CN-122024397-A - Unattended system

CN122024397ACN 122024397 ACN122024397 ACN 122024397ACN-122024397-A

Abstract

The embodiment of the application discloses an on-duty system, which comprises video monitoring equipment, a cloud image analysis processing center, a scene equipment deployment and maintenance cost, an AI language big model, a linkage alarm equipment and a fire linkage alarm equipment, wherein the video monitoring equipment firstly performs light encoding on a real-time video stream, effectively reduces the volume of video data, reduces the occupied network bandwidth, improves the cloud transmission efficiency, the cloud image analysis processing center relies on the distributed computing capability of a cloud computing power cluster to realize simultaneous capture and analysis of multi-area fire alarm characteristics, replaces traditional edge computing equipment, reduces the deployment and maintenance cost of the scene equipment, can realize unified management and model iterative optimization of fire alarm characteristic data, the AI language big model converts the extracted fire alarm image characteristics of the cloud into structured and semantic natural language alarm information, breaks through the information limitation of traditional simple audible and visual alarm, provides accurate fire key information for fire control treatment, and the linkage alarm equipment solves the problem of equipment isomerism through protocol conversion middleware, realizes unified scheduling of different brands of alarm equipment and different types, and improves the fire linkage response efficiency.

Inventors

CHEN LUBIN
WANG SEN

Assignees

深圳市小鹰视界智能有限公司

Dates

Publication Date: 20260512
Application Date: 20260303

Claims (10)

1. The unattended system is characterized by comprising an AI language big model, video monitoring equipment, a cloud image analysis processing center and linkage alarm equipment, wherein the video monitoring equipment is connected with the cloud image analysis processing center through network communication, the cloud image analysis processing center is connected with the AI language big model through network communication, the AI language big model is connected with the linkage alarm equipment through network communication, the video monitoring equipment is provided with an audible and visual alarm component, and the linkage alarm equipment comprises a plurality of protocol conversion middleware; The video monitoring equipment is used for collecting real-time video streams of the current monitoring area, and uploading the real-time video streams to a cloud image analysis processing center after light-weight encoding; The AI language big model is used for analyzing target image characteristics extracted by the cloud image analysis processing center, determining a fire behavior of an illegal fire event and generating natural language alarm information, wherein the fire behavior is determined by the cloud image analysis processing center based on the fire characteristic capture of a real-time video stream and the data analysis of a cloud computing power cluster.
2. The unattended system according to claim 1, wherein the video monitoring device is configured to determine a first area on a current frame image of a real-time video stream to determine a first target identifier of a first focusing target on the current frame image, wherein the first target identifier is a positioning identifier of second identifier information of a fire monitoring behavior and a fire feature of any one of preset fire monitoring behaviors, the second identifier information is used to determine a target frame image of a fire behavior of a fire target on the real-time video stream of the current monitoring area, and the high-definition panoramic camera encodes the first area frame data with the first target identifier and then separately uploads the encoded first area frame data to the cloud image analysis processing center.
3. The unattended system according to claim 2, wherein the first area is configured with a first bounding box and a second bounding box, wherein the first bounding box is a display box of a fire behavior on a current frame image, the second bounding box is a display box of a fire source in the fire behavior, and the second bounding box is configured with a millisecond-level timestamp; And carrying out time sequence marking on the target frame image through the time stamp, and uploading the target frame image with the time sequence marking to a cloud image analysis processing center after the high-definition panoramic camera is subjected to light weight compression, wherein the cloud image analysis processing center sequentially fuses the target frame image through the time stamp to generate structured real-time video stream data.
4. The unattended operation system according to claim 1, wherein the cloud image analysis processing center is configured with a cloud distributed storage module, the cloud distributed storage module stores data features of fire behaviors, the data features of the fire behaviors include behavior feature identifiers according to time sequence, initial identification information of the behavior feature identifiers is a first frame compressed image when the fire features appear in the real-time video stream as target fire, and sequential adjacent images of the first frame images are target frame compressed images corresponding to the data features of the fire behaviors, which are recognized by the cloud image analysis processing center, of the presence of the target fire and the fire features.
5. The unattended operation system according to claim 4, wherein the cloud image analysis processing center is provided with a cloud distributed visual feature extraction model, the model operates on a cloud computing power cluster and is used for determining fire alarm target image features corresponding to target frame images in structured real-time video stream data and calculating a first confidence value of the fire alarm features in the target frame images belonging to archived fire types and a second confidence of the fire alarm behavior features in the target frame images belonging to preset fire levels under corresponding fire types; And fusing the first confidence coefficient and the second confidence coefficient, determining a fused confidence coefficient value, and when the fused confidence coefficient value is higher than a first threshold value, taking the target frame image as a fire alarm judging image with fire alarm characteristics and pushing the image to the AI language big model.
6. The unattended system according to claim 1, wherein the cloud image analysis processing center is configured with a second threshold and a third threshold, wherein when the first confidence value is lower than the second threshold, the cloud image analysis processing center sends an instruction to the linkage alarm device, the linkage alarm device executes a fire recognition abnormal alarm, and when the second confidence value is lower than the third threshold and the first confidence value is higher than the second threshold, the cloud image analysis processing center sends an instruction to the linkage alarm device, and the linkage alarm device executes a fire level matching abnormal alarm.
7. The unattended operation system according to claim 1, wherein the AI language big model comprises a multi-mode input interface and a big language processing model, wherein the multi-mode input interface comprises a video analysis input interface and a fire control system data input interface, the video analysis input interface is used for receiving a first representation text pushed by a cloud image analysis processing center and based on a monitoring area level and a fire control responsibility label, the first representation text is a fire control treatment task text corresponding to a first fire control treatment type under the monitoring area level of a fire condition corresponding to a fire feature; The large language processing model is used for generating natural language alarm information which is provided with causal relation and used for executing fire alarm under the current fire alarm treatment task according to the fire type according to the first characterization text and the second characterization text.
8. The unattended system of claim 7 wherein the natural language alarm information comprises first alarm information based on time, fire location and monitoring area fused to cloud space-time positioning data and second alarm information based on fire status, risk level and fire operation and maintenance fused to cloud fire protection system real time data.
9. The unattended system according to claim 7, wherein the large language processing model further determines first fire alarm equipment information of a monitoring area corresponding to a current fire alarm processing task according to the current fire alarm processing task of a fire type by combining space positioning data of a cloud image analysis processing center, wherein the first fire alarm equipment information comprises video monitoring equipment information with an audible and visual alarm function and second fire alarm equipment information, wherein the distance between the second fire alarm equipment information and a fire source in the current monitoring area does not exceed a preset maximum alarm distance threshold value.
10. The unattended system according to claim 1, wherein the linkage alarm device is further configured with a field alarm control unit based on a first protocol stack and a remote linkage alarm control unit based on a second protocol stack, the first protocol stack and the second protocol stack support cloud communication protocol adaptation with a cloud image analysis processing center and an AI language big model, the first protocol stack is used for receiving a first protocol configuration instruction which is based on protocol conversion configuration and is transmitted by the AI language big model and is called by the device authority of a current monitoring area, the first protocol configuration instruction is a control instruction of an acousto-optic alarm signal of the linkage alarm device and a video monitoring device in the current monitoring area, the second protocol stack is used for receiving a second protocol configuration instruction which is based on protocol conversion configuration and is transmitted by the AI language big model and is used for receiving a control instruction of an acousto-optic alarm signal of a mobile alarm device and an emergency fire control response device of a remote fire control management end.

Description

Unattended system Technical Field The application relates to the technical field of monitoring, in particular to an unattended system. Background The traditional fire alarm monitoring system has a plurality of technical defects that a sensor is easy to be interfered by environment to generate false alarm, such as a smoke sensor is touched by oil smoke and steam, and the accuracy of the temperature sensor is influenced by dust and high temperature environment to cause invalid alarm frequently; The real-time video stream collected by the video monitoring equipment is original unstructured data, a large amount of network bandwidth is occupied by directly uploading the cloud end, the transmission efficiency is low, the targeted screening of massive video data is lacking, and the computational cost of subsequent data analysis is high; The type judgment, grading and disposal instruction generation of the fire event completely depend on manual experience, automatic analysis and accurate judgment of the fire behavior cannot be realized, the alarm information is only a simple sound-light signal, and key information such as the position, the type and the grade of the fire is lacked, so that the emergency response efficiency is low; The on-site fire alarm equipment and the emergency equipment protocol of the remote fire control management end are not uniform, the equipment heterogeneous problem is prominent, the cross-regional and cross-type fire alarm linkage alarm is difficult to realize, and a closed loop for on-site quick reminding and remote emergency treatment cannot be formed. Disclosure of Invention Aiming at the technical defects of traditional fire alarm monitoring, the application provides an unmanned system, which adopts a cloud image analysis processing center to realize centralized and automatic analysis of fire alarm characteristics, converts the fire alarm image characteristics into natural language alarm information containing key information by combining semantic analysis and reasoning capacity of an AI language large model, realizes accurate on-site and remote acousto-optic alarm through linkage alarm equipment, solves the problems of high occupation of video transmission bandwidth, fire alarm identification lag, fuzzy alarm information, poor equipment linkage compatibility and the like, and realizes automatic capture, accurate judgment, semantic alarm and multi-terminal linkage treatment of fire alarm behaviors. The embodiment of the application provides an unmanned system, which comprises an AI language big model, video monitoring equipment, a cloud image analysis processing center and linkage alarm equipment, wherein the video monitoring equipment is connected with the cloud image analysis processing center through network communication, the cloud image analysis processing center is connected with the AI language big model through network communication, the AI language big model is connected with the linkage alarm equipment through network communication, the video monitoring equipment is provided with an audible and visual alarm component, and the linkage alarm equipment comprises a plurality of protocol conversion middleware; The video monitoring equipment is used for collecting real-time video streams of the current monitoring area, and uploading the real-time video streams to a cloud image analysis processing center after light-weight encoding; The AI language big model is used for analyzing target image characteristics extracted by the cloud image analysis processing center, determining a fire behavior of an illegal fire event and generating natural language alarm information, wherein the fire behavior is determined by the cloud image analysis processing center based on the fire characteristic capture of a real-time video stream and the data analysis of a cloud computing power cluster. Optionally, in some embodiments of the present application, the video monitoring device is configured to determine a first area on a current frame image of a real-time video stream, so as to determine a first target identifier of a first focusing target on the current frame image, where the first target identifier is a positioning identifier of second identifier information of any one of preset fire monitoring behaviors and fire characteristics, the second identifier information is used to determine a target frame image of a fire behavior of a fire target on the real-time video stream of the current monitoring area, and the high-definition panoramic camera encodes the first area frame data with the first target identifier and then separately uploads the encoded first area frame data to the cloud image analysis processing center. Optionally, in some embodiments of the present application, the first area is configured with a first bounding box and a second bounding box, where the first bounding box is a display box of a fire behavior on the current frame image, the second bounding box is a display box of a fire s