Smart park safety management system

Abstract

The invention, which relates to the technical field of intelligent security management, discloses a smart park security management system comprising a data sensing module, an edge calculation module, an intelligent analysis module and a command execution module. According to the system, related safety data of the park are collected in real time in a multi-source mode through the data sensing module, the data collected in real time are locally preprocessed through the edge calculation module, the global safety situation in the park is sensed, then the data collected in real time are analyzed and processed through the intelligent analysis module, graded early warning signals are generated, and the safety of the park is improved. And finally, corresponding early warning emergency operation is executed through a command execution module, so that real-time monitoring, intelligent early warning and automatic emergency response of the safety of the smart park are realized, the safety management efficiency and accuracy are improved, the problems of response lag, data island and the like in the safety management of the traditional park are solved, and the safety of the park is improved. And full-process management from data acquisition, intelligent analysis to automatic disposal can be realized, and the safety protection capability of the park is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent security management, and particularly to a security management system for a smart park. Background Art

[0002] A park refers to a designated area planned and unified centrally, where certain specific industries or forms of enterprises, companies, etc. are specially set up for unified management. Typical examples include industrial parks, free trade parks, industrial parks, and animation parks. A smart park refers to an organic entity and sustainable development space that applies digital technologies and is a deep integration of humans, machines, things, and events based on comprehensive perception and ubiquitous connection, with characteristics such as active service and intelligent evolution capabilities. It realizes timely, interactive, and integrated information perception, transmission, and processing within the park through new-generation information and communication technologies, aiming to improve the industrial agglomeration ability, enterprise economic competitiveness, and sustainable development of the park.

[0003] The security management of a smart park is an important part of the construction of a smart park. It involves multiple aspects such as security monitoring, accident prevention, and emergency response within the park. The security management of a smart park is crucial for ensuring the safety of personnel and property within the park and maintaining normal production and operation order. Through intelligent means, the occurrence of security accidents can be effectively prevented and reduced, and the overall security level of the park can be improved.

[0004] With the acceleration of the urbanization process, as an important part of a modern city, the security management requirements of a smart park are becoming increasingly complex. Traditional park security management relies on manual inspections, video monitoring, or independent subsystems such as fire alarms. However, the data of each subsystem cannot be effectively interconnected, making it difficult to perceive the overall security situation within the park, and thus unable to achieve comprehensive security management of the park. Moreover, relying on manual judgment of abnormal events has low efficiency and is prone to missed judgments, and the security needs to be improved. In addition, there is a lack of intelligent prediction of potential risks, and the degree of intelligence is relatively low. Therefore, the present invention proposes a security management system for a smart park to solve the problems existing in the prior art. Summary of the Invention

[0005] Aiming at the above problems, the purpose of the present invention is to propose a security management system for a smart park, which solves the problems that it is difficult to perceive the overall security situation within the park in traditional park security management, relying on manual judgment of abnormal events, with low efficiency and being prone to missed judgments, and lacking intelligent prediction of potential risks.

[0006] To achieve the object of the present invention, the present invention is implemented through the following technical solutions: A smart park security management system includes a data perception module, an edge computing module, an intelligent analysis module, and a command execution module. The data perception module includes an environmental data acquisition unit for perceiving environmental data in the park, a device operation parameter acquisition unit for perceiving the operation parameters of devices in the park, and a personnel behavior data acquisition unit for perceiving the behavior of personnel in the park;

[0007] The edge computing module is deployed at the gateway of each partition in the park and receives data from the data perception module. The edge computing module includes a data preprocessing unit for preprocessing each perceived data and a target detection unit for identifying abnormal event metadata;

[0008] The intelligent analysis module is connected to the edge computing module through an optical fiber network. The intelligent analysis module includes a multi-source data fusion unit for fusing data from the edge computing module, a dynamic risk assessment unit for dynamically assessing the risk level of the park, and an emergency decision-making unit for generating emergency response plans for the park;

[0009] The command execution module receives the emergency response plan decision for the park from the intelligent analysis module. The command execution module includes a standardized API interface for interconnecting with the park's security equipment, fire protection system, access control controller, and broadcast terminal, and an emergency operation execution unit for sending emergency operation execution commands to the park's security equipment, fire protection system, access control controller, and broadcast terminal.

[0010] A further improvement lies in that: the environmental data acquisition unit includes a temperature and humidity sensor, a smoke concentration sensor, a monitoring camera, and a millimeter-wave radar. The device operation parameter acquisition unit includes an electric load sensor, a pressure sensor, and a voiceprint recognition sensor. The personnel behavior data acquisition unit is a personnel ID card integrated with a UWB positioning module.

[0011] A further improvement lies in that: the dynamic risk assessment unit is based on a deep reinforcement learning framework. The input data includes a historical accident dataset, a real-time sensor data stream, device operation and maintenance logs, and external environmental parameters. The risk assessment strategy is iteratively optimized through the Q-learning algorithm to build a dynamic risk assessment model. The risk probability calculation formula is defined as:

[0012] P risk = α·S env + β·S device + γ·S human

[0013] where S env is the environmental risk coefficient, S device is the device risk coefficient, S humanis the risk coefficient of personnel behavior, α, β, and γ are dynamic weight parameters. The output result of the dynamic risk assessment model is risk level labels of different colors including blue, yellow, and red and their corresponding confidence levels, triggering different early warning response mechanisms.

[0014] The further improvement lies in that: a transfer learning mechanism is introduced for training during the construction of the dynamic risk assessment model. The specific steps are as follows: in the pre-training stage, a public safety accident database is used to generate a basic risk feature library, and in the online learning stage, the local data of the park edge nodes is aggregated through federated learning technology to update the model parameters.

[0015] The further improvement lies in that: the linkage logic between the command execution module and the intelligent analysis module includes: when the early warning level is red, automatically cut off the power supply of the associated area in the park, start the fire sprinkler device, and play the evacuation instructions through the broadcast; when the early warning level is yellow, push the warning information to the mobile terminals of the security personnel and lock the access control system of the relevant areas in the park; when the early warning level is blue, mark the potential risk points on the park electronic map to prompt manual recheck.

[0016] The further improvement lies in that: the target detection unit is a target detection model improved based on the MobileNetV3 architecture, which is used to process the video stream data as follows: real-time identify events such as flames, smoke, abnormal aggregation of people or intrusion into restricted areas, filter the static background interference through the frame difference method, extract the coordinates and timestamps of the dynamic abnormal areas, and upload the metadata of the abnormal events to the intelligent analysis module, and locally store the original video data.

[0017] The further improvement lies in that: the multi-source data fusion unit is based on the synchronization method of video streams, sensor data, and device logs aligned by timestamps, uses knowledge graph technology to establish an entity relationship network of devices, regions, and people, annotates the associated risk conduction paths, and performs Bayesian probability arbitration on the conflicting data.

[0018] The further improvement lies in that: the data preprocessing unit is built with a lightweight AI model and a data filtering algorithm to perform local preprocessing on the raw data obtained by the data perception module, extract the abnormal event features and compress the redundant data.

[0019] The beneficial effects of the present invention are as follows: The present invention collects relevant safety data of the park in real time through the multi-source data perception module, and performs local preprocessing on the real-time collected data through the edge computing module to realize the perception of the global safety situation in the park. Then, the intelligent analysis module analyzes and processes the real-time collected data and generates hierarchical warning signals. Finally, the command execution module executes the corresponding warning emergency operations to realize the real-time monitoring, intelligent warning and automatic emergency response of the intelligent park safety, improve the safety management efficiency and accuracy, solve the problems of lagging response and data islands in traditional park safety management, and can realize the whole-process management from data collection, intelligent analysis to automatic disposal, significantly improving the safety protection ability of the park. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a schematic structural diagram of the intelligent park safety management system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0022] With the continuous development of technology, the form of the park is also constantly evolving and developing. The business carried by the park is becoming more and more complex, the objects of park management are increasing, and the management scope of the park is getting larger. Due to the limited infrastructure and service capabilities of the park, the contradiction between the limited infrastructure and service capabilities and the growing production and living needs of people has become increasingly prominent. There are great deficiencies in the safety, cost and efficiency of the park. Therefore, it is necessary to assist park management personnel in park safety management work through the corresponding park safety management system.

[0023] According to Figure 1 As shown, this embodiment provides an intelligent park safety management system, which is composed of a data perception module for collecting relevant safety data of the park in real time, an edge computing module for locally preprocessing the real-time collected data, an intelligent analysis module for generating hierarchical warning signals according to the real-time collected data, and a command execution module for executing warning emergency operations, wherein:

[0024] The data perception module consists of multi-modal Internet of Things devices distributed in the smart park, including an environmental data collection unit, a device operation parameter collection unit, and a personnel behavior data collection unit. The environmental data collection unit is used to sense the environmental data in the park, the device operation parameter collection unit is used to sense the device operation parameters in the park, and the personnel behavior data collection unit is used to sense the personnel behavior in the park;

[0025] The edge computing module is deployed at the gateway of each partition in the park and receives data from the data perception module. The edge computing module consists of a data preprocessing unit and a target detection unit. The data preprocessing unit is used to preprocess the sensed data, and the target detection unit is used to identify the metadata of abnormal events;

[0026] The intelligent analysis module is connected to the edge computing module through an optical fiber network. The intelligent analysis module consists of a multi-source data fusion unit, a dynamic risk assessment unit, and an emergency decision-making unit. The multi-source data fusion unit is used to fuse the data from the edge computing module, the dynamic risk assessment unit is used to dynamically evaluate the risk level of the park, and the emergency decision-making unit is used to generate the emergency plan decision for the park;

[0027] The command execution module receives the emergency plan decision of the park from the intelligent analysis module. The command execution module consists of a standardized API interface and an emergency operation execution unit. The command execution module is interconnected with the park security equipment, fire protection system, access control controller, and broadcast terminal through the standardized API interface. The emergency operation execution unit sends the corresponding emergency operation execution commands to the park security equipment, fire protection system, access control controller, and broadcast terminal according to the emergency plan decision generated by the emergency decision-making unit.

[0028] The environmental data collection unit of this embodiment consists of temperature and humidity sensors, smoke concentration sensors, monitoring cameras, and millimeter-wave radars deployed on the key paths in the park (such as fire escape routes, dangerous goods transportation paths, perimeters of high-voltage equipment areas, and entrances to chemical storage areas). Among them, the temperature and humidity sensors are used to monitor the temperature and humidity data in real time, the smoke concentration sensors are used to monitor the smoke concentration in the air in real time, the monitoring cameras are used to monitor and photograph the environment in the park in real time, and the millimeter-wave radars are used to detect hidden dangerous goods by penetrating obstacles;

[0029] The device operation parameter collection unit of this embodiment consists of a power load sensor, a pressure sensor, and a voiceprint recognition sensor. Among them, the power load sensor is installed on the park devices and detects the device power load in real time, the pressure sensor is installed in the pipeline and detects the pipeline pressure in real time, and the voiceprint recognition sensor is installed in the pipe gallery and judges mechanical failures by analyzing the device operation noise spectrum;

[0030] The personnel behavior data collection unit is a personnel ID card integrated with a UWB (Ultra-Wideband) positioning module, which is worn by the mobile personnel in the park to track the location and movement trajectory of the personnel in real time, so as to obtain the personnel density and trajectory in the park.

[0031] The dynamic risk assessment unit is based on the Deep Reinforcement Learning (DRL) framework. The input data includes historical accident datasets, real-time sensor data streams, equipment operation and maintenance logs, and external environment parameters (weather, holiday pedestrian flow density). Through the Q-learning algorithm, the risk assessment strategy is iteratively optimized to build a dynamic risk assessment model. The risk probability calculation formula is defined as:

[0032] P risk = α·S env + β·S device + γ·S human

[0033] Where S env is the environmental risk coefficient, S device is the equipment risk coefficient, S human is the personnel behavior risk coefficient, and α, β, γ are dynamic weight parameters. The output result of the dynamic risk assessment model is the risk level label (blue / yellow / red) and the corresponding confidence level, triggering different warning response mechanisms;

[0034] During the construction process of the dynamic risk assessment model, a transfer learning mechanism is introduced for training. The specific steps are as follows: In the pre-training stage, a public safety accident database (such as factory explosions, stampedes) is used to generate a basic risk feature library. In the online learning stage, the local data of multiple park edge nodes is aggregated through federated learning technology to update the model parameters.

[0035] The linkage logic between the command execution module and the intelligent analysis module includes: when the warning level is red, automatically cut off the power supply in the associated area, start the fire sprinkler device, and broadcast evacuation instructions; when the warning level is yellow, push warning information to the mobile terminals of security personnel and lock the access control system of the relevant area; when the warning level is blue, mark potential risk points on the digital twin interface of the park to prompt manual review.

[0036] The target detection unit is a target detection model improved based on the MobileNetV3 architecture, which is used to process the video stream data as follows: real-time identification of flames, smoke, abnormal personnel gatherings or intrusion into restricted areas. The static background interference is filtered by the frame difference method, and the coordinates and timestamps of the dynamic abnormal areas are extracted. Only the metadata of abnormal events (type, location, confidence level) are uploaded to the cloud platform, and the original video data is stored locally.

[0037] The multi-source data fusion unit uses a synchronization method based on time-stamp aligned video streams, sensor data, and equipment logs, and uses knowledge graph technology to establish an entity relationship network of equipment, area, and personnel, mark the associated risk transmission paths, and perform Bayesian probabilistic arbitration on conflicting data (such as the difference in fire judgment between cameras and infrared sensors).

[0038] The data preprocessing unit has a built-in lightweight AI model and data filtering algorithm to perform local preprocessing on the raw data obtained by the data perception module, extract abnormal event features and compress redundant data. The specific steps are as follows: receive heterogeneous data streams from different sources, first remove the outliers, and then align sensor data with different sampling rates (such as 1Hz temperature and humidity sensors and 30fps video streams) through interpolation compensation to ensure that the timestamp synchronization error is ≤20ms, perform frequency domain feature extraction on numerical data, and perform multimodal fusion on the extracted feature data, input it into the lightweight random forest model, output the comprehensive abnormality probability, and then use Delta encoding + Zstandard compression to encapsulate the compressed data into a standardized data packet.

[0039] When it is necessary to manage the security of the smart park, the data perception module collects the park environment, park equipment and personnel data in real time through the multimodal IoT devices deployed in the park, and then uses the edge computing modules deployed at the gateways of each partition of the park to receive the data from the data perception module and pre-process the data in real time. The pre-processed data is uploaded to the intelligent analysis module via the optical fiber network, and multi-source data is first integrated, and then a dynamic risk assessment is performed, and then an emergency plan is generated. Finally, the command execution module sends control instructions to the execution devices such as the park security equipment, fire protection system, access control controller and broadcast terminal through the MQTT protocol to complete the security management of the smart park.

[0040] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A smart park security management system, including a data perception module, an edge computing module, an intelligent analysis module and a command execution module, characterized in that: The data sensing module includes an environmental data acquisition unit for sensing environmental data in the park, an equipment operation parameter acquisition unit for sensing equipment operation parameters in the park, and a personnel behavior data acquisition unit for sensing personnel behavior in the park; The edge computing module is deployed at each partition gateway of the park and receives data from the data perception module. The edge computing module includes a data preprocessing unit for preprocessing each perceived data and a target detection unit for identifying metadata of abnormal events; The intelligent analysis module is connected to the edge computing module through an optical fiber network, and the intelligent analysis module includes a multi-source data fusion unit for fusing data from the edge computing module, a dynamic risk assessment unit for dynamically assessing the risk level of the park, and an emergency decision unit for generating emergency plan decisions for the park; The command execution module receives the campus emergency plan decision from the intelligent analysis module. The command execution module includes a standardized API interface for interconnecting with the campus security equipment, fire protection system, access control controller and broadcast terminal, and an emergency operation execution unit for sending emergency operation execution commands to the campus security equipment, fire protection system, access control controller and broadcast terminal.

2. The smart park security management system according to claim 1, characterized in that: The environmental data acquisition unit includes a temperature and humidity sensor, a smoke concentration sensor, a surveillance camera and a millimeter wave radar; the equipment operation parameter acquisition unit includes a power load sensor, a pressure sensor and a voiceprint recognition sensor; and the personnel behavior data acquisition unit is a personnel badge with an integrated UWB positioning module.

3. The smart park security management system according to claim 1 is characterized by: The dynamic risk assessment unit is based on a deep reinforcement learning framework. The input data includes historical accident data sets, real-time sensor data streams, equipment operation and maintenance logs, and external environmental parameters. The risk assessment strategy is iteratively optimized through the Q-learning algorithm to build a dynamic risk assessment model. The risk probability calculation formula is defined as: P risk =α·S env +β·S device +γ·S human Among them, S env is the environmental risk factor, S device is the equipment risk factor, S human is the personnel behavior risk coefficient, α, β, and γ are dynamic weight parameters, and the output of the dynamic risk assessment model is risk level labels of different colors including blue, yellow, and red, and the corresponding confidence levels, which trigger different early warning response mechanisms.

4. The smart park security management system according to claim 3 is characterized by: During the construction of the dynamic risk assessment model, a transfer learning mechanism is introduced for training. The specific steps are: in the pre-training stage, a public safety accident database is used to generate a basic risk feature library; in the online learning stage, local data of the edge nodes of the park are aggregated through federated learning technology to update the model parameters.

5. The smart park security management system according to claim 1, characterized in that: The linkage logic between the command execution module and the intelligent analysis module includes: when the warning level is red, the power supply to the related areas in the park is automatically cut off, the fire sprinkler system is started, and the evacuation instructions are broadcasted through the radio; when the warning level is yellow, the alarm information is pushed to the mobile terminal of the security personnel, and the access control system of the relevant areas in the park is locked; when the warning level is blue, the potential risk points are marked on the electronic map of the park, prompting manual review.

6. The smart park security management system according to claim 1, characterized in that: The target detection unit is a target detection model improved based on the MobileNetV3 architecture, which is used to perform the following processing on the video stream data: real-time identification of flames, smoke, abnormal gathering of people or intrusion into restricted areas, filtering static background interference through frame difference method, extracting dynamic abnormal area coordinates and timestamps, and uploading abnormal event metadata to the intelligent analysis module, and storing the original video data locally.

7. The smart park security management system according to claim 1, characterized in that: The multi-source data fusion unit uses a synchronization method for video streams, sensor data, and device logs based on timestamp alignment, uses knowledge graph technology to establish an entity relationship network of equipment, regions, and personnel, labels associated risk transmission paths, and performs Bayesian probabilistic arbitration on conflicting data.

8. The smart park security management system according to claim 1, characterized in that: The data preprocessing unit has a built-in lightweight AI model and data filtering algorithm to perform localized preprocessing on the original data obtained by the data perception module, extract abnormal event features and compress redundant data.

Images