Cloud edge end collaborative intelligent security system and method based on decision-perception-execution closed loop

Abstract

The present application relates to a cloud edge end collaborative intelligent security system and method based on a decision-aware-execution closed loop, and a video monitoring processing device, which relates to the field of image communication.The present application comprises: a data-aware node arranged in the cloud, used for receiving field-aware data of a current engineering construction security area at each past time segment before the current time; a distributed edge computing node set arranged in the edge, used for intelligently predicting the type, occurrence location and occurrence time of future security risks in the current engineering construction security area.The present application aims at the technical problems of incomplete video monitoring processing prediction data and low-efficiency architecture calculation of the existing construction security area, and through the design of an intelligent security prediction model, the introduction of multi-source heterogeneous data, and the synchronous prediction of the type, occurrence location and occurrence time of future security risks on the efficient computing architecture of the "cloud-edge-end" three-layer collaboration, the above technical problems are solved.

Description

Technical Field

[0001] The video surveillance processing device proposed in this invention relates to the field of image communication, and in particular to a cloud-edge-device collaborative intelligent security system and method based on a decision-perception-execution closed loop. Background Technology

[0002] Construction security areas of various projects are important application sites for video surveillance processing equipment. By installing video surveillance processing equipment in these areas for real-time monitoring, the acquired real-time monitoring images are transmitted via wired or wireless transmission to local or remote image analysis devices for analysis. This allows for the identification of different types of anomalies or malfunctions in the construction security area, providing crucial reference data for construction security management departments to formulate corresponding response measures. This helps prevent the further spread of anomalies or malfunctions and ensures the safety of construction progress, personnel, and equipment.

[0003] As an example, Chinese invention patent publication CN113506416A proposes a method and system for early warning of engineering anomalies based on intelligent visual analysis. The system includes a client platform, a data transmission platform, and an intelligent visual analysis platform. The client remotely accesses the intelligent visual analysis platform locally via a wireless LAN or broadband connection. The data transmission platform facilitates data transmission between the client and the intelligent visual analysis platform. The intelligent visual analysis platform includes a monitoring camera, a server, and a monitor. The camera captures on-site monitoring video of the water conservancy project as input to the system. The server processes the images in the video data, including target recognition, human posture detection, and abnormal behavior determination. The processing results are displayed on the monitor in real time. This invention effectively detects and warns of potential safety accidents that may occur to workers at water conservancy project sites, achieving the effect of preventing safety accidents.

[0004] As an example, Chinese invention patent publication CN116560259A discloses a monitoring device and system, belonging to the field of intelligent monitoring and safety engineering. The monitoring device includes: a power management module for acquiring external power signals and safely processing them to provide the processed power signals to multiple modules; the power management module has multiple power supply lines to address the power consumption of the modules; an image acquisition module for acquiring image information from various locations at the construction site; a data processing module for processing the image information acquired by the image acquisition module to obtain corresponding image data; and a remote communication module for uploading the image data processed by the data processing module to a backend server, enabling the backend server to display the construction site in real time based on the image data and identify any abnormalities at the construction site. The monitoring device of this invention meets explosion-proof requirements, adapts to complex environments, and ensures the safety of the construction site.

[0005] Therefore, it is evident that in existing technologies, when analyzing images transmitted from the construction security area, the analysis only identifies different types of anomalies or malfunctions that have already occurred in the construction security area, and cannot predict future anomalies or malfunctions. More importantly, it cannot reliably predict the specific location and time of future anomalies or malfunctions, making it difficult to provide a spatiotemporal reference for subsequent precise handling. In addition, existing technologies also suffer from technical problems such as insufficient data for analysis, the inability of the analysis model to adaptively optimize according to different construction security areas, and the lack of security areas. Summary of the Invention

[0006] To address the technical problems in existing technologies, this invention provides a cloud-edge-device collaborative intelligent security system and method based on a decision-perception-execution closed loop. It constructs different intelligent security prediction models for different construction security areas, introduces multi-source heterogeneous data such as visual, environmental, and access control data, and reliably and synchronously predicts the type, specific location, and specific time of future anomalies or faults. Simultaneously, it establishes a complete "decision-perception-prediction-execution" closed loop, connecting the entire process from strategy configuration, data perception, intelligent prediction to execution-driven processes, forming a self-consistent intelligent security ecosystem. This complete closed loop is built on a highly efficient computing architecture of "cloud-edge-device" three-layer collaboration, fully leveraging the powerful computing capabilities of the cloud, the low-latency processing advantages of the edge, and the flexible deployment characteristics of terminal execution units to form a hierarchical collaborative system.

[0007] According to one aspect of the present invention, a cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop is provided, the system comprising: The decision configuration node, located in the cloud, is used to configure various response and handling strategies for different security risks. The data sensing node, set up in the cloud, is used to receive the on-site sensing data corresponding to each past time segment of the current construction security area before the current moment. The on-site sensing data corresponding to each time segment is the visual sensing content, environmental sensing content, and access control sensing content of the current construction security area within the time segment. A distributed edge computing node set, located at the edge, is connected to decision configuration nodes and data perception nodes respectively. It is used to use the intelligent security prediction model corresponding to the current construction security area. Based on the occupation duration of each time segment, the on-site perception data corresponding to each past time segment of the current construction security area before the current time, the three-dimensional coordinate values ​​of the overhead imaging mechanism directly above the center of the current construction security area, the area, height difference, number of full personnel and number of designated parking spaces of the current construction security area, it intelligently predicts the security hazards that will occur in the current construction security area within the current time segment, as well as the three-dimensional point cloud data of the location of the occurrence and the time of occurrence. The current time segment starts from the current time. The execution drive node is set up within the current construction security area and connected to the distributed edge computing node set. It is used to drive each terminal execution unit within the current construction security area to perform its corresponding response and handling actions based on intelligent prediction results and various response and handling strategies corresponding to various security hazards. Among them, the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks.

[0008] According to another aspect of the present invention, a cloud-edge-device collaborative intelligent security method based on a decision-perception-execution closed loop is provided, the method comprising: Configure various response and handling strategies for different security risks in the cloud; The system receives on-site perception data from the cloud for each past time segment of the current construction security area before the current moment. The on-site perception data for each time segment includes the visual perception content, environmental perception content, and access control perception content of the current construction security area within that time segment. At the edge, the intelligent security prediction model corresponding to the current construction security area is used. Based on the occupation duration of each time segment, the on-site perception data corresponding to each past time segment before the current construction security area, the three-dimensional coordinate values ​​of the overhead imaging mechanism directly above the center of the current construction security area, the area, height difference, number of full personnel and number of designated parking spaces of the current construction security area, the intelligent prediction of the security hazards that will occur in the current construction security area in the current time segment, as well as the three-dimensional point cloud data of the location of occurrence and the time of occurrence. The current time segment starts from the current moment. Based on intelligent prediction results and various response and handling strategies corresponding to different security risks within the current construction security area, each terminal execution unit within the current construction security area is driven to perform its respective response and handling actions. Among them, the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks.

[0009] Therefore, the technical effects of the present invention are as follows: 1. Realize the transformation of security from "passive response" to "proactive prevention": Predict dangerous events in advance through intelligent prediction models, move the security checkpoint forward, and change post-event handling to pre-event prevention. In particular, it provides accurate spatial-temporal joint prediction capabilities: It not only predicts "what will happen", but also accurately predicts "where it will happen" (3D point cloud coordinates) and "when it will happen", providing a spatiotemporal reference for accurate handling. 2. Based on the fusion analysis of multi-source heterogeneous data such as vision, environment, and access control, it comprehensively and accurately identifies various potential risks, overcoming the problem of insufficient information from a single perception dimension; 3. Achieve scene-adaptive optimization of the prediction model: enable the model complexity to be dynamically adjusted according to physical characteristics such as the height difference of the construction area, so as to avoid over-computation while ensuring prediction accuracy and achieve a balance between accuracy and efficiency. 4. Establish a complete closed loop of "decision-perception-prediction-execution": Connect the entire process from strategy configuration, data perception, intelligent prediction to execution-driven, forming a self-consistent intelligent security ecosystem. At the same time, the complete closed loop is built on a highly efficient computing architecture of "cloud-edge-device" three-layer collaboration, which can give full play to the powerful computing capabilities of the cloud, the low-latency processing advantages of the edge, and the flexible deployment characteristics of the terminal execution unit, forming a well-defined collaborative system. Attached Figure Description

[0010] Figure 1 Technical process diagram.

[0011] Figure 2-6 System structure (specific implementation methods 1-5).

[0012] Figure 7 Method flow (specific implementation method 6). Detailed Implementation

[0013] like Figure 1 The diagram illustrates the technical flow of a cloud-edge-device collaborative intelligent security system and method based on a decision-perception-execution closed loop, according to the present invention. The video surveillance processing device proposed in this invention relates to the field of image communication.

[0014] The specific technical process of this invention is as follows: Technical Process A: Configure various response and handling strategies for different security risks in the cloud; More specifically, various security hazards include unauthorized entry by personnel, unauthorized vehicle intrusion, theft and vandalism, fire and explosion, construction machinery injuries, falls from heights and being struck by objects, and toxic gas pollution. Technical Process B: Design customized intelligent security prediction models with different structures for different construction security areas of different projects in the cloud. For example, a customized intelligent security prediction model with a specific structure was designed for the current construction security area of ​​the project. Figure 1 As shown; More specifically, the customized structure of the intelligent security prediction model corresponding to the current engineering construction security area is mainly reflected in the following aspects: Firstly: The intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks; Secondly: In the intelligent security prediction model corresponding to the current construction security area, the preset number follows the numerical trend of the height difference of the current construction security area. As an example, the height difference of the current construction security area is 100, and the preset number is 2; the height difference of the current construction security area is 150, and the preset number is 3; the height difference of the current construction security area is 200, and the preset number is 4; the height difference of the current construction security area is 300, and the preset number is 6, and so on. Thirdly: The intelligent security prediction models corresponding to the current construction security area use the same deep neural network structure, which includes a single input layer, a single output layer and multiple hidden layers between the single input layer and the single output layer. The number of hidden layers is positively correlated with the number of full personnel in the current construction security area, and each hidden layer uses the Maxout function as the activation function. As an example, the current construction security area has a full capacity of 600 people and 3 hidden layers; the current construction security area has a full capacity of 800 people and 4 hidden layers; the current construction security area has a full capacity of 1000 people and 5 hidden layers; the current construction security area has a full capacity of 1200 people and 6 hidden layers, and so on. Fourthly: The number of times the intelligent security prediction model is trained in the current construction security area is directly proportional to the area occupied by the current construction security area; Fifthly: In each training session, the known security hazards that occurred in the current construction security area within a certain past time segment, along with the 3D point cloud data of their locations and the times of occurrence, are used as the output data of the model. The occupancy duration of each time segment, the on-site perception data corresponding to each past time segment before the current construction security area, the 3D coordinates of the overhead imaging mechanism directly above the center of the current construction security area, the area, height difference, number of personnel at full capacity, and number of designated parking spaces of the current construction security area are used as the input data of the model to complete this training. In this way, the customized structural design in the above five aspects ensures the effectiveness and stability of the intelligent security prediction model corresponding to the current engineering construction security area. Technical Process C: To simultaneously predict the security hazards that will occur in the current construction security area within the current time segment, as well as the 3D point cloud data of their locations and the time of occurrence, multi-source heterogeneous data is introduced; More specifically, the multi-source heterogeneous data includes the duration of each time segment, the on-site perception data corresponding to each past time segment before the current construction security area, the three-dimensional coordinate values ​​of the overhead imaging mechanism directly above the center of the current construction security area, the area, height difference, number of full personnel, and number of designated parking spaces of the current construction security area, and the on-site perception data corresponding to each time segment are the visual perception content, environmental perception content, and access control perception content of the current construction security area within the time segment. like Figure 1 As shown, the multi-source heterogeneous data includes various on-site perception data, various regional association information of the current construction security area, the occupancy duration of each time segment, and three-dimensional coordinate values. Among them, the regional association information of the current construction security area includes the area occupied, height difference, number of full personnel, and number of designated parking spaces of the current construction security area. In more detail, the visual perception content of the current construction security area in each time segment is the visual filtering information corresponding to each timestamp at a uniform interval in the time segment; the environmental perception content of the current construction security area in each time segment is the environmental perception information corresponding to each timestamp at a uniform interval in the time segment; and the access control perception content of the current construction security area in each time segment is the access control perception information corresponding to each timestamp at a uniform interval in the time segment. In more detail, the visual filtering information corresponding to each timestamp of the current construction security area is the depth of field value, grayscale gradient value, red-green component value, black-and-white component value, yellow-blue component value, horizontal coordinate value, and vertical coordinate value of each edge pixel in the central overhead view of the current construction security area at the timestamp. The central overhead view of the current construction security area at the timestamp is obtained by an overhead imaging mechanism located directly above the center of the current construction security area at the timestamp. In more detail, the environmental perception information corresponding to each timestamp in the current construction security area consists of the output values ​​of various sensors installed in the current construction security area at the timestamp, as well as the positioning data of various sensors in the current construction security area at the timestamp. The various sensors in the current construction security area include vibration sensors, smoke sensors, noise sensors, gas concentration sensors, temperature sensors, and humidity sensors. To be even more detailed, the access control sensing information corresponding to each timestamp in the current construction security area is the difference in the number of people entering and exiting and the difference in the number of vehicles entering and exiting, as counted by the access control mechanism of the current construction security area at the timestamp, and the duration of each time segment is equal; In this way, through the targeted screening of the above-mentioned multi-source heterogeneous data and the specific design of multi-layer data structures, the effectiveness and stability of the intelligent prediction results of the intelligent security prediction model corresponding to the current engineering construction security area are further guaranteed. Technical Process D: Using a distributed set of edge computing nodes set up at the edge, and employing the intelligent security prediction model corresponding to the current construction security area with the customized structural design in Technical Process B, based on the multi-source heterogeneous data specifically selected in Technical Process C, synchronously and intelligently predicts the security hazards that will occur in the current construction security area within the current time segment, as well as the 3D point cloud data of their locations and the time of occurrence. Figure 1 As shown, the current time segment starts from the current time and is actually a future time segment; In this way, it is not only possible to predict the types of security risks that will occur in the current construction security area within the current time segment, but also to make simultaneous and targeted predictions of the specific location and time of occurrence of the security risks. Technical Process E: Based on the intelligent prediction results of Technical Process D and the various response and handling strategies corresponding to the various security hazards configured in Technical Process A, the terminal execution units in the current engineering construction security area are driven to perform their respective response and handling actions, thus completing the terminal device actions of the cloud-edge-end. More specifically, based on the intelligent prediction results and the various response and handling strategies corresponding to various security hazards, each terminal execution unit in the current construction security area is driven to perform its respective response and handling actions for the security hazards that will occur in the current time segment before the time of occurrence of the security hazard arrives. The response and handling strategy corresponding to each security hazard includes the respective response and handling actions that each terminal execution unit in the current construction security area needs to perform for that security hazard. In more detail, each terminal execution unit needs to perform corresponding response actions for each type of security hazard, including one or more of the following: adjustment of associated lighting fixtures, disconnection of associated equipment, activation of associated equipment, control of associated access control switches, and linkage of associated alarm equipment. Furthermore, the associated lighting fixtures, associated equipment, associated access control, and associated alarm equipment are determined based on the three-dimensional point cloud data of the location where each type of security hazard occurs. Therefore, this invention, through the coordinated operation of the aforementioned multiple technical processes, completes the deployment of a cloud-edge-device collaborative intelligent security architecture based on a decision-making-perception-execution closed loop. This improves the architecture's computational efficiency, fully leverages the powerful computing capabilities of the cloud, the low-latency processing advantages of the edge, and the flexible deployment characteristics of the terminal execution units. Most importantly, by designing intelligent security prediction models with different customized structures for the construction security areas of different projects, and based on multi-source heterogeneous data with targeted selection of customized data structures, it achieves synchronous intelligent prediction of the type, location, and time of future security hazards in the corresponding project's construction security area. It not only predicts the type of future security hazards but also simultaneously predicts the location and time of future security hazards, thereby providing more comprehensive reference data for subsequent security department responses and improving the pertinence and effectiveness of response measures.

[0015] The key points of this invention are: the deployment of a cloud-edge-device collaborative intelligent security architecture based on a decision-perception-execution closed loop; the design of intelligent security prediction models with different customized structures for the construction security areas of different projects; the customized design of data structures for multi-source heterogeneous data; and the targeted screening of multi-source heterogeneous data.

[0016] The cloud-edge-device collaborative intelligent security system and method based on a decision-perception-execution closed loop of the present invention will be described in detail below with reference to specific implementation methods. Detailed Implementation Method 1

[0018] Figure 2 This is an internal structure diagram of a cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop, as shown in Specific Embodiment 1 of the present invention.

[0019] like Figure 2 As shown, the cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop includes the following components: The decision configuration node, located in the cloud, is used to configure various response and handling strategies for different security risks. More specifically, various security hazards include unauthorized entry by personnel, unauthorized vehicle intrusion, theft and vandalism, fire and explosion, construction machinery injury, falls from heights and being struck by objects, and toxic gas pollution. Different security hazards correspond to different hazard codes, and all hazard codes can be encoded using a binary code mode. The data sensing node, set up in the cloud, is used to receive the on-site sensing data corresponding to each past time segment of the current construction security area before the current moment. The on-site sensing data corresponding to each time segment is the visual sensing content, environmental sensing content, and access control sensing content of the current construction security area within the time segment. As an example, the current time is 5:00 PM. The previous time segments before the current time are 4:00 PM to 4:10 PM, 4:10 PM to 4:20 PM, 4:20 PM to 4:30 PM, 4:30 PM to 4:40 PM, 4:40 PM to 4:50 PM, and 4:50 PM to 5:00 PM, for a total of 6 previous time segments. A distributed edge computing node set, located at the edge, is connected to decision configuration nodes and data perception nodes respectively. It is used to use the intelligent security prediction model corresponding to the current construction security area. Based on the occupation duration of each time segment, the on-site perception data corresponding to each past time segment of the current construction security area before the current time, the three-dimensional coordinate values ​​of the overhead imaging mechanism directly above the center of the current construction security area, the area, height difference, number of full personnel and number of designated parking spaces of the current construction security area, it intelligently predicts the security hazards that will occur in the current construction security area within the current time segment, as well as the three-dimensional point cloud data of the location of the occurrence and the time of occurrence. The current time segment starts from the current time. As an example, the current time is 5:00 PM, and the current time segment is from 5:00 PM to 5:10 PM. That is, the current time segment is a future time segment, and each time segment occupies a duration of 10 minutes. As an example, the current construction security area can cover an area of ​​50,000 square meters, and the height difference of the current construction security area is 100 meters. The highest point of the current construction security area is 70 meters above sea level, and the lowest point of the current construction security area is -30 meters above sea level. The execution drive node is set up within the current construction security area and connected to the distributed edge computing node set. It is used to drive each terminal execution unit within the current construction security area to perform its corresponding response and handling actions based on intelligent prediction results and various response and handling strategies corresponding to various security hazards. As can be seen, each terminal execution unit here performs its corresponding response and handling actions, thus completing the actions of the terminal devices in the cloud-edge-end; More specifically, cloud-edge-device is a distributed architecture in which cloud computing, edge computing, and terminal devices work together. It aims to provide efficient business support through layered data processing. This architecture moves computing power from the center to the edge, balancing processing efficiency, real-time performance, and cost. It is the core underlying architecture of the physical AI era. Among them, the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks; More specifically, the intelligent security prediction model corresponding to the current engineering construction security area is a complex structure formed by multiple deep neural networks with the same structure connected in series. Depending on the number of superimposed layers, that is, depending on the value of the preset number, the final intelligent security prediction model will be different, which reflects a customized structural design of different intelligent security prediction models corresponding to different engineering construction security areas. Among them, the various response and handling strategies corresponding to various security hazards, based on the intelligent prediction results, drive each terminal execution unit in the current construction security area to perform its corresponding response and handling actions. This includes: based on the intelligent prediction results and various response and handling strategies corresponding to various security hazards, drive each terminal execution unit in the current construction security area to perform its corresponding response and handling actions for the security hazards that will occur in the current time segment before the time of occurrence of the security hazard arrives. Each response and handling strategy for a security hazard includes the corresponding response and handling actions that each terminal execution unit in the current construction security area needs to perform for that security hazard. The response and handling strategy for each type of security hazard includes the corresponding response and handling actions to be performed by each terminal execution unit in the current construction security area for that type of security hazard. These actions include one or more of the following: adjustment of associated lighting fixtures, disconnection of associated equipment, activation of associated equipment, control of associated access control switches, and linkage of associated alarm devices. Furthermore, the associated lighting fixtures, associated equipment, associated access control, and associated alarm devices are determined based on the three-dimensional point cloud data of the location where each type of security hazard occurs. More specifically, different response and handling actions correspond to different action type code values, and all action type code values ​​can be encoded in binary value mode; Among them, various security hazards include unauthorized personnel entry, unauthorized vehicle intrusion, theft and vandalism, fire and explosion, construction machinery injury, falls from height and being struck by objects, and toxic gas pollution. The height difference of the current construction security area is the difference between the altitude values ​​of the highest and lowest positions in the current construction security area, and the duration of each time segment is equal. Among them, the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks, which includes: the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks, and the preset number follows the numerical change trend of the height difference of the current construction security area. As an example, the preset number follows the numerical trend of the height difference of the current construction security area, including: the current construction security area has a height difference of 100, the preset number is 2; the current construction security area has a height difference of 150, the preset number is 3; the current construction security area has a height difference of 200, the preset number is 4; the current construction security area has a height difference of 300, the preset number is 6, and so on. Among them, the intelligent security prediction model corresponding to the current construction security area uses the same deep neural network structure, which includes a single input layer, a single output layer and multiple hidden layers located between the single input layer and the single output layer. The number of hidden layers is positively correlated with the number of full personnel in the current construction security area, and each hidden layer uses the Maxout function as the activation function. It can be seen that the number of hidden layers leads to different final intelligent security prediction models, reflecting another customized structural design of different intelligent security prediction models corresponding to different security areas in different engineering construction. As an example, the positive correlation between the number of hidden layers and the number of personnel at full capacity in the current construction security area includes: 600 personnel at full capacity in the current construction security area, 3 hidden layers; 800 personnel at full capacity in the current construction security area, 4 hidden layers; 1000 personnel at full capacity in the current construction security area, 5 hidden layers; 1200 personnel at full capacity in the current construction security area, 6 hidden layers, and so on. Furthermore, the number of times the intelligent security prediction model corresponding to the current construction security area is trained is proportional to the area of ​​the current construction security area. In each training session, the known security hazards that occurred in the current construction security area within a certain past time segment, as well as the 3D point cloud data of their locations and the time of occurrence, are used as the output data of the model. The occupancy duration of each time segment, the on-site perception data corresponding to each past time segment before the certain past time segment, the 3D coordinate values ​​of the overhead imaging mechanism directly above the center of the current construction security area, the area of ​​the current construction security area, the height difference, the number of full personnel, and the number of designated parking spaces are used as the input data of the model to complete this training. More specifically, programmable logic devices can be used to complete the testing and simulation of each training iteration of the intelligent security prediction model corresponding to the current engineering construction security area. Detailed Implementation Method 2 Figure 3 This is an internal structure diagram of a cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop, as shown in Embodiment 2 of the present invention.

[0021] like Figure 3 As shown, compared to Figure 2 The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop also includes: The model building nodes are set up in the cloud and connected to a set of distributed edge computing nodes. They are used to build a smart security prediction model corresponding to the current construction security area. The model consists of a preset number of superimposed deep neural networks. Among them, the model building node also has a built-in storage unit, which is used to store the model parameters of the intelligent security prediction model corresponding to the current engineering construction security area; Among them, the model building node is also connected to the distributed edge computing node set, which is used to send the intelligent security prediction model corresponding to the current engineering construction security area to the distributed edge computing node set; More specifically, sending the intelligent security prediction model corresponding to the current construction security area to the distributed edge computing node set includes: sending each set of model parameters of the intelligent security prediction model corresponding to the current construction security area to the distributed edge computing node set to complete the model sending of the intelligent security prediction model corresponding to the current construction security area. Detailed Implementation Method 3 Figure 4 This is a diagram illustrating the internal structure of a cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop, according to a specific embodiment 3 of the present invention.

[0023] like Figure 4 As shown, compared to Figure 2 The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop also includes: The event storage node, located in the cloud and connected to a distributed edge computing node set, is used to receive and store the three-dimensional point cloud data of the security hazards that will occur in the current construction security area within the current time segment, as well as the location and time of occurrence of the events. As an example, the event storage node, set up in the cloud and connected to a distributed edge computing node set, is used to receive and store the three-dimensional point cloud data of the security hazards that will occur in the current construction security area within the current time segment, as well as the location and time of occurrence of the events. The event storage node can be a big data storage node, a cloud computing storage node, or a blockchain storage node. Among them, the event storage nodes use different physical storage addresses to store the 3D point cloud data of the security hazards that will occur in the current construction security area within each time segment, as well as the location and time of occurrence of the events. Detailed Implementation Method 4 Figure 5 This is an internal structure diagram of a cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop, as shown in Embodiment 4 of the present invention.

[0025] like Figure 5 As shown, compared to Figure 2 The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop also includes: Emergency transmission nodes, located in the cloud and connected to a distributed edge computing node set, are used to receive 3D point cloud data of the security hazards that will occur in the current construction security area within the current time segment, as well as the location and time of occurrence of the hazards. Among them, the emergency transmission node is also used to wirelessly transmit the three-dimensional point cloud data of the security hazards that will occur in the current construction security area within the current time segment, as well as the location and time of occurrence, to the handheld terminal of the security management personnel closest to the location of the security hazards that will occur in the current time segment within the current construction security area via a wireless communication link. As an example, the three-dimensional point cloud data of the security hazards that will occur in the current construction security area within the current time segment, as well as the location and time of occurrence, can be wirelessly transmitted via frequency division duplex communication mechanism to the handheld terminal of the security management personnel closest to the location of the security hazard that will occur within the current construction security area within the current time segment. Detailed Implementation Method 5 Figure 6 This is a diagram illustrating the internal structure of a cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop, according to a specific embodiment 5 of the present invention.

[0027] like Figure 6 As shown, compared to Figure 2 The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop also includes: The nodes are set up in segments, located in the cloud and connected to the data sensing nodes and the distributed edge computing node set respectively, to set the duration of each time segment. Among them, the segmented setting node is set in the cloud and connected to the data sensing node and the distributed edge computing node set respectively. It is used to set the occupation duration of each time segment, including: the occupation duration of each time segment is positively correlated with the number of full personnel in the current construction security area; As an example, the current construction security area has a full capacity of 500 people, and each time segment is set to occupy 5 minutes; the current construction security area has a full capacity of 700 people, and each time segment is set to occupy 7 minutes; the current construction security area has a full capacity of 1000 people, and each time segment is set to occupy 10 minutes, and so on.

[0028] Next, we will continue to describe the various specific embodiments of the present invention.

[0029] Optionally, within the above-described specific embodiments, in the cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop: The visual perception content of the current construction security area in each time segment is the visual filtering information corresponding to each time stamp at a uniform interval in the time segment. The environmental perception content of the current construction security area in each time segment is the environmental perception information corresponding to each time stamp at a uniform interval in the time segment. The access control perception content of the current construction security area in each time segment is the access control perception information corresponding to each time stamp at a uniform interval in the time segment. As an example, when each time segment is 10 minutes, the interval between any two adjacent timestamps is 2 minutes in the evenly spaced timestamps; when each time segment is 20 minutes, the interval between any two adjacent timestamps is 4 minutes in the evenly spaced timestamps; and when each time segment is 30 minutes, the interval between any two adjacent timestamps is 6 minutes in the evenly spaced timestamps. Among them, the visual filtering information corresponding to each timestamp of the current construction security area is the depth value, grayscale gradient value, red-green component value, black-and-white component value, yellow-blue component value, horizontal coordinate value, and vertical coordinate value of each edge pixel in the central overhead view of the current construction security area at the timestamp. The central overhead view of the current construction security area at the timestamp is obtained by an overhead imaging mechanism located directly above the center of the current construction security area at the timestamp. As an example, the overhead imaging mechanism located directly above the center of the current construction security area can be a fixed-position overhead imaging mechanism or a mobile-position overhead imaging mechanism, such as the aerial photography mechanism of a drone. More specifically, in the overhead view at the timestamp, the value of any of the red-green, black-and-white, and yellow-and-blue components of each pixel is between 0 and 255. Among them, the environmental perception information corresponding to each timestamp of the current construction security area consists of the output values ​​of various sensors installed in the current construction security area at the timestamp and the positioning data of various sensors in the current construction security area at the timestamp. The various sensors in the current construction security area include vibration sensors, smoke sensors, noise sensors, gas concentration sensors, temperature sensors and humidity sensors. Among them, the access control perception information corresponding to each timestamp in the current construction security area is the difference in the number of people entering and exiting and the difference in the number of vehicles entering and exiting, as counted by the access control agency of the current construction security area at the timestamp, and the duration of each time segment is equal.

[0030] And, optionally, in the cloud-edge-device collaborative intelligent security system based on the decision-perception-execution closed loop described above: Each pixel in the central overhead shot has an L component value (red and green component value), an A component value (black and white component value), and a B component value (yellow and blue component value) in the LAB color space. In the central overhead shot, the grayscale gradient value of each pixel is the standard deviation of the set of grayscale values ​​formed by the pixel's grayscale value and the grayscale values ​​of its neighboring pixels. More specifically, programmable logic devices can be selected to perform numerical calculations of the grayscale gradient values ​​for each pixel; Among them, the number of past time segments before the current moment in the selected current construction security area is directly proportional to the area occupied by the current construction security area; As an example, the current construction security area covers an area of ​​50,000 square meters, and the number of past time segments before the current moment for the selected current construction security area is 6; the current construction security area covers an area of ​​100,000 square meters, and the number of past time segments before the current moment for the selected current construction security area is 12; the current construction security area covers an area of ​​150,000 square meters, and the number of past time segments before the current moment for the selected current construction security area is 18, and so on. Detailed Implementation Method 6 Figure 7 The flowchart illustrates the steps of a cloud-edge-device collaborative intelligent security method based on a decision-perception-execution closed loop, as shown in Embodiment 6 of the present invention.

[0032] like Figure 7 As shown, the cloud-edge-device collaborative intelligent security method based on a decision-perception-execution closed loop includes the following steps: Configure various response and handling strategies for different security risks in the cloud; More specifically, various security hazards include unauthorized entry by personnel, unauthorized vehicle intrusion, theft and vandalism, fire and explosion, construction machinery injury, falls from heights and being struck by objects, and toxic gas pollution. Different security hazards correspond to different hazard codes, and all hazard codes can be encoded using a binary code mode. The system receives on-site perception data from the cloud for each past time segment of the current construction security area before the current moment. The on-site perception data for each time segment includes the visual perception content, environmental perception content, and access control perception content of the current construction security area within that time segment. As an example, the current time is 5:00 PM. The previous time segments before the current time are 4:00 PM to 4:10 PM, 4:10 PM to 4:20 PM, 4:20 PM to 4:30 PM, 4:30 PM to 4:40 PM, 4:40 PM to 4:50 PM, and 4:50 PM to 5:00 PM, for a total of 6 previous time segments. At the edge, the intelligent security prediction model corresponding to the current construction security area is used. Based on the occupation duration of each time segment, the on-site perception data corresponding to each past time segment before the current construction security area, the three-dimensional coordinate values ​​of the overhead imaging mechanism directly above the center of the current construction security area, the area, height difference, number of full personnel and number of designated parking spaces of the current construction security area, the intelligent prediction of the security hazards that will occur in the current construction security area in the current time segment, as well as the three-dimensional point cloud data of the location of occurrence and the time of occurrence. The current time segment starts from the current moment. As an example, the current time is 5:00 PM, and the current time segment is from 5:00 PM to 5:10 PM. That is, the current time segment is a future time segment, and each time segment occupies a duration of 10 minutes. As an example, the current construction security area can cover an area of ​​50,000 square meters, and the height difference of the current construction security area is 100 meters. The highest point of the current construction security area is 70 meters above sea level, and the lowest point of the current construction security area is -30 meters above sea level. Based on intelligent prediction results and various response and handling strategies corresponding to different security risks within the current construction security area, each terminal execution unit within the current construction security area is driven to perform its respective response and handling actions. As can be seen, each terminal execution unit here performs its corresponding response and handling actions, thus completing the actions of the terminal devices in the cloud-edge-end; More specifically, cloud-edge-device is a distributed architecture in which cloud computing, edge computing, and terminal devices work together. It aims to provide efficient business support through layered data processing. This architecture moves computing power from the center to the edge, balancing processing efficiency, real-time performance, and cost. It is the core underlying architecture of the physical AI era. Among them, the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks; More specifically, the intelligent security prediction model corresponding to the current engineering construction security area is a complex structure formed by multiple deep neural networks with the same structure connected in series. Depending on the number of superimposed layers, that is, depending on the value of the preset number, the final intelligent security prediction model will be different, which reflects a customized structural design of different intelligent security prediction models corresponding to different engineering construction security areas. Among them, the various response and handling strategies corresponding to various security hazards, based on the intelligent prediction results, drive each terminal execution unit in the current construction security area to perform its corresponding response and handling actions. This includes: based on the intelligent prediction results and various response and handling strategies corresponding to various security hazards, drive each terminal execution unit in the current construction security area to perform its corresponding response and handling actions for the security hazards that will occur in the current time segment before the time of occurrence of the security hazard arrives. Each response and handling strategy for a security hazard includes the corresponding response and handling actions that each terminal execution unit in the current construction security area needs to perform for that security hazard. The response and handling strategy for each type of security hazard includes the corresponding response and handling actions to be performed by each terminal execution unit in the current construction security area for that type of security hazard. These actions include one or more of the following: adjustment of associated lighting fixtures, disconnection of associated equipment, activation of associated equipment, control of associated access control switches, and linkage of associated alarm devices. Furthermore, the associated lighting fixtures, associated equipment, associated access control, and associated alarm devices are determined based on the three-dimensional point cloud data of the location where each type of security hazard occurs. More specifically, different response and handling actions correspond to different action type code values, and all action type code values ​​can be encoded in binary value mode; Among them, various security hazards include unauthorized personnel entry, unauthorized vehicle intrusion, theft and vandalism, fire and explosion, construction machinery injury, falls from height and being struck by objects, and toxic gas pollution. The height difference of the current construction security area is the difference between the altitude values ​​of the highest and lowest positions in the current construction security area, and the duration of each time segment is equal. Among them, the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks, which includes: the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks, and the preset number follows the numerical change trend of the height difference of the current construction security area. As an example, the preset number follows the numerical trend of the height difference of the current construction security area, including: the current construction security area has a height difference of 100, the preset number is 2; the current construction security area has a height difference of 150, the preset number is 3; the current construction security area has a height difference of 200, the preset number is 4; the current construction security area has a height difference of 300, the preset number is 6, and so on. Among them, the intelligent security prediction model corresponding to the current construction security area uses the same deep neural network structure, which includes a single input layer, a single output layer and multiple hidden layers located between the single input layer and the single output layer. The number of hidden layers is positively correlated with the number of full personnel in the current construction security area, and each hidden layer uses the Maxout function as the activation function. It can be seen that the number of hidden layers leads to different final intelligent security prediction models, reflecting another customized structural design of different intelligent security prediction models corresponding to different security areas in different engineering construction. As an example, the positive correlation between the number of hidden layers and the number of personnel at full capacity in the current construction security area includes: 600 personnel at full capacity in the current construction security area, 3 hidden layers; 800 personnel at full capacity in the current construction security area, 4 hidden layers; 1000 personnel at full capacity in the current construction security area, 5 hidden layers; 1200 personnel at full capacity in the current construction security area, 6 hidden layers, and so on. Furthermore, the number of times the intelligent security prediction model corresponding to the current construction security area is trained is proportional to the area of ​​the current construction security area. In each training session, the known security hazards that occurred in the current construction security area within a certain past time segment, as well as the 3D point cloud data of their locations and the time of occurrence, are used as the output data of the model. The occupancy duration of each time segment, the on-site perception data corresponding to each past time segment before the certain past time segment, the 3D coordinate values ​​of the overhead imaging mechanism directly above the center of the current construction security area, the area of ​​the current construction security area, the height difference, the number of full personnel, and the number of designated parking spaces are used as the input data of the model to complete this training. More specifically, programmable logic devices can be selected to complete the testing and simulation of each training iteration of the intelligent security prediction model corresponding to the current engineering construction security area. Furthermore, in the cloud-edge-device collaborative intelligent security system and method based on the decision-perception-execution closed loop according to the present invention: The preset number follows the numerical trend of the height difference of the current construction security area, including: the height difference of the current construction security area is a dynamic value that changes over time, and the first numerical change curve represents the numerical trend of the height difference of the current construction security area, and the second numerical change curve represents the numerical trend of the preset number. The preset number follows the numerical change trend of the height difference of the current construction security area, which also includes: normalizing the curve length of the first numerical change curve and the second numerical change curve to obtain the first normalized curve and the second normalized curve respectively, and the first normalized curve and the second normalized curve completely overlap. As an example, the height difference of the current construction security area is a dynamic value that changes over time. The first numerical change curve represents the trend of the height difference of the current construction security area, and the second numerical change curve represents the trend of a preset number of numerical changes. The simulation and testing of the first and second numerical change curves can be completed using the MATLAB toolbox.

[0033] The foregoing description of exemplary embodiments of the present invention is provided for illustrative and descriptive purposes. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Many modifications and variations will obviously be apparent to those skilled in the art. Exemplary embodiments were chosen and described in order to best illustrate the principles of the invention and its practical application, thereby enabling others skilled in the art to understand the various specific embodiments and variations of the invention suitable for the contemplated particular purpose. The scope of the invention is intended to be defined by the appended claims and their equivalents.

Claims

1. A cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop, characterized in that: The system includes: a decision configuration node, located in the cloud, used to configure various response and handling strategies corresponding to various security risks; and a data sensing node, located in the cloud, used to receive on-site sensing data corresponding to each past time segment of the current construction security area before the current moment, wherein the on-site sensing data corresponding to each time segment is the visual sensing content, environmental sensing content, and access control sensing content of the current construction security area within the time segment. A distributed edge computing node set, located at the edge, connects to decision configuration nodes and data sensing nodes respectively. It uses the intelligent security prediction model corresponding to the current construction security area to intelligently predict security hazards that will occur in the current construction security area within the current time segment, along with the 3D point cloud data of the location and the time of occurrence. This prediction is based on the occupation duration of each time segment, the on-site sensing data corresponding to each past time segment before the current moment, the 3D coordinates of the overhead imaging device directly above the center of the current construction security area, the area, height difference, number of personnel at full capacity, and number of designated parking spaces of the current construction security area. The current time segment begins at the current moment. An execution drive node, located within the current construction security area and connected to the distributed edge computing node set, drives each terminal execution unit within the current construction security area to perform its corresponding response and handling actions based on the intelligent prediction results and various response and handling strategies corresponding to different security hazards. The intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks.

2. The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop as described in claim 1, characterized in that: Based on intelligent prediction results and various response and handling strategies corresponding to different security hazards, the terminal execution units within the current construction security area are driven to perform their respective response and handling actions. This includes: based on intelligent prediction results and various response and handling strategies corresponding to different security hazards, the terminal execution units within the current construction security area are driven to perform their respective response and handling actions for security hazards that will occur within the current time segment before the arrival of the expected occurrence time of the security hazard. Each response and handling strategy for a security hazard includes the respective response and handling actions that each terminal execution unit within the current construction security area needs to perform for that security hazard. The response and handling strategy for each type of security hazard includes the corresponding response and handling actions to be performed by each terminal execution unit in the current construction security area for that type of security hazard. These actions include one or more of the following: adjustment of associated lighting fixtures, disconnection of associated equipment, activation of associated equipment, control of associated access control switches, and linkage of associated alarm devices. Furthermore, the associated lighting fixtures, associated equipment, associated access control, and associated alarm devices are determined based on the three-dimensional point cloud data of the location where each type of security hazard occurs. Among them, various security hazards include unauthorized personnel entry, unauthorized vehicle intrusion, theft and vandalism, fire and explosion, construction machinery injury, falls from height and being struck by objects, and toxic gas pollution. The height difference of the current construction security area is the difference between the altitude values ​​of the highest and lowest positions within the current construction security area, and the duration of each time segment is equal.

3. The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop as described in claim 2, characterized in that: The intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks, which includes: the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks, and the preset number follows the numerical trend of the height difference of the current construction security area. Among them, the intelligent security prediction model corresponding to the current construction security area uses the same deep neural network structure, which includes a single input layer, a single output layer and multiple hidden layers located between the single input layer and the single output layer. The number of hidden layers is positively correlated with the number of full personnel in the current construction security area, and each hidden layer uses the Maxout function as the activation function. The number of times the intelligent security prediction model corresponding to the current construction security area is trained is proportional to the area of ​​the current construction security area. In each training session, the known security hazards that occurred in the current construction security area within a certain past time segment, along with the 3D point cloud data of their locations and the time of occurrence, are used as the model's output data. The occupancy duration of each time segment, the on-site perception data corresponding to each past time segment before the aforementioned past time segment, the 3D coordinates of the overhead imaging mechanism directly above the center of the current construction security area, the area of ​​the current construction security area, the height difference, the number of people at full capacity, and the number of designated parking spaces are used as the model's input data to complete this training session.

4. The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop as described in claim 3, characterized in that, The system also includes: The model building nodes are set up in the cloud and connected to a set of distributed edge computing nodes. They are used to build a smart security prediction model corresponding to the current construction security area. The model consists of a preset number of superimposed deep neural networks. Among them, the model building node also has a built-in storage unit, which is used to store the model parameters of the intelligent security prediction model corresponding to the current engineering construction security area; The model building node is also connected to the distributed edge computing node set, which is used to send the intelligent security prediction model corresponding to the current engineering construction security area to the distributed edge computing node set.

5. The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop as described in claim 3, characterized in that, The system also includes: The event storage node, located in the cloud and connected to a distributed edge computing node set, is used to receive and store the three-dimensional point cloud data of the security hazards that will occur in the current construction security area within the current time segment, as well as the location and time of occurrence of the events. Among them, the event storage nodes use different physical storage addresses to store the 3D point cloud data of the security hazards that will occur in the current construction security area within each time segment, as well as the location and time of occurrence of the events.

6. The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop as described in claim 3, characterized in that, The system also includes: Emergency transmission nodes, located in the cloud and connected to a distributed edge computing node set, are used to receive 3D point cloud data of the security hazards that will occur in the current construction security area within the current time segment, as well as the location and time of occurrence of the hazards. The emergency transmission node is also used to wirelessly transmit, via wireless communication link, the three-dimensional point cloud data of the security hazards that will occur in the current construction security area within the current time segment, as well as the location and time of occurrence, to the handheld terminal of the security management personnel closest to the location of the security hazard that will occur in the current time segment within the current construction security area.

7. The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop as described in claim 3, characterized in that, The system also includes: The nodes are set up in segments, located in the cloud and connected to the data sensing nodes and the distributed edge computing node set respectively, to set the duration of each time segment. Among them, the segmented setting nodes are set in the cloud and connected to the data sensing nodes and the distributed edge computing node set respectively. They are used to set the occupation duration of each time segment, including: the occupation duration of each time segment is positively correlated with the number of full personnel in the current construction security area.

8. The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop as described in any one of claims 3-7, characterized in that: The visual perception content of the current construction security area in each time segment is the visual filtering information corresponding to each time stamp at a uniform interval in the time segment. The environmental perception content of the current construction security area in each time segment is the environmental perception information corresponding to each time stamp at a uniform interval in the time segment. The access control perception content of the current construction security area in each time segment is the access control perception information corresponding to each time stamp at a uniform interval in the time segment. Among them, the visual filtering information corresponding to each timestamp of the current construction security area is the depth value, grayscale gradient value, red-green component value, black-and-white component value, yellow-blue component value, horizontal coordinate value, and vertical coordinate value of each edge pixel in the central overhead view of the current construction security area at the timestamp. The central overhead view of the current construction security area at the timestamp is obtained by an overhead imaging mechanism located directly above the center of the current construction security area at the timestamp. Among them, the environmental perception information corresponding to each timestamp of the current construction security area consists of the output values ​​of various sensors installed in the current construction security area at the timestamp and the positioning data of various sensors in the current construction security area at the timestamp. The various sensors in the current construction security area include vibration sensors, smoke sensors, noise sensors, gas concentration sensors, temperature sensors and humidity sensors. Among them, the access control sensing information corresponding to each timestamp in the current construction security area is the difference in the number of people entering and exiting and the difference in the number of vehicles entering and exiting, as counted by the access control agency of the current construction security area at the timestamp, and the duration of each time segment is equal.

9. The cloud-edge-device collaborative intelligent security system based on a decision-perception-execution closed loop as described in any one of claims 3-7, characterized in that: Each pixel in the central overhead shot has an L component value (red and green component value), an A component value (black and white component value), and a B component value (yellow and blue component value) in the LAB color space. In the central overhead view, the grayscale gradient value of each pixel is the standard deviation of the set of grayscale values ​​formed by the grayscale value of the pixel and the grayscale values ​​of its neighboring pixels. The number of past time segments before the current moment in the selected current construction security area is directly proportional to the area occupied by the current construction security area.

10. A cloud-edge-device collaborative intelligent security method based on a decision-perception-execution closed loop, characterized in that: The method includes: Configure various response and handling strategies for different security risks in the cloud; The system receives on-site perception data from the cloud for each past time segment of the current construction security area before the current moment. The on-site perception data for each time segment includes the visual perception content, environmental perception content, and access control perception content of the current construction security area within that time segment. At the edge, the intelligent security prediction model corresponding to the current construction security area is used. Based on the occupation duration of each time segment, the on-site perception data corresponding to each past time segment before the current construction security area, the three-dimensional coordinate values ​​of the overhead imaging mechanism directly above the center of the current construction security area, the area, height difference, number of full personnel and number of designated parking spaces of the current construction security area, the intelligent prediction of the security hazards that will occur in the current construction security area in the current time segment, as well as the three-dimensional point cloud data of the location of occurrence and the time of occurrence. The current time segment starts from the current moment. Based on intelligent prediction results and various response and handling strategies corresponding to different security risks within the current construction security area, each terminal execution unit within the current construction security area is driven to perform its respective response and handling actions. Among them, the intelligent security prediction model corresponding to the current construction security area is a preset number of superimposed deep neural networks.

Images