitive early warning system according to claim 1, wherein the hierarchical early warning sub-module is configured to divide early warning information into three levels of primary emergency, secondary importance and tertiary attention, and the multi-channel pushing sub-module is configured to synchronously push the early warning information through a security terminal, a mobile APP, a short message and an audible and visual alarm.
6. The multi-agent large model security event cognitive early warning system according to claim 1, further comprising a man-machine interaction module, wherein the man-machine interaction module is connected with the large model core processing module and is configured to receive manual intervention instructions and correct and confirm early warning results.
7. The multi-agent large model security event cognitive early warning system according to claim 3, wherein the risk assessment engine adopts a weighted scoring algorithm, and the weight parameters comprise event urgency, influence range, occurrence probability and historical treatment difficulty, and the weight of each parameter can be dynamically adjusted according to security scene types.
8. The multi-agent large-model security event cognitive early warning system according to claim 2, wherein the data perception agents comprise high-definition cameras, sound pick-up devices, infrared sensors, metal detectors and face recognition terminals, and each perception device realizes data local preliminary processing through an edge computing unit.
9. The multi-agent large model security event cognition early warning system according to claim 1, wherein the data interaction module supports 5G and WiFi6 dual mode communication, is configured to automatically switch to a local cache mode when a network is interrupted, and synchronizes data to the large model core processing module after the network is restored.
10. The multi-agent large model security event cognitive early warning system according to claim 8, wherein the edge calculation unit is configured to perform real-time target detection on video data, and only transmit video clips containing abnormal targets to the large model core processing module, so as to reduce data transmission quantity.

Description

Multi-agent large-model security event cognition early-warning system Technical Field The invention relates to the technical field of intersection of artificial intelligence and security technologies, in particular to a security event cognition early warning system based on multi-agent coordination and large model cognition capability, which is suitable for various complex security scenes of parks, transportation hubs and large venues. Background The current security early warning system is gradually changed from a traditional manual on-duty mode to an intelligent monitoring mode, and mainstream products are generally integrated with video monitoring, infrared detection and personnel identification hardware equipment, and a basic abnormality detection function is realized by combining a simple algorithm model. The system judges the collected security data through preset rules (such as regional invasion and exceeding of personnel aggregation density), and sends out an early warning signal when a preset threshold value is triggered, so that the system is widely applied to market and factory scenes. And a deep learning algorithm is also introduced into part of the high-end systems, so that the accuracy of target identification is improved, and the preliminary identification of specific dangerous goods and abnormal behaviors can be realized. The following problems also exist for the prior art: 1. In the existing system, a single intelligent module is adopted to process multi-source security data, so that classification processing and collaborative analysis of the data cannot be realized, and the fusion degree of video, audio and sensing multi-type data is low; 2. the early warning logic of the existing system depends on fixed rules or simple models, lacks depth cognition capability on complex scenes, and cannot effectively identify novel security events outside the rules. Disclosure of Invention The invention aims to provide a multi-agent large-model security event cognition early warning system, which realizes the accurate identification and real-time early warning of security events by cooperatively processing multi-source data through multi-agents and combining the deep cognition capability of a large model so as to solve the problems in the background technology. In order to solve the technical problems, the invention is realized by the following technical scheme: The invention relates to a multi-agent large-model security event cognition early warning system which comprises a multi-agent cluster, a large-model core processing module, a data interaction module and an early warning output module, wherein the multi-agent cluster is connected with the data interaction module; the multi-agent cluster and the large model core processing module are in bidirectional communication through the data interaction module, and the output end of the large model core processing module is connected with the early warning output module; The multi-agent cluster is used for collecting multi-source security data and performing preliminary pretreatment, the large model core processing module is used for performing event cognition and risk assessment on the preprocessed security data, and the early warning output module is used for generating and outputting early warning information according to an assessment result; The multi-agent cluster comprises a data sensing agent, a data transmission agent and a data filtering agent; the large model core processing module comprises an event cognition engine, a risk assessment engine and a model optimization unit; the early warning output module comprises a grading early warning sub-module and a multi-channel pushing sub-module. Further, the data sensing agent is configured to collect video data, audio data, environmental sensing data and personnel identity data, the data transmission agent is configured to transmit the collected data by using an encryption transmission protocol, and the data filtering agent is configured to reject redundant data and abnormal interference data. Further, the event cognition engine is configured to identify security event types and key features based on a pre-trained large model, the risk assessment engine is configured to calculate a risk level by combining historical event data and real-time scene parameters, and the model optimization unit is configured to update model parameters according to early warning feedback results. Further, the event cognition engine comprises a feature extraction sub-module and a semantic understanding sub-module, wherein the feature extraction sub-module is used for extracting visual features, audio features and text features of multi-source data by adopting a deep learning algorithm, and the semantic understanding sub-module is used for analyzing abnormal behavior description and dangerous object identification semantic information. Further, the hierarchical early warning sub-module is arranged to divide early warning informat