CN-121997702-A - Real-time safety protection method based on sensor array

CN121997702ACN 121997702 ACN121997702 ACN 121997702ACN-121997702-A

Abstract

The invention belongs to the technical field of safety protection, and discloses a real-time safety protection method based on a sensor array, which is based on a multi-mode sensor array system combining digital twin modeling, scene adaptation design and accurate deployment, is matched with a scene self-adaptive mixed domain attention reinforcement learning fusion model, extracts sensing data high-dimensional characteristics from four dimensions of space, time, frequency and semantics, dynamically distributes data weights of all sensors according to scene characteristics, strengthens threat association characteristics and simultaneously suppresses environmental interference; the virtual threat sample and the real sample generated by combining the digital twin model are combined to construct a scene-division threat sample data set, and the light model is optimized through migration learning, model pruning and incremental training, so that the accuracy of threat identification under a complex scene is improved, meanwhile, the adaptation capability of the system to novel threats is enhanced, the identification deviation caused by scene difference is reduced, and the accurate monitoring requirements of different protection scenes are met.

Inventors

LIU YUZHU
WU SHUPING

Assignees

湖州职业技术学院

Dates

Publication Date: 20260508
Application Date: 20251216

Claims (8)

1. The real-time safety protection method based on the sensor array is characterized by comprising a deployment stage, an acquisition preprocessing stage, a data fusion stage, a threat identification stage, an active protection pre-judgment stage, a positioning and tracing stage and an optimization iteration stage; the deployment phase comprises deploying a multi-mode sensor array and establishing a digital twin model, verifying the non-blind area coverage effect of a deployment scheme, and compensating environmental errors through a cross calibration mechanism; the acquisition preprocessing stage comprises the steps of adopting a three-stage cooperative computing architecture to realize synchronous data acquisition of a sensor, performing denoising and encryption, and then uploading a certificate, packaging the data into a unified format and transmitting the unified format to an edge gateway; The data fusion stage comprises the steps of extracting high-dimensional characteristics of sensing data through multiple dimensions of a scene self-adaptive fusion model based on the real-time scene state of a digital twin model, dynamically distributing data weights of all sensors, and generating a global characteristic vector of an adaptive scene; the threat identification stage is to train, optimize and lighten a deep learning model based on a scene threat sample data set so as to realize the real-time identification of scene threats; An active protection prejudging stage, namely establishing an active protection system, prejudging threat development trend by combining threat types and time sequence characteristics, establishing a hierarchical dynamic protection strategy library, and adjusting protection actions according to threat evolution situation; Establishing a Beidou, UWB and inertial navigation fusion positioning model, combining a digital twin scene topological feature correction positioning algorithm, and completing threat full-link tracing based on time sequence evidence storage data; and in the optimization iteration stage, scene adaptation effect data and active protection effect data are acquired, system parameters are optimized based on the evaluation result, and the supplementary data are used for digital twin model optimization.
2. The sensor array-based real-time safety protection method according to claim 1, wherein the deployment stage establishes a digital twin modeling, scene adaptation design and accurate deployment implementation trinity multi-modal sensor array deployment system, and the multi-modal sensor comprises a vibration sensor, an acoustic sensor, an infrared thermal imaging sensor, a microwave radar sensor, a Beidou and UWB dual-mode positioning sensor and an environment perception sensor; the digital twin model integrates scene topology parameters, threat type distribution, environmental interference characteristics and historical protection data, and builds a scene characteristic database, wherein the key parameter dimensions comprise electromagnetic interference levels and signal shielding areas; optimizing risk partitions of the protection area and deployment point positions of the sensor through an optimization algorithm, and verifying the non-blind area coverage effect of the deployment scheme; the multi-sensor cross calibration mechanism utilizes the complementarity of the data of each sensor, dynamically corrects acquisition parameters by combining environment perception data and compensates environment errors.
3. The real-time safety protection method based on the sensor array according to claim 2, wherein the acquisition preprocessing stage realizes synchronous acquisition of the sensors through a time synchronization protocol, and the three-level cooperative computing architecture comprises an edge layer, a fog node layer and a cloud node layer; The cloud node layer realizes distributed storage backup and key data storage of the data, and the data is transmitted to the edge gateway after the preprocessing is completed.
4. The sensor array-based real-time safety protection method according to claim 3, wherein the fusion model of the data fusion stage comprises a scene recognition, feature matching and weight optimization three-stage module, wherein the current safety scene type is judged by combining a digital twin model real-time scene state through a scene recognition algorithm; Extracting high-dimensional characteristics of sensor data from four dimensions of a space domain, a time domain, a frequency domain and a semantic domain by a mixed domain attention module; The threat related features and the environment interference suppression features are strengthened through the attention mechanism, global feature vectors are generated through fusion, and the weight distribution strategy is optimized autonomously.
5. The sensor array-based real-time security method of claim 4, wherein the split scene threat sample dataset is constructed by fusion of virtual threat samples generated by a digital twin model with real samples; performing scene-division training optimization on the lightweight model by adopting a transfer learning method, and reducing the calculation complexity of the model; And a scene self-adaptive updating mechanism is additionally arranged, and the model is subjected to incremental training through the newly-added threat sample and the digital twin generated iterative virtual sample, so that the recognition capability of the model on the novel threat is improved.
6. The real-time safety protection method based on the sensor array according to claim 5, wherein the active protection system comprises four modules of threat prediction, virtual simulation, grade evaluation and dynamic adaptation protection strategies, wherein the future development trend of the threat is predicted by combining a time sequence prediction model with a digital twin model, and threat evolution simulation is carried out in the digital twin model so as to simulate the implementation effects of different protection strategies; Based on threat types, prejudging results, simulation effects and scene characteristics, a scene adaptation type threat level assessment system is established, threats are divided into three levels of potential threats, ongoing threats and emergency threats, a reinforcement learning algorithm is introduced to dynamically adjust a protection action sequence according to real-time evolution situations of the threats, and active prejudging is achieved.
7. The real-time safety protection method based on the sensor array according to claim 6, wherein the fusion positioning model combines the spatial distribution information of the sensor array, the digital twin scene topological characteristics and the multi-mode time sequence fusion data; threat tracing is based on time sequence data and positioning data of blockchain evidence, a threat diffusion path is traced, a tracing report containing a time stamp, positioning coordinates and a threat evolution process is generated, and the tracing data is fed back to a digital twin model to optimize a protection strategy.
8. The sensor array-based real-time safety protection method according to claim 7, wherein the optimization iteration stage establishes a full-link closed-loop optimization system, the scene adaptation effect data comprises data fusion precision, threat identification accuracy and positioning precision, and the active protection effect data comprises threat treatment duration, whether secondary loss is avoided or not and pre-judging accuracy; optimizing sensor deployment parameters, fusing model weight rules, threat prejudging model parameters and a protection strategy library according to the pertinence of the evaluation result; the treated threat data and the protective effect data are supplemented to a scene characteristic database for iterative optimization of the digital twin model and calibration updating of the sensor array.

Description

Real-time safety protection method based on sensor array Technical Field The invention belongs to the technical field of safety protection, and particularly relates to a real-time safety protection method based on a sensor array. Background The sensor safety protection technology is widely applied to the fields of perimeter security protection, industrial monitoring, public safety and the like, threat monitoring is realized by collecting and analyzing multi-source sensing signals, and key guarantee is provided for personnel and property safety. At present, although the technology is widely applied, the following technical problems still exist in complex scene adaptation and active protection response: Most of the prior art adopts a fixed weight fusion strategy, threat characteristic differences under different protection scenes such as indoor closed space, outdoor open perimeter, industrial high-risk area and the like are not fully considered, so that fusion data are difficult to accurately fit scene requirements, and finally the threat identification accuracy is reduced; The prior art often starts the protection action after threat occurs and triggers the alarm threshold value, lacks the prejudgement ability to threat development trend, and the protection flow is mostly preset fixed mode, can't be according to threat evolution condition dynamic adjustment, leads to the protection untimely, the treatment efficiency is lower, causes the secondary loss easily. In addition, in the aspects of threat positioning and tracing, the prior art can only realize rough regional alarm, cannot accurately determine the threat occurrence position and is difficult to trace the threat diffusion path. Disclosure of Invention The invention aims to provide a real-time safety protection method based on a sensor array, so as to solve the problems in the background technology. In order to achieve the aim, the invention provides a real-time safety protection method based on a sensor array, which comprises a deployment stage, an acquisition preprocessing stage, a data fusion stage, a threat identification stage, an active protection prejudgment stage, a positioning tracing stage and an optimization iteration stage; Preferably, the deployment stage establishes a digital twin modeling, scene adaptation design and accurate deployment implementation trinity multi-modal sensor array deployment system, wherein the multi-modal sensor comprises a vibration sensor, an acoustic sensor, an infrared thermal imaging sensor, a microwave radar sensor, a Beidou and UWB dual-mode positioning sensor and an environment sensing sensor; firstly, a scene digital twin model is established, scene topology parameters, threat type distribution, environment interference characteristics and historical protection data are integrated, real-time correspondence between a physical scene and a virtual model is realized, a scene characteristic database is established based on the digital twin model, and key parameter dimensions including electromagnetic interference levels and signal shielding areas are established; Secondly, adopting an improved genetic and particle swarm hybrid algorithm to perform risk partition and sensor deployment point position optimization on a protection area, verifying the non-blind area coverage effect of a deployment scheme through virtual roaming, and matching the sensor type with deployment density to scene threat distribution; the specific method and the blind zone judgment standard of the virtual roaming verification are as follows: the roaming path planning comprises the steps of generating evenly distributed roaming path points based on scene topology parameters of a digital twin model, wherein the distance between the path points is less than or equal to 0.5 meter, covering the whole protection area, and each path point simulates a sensor signal receiving scene; The coverage intensity threshold value is that the signal receiving intensity of the sensor of each path point is more than or equal to-85 dBm, the electromagnetic interference level acquired by the environment sensing sensor is dynamically corrected, and the threshold value is adjusted downwards by 5dBm when the interference level is increased by 1 level; The dead zone judging rule is that if the signal receiving intensity of 3 or more continuous path points is less than a threshold value or the area of a single dead zone area is more than 0.5m2, dead zones exist in the deployment scheme, and sensor deployment points need to be optimized again; And (3) verifying and iterating, namely repeating virtual roaming verification after each optimization until the dead zone area ratio is less than or equal to 0.1% and no continuous dead zone path point exists. The specific fusion logic and parameters of the improved genetic and particle swarm hybrid algorithm (IGA-PSO) are as follows: Algorithm objective to minimize the area of the monitoring blind area Cost of deployment with senso