Urban inland inundation early warning method and system based on neural network, and medium

Abstract

The invention discloses an urban waterlogging early warning method and system based on a neural network and a medium, and the method comprises the steps: obtaining urban meteorological data and network public opinion data of a target city, and obtaining a waterlogging risk region set according to the urban meteorological data and the network public opinion data; scheduling cameras in the waterlogging risk area set to obtain current monitoring data; training based on a preset neural network to obtain a visual large model, and further obtaining a regional waterlogging level label and an early warning range according to the current monitoring data through the visual large model; and issuing corresponding early warning information according to the regional waterlogging level label and the early warning range. The method can effectively reduce the consumption of computing resources, remarkably enhances the accuracy and timeliness of urban inland inundation early warning, and can be widely applied to the technical field of artificial intelligence.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a neural network-based urban waterlogging early warning method, system and medium. Background Art

[0002] Urban waterlogging is often the result of the dual effects of extreme climate events and urbanization, including frequent rainstorms, inadequate planning and maintenance of drainage systems, and reduced rainwater absorption capacity due to a high proportion of hardened urban surfaces, which together make it difficult to discharge rainwater runoff in a timely manner, leading to serious waterlogging in streets, tunnels, residential areas and other areas. Existing urban waterlogging early warning methods, such as prediction models based on meteorological data analysis and groundwater level monitoring technology, have provided assistance to waterlogging management to some extent, but these traditional methods generally have problems such as poor prediction accuracy, slow response and monitoring blind spots, and lack of efficient integration between them. Summary of the invention

[0003] In order to solve the above technical problems, the purpose of the present invention is to provide a neural network-based urban waterlogging early warning method, system and medium with higher prediction accuracy and faster response speed.

[0004] To achieve the above purpose, one aspect of an embodiment of the present application proposes a method for early warning of urban waterlogging based on a neural network, comprising the following steps:

[0005] Obtaining urban meteorological data and online public opinion data of a target city, and obtaining a set of waterlogging risk areas according to the urban meteorological data and the online public opinion data;

[0006] Dispatching cameras in the waterlogging risk area set to obtain current monitoring data;

[0007] A visual large model is obtained based on a preset neural network training, and then a regional waterlogging grade label and a warning range are obtained through the visual large model according to the current monitoring data;

[0008] Corresponding warning information is issued according to the regional waterlogging grade label and the warning range.

[0009] In some embodiments, the urban waterlogging early warning method further includes:

[0010] The map information of the target city is obtained, and the target city is rasterized according to the map information to obtain a raster map of the target city.

[0011] In some embodiments, the acquiring of urban meteorological data and online public opinion data of the target city, and obtaining a set of waterlogging risk areas according to the urban meteorological data and the online public opinion data, specifically includes:

[0012] Acquire the urban meteorological data, construct a precipitation prediction model according to the urban meteorological data, and obtain a first waterlogging risk area set according to the precipitation prediction model;

[0013] Acquire the network public opinion data, and obtain a second waterlogging risk area set and a third waterlogging risk area set according to the network public opinion data;

[0014] The first waterlogging risk area set, the second waterlogging risk area set and the third waterlogging risk area set are fused to obtain the waterlogging risk area set.

[0015] In some embodiments, constructing a precipitation prediction model according to the urban meteorological data, and obtaining a first set of waterlogging risk areas according to the precipitation prediction model specifically includes:

[0016] Acquire historical rainfall data and historical meteorological data of the target city, and pre-process the historical rainfall data and the historical meteorological data;

[0017] The pre-processed historical rainfall data and the historical meteorological data are merged with the target city grid map to obtain parameter features;

[0018] The back propagation neural network is trained by the parameter characteristics to obtain the precipitation prediction model;

[0019] Inputting the urban meteorological data into the precipitation prediction model to obtain a regional precipitation sequence;

[0020] The first set of urban flood risk areas is obtained according to the total number of areas in the target city, the regional precipitation sequence and the preset risk area division cardinality.

[0021] In some embodiments, the network public opinion data includes historical public opinion data and current public opinion data, and obtaining the network public opinion data and obtaining the second waterlogging risk area set and the third waterlogging risk area set according to the network public opinion data specifically include:

[0022] Acquiring the historical public opinion data and the current public opinion data;

[0023] Preprocessing the historical public opinion data, performing cluster analysis on the preprocessed historical public opinion data, and obtaining the frequency of mentions of waterlogging;

[0024] Performing sentiment analysis on the preprocessed historical public opinion data to obtain a sentiment score;

[0025] Obtaining a regional risk index according to the frequency of mentioning waterlogging and the sentiment score;

[0026] According to the historical public opinion data, the regional risk index and a preset first risk index threshold, a historical risk area is obtained, and then the historical risk area is mapped to the target city grid map to obtain the second waterlogging risk area set;

[0027] According to the current public opinion data, the regional risk index and a preset second risk index threshold, the current risk area is obtained, and then the current risk area is mapped to the target city grid map to obtain the third waterlogging risk area set.

[0028] In some embodiments, the visual large model is obtained based on the preset neural network training, specifically including:

[0029] Acquire historical waterlogging videos and historical waterlogging images of the target city;

[0030] Performing graded labeling on the historical waterlogging videos and the historical waterlogging images to obtain training data containing waterlogging grade labels;

[0031] The neural network is trained using the training data to obtain the large visual model.

[0032] In some embodiments, obtaining a regional waterlogging grade label and a warning range according to the current monitoring data through the visual big model specifically includes:

[0033] Analyze the water depth, water flow dynamics, and traffic obstruction status in the current monitoring data through the visual big model to obtain a waterlogging grade label in the area;

[0034] Determine the warning range according to the location information of the camera, the target city grid map and the regional waterlogging level label;

[0035] The regional waterlogging level labels include the first level, the second level and the third level.

[0036] To achieve the above purpose, another aspect of the embodiment of the present application proposes an urban flood warning system based on a neural network, comprising:

[0037] A risk area perception module is used to obtain urban meteorological data and online public opinion data of a target city, and obtain a waterlogging risk area set based on the urban meteorological data and the online public opinion data;

[0038] A risk area monitoring module is used to schedule cameras in the waterlogging risk area set to obtain current monitoring data;

[0039] A risk area identification module is used to obtain a visual large model based on a preset neural network training, and then obtain a regional waterlogging grade label and a warning range according to the current monitoring data through the visual large model;

[0040] The risk warning module is used to issue corresponding warning information according to the regional waterlogging grade label and the warning range.

[0041] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application proposes an electronic device, which includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the urban waterlogging early warning method based on a neural network as described above is realized.

[0042] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium used for computer-readable storage, and the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the urban waterlogging warning method based on neural network as described above.

[0043] The beneficial effects of the present invention are as follows: the urban waterlogging early warning method, system and medium based on neural network of the present invention first obtain the urban meteorological data and network public opinion data of the target city, obtain the waterlogging risk area set according to the urban meteorological data and network public opinion data, and then dispatch the cameras in the waterlogging risk area set to obtain the current monitoring data, then obtain the visual large model based on the preset neural network training, and then obtain the regional waterlogging level label and warning range according to the current monitoring data through the visual large model, and finally release the corresponding warning information according to the regional waterlogging level label and warning range. On the one hand, the present invention integrates multi-source data such as urban meteorological data and network public opinion data, screens the waterlogging risk area set by data analysis, and then only dispatches the cameras in the waterlogging risk area set, which can broaden the data source of urban waterlogging early warning, improve the comprehensiveness of early warning, and reduce the cost of urban full-domain monitoring; on the other hand, the visual large model obtained by neural network training patrols the urban area, so as to accurately locate the early warning range and regional waterlogging level label in the city, and can realize intelligent screening, grading and domain monitoring of risk areas, effectively reduce the consumption of computing resources, and significantly enhance the accuracy and timeliness of urban waterlogging early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solution in the embodiments of the present invention, the following introduction is made to the drawings required for use in the embodiments of the present invention. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solution of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 A flowchart of a method for early warning of urban waterlogging based on a neural network provided by an embodiment of the present invention;

[0046] Figure 2 A schematic diagram of dividing waterlogging risk areas based on a target city grid provided in an embodiment of the present invention;

[0047] Figure 3 A step flow chart of step S101 provided in an embodiment of the present invention;

[0048] Figure 4 A step flow chart of step S1011 provided in an embodiment of the present invention;

[0049] Figure 5 A step flow chart of step S1012 provided in an embodiment of the present invention;

[0050] Figure 6 The flood risk area S provided in the embodiment of the present invention xy Schematic diagram of street camera point warning within the grid;

[0051] Figure 7 A flowchart of the steps of training a large visual model provided by an embodiment of the present invention;

[0052] Figure 8 A flowchart of steps for determining a regional waterlogging grade label provided by an embodiment of the present invention;

[0053] Fig. 9 A schematic diagram of a process flow of a neural network-based urban waterlogging early warning method provided by an embodiment of the present invention;

[0054] Fig.10 A schematic diagram of a structure of a neural network-based urban waterlogging early warning system provided by an embodiment of the present invention;

[0055] Fig.11 A schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0057] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0058] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0059] Before describing the embodiments of the present application in detail, some nouns and terms involved in the embodiments of the present application are first described. The nouns and terms involved in the embodiments of the present application are subject to the following explanations.

[0060] Neural Network: A type of deep learning that consists of a hierarchical structure of interconnected neuron nodes. Each connection has a weight that can be adjusted through deep learning and can learn complex mapping relationships between inputs and outputs.

[0061] Vision Big Model: An artificial neural network model with a huge parameter size used in the field of computer vision, which can process and analyze image and video data and perform tasks such as image recognition, object detection, image segmentation, image generation, etc.

[0062] Multi-layer BP neural network: The full name is Backpropagation Neural Network, which is an artificial neural network model widely used in machine learning, data mining and pattern recognition. Its working principle is based on a multi-layer feedforward network structure, and the network is trained through the Error Backpropagation Algorithm to achieve learning and solving complex problems.

[0063] Urban waterlogging is often the result of the dual effects of extreme climate events and urbanization, including frequent rainstorms, inadequate planning and maintenance of drainage systems, and reduced rainwater absorption capacity due to a high proportion of hardened urban surfaces, which together make it difficult to discharge rainwater runoff in a timely manner, leading to serious waterlogging in streets, tunnels, residential areas and other areas. Existing urban waterlogging early warning methods, such as prediction models based on meteorological data analysis and groundwater level monitoring technology, have provided assistance to waterlogging management to some extent, but these traditional methods generally have problems such as poor prediction accuracy, slow response and monitoring blind spots, and lack of efficient integration between them.

[0064] To this end, the embodiment of the present invention proposes a method for early warning of urban waterlogging based on a neural network. First, the urban meteorological data and network public opinion data of the target city are obtained, and the waterlogging risk area set is obtained according to the urban meteorological data and network public opinion data, and then the cameras in the waterlogging risk area set are dispatched to obtain the current monitoring data, and then the visual large model is obtained based on the preset neural network training, and then the regional waterlogging level label and warning range are obtained according to the current monitoring data through the visual large model, and finally the corresponding warning information is issued according to the regional waterlogging level label and warning range. On the one hand, the present invention integrates multi-source data such as urban meteorological data and network public opinion data, and screens the waterlogging risk area set for data analysis, and then only dispatches the cameras in the waterlogging risk area set, which can broaden the data source of urban waterlogging warning, improve the comprehensiveness of warning, and reduce the cost of urban full-domain monitoring; on the other hand, the visual large model obtained by neural network training patrols the urban area, so as to accurately locate the warning range and regional waterlogging level label in the city, and can realize intelligent screening, grading and domain monitoring of risk areas, effectively reduce the consumption of computing resources, and significantly enhance the accuracy and timeliness of urban waterlogging warning. The urban waterlogging early warning method can be applied to scenarios such as urban management and emergency response, smart city construction, and urban waterlogging risk assessment, but is not limited to these.

[0065] Reference Figure 1 , Figure 1A flowchart of a method for early warning of urban waterlogging based on a neural network provided by an embodiment of the present invention. An embodiment of the present invention provides a method for early warning of urban waterlogging based on a neural network, and the method includes steps S101 to S104:

[0066] S101, obtaining urban meteorological data and online public opinion data of the target city, and obtaining a set of waterlogging risk areas based on the urban meteorological data and online public opinion data;

[0067] Specifically, by deeply preprocessing and feature engineering the real-time and historical information of urban meteorological data and online public opinion data, we can deeply explore the hidden patterns and correlations in the data. This process can not only identify the public's emotional tendencies and attention to specific meteorological events, but also predict potential risk trends and abnormal fluctuations, and achieve accurate positioning of key risk areas. At the same time, by integrating geographic information system technology, through spatial analysis and statistical methods, the risk areas analyzed are cross-validated with the geographic information database to achieve accurate positioning. In this way, the longitude and latitude coordinates of each risk area are calculated, and a complete set of waterlogging risk areas is formed.

[0068] It should be noted that the embodiments of the present invention can broaden the data sources of urban waterlogging warnings and improve the comprehensiveness of warnings by integrating multi-source data such as urban meteorological data and online public opinion data and filtering the waterlogging risk area set through data analysis.

[0069] As an optional implementation, the urban waterlogging early warning method further includes the following steps S01:

[0070] S01. Obtain map information of a target city, and perform regional rasterization on the target city according to the map information to obtain a raster map of the target city.

[0071] Specifically, Figure 2 The figure shows a schematic diagram of waterlogging risk area division based on the target city grid. First, the map information of the target city is obtained, and then the target city area is rasterized based on the map information to form a kilometer-level grid division, such as: 5km*5km, and the upper left corner is used as the starting point, and j grids are divided horizontally and i grids are divided vertically to complete the grid coding S 00 -S ij , and get the target city grid map.

[0072] Reference Figure 3 , Figure 3 A step flow chart of step S101 provided in an embodiment of the present invention is further provided as an optional implementation mode, wherein the step of obtaining urban meteorological data and online public opinion data of the target city and obtaining a set of urban flood risk areas according to the urban meteorological data and online public opinion data can be further divided into the following steps S1011 to S1013:

[0073] S1011, obtaining urban meteorological data, constructing a precipitation prediction model according to the urban meteorological data, and obtaining a first waterlogging risk area set according to the precipitation prediction model;

[0074] Reference Figure 4 , Figure 4 A step flow chart of step S1011 provided in an embodiment of the present invention is further provided as an optional implementation mode. A precipitation prediction model is constructed according to urban meteorological data. The step of obtaining a first set of waterlogging risk areas according to the precipitation prediction model can be further divided into the following steps S10111 to S10115:

[0075] S10111. Obtain historical rainfall data and historical meteorological data of the target city, and pre-process the historical rainfall data and historical meteorological data;

[0076] In some optional embodiments, hourly (minute-level for higher accuracy) historical rainfall data and historical meteorological data of the target city in the past year or the past n years are collected, and the collected historical rainfall data and historical meteorological data are deeply cleaned and standardized. For example, invalid, erroneous or incomplete records are removed, and then the numerical data is scaled using Z-score standardization, and the time series data is converted into a supervised learning format using a sliding window method to obtain the pre-processed historical rainfall data and historical meteorological data.

[0077] Exemplarily, the historical meteorological data includes data such as temperature, humidity, wind speed, wind direction, and air pressure, but is not limited thereto.

[0078] S10112, fusing the pre-processed historical rainfall data and historical meteorological data with the target city grid map to obtain parameter features;

[0079] In some optional embodiments, the target city grid map is encoded and the central longitude and latitude coordinates are taken as the grid location description. The historical rainfall data and historical meteorological data are distributed and matched to the grid center, the urban area meteorological station and geographic information data are integrated, and the input containing historical rainfall and meteorological parameters is constructed. The specific parameter features are: X1 grid code, X2 time point, X3 temperature, X4 humidity, X5 wind speed, X6 wind direction, X7 air pressure, X8 time point and X9 precipitation, etc.

[0080] S10113, training the back propagation neural network through parameter features to obtain a precipitation prediction model;

[0081] S10114, inputting the urban meteorological data into the precipitation prediction model to obtain a regional precipitation sequence;

[0082] Specifically, in terms of model architecture design, the embodiment of the present invention adopts a multi-layer BP neural network, in which the input layer corresponds to the pre-processed historical rainfall data and historical meteorological data, receives the original data and passes it to the first layer of the network. Assume that the input vector is x = (x1, x2, ..., x n ), where n represents the number of features or attributes.

[0083] Furthermore, the hidden layer is designed to be 3 layers, with the number of nodes in each layer being several times that of the input layer. The ReLU activation function f(x)=max(0,x) is used to enhance the nonlinear expression, and the output layer predicts a single rainfall value with a linear activation function f(x)=x. Among them, the network is trained using the mean square error (MSE) loss function and the Adam optimizer, and batch learning and early stopping strategies are implemented to avoid overfitting. At the same time, the prediction performance of the model can be improved by adjusting the neural network structure, learning strategy, regularization method, data preprocessing, hyperparameter setting, and using appropriate loss functions and evaluation indicators. After training, verifying and testing the hierarchical evaluation model, we finally obtained a precipitation prediction model for real-time precipitation data prediction, and obtained the precipitation sequence p={p1,p2,...,p N}, where p i represents the precipitation in the ith region, and N is the total number of regions in the target city.

[0084] S10115. Obtain a first set of urban flood risk areas based on the total number of areas in the target city, the regional precipitation sequence, and a preset risk area division cardinality.

[0085] Specifically, the N / M regions with the largest precipitation are found by the following formula, namely: Among them, p i is the i-th maximum precipitation in the sorted sequence, N is the total number of regions in the target city, and m is the cardinality of risk area division:

[0086]

[0087] For example, when m=3, the above formula can be expressed as:

[0088]

[0089] The above formula allows us to find the one-third area with the largest precipitation as the risk area. The grid set is represented as: A = {S1, S2, ..., S n}, where S i Represents the grid of the i-th region. Through this precipitation prediction model, the weather conditions of the target city can be monitored in real time, and the first set of urban flood risk areas A can be obtained through prediction and analysis. The first set of urban flood risk areas A is as follows: Figure 2 The area shown by the medium blue grid.

[0090] S1012, obtaining network public opinion data, and obtaining a second set of urban waterlogging risk areas and a third set of urban waterlogging risk areas according to the network public opinion data;

[0091] Reference Figure 5 , Figure 5 A step flow chart of step S1012 provided in an embodiment of the present invention, further as an optional implementation mode, the network public opinion data includes historical public opinion data and current public opinion data, and the step of acquiring the network public opinion data and obtaining the second waterlogging risk area set and the third waterlogging risk area set according to the network public opinion data can be further divided into the following steps S10121 to S10126:

[0092] S10121. Obtain historical public opinion data and current public opinion data;

[0093] For example, web crawler technology can be used to widely collect text, picture and video information related to waterlogging, accumulation of water, heavy rain, traffic interruption, etc. in the target city from social media (such as Weibo, WeChat, Tieba and forums, etc.), news platforms and other network channels in the past two years or the past n years to obtain historical public opinion data; at the same time, real-time monitoring of text, picture and video information related to the current waterlogging, accumulation of water, heavy rain, traffic interruption, etc. in the target city from social media, news platforms and other network channels can be performed to obtain current public opinion data.

[0094] S10122, preprocessing the historical public opinion data, performing cluster analysis on the preprocessed historical public opinion data, and obtaining the frequency of mentions of waterlogging;

[0095] In some optional embodiments, OCR technology can be applied to convert text information in images and videos in historical public opinion data into processable data, and then pre-process these processable data. The pre-processing includes steps such as word segmentation, denoising, and sentiment pre-labeling to remove noise data such as advertisements and spam in the historical public opinion data, and perform Jieba word segmentation and standardization on the denoised historical public opinion data.

[0096] Next, sentiment analysis is performed on the preprocessed historical public opinion data to remove neutral words in the data and retain positive and negative words for subsequent sentiment analysis. First, the TF-IDF technology can be used to screen keywords and topics that are highly related to urban waterlogging in the target city. Then, the K-means clustering algorithm is combined to classify and aggregate the preprocessed historical public opinion data according to geographic location, time axis, and sentiment polarity to identify hot spots and public sentiment trends related to urban waterlogging. Based on the results of cluster analysis, the frequency of mentions of urban waterlogging F_Index is obtained.

[0097] S10123, performing sentiment analysis on the pre-processed historical public opinion data to obtain a sentiment score;

[0098] In some optional embodiments, a machine learning model SVM can be used to quantify the sentiment tendency of each piece of historical public opinion data to implement sentiment analysis and obtain a sentiment score S_Score. If the proportion of negative sentiment in the data is higher, the sentiment score S_Score is higher.

[0099] S10124, obtain the regional risk index based on the frequency of mentions of waterlogging and the sentiment score;

[0100] Specifically, the comprehensive assessment of the risk area can be quantified by the regional risk index R obtained by the following formula:

[0101] R=(α×F_Index+β×S_Score+γ×T)×μ

[0102] Among them, F_Index represents the frequency of mentions of urban flooding, S_Score represents the sentiment score, T represents time sensitivity (such as the number and urgency of recent reports), α, β and γ are weight factors, and μ is the normalization coefficient, which is balanced and adjusted according to the actual effect.

[0103] S10125. Obtain historical risk areas according to historical public opinion data, regional risk indexes, and a preset first risk index threshold, and then map the historical risk areas to a target city grid map to obtain a second waterlogging risk area set;

[0104] Those skilled in the art will appreciate that the first risk index threshold value may be selected according to actual needs. For example, the first risk index threshold value may be values ​​such as 0.9, 0.8, and 0.7, but is not limited thereto.

[0105] Specifically, GIS technology is used to combine the geographical location and event severity in the historical public opinion data, and the second set of waterlogging risk areas is identified based on the calculated regional risk index. The regional risk index selects the first risk index threshold, and for example, R>0.8 is taken as the historical risk area, and the historical risk area is mapped to the grid S 00 -S ij , we get the second set of waterlogging risk areas, and the grid set is represented as: B1 = {S1, S2, ..., S n}, where S i The raster represents the i-th area, which usually includes low-lying areas, poorly drained neighborhoods, and areas that have historically experienced severe flooding.

[0106] S10126. According to the current public opinion data, the regional risk index and the preset second risk index threshold, the current risk area is obtained, and then the current risk area is mapped to the target city raster map to obtain a third set of urban flood risk areas.

[0107] Those skilled in the art can understand that the second risk index threshold can be selected according to actual needs, and the second risk index threshold must be smaller than the first risk index threshold. Exemplarily, the second risk index threshold can be 0.6, 0.5, 0.4, etc., but is not limited thereto.

[0108] Specifically, the current public opinion data is processed, and the regional risk index selects the second risk index threshold. For example, R>0.6 is taken as the current risk area, the regional coverage is increased, and the current risk area is mapped to the grid S 00 -S ij In the example, the grid set is represented as: B2 = {S1, S2, ..., S n}, where S i The above steps can be used to integrate the historical key risk area set (i.e., the second waterlogging risk area set B1), and monitor the public opinion in real time to obtain the real-time risk area set (i.e., the third waterlogging risk area set B2). The second waterlogging risk area set B1 and the third waterlogging risk area set B2 are as follows: Figure 2 Shown in purple grid.

[0109] S1013. Fuse the first waterlogging risk area set, the second waterlogging risk area set, and the third waterlogging risk area set to obtain a waterlogging risk area set.

[0110] Specifically, the first waterlogging risk area set A, the second waterlogging risk area set B1, and the third waterlogging risk area set B2 obtained by analysis are fused through the following formula to obtain the waterlogging risk area set C. The waterlogging risk area set C is as follows: Figure 2 The medium blue, purple, and blue + purple grids show:

[0111] C=A+B1+B2-A∩B2-B1∩B2-B2∩A+A∩B1∩B2

[0112] S102, dispatching cameras in the waterlogging risk area set to obtain current monitoring data;

[0113] Specifically, the embodiment of the present invention only schedules the multi-domain cameras of the grid structure in the waterlogging risk area set C (select S1...S n The camera in the grid) pulls the current monitoring data and completes the stream extraction, such as Figure 6 Shown is the flood risk area S xySchematic diagram of street camera point warning within the grid. By monitoring the waterlogging risk in the local key areas of the target city, the cost of full-area retrieval monitoring can be greatly reduced, and the efficiency of waterlogging risk inspection can be effectively improved. Among them, the camera configuration supports all-weather, multi-angle monitoring, combined with night vision and clear imaging technology in bad weather, it can accurately capture signs of waterlogging under various conditions.

[0114] It should be noted that the specific frame extraction frequency in the stream frame is related to the computing power. Under the condition of certain computing power, the fewer cameras are called up for identification, the higher the frequency, the more real-time the response, and the more accurate and sensitive the waterlogging risk patrol and warning will be.

[0115] S103, obtaining a visual large model based on a preset neural network training, and then obtaining a regional waterlogging grade label and a warning range through the visual large model according to current monitoring data;

[0116] It should be noted that the embodiment of the present invention patrols the urban area through a large visual model obtained based on neural network training, so as to accurately locate the warning range in the city and the regional waterlogging level label, and can realize intelligent screening, grading and domain monitoring of risk areas, effectively reduce the consumption of computing resources, and significantly enhance the accuracy and timeliness of urban waterlogging warnings.

[0117] Reference Figure 7 , Figure 7 A flowchart of a step of training a visual large model provided in an embodiment of the present invention, further as an optional implementation, the step of obtaining a visual large model based on a preset neural network training can be further divided into the following steps S1031 to S1033:

[0118] S1031. Acquire historical waterlogging videos and historical waterlogging images of the target city;

[0119] S1032, grade-labeling the historical waterlogging videos and historical waterlogging images to obtain training data containing waterlogging grade labels;

[0120] Specifically, based on the urban video surveillance network, through large-scale data collection, historical waterlogging videos and images are collected and manually annotated. The hierarchical annotation content needs to cover water accumulation of different depths, types of submerged objects (such as vehicles, sidewalks, etc.) and the dynamic evolution process of waterlogging. For example, red R (i.e. the first level): red, severe waterlogging, the depth of water accumulation exceeds half of the vehicle tires or even higher, resulting in vehicles and pedestrians being unable to pass; yellow Y (i.e. the second level): yellow, moderate waterlogging, the depth of water accumulation is sufficient to cover the bottom to the middle of the vehicle tires, resulting in congestion and slow passage of vehicles and pedestrians; green G (i.e. the third level): green, slight waterlogging, vehicles and pedestrians are still unobstructed, thereby increasing the diversity and comprehensiveness of the annotated data and constructing training data containing waterlogging grade labels.

[0121] S1033. Train the neural network using training data to obtain a large visual model.

[0122] Furthermore, based on the training data constructed above, a neural network is used for customized training. The neural network can use the DINOv2 visual large model, but is not limited to this. A transfer learning strategy is adopted to freeze some of the underlying feature extraction layers of the model, and only fine-tune the high-level classification and detection heads. When adjusting the model parameters, loss functions that are sensitive to water body texture, color, and dynamic features, such as IoU (Intersection over Union) loss and focal loss (FocalLoss), are introduced to optimize the model's target detection performance in complex environments. At the same time, data enhancement and regularization techniques are implemented to improve the generalization performance of the model, such as random rotation, shearing, brightness adjustment, etc., to simulate waterlogging scenes under different lighting, viewing angles, and water quality conditions. Regularization techniques can use Dropout, weight decay, etc. to avoid model overfitting. In the training process, dynamic learning rate adjustment strategies such as cosine annealing or one-cycle policy are used to intelligently adjust the learning rate according to the training progress, so as to converge quickly in the early stage of training and make fine adjustments in the later stage. Finally, the optimized model is deployed to obtain a large visual model that can realize real-time frame extraction, retrieval, analysis and processing of the city's full-area surveillance video.

[0123] Reference Figure 8 , Figure 8 A flowchart of a step for determining a regional waterlogging grade label provided in an embodiment of the present invention, further as an optional implementation, the step of obtaining a regional waterlogging grade label and a warning range according to current monitoring data through a visual big model can be further divided into the following steps S1034 and S1035:

[0124] S1034, analyzing the water depth, water flow dynamics, and traffic obstruction status in the current monitoring data through a visual big model to obtain a waterlogging grade label in the area;

[0125] Among them, the regional waterlogging level labels include the first level, the second level and the third level.

[0126] Specifically, based on the current monitoring data obtained, the visual large model trained by deep learning can quickly identify the depth of water accumulation, water flow dynamics, pedestrian and vehicle obstruction, etc., to achieve real-time analysis of the waterlogging situation in the monitoring screen. The model outputs the regional waterlogging level label, thereby realizing the rapid classification of waterlogging conditions, achieving the following effects: Figure 6 As shown, R represents the first level, Y represents the second level, and G represents the third level.

[0127] S1035, determining the warning range according to the location information of the camera, the target city grid map, and the regional waterlogging grade label;

[0128] Specifically, by combining the precise location information of each camera, the target city grid map and the GIS geographic information system, the regional waterlogging level labels identified by the visual large model are mapped to the city map in real time to determine the warning range. Furthermore, a dynamic waterlogging risk heat map can be generated, allowing managers to intuitively identify which areas are or will soon face the threat of waterlogging.

[0129] S104, issuing corresponding warning information according to the regional waterlogging grade label and warning range;

[0130] Optionally, after the system issues warning information by accurately locating the risk area, the relevant departments can automatically trigger the corresponding response mechanism according to the regional waterlogging level label. For example, the warning information can be divided into red warning, yellow warning and green warning according to the regional waterlogging level label. The red warning can immediately notify the relevant departments to carry out emergency drainage operations and evacuate people; the yellow warning prompts to strengthen monitoring and prepare emergency resources; the green status maintains regular monitoring, so as to reasonably allocate social resources.

[0131] Furthermore, when warning information is released, it can also automatically integrate multi-source information such as weather forecasts, historical waterlogging data, and the status of urban drainage systems to form a comprehensive risk assessment report to provide a scientific basis for decision makers. In addition, a multi-channel warning release mechanism can be established, including but not limited to mobile phone APP push, social media announcements, broadcast notifications and other network media, so as to quickly and accurately convey warning information to the public, enhance the public's awareness of disaster prevention and self-protection capabilities, and thus form an efficient and accurate urban waterlogging warning and response system.

[0132] In summary, the embodiment of the present invention relies on advanced big data analysis and artificial intelligence framework, integrates urban meteorological big data, social media intelligence resources, and uses machine learning algorithms and the complex pattern recognition capabilities of deep neural networks, supplemented by the spatial analysis function of geographic information systems (GIS), to jointly construct a highly intelligent urban waterlogging early warning method. The processing flow of the urban waterlogging early warning method based on neural networks is as follows: Fig. 9 As shown:

[0133] The first step is to use precise geospatial analysis technology to divide the urban space into uniform grid units (such as 5km*5km), and assign a unique code to each grid unit to lay the foundation for subsequent high-precision analysis. Through in-depth mining of urban meteorological data and analysis of online public opinion data related to waterlogging in social media, the waterlogging risk area set C is accurately identified and defined, achieving a preliminary mapping of the spatial distribution of risks.

[0134] The second step is to efficiently dispatch and deploy high-definition video surveillance resources for key areas in the waterlogging risk area set C. Using historical video data and annotated data, a neural network-driven visual large model is trained to optimize waterlogging feature recognition and quickly identify key indicators such as waterlogging, water flow dynamics, and traffic obstruction in complex visual scenes. This visual large model reduces computational complexity, improves the efficiency and accuracy of monitoring local high-risk areas in the city, and achieves real-time and refined management of risk warnings.

[0135] The third step is to use the real-time analysis capabilities of the visual big model to instantly output regional waterlogging level labels (i.e., first level R, second level Y, and third level G) and warning ranges. These data are then integrated into the GIS platform to dynamically generate waterlogging risk heat maps, which facilitate intuitive display of the spatiotemporal distribution of urban waterlogging risks, provide intuitive decision-making support for emergency management departments, and ensure that resources can be quickly and accurately deployed to where they are most needed.

[0136] The fourth step is to release early warning information in combination with the real-time output of the visual big model, and notify relevant departments to make emergency response strategies in a timely manner, such as starting drainage facilities, dispatching rescue teams and materials, etc. At the same time, through an integrated multi-channel information release system, such as mobile application push, social media, radio and television, etc., timely release waterlogging warning information to the public, enhance the public's self-protection awareness and action ability, and ensure the safe and stable operation of the city and the orderly life of residents.

[0137] The above is a description of the urban waterlogging early warning method based on a neural network according to an embodiment of the present invention. It can be appreciated that compared with the urban waterlogging early warning method in the prior art, the embodiment of the present invention has the following advantages:

[0138] First, by integrating multi-source data such as urban meteorological data and online public opinion data, and analyzing and screening the waterlogging risk area set, it can broaden the data sources of urban waterlogging warning and improve the comprehensiveness of warning;

[0139] Second, only dispatching cameras in the flood risk area can improve patrol efficiency and reduce the cost of city-wide monitoring under a certain computing power;

[0140] 3. The visual large model obtained through neural network training can patrol the urban area to accurately locate the warning range and regional waterlogging level labels in the city, which can realize intelligent screening, classification and domain monitoring of risk areas, effectively reduce the consumption of computing resources, and significantly enhance the accuracy and timeliness of urban waterlogging warnings;

[0141] 4. Release responsive warning information based on regional waterlogging grade labels to facilitate relevant departments to match response mechanisms and rationally allocate social resources.

[0142] Reference Fig.10 The embodiment of the present invention further provides an urban waterlogging early warning system based on a neural network, comprising:

[0143] The risk area perception module is used to obtain the urban meteorological data and online public opinion data of the target city, and obtain the urban waterlogging risk area set based on the urban meteorological data and online public opinion data;

[0144] The risk area monitoring module is used to dispatch cameras in the waterlogging risk area set to obtain current monitoring data;

[0145] The risk area identification module is used to obtain a large visual model based on the preset neural network training, and then obtain the regional waterlogging level label and warning range according to the current monitoring data through the large visual model;

[0146] The risk warning module is used to issue corresponding warning information according to the regional waterlogging level label and warning range.

[0147] The contents of the above-mentioned neural network-based urban waterlogging warning method embodiment are all applicable to the present neural network-based urban waterlogging warning system embodiment. The functions specifically implemented by the present neural network-based urban waterlogging warning system embodiment are the same as those of the above-mentioned neural network-based urban waterlogging warning method embodiment, and the beneficial effects achieved are also the same as those achieved by the above-mentioned neural network-based urban waterlogging warning method embodiment.

[0148] The embodiment of the present invention also provides an electronic device, the electronic device comprising: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, and when the program is executed by the processor, the above-mentioned urban waterlogging early warning method based on neural network is realized. The electronic device can be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0149] like Fig.11 FIG. 1 is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention, referring to FIG. Fig.11 , an embodiment of the present invention provides an electronic device, including:

[0150] The processor 1001 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention.

[0151] The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 can store an operating system and other application programs. When the technical solution provided in the embodiments of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1002, and the processor 1001 calls and executes the urban waterlogging early warning method based on a neural network in the embodiment of the present invention;

[0152] Input / output interface 1003, used to implement information input and output;

[0153] The communication interface 1004 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0154] A bus 1005 , which transmits information between various components of the device (e.g., the processor 1001 , the memory 1002 , the input / output interface 1003 , and the communication interface 1004 );

[0155] The processor 1001 , the memory 1002 , the input / output interface 1003 and the communication interface 1004 are connected to each other in communication within the device via the bus 1005 .

[0156] An embodiment of the present invention also provides a storage medium, which is a computer-readable storage medium used for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned urban waterlogging early warning method based on neural network.

[0157] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0158] The embodiment of the present invention also discloses a computer program product or a computer program, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 1 The method shown.

[0159] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the above-mentioned boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0160] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified to the contrary, one or more of the above-mentioned functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0161] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the above methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0162] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0163] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the above-mentioned program is printed, since the above-mentioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or processing in other suitable ways as necessary, and then stored in a computer memory.

[0164] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0165] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0166] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0167] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for early warning of urban waterlogging based on a neural network, characterized in that: The following steps are involved: Obtaining urban meteorological data and online public opinion data of a target city, and obtaining a set of waterlogging risk areas according to the urban meteorological data and the online public opinion data; Dispatching cameras in the waterlogging risk area set to obtain current monitoring data; A visual large model is obtained based on a preset neural network training, and then a regional waterlogging grade label and a warning range are obtained through the visual large model according to the current monitoring data; Corresponding warning information is issued according to the regional waterlogging grade label and the warning range.

2. The urban waterlogging early warning method based on neural network according to claim 1 is characterized in that: The urban waterlogging early warning method further comprises: The map information of the target city is obtained, and the target city is rasterized according to the map information to obtain a raster map of the target city.

3. The urban waterlogging early warning method based on neural network according to claim 2 is characterized in that: The step of obtaining urban meteorological data and online public opinion data of the target city and obtaining a set of waterlogging risk areas according to the urban meteorological data and the online public opinion data specifically includes: Acquire the urban meteorological data, construct a precipitation prediction model according to the urban meteorological data, and obtain a first waterlogging risk area set according to the precipitation prediction model; Acquire the network public opinion data, and obtain a second waterlogging risk area set and a third waterlogging risk area set according to the network public opinion data; The first waterlogging risk area set, the second waterlogging risk area set and the third waterlogging risk area set are fused to obtain the waterlogging risk area set.

4. The urban waterlogging early warning method based on neural network according to claim 3 is characterized in that: The step of constructing a precipitation prediction model according to the urban meteorological data and obtaining a first set of waterlogging risk areas according to the precipitation prediction model specifically includes: Acquire historical rainfall data and historical meteorological data of the target city, and pre-process the historical rainfall data and the historical meteorological data; The pre-processed historical rainfall data and the historical meteorological data are merged with the target city grid map to obtain parameter features; The back propagation neural network is trained by the parameter characteristics to obtain the precipitation prediction model; Inputting the urban meteorological data into the precipitation prediction model to obtain a regional precipitation sequence; The first set of urban flood risk areas is obtained according to the total number of areas in the target city, the regional precipitation sequence and the preset risk area division cardinality.

5. The urban waterlogging early warning method based on neural network according to claim 3 is characterized in that: The network public opinion data includes historical public opinion data and current public opinion data, and the acquiring of the network public opinion data and obtaining the second waterlogging risk area set and the third waterlogging risk area set according to the network public opinion data specifically include: Acquiring the historical public opinion data and the current public opinion data; Preprocessing the historical public opinion data, performing cluster analysis on the preprocessed historical public opinion data, and obtaining the frequency of mentions of waterlogging; Performing sentiment analysis on the preprocessed historical public opinion data to obtain a sentiment score; Obtaining a regional risk index according to the frequency of mentioning waterlogging and the sentiment score; According to the historical public opinion data, the regional risk index and a preset first risk index threshold, a historical risk area is obtained, and then the historical risk area is mapped to the target city grid map to obtain the second waterlogging risk area set; According to the current public opinion data, the regional risk index and a preset second risk index threshold, the current risk area is obtained, and then the current risk area is mapped to the target city grid map to obtain the third waterlogging risk area set.

6. The urban waterlogging early warning method based on neural network according to claim 1, characterized in that: The visual large model is obtained based on the preset neural network training, specifically including: Acquire historical waterlogging videos and historical waterlogging images of the target city; Performing graded labeling on the historical waterlogging videos and the historical waterlogging images to obtain training data containing waterlogging grade labels; The neural network is trained using the training data to obtain the large visual model.

7. The urban waterlogging early warning method based on neural network according to claim 2 is characterized in that: The obtaining of the regional waterlogging grade label and warning range according to the current monitoring data by the visual big model specifically includes: Analyze the water depth, water flow dynamics, and traffic obstruction status in the current monitoring data through the visual big model to obtain a waterlogging grade label in the area; Determine the warning range according to the location information of the camera, the target city grid map and the regional waterlogging level label; The regional waterlogging level labels include the first level, the second level and the third level.

8. A neural network-based urban flood warning system, characterized in that: include: A risk area perception module is used to obtain urban meteorological data and online public opinion data of a target city, and obtain a waterlogging risk area set based on the urban meteorological data and the online public opinion data; A risk area monitoring module is used to schedule cameras in the waterlogging risk area set to obtain current monitoring data; A risk area identification module is used to obtain a visual large model based on a preset neural network training, and then obtain a regional waterlogging grade label and a warning range according to the current monitoring data through the visual large model; The risk warning module is used to issue corresponding warning information according to the regional waterlogging grade label and the warning range.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the steps of the urban waterlogging early warning method based on a neural network as described in any one of claims 1 to 7 are realized.

10. A storage medium, the storage medium being a computer-readable storage medium, used for computer-readable storage, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the urban waterlogging early warning method based on a neural network as described in any one of claims 1 to 7.

Images