Digital cell management method based on multi-platform collaboration

Abstract

The invention discloses a digital cell management method based on multi-platform collaboration, and belongs to the field of digital cell management, and the management method comprises the following specific steps: I, collecting and integrating cell internal and external data from different data sources; according to the method, the working path of each worker can be intelligently and dynamically adjusted according to the real-time community condition, the redundancy and repetition on the path are reduced, the task completion efficiency is improved, the excessive workload of individual personnel is effectively reduced, and the actual utilization rate of personnel and equipment is optimized, so that the overall efficiency is improved; more diversified path planning and resource scheduling targets are achieved, high-risk scenes can be rapidly identified and accurately responded, resource waste is reduced, cell management is more flexible and real-time, platform data play a synergistic role in emergency decision making, and the emergency decision making efficiency is improved. The problem of response lag or resource dispersion caused by independent operation of each platform is avoided, and repeated investment and loss caused by improper resource allocation are reduced.

Description

Technical Field

[0001] The present invention relates to the field of digital cell management, and in particular to a digital cell management method based on multi-platform collaboration. Background Art

[0002] In the context of the continuous development of digitalization and intelligence, the management mode of communities, as an important component of cities, is undergoing a profound transformation from traditional manual management to digital intelligent management. Digital community management has gradually become a key means to improve community safety, optimize resource allocation and improve service efficiency. With the rapid development of smart cities and Internet of Things (IoT) technologies, traditional community management usually relies on manual inspections and static monitoring, which makes it difficult to achieve real-time, comprehensive and accurate management and early warning. Modern digital community management methods integrate multiple data sources and realize comprehensive perception and dynamic management of various types of information such as community environment, equipment, and personnel flow through technical means such as data analysis, deep learning, and knowledge graphs. However, due to the variety of data types within the community (such as structured data, video surveillance images, personnel entry and exit records, etc.), and the different data standards and transmission methods between platforms, traditional community management is difficult to achieve efficient collaboration; therefore, it is particularly important to invent a digital community management method based on multi-platform collaboration.

[0003] The existing digital community management methods are unable to dynamically adjust the work paths of each staff member according to the real-time community situation. There is a high degree of redundancy and repetition on the paths, which reduces the efficiency of task completion, increases the excessive workload of individual personnel, and reduces the actual utilization rate of personnel and equipment. In addition, the existing digital community management methods are unable to quickly identify high-risk scenarios and respond accurately, which increases resource waste and cannot ensure that the data of each platform play a synergistic role in emergency decision-making. It is easy to have delayed response or resource dispersion caused by the independent operation of each platform. To this end, we propose a digital community management method based on multi-platform collaboration. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose a digital cell management method based on multi-platform collaboration.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A digital community management method based on multi-platform collaboration, the specific steps of the management method are as follows:

[0007] Ⅰ. Collect and integrate data from different data sources inside and outside the community;

[0008] II. Identify and annotate abnormal behaviors in the integrated internal and external data of the community;

[0009] III. Real-time optimization of the work routes of community staff based on the integration and analysis results;

[0010] IV. Dynamically simulate events in each community and estimate the risk of events in each community;

[0011] V. Dynamically adjust community management decisions based on risk prediction and path planning results;

[0012] VI. Continuously optimize based on actual feedback from management decisions.

[0013] As a further solution of the present invention, the specific steps of collecting and integrating the internal and external data of the cell from different data sources in step I are as follows:

[0014] S1.1: Collect resident information, device status, surveillance video data, and environmental sensor data from various data sources inside and outside the community, including resident terminals, device sensors, and cameras, and pre-process each set of collected data;

[0015] S1.2: Identify entities in each group of pre-processed data inside and outside the community through named entity recognition technology, directly extract the attributes corresponding to each group of entities from structured data, images and video data, and then extract detailed information describing the entities from the text;

[0016] S1.3: Use data mining algorithms to extract the relationships between groups of entities, and use the extracted entities as nodes and the relationships between entities as edges to construct a corresponding community relationship graph, and use a graph database to represent and store the community relationship graph.

[0017] As a further solution of the present invention, the specific steps of the identification and marking in step II are as follows:

[0018] S2.1: query and extract corresponding video data and image data from the community relationship map, decompose the extracted video data frame by frame to obtain corresponding image data, scale, crop, denoise and standardize each group of image data, and then randomly transform the processed image data;

[0019] S2.2: Load the pre-trained CNN model and input each group of processed image data into the CNN model. The CNN model receives each group of image data, convolves the image data with the convolution kernel in the convolution layer, extracts the local features of the image, and then performs average pooling on the convolved image data through the pooling layer to reduce the dimension of the feature map and retain important features.

[0020] S2.3: The features of the image are extracted layer by layer through multiple sets of convolutional layers and pooling layers, and then the final feature data is input into the fully connected layer. Through weighting and biasing, the features extracted by the convolutional layer and the pooling layer are connected to the output layer. Then the output layer uses the Softmax activation function to convert the output data into the probability of each category;

[0021] S2.4: Select the category with the highest probability and determine whether the behavior in the current frame or video segment is abnormal. When abnormal behavior is identified, the relevant frames or areas will be marked and displayed visually on the monitoring interface, and the management personnel will be reminded through the alarm mechanism.

[0022] As a further solution of the present invention, the specific steps of real-time optimization of the work routes of the community staff in step III are as follows:

[0023] S3.1: Collect the community floor plan data, determine the work paths of security, cleaning and maintenance staff in the community based on the community relationship map, and regard the community management task as a graph. According to different tasks, each node in the graph represents a work point, and the edges between nodes represent the path from one work point to another.

[0024] S3.2: Initialize the number of individuals, positions, and pheromone concentrations of each group of paths, and calculate the heuristic information of the corresponding path based on the distance between each group of nodes. Then, calculate the corresponding selection probability of each path based on the current pheromone concentration and heuristic information of each path;

[0025] S3.3: Each group of individuals adopts a random selection strategy based on the selection probability of the current path, starting from the starting node, and gradually selects the next node until the corresponding work goal is completed to construct a complete path. When each group of individuals walks through a path, the pheromone concentration on the path is reduced according to the preset attenuation factor. When all individuals complete the task, the pheromone concentration on the path is updated according to the quality of the path found by the individuals;

[0026] S3.4: Repeat the path construction and pheromone concentration update. After each iteration, compare the quality of the current optimal path with the historical optimal path until the maximum number of iterations is reached or the path quality converges to a preset range. Stop the iteration, output the optimized work path and task scheduling plan for each group of staff, and implement them in practical applications.

[0027] As a further solution of the present invention, the specific steps of estimating the event risk of each cell in step IV are as follows:

[0028] S4.1: According to the demand and historical data, set the preliminary model of state space and event occurrence, and calculate the probability of different emergencies under each state according to the historical data and the previous conditions of the event, collect the current situation of the community, the current time, the equipment status and the weather conditions, and integrate the collected factors to obtain the current community status;

[0029] S4.2: Take the current cell state as the root node, and generate corresponding child nodes after the state change according to different emergencies that may occur under the root node. At the same time, record the probability of the event occurring and the operation information required to reach the state in the child node, and form a search tree with the generated child nodes and the root node;

[0030] S4.3: Select child nodes layer by layer based on the UCB strategy until the selected child node is a leaf node. Then, based on the state information recorded by the leaf node, expand the leaf node and generate multiple groups of new child nodes. Randomly simulate different paths starting from the root node and estimate the long-term return or risk of the path based on the probability distribution of historical data and risk model.

[0031] S4.4: After completing a simulation, the simulation results are passed back to all parent nodes of the tree, and the evaluation value of each node is updated. The selection, expansion, simulation and backtracking steps are repeated until the preset number of iterations is reached. The path with the highest evaluation value in the current tree is selected as the final decision, and based on the final decision, dynamic risk prediction and adjustment of emergencies are carried out to optimize resource allocation in real time.

[0032] As a further solution of the present invention, the specific steps of dynamically adjusting the cell management decision in step V are as follows:

[0033] S5.1: Collect various risk events that the emergency response system in the community may encounter, as well as different resource allocation strategies or emergency measures that can be taken at different risk times, and then calculate the prior probability of the corresponding risk event based on the historical risk event information of the community;

[0034] S5.2: Based on the previous risk prediction and path planning, calculate the conditional probability of risk events under different resource configurations or emergency response strategies. Based on the different financial losses, equipment damage, and personnel safety threats caused by various risk events, define the corresponding loss function to calculate the expected loss of each decision strategy;

[0035] S5.3: After calculating the expected losses of different strategies, select the strategy with the smallest expected loss as the optimal decision and implement the optimal strategy in actual applications. During the actual operation, regularly collect data and update the event status and conditional probability. When the conditional probability, prior probability or loss situation of the event changes, recalculate the expected loss and adjust the optimal decision strategy.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The digital community management method based on multi-platform collaboration collects community floor plan data, determines the work paths of security, cleaning and maintenance staff in the community based on the community relationship map, and regards the community management task as a graph. The number of individuals, positions and pheromone concentrations of each group of paths are initialized, and the heuristic information of the corresponding path is calculated based on the distance between each group of nodes. Then, the corresponding selection probability of each path is calculated according to the current pheromone concentration and heuristic information of each path. Each group of individuals adopts a random selection strategy based on the selection probability of the current path, starting from the starting node, and gradually selects the next node until the corresponding work goal is completed, so as to construct a complete path. When each group of individuals walks through a path, the pheromone concentration on the path is reduced according to the preset attenuation factor. When all individuals complete the task, the pheromone concentration on the path is updated according to the quality of the path found by the individuals, and the path construction and pheromone concentration update are repeated. After each iteration, the quality of the current optimal path is compared with the historical optimal path until the maximum number of iterations is reached or the path quality converges to the preset range. The iteration is stopped, and the optimized work path and task scheduling plan of each group of staff are output. At the same time, it is implemented in actual applications, which can intelligently adjust the work path of each staff member according to the real-time community situation, reduce redundancy and repetition on the path, improve the efficiency of task completion, effectively reduce the excessive workload of individual personnel, optimize the actual utilization of personnel and equipment, thereby improving the overall efficiency and achieving more diversified path planning and resource scheduling goals.

[0038] 2. The digital community management method based on multi-platform collaboration collects various risk events that the emergency response system in the community will encounter, as well as different resource allocation strategies or emergency measures that can be taken at different risk times. Then, based on the historical risk event information of the community, the prior probability of the corresponding risk event is calculated. According to the previous risk prediction and path planning, the conditional probability of the occurrence of risk events under different resource configurations or emergency response strategies is calculated. Based on the different financial losses, equipment damage and personnel safety threats caused by various risk events, the corresponding loss function is defined to calculate the expected loss of each decision strategy. After calculating the expected losses of different strategies, the strategy with the smallest expected loss is selected as the optimal decision, and the optimal strategy is implemented in practical applications. In the actual operation process, data is collected regularly and the event status and conditional probability are updated. When the conditional probability, prior probability or loss situation of the event changes, the expected loss is recalculated and the optimal decision strategy is adjusted. It can quickly identify high-risk scenarios and respond accurately, reduce resource waste, make community management more flexible and real-time, and enable the data of each platform to play a synergistic role in emergency decision-making, avoid the problem of delayed response or resource dispersion caused by the independent operation of each platform, and reduce repeated investment and losses caused by improper resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0040] Figure 1 This is a flowchart of the digital cell management method based on multi-platform collaboration proposed by the present invention. DETAILED DESCRIPTION

[0041] Example 1

[0042] Reference Figure 1 , a digital community management method based on multi-platform collaboration, the specific steps of the management method are as follows:

[0043] Collect and integrate internal and external data from different data sources.

[0044] Specifically, resident information, equipment status, surveillance video data and environmental sensor data of each group of data inside and outside the community are collected from various data sources such as resident terminals, equipment sensors and cameras, and each group of collected data is preprocessed. The entities in each group of data inside and outside the community after preprocessing are identified through named entity recognition technology, and the attributes corresponding to each group of entities are directly extracted from structured data, images and video data. Detailed information describing the entity is extracted from the text, and the relationship between each group of entities is extracted using a data mining algorithm. The extracted entities are used as nodes, and the relationships between entities are used as edges to construct a corresponding community relationship map, and the community relationship map is represented and stored using a graph database.

[0045] Identify and annotate abnormal behaviors in the integrated internal and external data of the community.

[0046] Specifically, query and extract the corresponding video data and image data from the community relationship map, decompose the extracted video data frame by frame to obtain the corresponding image data, scale, crop, denoise and standardize each group of image data, and then randomly transform the processed image data, load the pre-trained CNN model, and input each group of processed image data into the CNN model. The CNN model receives each group of image data, convolves the image data with the convolution kernel in the convolution layer, extracts the local features of the image, and then performs average pooling on the convolved image data through the pooling layer. , reduce the dimension of the feature map, retain important features, extract the features of the image layer by layer through multiple sets of convolutional layers and pooling layer structures, and then input the final feature data into the fully connected layer. Through weighting and bias, the features extracted by the convolutional layer and the pooling layer are connected to the output layer. Then the output layer uses the Softmax activation function to convert the output data into the probability of each category, selects the category with the highest probability, and determines whether the behavior in the current frame or video segment is abnormal. When abnormal behavior is identified, the relevant frames or areas will be marked, displayed visually on the monitoring interface, and the management personnel will be reminded through the alarm mechanism.

[0047] Optimize the work routes of community staff in real time based on the integration and analysis results.

[0048] Specifically, we collect the floor plan data of the community, determine the work paths of the security, cleaning and maintenance staff in the community based on the community relationship map, and regard the management task of the community as a graph. According to different tasks, each node in the graph represents a work point, and the edge between the nodes represents the path from one work point to another. Initialize the number and position of individuals and the pheromone concentration of each group of paths, and calculate the heuristic information of the corresponding path based on the distance between each group of nodes. Then, according to the current pheromone concentration and heuristic information of each path, the corresponding selection probability is calculated. Each group of individuals adopts a random selection strategy from the start according to the selection probability of the current path. Starting from the starting node, the next node is gradually selected until the corresponding work goal is completed to construct a complete path. When each group of individuals walks through a path, the pheromone concentration on the path is reduced according to the preset attenuation factor. When all individuals complete the task, the pheromone concentration on the path is updated according to the quality of the path found by the individuals. The path construction and pheromone concentration update are repeated. After each iteration, the quality of the current optimal path is compared with the historical optimal path until the maximum number of iterations is reached or the path quality converges to the preset range. The iteration is stopped, and the optimized work path and task scheduling plan of each group of staff are output and implemented in practical applications.

[0049] Example 2

[0050] Reference Figure 1 , a digital community management method based on multi-platform collaboration, the specific steps of the management method are as follows:

[0051] Dynamically simulate events in each community and estimate the risk of events in each community.

[0052] Specifically, according to the needs and historical data, the initial model of state space and event occurrence is set, and the probability of different emergencies occurring under each state is calculated according to the historical data and the previous conditions of the event. The current situation of the cell, the current time, the equipment status and the weather conditions are collected, and the collected factors are integrated to obtain the current cell state. The current cell state is used as the root node, and the corresponding child nodes after the state change are generated according to the different emergencies that may occur under the root node. At the same time, the probability of the event occurring and the operation information required to reach the state are recorded in the child nodes, and the generated child nodes and the root node are combined into a set of search trees, and the UCB strategy is selected layer by layer. Child nodes, until the selected child node is a leaf node, based on the status information recorded by the leaf node, the leaf node is expanded, and multiple groups of new child nodes are generated, starting from the root node, different paths are randomly simulated, and the long-term return or risk of the path is estimated based on the probability distribution of historical data and risk model. After completing a simulation, the simulation result is passed back to all parent nodes of the tree, and the evaluation value of each node is updated. The selection, expansion, simulation and backtracking steps are repeated until the preset number of iterations is reached, and the path with the highest evaluation value in the current tree is selected as the final decision, and based on the final decision, dynamic risk prediction and adjustment of emergencies are carried out to optimize resource allocation in real time.

[0053] Dynamically adjust cell management decisions based on risk prediction and path planning results.

[0054] Specifically, various risk events that the emergency response system in the community may encounter, as well as different resource allocation strategies or emergency measures that can be taken at different risk times, are collected. Then, based on the historical risk event information of the community, the prior probability of the corresponding risk event is calculated. Based on the previous risk forecast and path planning, the conditional probability of risk events occurring under different resource configurations or emergency response strategies is calculated. Based on the different financial losses, equipment damage, and personnel safety threats caused by various risk events, the corresponding loss function is defined to calculate the expected loss of each decision strategy. After calculating the expected losses of different strategies, the strategy with the smallest expected loss is selected as the optimal decision, and the optimal strategy is implemented in actual applications. During the actual operation process, data is collected regularly and the event status and conditional probability are updated. When the conditional probability, prior probability, or loss situation of the event changes, the expected loss is recalculated and the optimal decision strategy is adjusted.

[0055] Continuously optimize based on actual feedback from management decisions.

Claims

1. A digital community management method based on multi-platform collaboration, characterized in that: The specific steps of this management method are as follows: Ⅰ. Collect and integrate data from different data sources inside and outside the community; II. Identify and annotate abnormal behaviors in the integrated internal and external data of the community; III. Real-time optimization of the work routes of community staff based on the integration and analysis results; IV. Dynamically simulate events in each community and estimate the risk of events in each community; V. Dynamically adjust community management decisions based on risk prediction and path planning results; VI. Continuously optimize based on actual feedback from management decisions.

2. The digital community management method based on multi-platform collaboration according to claim 1 is characterized in that: The specific steps of collecting and integrating the internal and external data of the community from different data sources in step I are as follows: S1.1: Collect resident information, device status, surveillance video data, and environmental sensor data from various data sources inside and outside the community, including resident terminals, device sensors, and cameras, and pre-process each set of collected data; S1.2: Identify entities in each group of pre-processed data inside and outside the community through named entity recognition technology, directly extract the attributes corresponding to each group of entities from structured data, images and video data, and then extract detailed information describing the entities from the text; S1.3: Use data mining algorithms to extract the relationships between groups of entities, and use the extracted entities as nodes and the relationships between entities as edges to construct a corresponding community relationship graph, and use a graph database to represent and store the community relationship graph.

3. The digital community management method based on multi-platform collaboration according to claim 2 is characterized in that: The specific steps of identifying the annotations described in step II are as follows: S2.1: query and extract corresponding video data and image data from the community relationship map, decompose the extracted video data frame by frame to obtain corresponding image data, scale, crop, denoise and standardize each group of image data, and then randomly transform the processed image data; S2.2: Load the pre-trained CNN model and input each group of processed image data into the CNN model. The CNN model receives each group of image data, convolves the image data with the convolution kernel in the convolution layer, extracts the local features of the image, and then performs average pooling on the convolved image data through the pooling layer to reduce the dimension of the feature map and retain important features. S2.3: The features of the image are extracted layer by layer through multiple sets of convolutional layers and pooling layers, and then the final feature data is input into the fully connected layer. Through weighting and biasing, the features extracted by the convolutional layer and the pooling layer are connected to the output layer. Then the output layer uses the Softmax activation function to convert the output data into the probability of each category; S2.4: Select the category with the highest probability and determine whether the behavior in the current frame or video segment is abnormal. When abnormal behavior is identified, the relevant frames or areas will be marked and displayed visually on the monitoring interface, and the management personnel will be reminded through the alarm mechanism.

4. The digital community management method based on multi-platform collaboration according to claim 3 is characterized in that: The specific steps for real-time optimization of the work routes of community staff in step III are as follows: S3.1: Collect the community floor plan data, determine the work paths of security, cleaning and maintenance staff in the community based on the community relationship map, and regard the community management task as a graph. According to different tasks, each node in the graph represents a work point, and the edges between nodes represent the path from one work point to another. S3.2: Initialize the number of individuals, positions, and pheromone concentrations of each group of paths, and calculate the heuristic information of the corresponding path based on the distance between each group of nodes. Then, calculate the corresponding selection probability of each path based on the current pheromone concentration and heuristic information of each path; S3.3: Each group of individuals adopts a random selection strategy based on the selection probability of the current path, starting from the starting node, and gradually selects the next node until the corresponding work goal is completed to construct a complete path. When each group of individuals walks through a path, the pheromone concentration on the path is reduced according to the preset attenuation factor. When all individuals complete the task, the pheromone concentration on the path is updated according to the quality of the path found by the individuals; S3.4: Repeat the path construction and pheromone concentration update. After each iteration, compare the quality of the current optimal path with the historical optimal path until the maximum number of iterations is reached or the path quality converges to a preset range. Stop the iteration, output the optimized work path and task scheduling plan for each group of staff, and implement them in practical applications.

5. The digital community management method based on multi-platform collaboration according to claim 4 is characterized in that: The specific steps for estimating the event risk of each community described in step IV are as follows: S4.1: According to the demand and historical data, set the preliminary model of state space and event occurrence, and calculate the probability of different emergencies under each state according to the historical data and the previous conditions of the event, collect the current situation of the community, the current time, the equipment status and the weather conditions, and integrate the collected factors to obtain the current community status; S4.2: Take the current cell state as the root node, and generate corresponding child nodes after the state change according to different emergencies that may occur under the root node. At the same time, record the probability of the event occurring and the operation information required to reach the state in the child node, and form a search tree with the generated child nodes and the root node; S4.3: Select child nodes layer by layer based on the UCB strategy until the selected child node is a leaf node. Then, based on the state information recorded by the leaf node, expand the leaf node and generate multiple groups of new child nodes. Randomly simulate different paths starting from the root node and estimate the long-term return or risk of the path based on the probability distribution of historical data and risk model. S4.4: After completing a simulation, the simulation results are passed back to all parent nodes of the tree, and the evaluation value of each node is updated. The selection, expansion, simulation and backtracking steps are repeated until the preset number of iterations is reached. The path with the highest evaluation value in the current tree is selected as the final decision, and based on the final decision, dynamic risk prediction and adjustment of emergencies are carried out to optimize resource allocation in real time.

6. The digital community management method based on multi-platform collaboration according to claim 4 is characterized in that: The specific steps of dynamically adjusting the cell management decision in step V are as follows: S5.1: Collect various risk events that the emergency response system in the community may encounter, as well as different resource allocation strategies or emergency measures that can be taken at different risk times, and then calculate the prior probability of the corresponding risk event based on the historical risk event information of the community; S5.2: Based on the previous risk prediction and path planning, calculate the conditional probability of risk events under different resource configurations or emergency response strategies. Based on the different financial losses, equipment damage, and personnel safety threats caused by various risk events, define the corresponding loss function to calculate the expected loss of each decision strategy; S5.3: After calculating the expected losses of different strategies, select the strategy with the smallest expected loss as the optimal decision and implement the optimal strategy in actual applications. During the actual operation, regularly collect data and update the event status and conditional probability. When the conditional probability, prior probability or loss situation of the event changes, recalculate the expected loss and adjust the optimal decision strategy.

Images