Urban community emergency evacuation cloud guidance system based on deep learning model
Through the urban community emergency evacuation cloud-oriented system based on deep learning model, the real-time evacuation path planning is used using SegNet network and enhanced A-star algorithm, which solves the problems of slow response speed and lag in the existing technology, and realizes the complex environment The rapid and accurate evacuation guidance under this article improves the response speed to emergency response and the safety of evacuation routes.
Patent Information
- Application Number
- CN202410958284.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-17
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When faced with emergencies, the existing urban community emergency evacuation technology has slow response speed and delayed information transmission, making it difficult to achieve real-time evacuation guidance. The development and maintenance cost of robot guidance systems is high, the calculation amount is large, the accuracy needs to be improved, and it cannot be quickly obtained. and understand key evacuation guidelines.
The urban community emergency evacuation cloud-oriented system based on deep learning models, including community perception terminals, community security information bases and cloud decision-making systems, uses SegNet network and enhanced A-star algorithm for real-time object detection, spatial blocking analysis and path planning. Fast training and real-time inference are achieved through cloud servers, providing instant evacuation instructions.
Quickly and accurately identify road obstacles and congestion in complex earthquake environments, dynamically adjust evacuation paths, improve the speed and accuracy of reaction speed and path search, ensure the safety and efficiency of evacuation routes, and reduce system development and maintenance costs.
Smart Images

Figure CN118917505A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of emergency management systems, and in particular to an urban community emergency evacuation cloud guidance system based on a deep learning model. Background Art
[0002] The dense population and complex social structure in urban areas significantly increase the difficulty of community emergency evacuation. In the face of emergencies, rapid and effective evacuation is the key to reducing casualties. Therefore, being able to monitor and analyze conditions within the community in real time and provide residents with the best evacuation route is a necessary way to ensure the efficiency and safety of the evacuation process.
[0003] At present, the research on evacuation guidance includes manual evacuation, robot guidance, and mobile phone information guidance. Among them, manual evacuation is mainly based on disaster management experience, through the formulation of detailed emergency evacuation plans, regular training and drills for community residents and emergency volunteers, and the evacuation work is coordinated by community leaders or emergency response personnel. In an emergency, emergency volunteers or staff guide residents to evacuate along the designated route on site; but it relies on existing emergency plans, with a high degree of dependence, slow response speed, and delayed information transmission. The crowd evacuation method guided by robots uses the robot's own sensors (such as cameras, infrared sensors, etc.) to perceive the surrounding environment in real time, and exchanges data and collaborates with the emergency command center and other robots through wireless communication technology. It can execute the evacuation instructions of the simulated evacuation model and provide residents with evacuation guidance and safety information using voice broadcasts, display screens, indicator lights, etc.; but the development, deployment and maintenance costs of the robot guidance system are high, and there are technical problems such as navigation errors and sensor failures in complex environments, and the application environment requirements are high. Mobile phone information guidance mainly focuses on the transmission of warning information, sending emergency warning information to residents through various channels such as mobile phone text messages, mobile applications, and social media. At the same time, some studies have also constructed crowd congestion models, calculated heat maps of evacuated crowds, and used GPS and map data to guide residents to evacuate safely. However, this method has a large amount of calculations and its accuracy needs to be improved. It is difficult to achieve information synchronization and it is impossible to quickly obtain and understand key evacuation instructions.
[0004] In addition, existing evacuation guidance technologies focus on the construction of evacuation simulation models, such as cellular automaton models, lattice gas models, social force models, fluid dynamics models, intelligent agent models, and game theory models, but pay less attention to solving data processing at the disaster site, information transmission and guidance of crowd evacuation. Although the chain navigation grid system based on cellular automata integrates real-time data collection technologies, such as counting cameras, Wi-Fi sensors, and GPS trackers, the research can improve the efficiency of evacuation simulation by 6%, but the real-time analysis and feedback capabilities are still unknown.
[0005] Therefore, there is an urgent need for a cloud-based guidance system for emergency evacuation safety in urban communities to improve the response speed to emergencies and provide a reference for optimizing emergency evacuation routes. Summary of the invention
[0006] The purpose of this invention is to provide an urban community emergency evacuation cloud guidance system based on a deep learning model, which can provide real-time evacuation guidance for community residents in a complex earthquake environment, ensure the safety of evacuation, and improve the response speed to emergencies.
[0007] To achieve the above objectives, the present invention provides an urban community emergency evacuation cloud guidance system based on a deep learning model, including a community perception terminal, a community safety information base, a cloud decision-making system and a community service terminal:
[0008] Community sensing terminal, using the smart community grid information port to collect various basic data;
[0009] Community safety information base, including a central database for storing historical and real-time data;
[0010] The cloud-based decision-making system, based on a composite convolutional neural network deep learning model, performs real-time object detection and spatial obstruction analysis on dangerous spaces, obtains planned paths, and forms evacuation instructions;
[0011] The community service terminal is used to transmit the formed evacuation instructions to community residents in real time through the integrated communication platform.
[0012] Preferably, various basic data are collected, including real-time data collected by sensors, cameras and GPS positioning, and all the collected basic data are summarized in the central database of the community security information base.
[0013] Preferably, the cloud decision-making system includes using a SegNet network model to perform real-time segmentation tasks on the data in the security information base, and perform path planning based on the real-time segmentation task results.
[0014] Preferably, based on the result of the real-time segmentation task, the enhanced A-star algorithm model is used to perform path analysis to obtain the planned path, as follows:
[0015] Initialize the starting point and target point, create the open list and the closed list;
[0016] Use Euclidean distance as the heuristic evaluation function and use the comprehensive cost function to search for path nodes;
[0017] According to the real-time segmentation results, the cost of the obstacle area is increased to perform path planning.
[0018] Preferably, the comprehensive cost function includes a heuristic evaluation function and an actual cost function, and the cost of the obstacle area is increased by adjusting the actual cost function.
[0019] Preferably, the path planning includes detecting the real-time segmented data, obtaining the blocked area where people are crowded or blocked by obstacles, issuing an early warning and sending the coordinates.
[0020] Preferably, the real-time segmentation task, path planning and detection of real-time segmentation data are all completed through the cloud server.
[0021] Preferably, based on the planned path and warning information, the evacuation strategy and emergency response plan are optimized, an emergency evacuation instruction is formed, and sent to community residents through the community service terminal.
[0022] Therefore, the present invention adopts the above-mentioned urban community emergency evacuation cloud guidance system based on deep learning model, which has the following technical effects:
[0023] (1) Using the SegNet network model for semantic segmentation, it can quickly and accurately identify and classify people, trees, facilities, building debris, etc. on the street in earthquake scenarios, and provide immediate feedback of danger information; through real-time data, it continuously monitors the community road space, and promptly updates the location and status of dangerous objects to achieve efficient dynamic feedback; it can identify obstacles and congestion on the road in real time, trigger the alarm mechanism in time, and feedback the specific coordinates of the blocked area, quickly notify relevant departments and personnel for processing, and improve the speed of spatial danger information feedback.
[0024] (2) Based on the enhanced A-star algorithm, it can quickly calculate the optimal path from the current location to the safe area, avoiding obstacles and dangerous areas identified in real time; dynamically adjust the path weight according to the real-time segmentation results, increase the cost of the blocked area, and ensure the selection of a safe and efficient evacuation route; use heuristic functions and optimization search strategies to improve the speed and accuracy of path search, and can quickly find the optimal evacuation path in a complex and changing street environment.
[0025] (3) The use of cloud servers can realize rapid training and real-time reasoning of complex models, improving the overall performance and response speed of the system. Cloud servers can also centrally manage and store large amounts of real-time data, ensuring data security and consistency, and facilitating subsequent analysis and optimization. In addition, they can integrate multiple channels, user-friendly interfaces, and personalized evacuation guidance to enhance cross-regional collaboration capabilities.
[0026] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1It is a schematic diagram of an urban community emergency evacuation cloud guidance system based on a deep learning model. DETAILED DESCRIPTION
[0028] The present invention can be explained in more detail by the following examples. The purpose of disclosing the present invention is to protect all changes and improvements within the scope of the present invention. The present invention is not limited to the following examples.
[0029] like Figure 1 As shown, the present invention provides an urban community emergency evacuation cloud guidance system based on a deep learning model, comprising:
[0030] Community sensing terminals use smart community grid information ports, such as community sensors, CCTV cameras, GPS positioning and community management databases, to collect various basic data such as environmental monitoring, population dynamics, infrastructure status, etc. to ensure the comprehensiveness and real-time nature of the data.
[0031] The collected data will be aggregated into a central database to form an urban community safety information base that can store historical and real-time data.
[0032] Based on the safety information base, a composite convolutional neural network (CNN) deep learning model is used to perform real-time object detection and spatial obstruction analysis in dangerous spaces.
[0033] (1) Collecting data
[0034] Image collection: Collect high-resolution images of streets, crowds, trees, facilities, building debris, etc. taken by community cameras in past earthquake scenes. The dataset uses an existing public dataset (Cityscapes) and supplements it with self-annotated street scene data.
[0035] Data labeling: Perform pixel-level labeling on images, labeling pixels of different categories (people, trees, facilities, building falls, etc.).
[0036] (2) Perform data enhancement operations on the collected images, such as rotation, flipping, scaling, and color transformation, to improve the generalization ability of the model; and normalize the enhanced images to normalize the image pixel values to the range of [0, 1] to improve the model training effect.
[0037] (3) Based on the SegNet deep learning architecture, the encoder-decoder architecture of VGG16 is used to perform real-time segmentation tasks on the preprocessed data. The encoder is used to extract image features, and the decoder is used to restore the spatial resolution of the image. The decoder part upsamples layer by layer and uses the pooling index of the corresponding encoding layer to perform nonlinear upsampling to restore the spatial resolution of the image. The model is trained using the training set data, and the model performance is evaluated on the validation set. The model is optimized by adjusting hyperparameters, increasing regularization, adding more data augmentation, etc.
[0038] Among them, adjusting hyperparameters includes adjusting different training batch sizes in the experiment, selecting the batch size with the best training speed and model performance, and on this basis, increasing or decreasing the number of convolutional layers and fully connected layers, and adjusting the number and size of convolution kernels in each layer; increasing regularization includes adding L2 regularization terms in convolutional layers and fully connected layers to limit the size of weight values; in addition, adding more data enhancements such as rotation, scaling, shearing, color transformation, and contrast adjustment to make the training data more diverse.
[0039] (4) After real-time segmentation, it identifies situations where crowd congestion and obstacles block more than 50% of the road, and feeds back alarm information and sends the blocked spatial coordinates.
[0040] Path analysis is performed based on the enhanced A-star algorithm (A Search Algorithm) model.
[0041] First, initialize the starting node start and the target node goal, and create an open list (to store the nodes to be processed) and a closed list (to store the processed nodes).
[0042] Then, the Euclidean distance is used as the heuristic evaluation function as follows:
[0043]
[0044] Where h(n) is the Euclidean distance, (x goal ,y goal ) is the target node coordinate, (x n ,y n ) are the coordinates of node n.
[0045] Finally, according to the real-time segmentation results, that is, the coordinates of the blocked space calculated in real time, the cost of the blocked area is increased, that is, the actual cost function is increased, and the path is dynamically adjusted to avoid it as much as possible in the path calculation, thereby improving the speed and accuracy of the path search and being able to quickly find the optimal evacuation path in a complex and changeable street environment. The details are as follows:
[0046] Select a node from the open list: Select the node with the smallest comprehensive cost function f(n) and check whether it reaches the target node:
[0047] If the current node is the target node, the search ends and the path is constructed;
[0048] If the current node is not the target node, the current node is expanded, the unprocessed adjacent nodes are searched and added to the open list, and the actual cost function g(n) and comprehensive cost function f(n) values of its adjacent nodes are calculated.
[0049] Among them, the comprehensive cost function f(n) and the actual cost function g(n) of node n are:
[0050] f(n)=g(n)+h(n)
[0051] g(n)=g(parent(n))+cost(parent(n),n)
[0052] In the formula, parent(n) is the parent node of node n, and cost(parent(n),n) represents the cost from the parent node to the current node.
[0053] During the path calculation process, if crowd congestion or obstacles blocking more than 50% are detected, an alarm is triggered and the coordinates of the blocked area are sent. The optimal path is replanned and sent to the mobile terminal or the voice device terminal within the community to ensure the selection of a safe and efficient evacuation route. Based on real-time positioning information, the current location is analyzed for dangerous information, and the optimal evacuation plan and immediate action instructions are provided for the evacuated crowd, guiding residents to quickly and safely reach the nearest shelter.
[0054] The cloud-based decision-making system releases evacuation guidance information to residents through multiple channels (such as mobile applications, text messages, broadcasts, etc.); among them, the mobile application provides an intuitive and easy-to-use interface that can display real-time updated evacuation routes and dangerous area maps to help users quickly understand and follow evacuation instructions. Residents can receive personalized evacuation routes and safety guidelines through mobile applications. At the same time, the system also supports sending alerts and instructions to residents who do not have smartphones through SMS services, which can ensure that every resident can get the best evacuation plan and improve evacuation efficiency and safety.
[0055] All of the above data processing and analysis tasks are completed on the cloud server. By utilizing the powerful computing power and resources of the cloud server, rapid training and real-time reasoning of complex models can be achieved, improving the overall performance and response speed of the system. Through the cloud server, a large amount of real-time data is centrally managed and stored to ensure the security and consistency of the data, which is convenient for subsequent analysis and optimization. The cloud server provides flexible resource expansion capabilities and can dynamically adjust computing and storage resources according to actual needs to ensure that the system can still run efficiently in an emergency. In addition, it can integrate multiple channels, user-friendly interfaces and personalized evacuation guidance to achieve cross-regional collaboration and improve the overall performance, reliability and emergency response capabilities of the system.
[0056] In summary, through the steps of information perception - AI monitoring and identification - path planning cloud computing - evacuation plan customization, safe evacuation guidance can be completed to provide support for urban community emergency responses.
[0057] Therefore, the present invention adopts the above-mentioned urban community emergency evacuation cloud guidance system based on the deep learning model, which can solve the inefficient evacuation caused by information asymmetry in emergency situations, improve the response speed to emergencies, integrate resources, optimize emergency evacuation routes, accurately guide evacuation, and greatly reduce casualties and property losses.
[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.
Claims
1. An urban community emergency evacuation cloud guidance system based on deep learning model, It is characterized in that Including community perception terminal, community safety information base, cloud decision-making system and community service terminal: Community sensing terminal, using the smart community grid information port to collect various basic data; Community safety information base, including a central database for storing historical and real-time data; The cloud-based decision-making system, based on a composite convolutional neural network deep learning model, performs real-time object detection and spatial obstruction analysis on dangerous spaces, obtains planned paths, and forms evacuation instructions; The community service terminal is used to transmit the formed evacuation instructions to community residents in real time through the integrated communication platform.
2. According to the urban community emergency evacuation cloud guidance system based on deep learning model according to claim 1, It is characterized in that Collect various basic data including real-time data collected by sensors, cameras and GPS positioning, and summarize all collected basic data into the central database of the community security information base.
3. According to claim 1, an urban community emergency evacuation cloud guidance system based on a deep learning model, It is characterized in that The cloud-based decision-making system includes the use of the SegNet network model to perform real-time segmentation tasks on the data in the security information base, and perform path planning based on the real-time segmentation task results.
4. According to claim 3, an urban community emergency evacuation cloud guidance system based on a deep learning model, It is characterized in that Based on the results of the real-time segmentation task, the enhanced A-star algorithm model is used to perform path analysis and obtain the planned path, as follows: Initialize the starting point and target point, create the open list and the closed list; Use Euclidean distance as the heuristic evaluation function and use the comprehensive cost function to search for path nodes; According to the real-time segmentation task results, the cost of the obstacle area is increased to perform path planning.
5. According to claim 4, an urban community emergency evacuation cloud guidance system based on a deep learning model, It is characterized in that The comprehensive cost function includes a heuristic evaluation function and an actual cost function. The cost of the obstacle area is increased by adjusting the actual cost function.
6. According to claim 4, an urban community emergency evacuation cloud guidance system based on a deep learning model, It is characterized in that Path planning includes detecting real-time segmented data, obtaining blocked areas due to crowd congestion or obstacles, issuing early warnings and sending coordinates.
7. An urban community emergency evacuation cloud guidance system based on a deep learning model according to any one of claims 3 to 6, It is characterized in that Real-time segmentation tasks, path planning, and detection of real-time segmentation data are all completed through the cloud server.
8. According to claim 6, an urban community emergency evacuation cloud guidance system based on a deep learning model, It is characterized in that Based on the planned path and warning information, the evacuation strategy and emergency response plan are optimized, emergency evacuation instructions are formed, and sent to community residents through the community service terminal.
Citation Information
Patent Citations
Crowd simulation evacuation method and system based on wolf pack algorithm and congestion detection
CN110598307A
Intelligent community emergency safety early warning management platform
CN112071014A
Emergency management method based on smart community
CN112148828A
Crowd emergency evacuation simulation method based on social force model and A star algorithm
CN116523160A
Crowd evacuation system based on real time perception, simulation, and warning
US11417106B1
Cited By
Population evacuation path optimization method and system based on AI intelligence
CN119338099A
Smart city service operation management method and system
CN122334861A
