Smart city digital management method and system based on live-action three dimensions

Abstract

The invention discloses a smart city digital management method and system based on real scene three-dimensional, and the method comprises the steps: dividing management units according to city scene data, generating a real scene three-dimensional city map, carrying out the function edge processing, obtaining city management features, constructing a city management platform, and carrying out the real scene three-dimensional city map. And directly inputting the live-action three-dimensional city map into the city management platform, encrypting and transmitting city management characteristics of different functional terminals to the city management platform, screening resident opinions uploaded by resident terminals to obtain city opinions, carrying out platform edge processing to obtain city fusion management characteristics, and sending the city fusion management characteristics to the city management platform. And constructing a smart city cloud early warning and prediction model according to the city fusion management features to obtain a city early warning and prediction result, and generating a live-action three-dimensional city information map. The method not only can improve the efficiency and accuracy of smart city digital management, but also has good interpretability, and can be directly applied to a smart city digital management system.

Description

Technical Field

[0001] The present invention relates to the field of urban management, and in particular to a digital management method and system for a smart city based on real-scene three-dimensional (3D) technology. Background Art

[0002] With the acceleration of global urbanization, urban management systems are facing unprecedented challenges. The concept of smart cities has emerged, using information technology to improve urban governance. Real-life 3D technology, as the key to smart city construction, restores real-life urban scenes through 3D modeling, providing an intuitive basis for urban management and playing an important role in urban planning, traffic management, emergency response and other fields.

[0003] Traditional urban management methods have numerous shortcomings: First, limited data collection methods make it difficult to comprehensively and accurately acquire multi-source, heterogeneous urban data, resulting in a lack of data support for urban management decisions. Second, insufficient data processing and analysis capabilities make it difficult to effectively tap into the potential value behind the data and achieve a deep understanding and precise control of complex urban systems. Furthermore, traditional technologies also have vulnerabilities in data security and privacy protection, which can easily lead to data leakage and abuse. Meanwhile, new technologies are constantly emerging, such as the use of drones and laser scanning technology for multi-source data collection, the integration of cloud computing and edge computing to enhance data processing capabilities, the use of encryption algorithms and distributed storage to ensure data security, and the generation of real-world three-dimensional maps by combining BIM models and satellite maps. This present invention combines these technologies to propose a real-world three-dimensional smart city digital management method and system, effectively overcoming the shortcomings of traditional technologies and achieving an intelligent, refined, and efficient urban management model. This has important practical significance for improving urban quality and promoting sustainable urban development. Summary of the Invention

[0004] The purpose of the present invention is to provide a digital management method and system for smart cities based on real-world three-dimensional images.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention comprises the following steps:

[0007] Obtaining urban scene data and urban management data, dividing management units according to the urban scene data, matching different modal urban scenes within the management units, and generating a real-life three-dimensional city map; the urban management data is collected by the corresponding functional departments;

[0008] Performing functional edge processing on the city management data to obtain city management features, generating path information according to path rules, and locally storing the city management data and the city management features according to the corresponding path information;

[0009] Constructing a city management platform, directly inputting the real-life three-dimensional city map into the city management platform, setting encryption transmission rules to encrypt and transmit city management features of different functional terminals to the city management platform, and setting display rules to filter resident opinions uploaded by resident terminals to obtain city opinions;

[0010] Performing platform edge processing to obtain city integration management characteristics, building a smart city cloud early warning prediction model based on the city integration management characteristics, and inputting the city management data to be analyzed into the smart city cloud early warning prediction model to obtain a city early warning prediction result;

[0011] A real-scene three-dimensional city information map is generated according to the city early warning prediction results, the city integration management characteristics, the city opinions and the real-scene three-dimensional city map; the real-scene three-dimensional city information map is composed of multiple management units.

[0012] Furthermore, the method for generating a real-life three-dimensional city map includes:

[0013] The urban scene data includes satellite street view maps, building information, drone scanning data and city light data;

[0014] Adaptive edge detection is performed using a Canny operator to calculate a dynamic threshold to extract boundary features of urban functional groups in satellite street view maps. A U-Net deep learning model is used to perform pixel classification on the edge-detected satellite street view maps to output vector boundaries of urban functional groups for semantic segmentation. The satellite street view maps are gridded using the H3 geographic grid system to determine management units. Image feature extraction is performed on the satellite street view maps corresponding to the management units to obtain topographic features. The urban functional groups include roads, rivers, parks, and buildings.

[0015] The city light data map is converted into city coordinates, and radiation correction and light intensity are performed to identify the population density. Linear regression is performed based on the light intensity to obtain the population density. The population density is associated with the corresponding management unit based on the city coordinates. The population density expression is:

[0016]

[0017] Among them, P d is the population density of the d grid, α1 and α2 are the linear regression weights of light intensity, τ1 and τ2 are the light intensity attenuation parameters, and L is the light intensity;

[0018] Importing building information into REVIT software to generate an initial BIM model, converting drone coordinates of drone point cloud data into city coordinates, aligning the initial BIM model with the drone point cloud data on the city coordinates, extracting SIFT key points from the point cloud data and the initial BIM model for feature matching, calculating the initial rigid body transformation matrix and optimizing it using an improved iterative closest point algorithm, calculating the Hausdorff distance between the corresponding surfaces of the point cloud data and the initial BIM model for difference detection and annotating the differences, and adjusting the initial BIM model based on the differences to obtain the city BIM model; the improved iterative closest point algorithm is combined with a kd tree algorithm and singular value decomposition;

[0019] A support vector machine classifier is used to classify buildings according to the urban BIM model to obtain the building use, and the building use is associated with the urban BIM model. The satellite street view maps in different management units are aligned with the urban BIM model according to the city coordinates, and the RGB information of the satellite street view maps is projected onto the surface of the urban BIM model to generate a real-life three-dimensional city map.

[0020] Furthermore, the method of locally storing the city management data and the city management features according to the corresponding path information includes:

[0021] Different types of urban management data are processed by corresponding functional departments to obtain different types of urban management characteristics; the urban management characteristics include perception network characteristics, demographic characteristics, resource demand characteristics, traffic characteristics, transaction characteristics, network characteristics and security characteristics;

[0022] According to the path rules, path information of different types of urban management features and urban management data is generated, and different types of urban management features and urban management data are stored locally in the functional departments according to the path information; the path rules specifically mean that the path information needs to cover the functional departments, data categories, coordinate areas and collection times.

[0023] Furthermore, the city management platform includes a management end, a resident end and a functional end; the management end is used for managers to authenticate their identities and obtain management authority; the management authority includes viewing real-life three-dimensional city information maps, marking and storing historical city opinions, and linking the functional end to view corresponding city management data based on path information of urban integrated management characteristics; the resident end is used for city residents to view the opinions of nearby residents and upload resident opinions; the functional end is used for different functional departments to transmit corresponding city management characteristics to the city management platform, and store corresponding city management characteristics and city management data locally.

[0024] Furthermore, the method of setting encryption transmission rules to encrypt and transmit city management features of different functional terminals to the city management platform includes:

[0025] Encrypting urban management features using encryption transmission rules; the encryption transmission rules include a hybrid layered encryption algorithm and a dynamic fragmentation transmission protocol;

[0026] The specific steps of the hybrid layered encryption algorithm are: using the BLAKE3 hash function to process the plaintext data to obtain a hash value, using AES encryption and GCM mode to symmetric encrypt the hash value, and using the elliptic curve encryption algorithm to asymmetric encrypt the symmetric encrypted ciphertext to obtain the encryption result. The expressions of symmetric encryption and asymmetric encryption are:

[0027]

[0028] C final =ECC secp521r1 (K sym ∥Nonce)∥C1

[0029] Where C1 is the ciphertext after symmetric encryption, C final is the ciphertext after asymmetric encryption, It is a symmetric encryption algorithm that combines AES encryption with GCM mode, K sym is a dynamically generated 256-bit session key, M is plaintext data, i.e., the city management features that need to be encrypted for transmission, and Hash blake3 (·) is the BLAKE3 hash function, Timestamp is the corresponding timestamp, ECC secp521r1 (·) is the elliptic curve encryption algorithm using the secp521r1 curve, K sym The session key is the same as that in the symmetric encryption phase, and Nonce is a one-time random number;

[0030] Generate public and private keys based on the hybrid layered encryption algorithm. The expression is:

[0031]

[0032] where K public is the public key, used to encrypt data or verify signatures, K private is the private key used to decrypt data or generate signatures, G is the base point of the elliptic curve, d base is the dynamic base private key, mod p is the modular operation of the parameter p in the elliptic curve equation, Hash sm3 (·) is the SM3 hash function, NodeID is the unique identifier of the node, and mod n is the modular operation of the elliptic curve of order n;

[0033] The specific dynamic sharding transmission protocol is as follows: the urban management features are divided into N shards according to management units, each shard is independently encrypted, and the shards are randomly routed to different edge nodes through the Kademlia distributed hash table. Only when M shards (M < N) are collected, the data is restored by threshold decryption. The expression for independent encryption of each shard is:

[0034]

[0035] K i = HKDF(K master , Salt i )

[0036] where S i is the i-th encrypted data shard, K i is the encryption key for the i-th shard, is symmetric encryption for the i-th shard, M i is the i-th plaintext data shard, CRC32(M i ) is the CRC32 check for the plaintext data shard M i , HKDF(·) is a key derivation function, K master is the master key, and Salt i is the salt value for the i-th shard;

[0037] A privacy protocol enhancement mechanism is introduced for data availability and invisibility and fine-grained access control; the privacy protocol enhancement mechanism includes zero-knowledge proof and dynamic attribute-based encryption; the zero-knowledge proof generates a validity proof by the functional end using the Groth16 protocol, writes the validity proof and the data hash value into the blockchain, and is quickly verified by the management platform; the dynamic attribute-based encryption defines data category permissions for each management unit, binds category policies during encryption, allows decryption when the data category permissions of the management unit are consistent with the category policies, and at the same time uses a covering tree to assign permission time limits to the category permissions;

[0038] The encrypted urban management features are transmitted to the management platform, and the management platform conducts validity, category, and timeliness checks on the encrypted urban management features. After passing the checks, the encrypted urban management features are decrypted using the public key and private key to obtain the corresponding urban management features of the category, and different category urban management features are associated with the corresponding management units according to the urban coordinates.

[0039] Furthermore, the method for setting display rules to filter the resident opinions uploaded by the resident side to obtain urban opinions includes:

[0040] City residents use the resident terminal to authenticate their identity and determine their management unit, then log in to the city management platform to obtain resident privileges. The authentication includes ID number verification and city coordinate verification. Resident privileges include viewing a real-life 3D city map, viewing the opinions of residents within their management unit, and liking and commenting on the opinions of residents within their management unit.

[0041] The display rules specifically include: setting an opinion timeliness threshold, displaying resident opinions within the opinion timeliness threshold on the resident side, setting a heat threshold, selecting resident opinions with heat greater than the heat threshold as city opinions to display on the management side, and updating the management side city opinions based on the opinion timeliness threshold; the heat of the resident opinions is calculated by weighting the number of likes and comments on the corresponding resident opinions.

[0042] Furthermore, the method of performing platform edge processing to obtain urban integration management features includes:

[0043] The platform edge processing includes edge correction and construction of perception factors;

[0044] The specific steps of the edge correction are: using clustering to classify resource demand characteristics into a first demand characteristic, a second demand characteristic, a third demand characteristic and a fourth demand characteristic, performing linear regression prediction on the first demand characteristic, the second demand characteristic and the third demand characteristic to obtain a first population prediction, a second population prediction and a third population prediction respectively, performing feature fusion on the first population prediction, the second population prediction and the third population prediction to obtain a population prediction characteristic, inputting the fourth demand characteristic into the resident food database to obtain a first population structure prediction characteristic, extracting network features to obtain a second population structure prediction characteristic, intersecting the first population structure prediction characteristic with the second population structure prediction characteristic to obtain a population structure prediction characteristic, performing feature splicing on the population structure prediction characteristic and the population prediction characteristic to obtain a demographic prediction characteristic, and performing feature fusion on the demographic prediction characteristic and the demographic characteristic to obtain a corrected demographic characteristic;

[0045] The first, second and third demand characteristics are related to water resources, electricity and gas respectively; the fourth demand characteristic is related to food structure; the resident food database reflects the relationship between food type and population gender, age and ethnicity; the first population structure prediction characteristics include population gender, age and ethnicity; the second population structure prediction characteristics include population gender, age and occupation;

[0046] The specific steps of constructing the perception factor are: inputting the perception network feature into the perception function to obtain the perception factor; the perception function expression is:

[0047]

[0048] Where Q is the perception factor, R int R is the short-term rainfall intensity score, spa is the spatial heterogeneity score, W is the drainage efficiency dynamic index, β1 is the drainage efficiency dynamic index power coefficient, β2 is the topological robustness factor power coefficient, β1 and β2 are obtained by fitting historical data, l j is the length of the j-th segment of the pipeline, is the health of the j-segment pipeline, is the material aging rate, t is the service life, M is the number of pipeline sections, F acc is the weather forecast accuracy, F0 is the standard weather forecast accuracy, λ1 and λ2 are the attenuation coefficients, T resp is the response delay time, T target Target response delay time, N sensor is the number of time-sensitive nodes, N total is the total number of monitoring nodes, I t is the current rainfall intensity, I max is the maximum rainfall intensity in history, T is the duration of rainfall, and T crit is the critical rainfall time, N is the number of management units, A i is the monitoring area of ​​management unit i, is the rainfall gradient of management unit i, A total is the total monitoring area of ​​the city, q real is the real-time flow rate of the pipe network, q design is the design flow rate of the pipeline network, γ is the blockage influence coefficient, obtained by fitting historical data, S clog is the pipe blockage rate, ΔH is the upstream and downstream water level difference, H crit is the critical safety water level;

[0049] The perception factors, modified demographic characteristics, topographical features, building uses, perception network characteristics, resource demand characteristics, traffic characteristics, transaction characteristics, network characteristics and security characteristics within the same management unit are combined to form the urban integration management characteristics.

[0050] Furthermore, the method for obtaining the city early warning forecast result includes:

[0051] The urban integration management features are randomly divided into training set and test set according to the ratio of 7:3;

[0052] Constructing a smart city cloud early warning prediction model, comprising a data early warning module, a sudden meteorological event prediction module, and a BP variable prediction module;

[0053] The data warning module detects abnormal patterns of urban integration management characteristics other than intra-unit perception factors through high-dimensional spatial mapping to obtain early risk warnings of corresponding categories. Specifically, it includes a graph attention network and a hypersphere detection algorithm.

[0054] The sudden meteorological event prediction module uses a spatiotemporal graph convolutional network to learn the propagation laws of different meteorological elements and predict extreme meteorological events based on real-time data;

[0055] The BP variable prediction module uses a causal inference network to learn the causal relationship of urban integration management features other than perception network features, and predicts the evolution trend of the target city's integration management features, including a causal discovery layer and an inference layer; the causal discovery layer uses conditional independence testing to construct a feature causal graph; the inference layer dynamically adjusts the input feature weights based on the gated recursive unit of the causal relationship; the BP variable prediction module uses random forest importance ranking to screen the top-20 key features, uses quantile loss and causal consistency constraints to evaluate the trend prediction error of urban integration management features, and uses a lookahead optimizer and gradient clipping to adjust the optimization efficiency;

[0056] Use the training set to train the model, use the test set to evaluate the model performance, and output the smart city cloud early warning prediction model;

[0057] Input the city management data to be analyzed into the smart city cloud early warning prediction model to obtain the city early warning prediction results

[0058] The second aspect is a smart city digital management system based on real-scene 3D, including:

[0059] City map module: used to obtain city scene data and city management data, divide the management units according to the city scene data, match different modal city scenes within the management units, and generate a real-life 3D city map;

[0060] Edge processing module: used to perform functional edge processing on the city management data to obtain city management features, generate path information according to path rules, store the city management data and the city management features locally according to the corresponding path information, and perform platform edge processing to obtain city integration management features;

[0061] Cloud early warning prediction model module: used to build a smart city cloud early warning prediction model based on the city integrated management characteristics, and input the city management data to be analyzed into the smart city cloud early warning prediction model to obtain the city early warning prediction results;

[0062] Management platform module: used to connect the management end with the resident end and the functional end, for managers to view real-life three-dimensional city information maps, mark and store historical city opinions, connect the functional end to view corresponding city management data based on the path information of urban integrated management characteristics, for city residents to view the opinions of nearby residents and upload residents' opinions, for different functional end departments to transmit corresponding city management characteristics to the city management platform, and store corresponding city management characteristics and city management data locally; used to notify managers and corresponding functional end departments based on the city early warning forecast results.

[0063] The beneficial effects of the present invention are:

[0064] The present invention is a digital management method and system for smart cities based on real-world 3D. Compared with the existing technology, the present invention has the following technical effects:

[0065] The present invention can improve the data preprocessing capabilities and enhance the adaptability of models in the digital management of smart cities through the steps of dividing management units, functional edge processing, building a city management platform, encrypted transmission, platform edge processing and model construction, and can improve the security of urban management data transmission, thereby improving the efficiency and accuracy of smart city digital management. It can optimize the digital management technology of smart cities, greatly save resources, improve work efficiency, and realize multi-dimensional and real-time viewing of urban management data, providing more reliable technical support for the digital management of smart cities, helping to form a government-resident collaborative governance model, and realizing dynamic deduction and risk warning of urban operation status. It can adapt to different smart city digital management systems and the digital management needs of smart cities of different users, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flowchart of the steps of a digital management method for a smart city based on real-scene three-dimensional technology of the present invention. DETAILED DESCRIPTION

[0067] The present invention will be further described below through specific examples. The illustrative examples and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0068] The present invention provides a digital management method and system for a smart city based on real-scene three-dimensional imagery, comprising the following steps:

[0069] like Figure 1 As shown, in this embodiment, the following steps are included:

[0070] Obtaining urban scene data and urban management data, dividing management units according to the urban scene data, matching different modal urban scenes within the management units, and generating a real-life three-dimensional city map; the urban management data is collected by the corresponding functional departments;

[0071] Performing functional edge processing on the city management data to obtain city management features, generating path information according to path rules, and locally storing the city management data and the city management features according to the corresponding path information;

[0072] Constructing a city management platform, directly inputting the real-life three-dimensional city map into the city management platform, setting encryption transmission rules to encrypt and transmit city management features of different functional terminals to the city management platform, and setting display rules to filter resident opinions uploaded by resident terminals to obtain city opinions;

[0073] Performing platform edge processing to obtain city integration management characteristics, building a smart city cloud early warning prediction model based on the city integration management characteristics, and inputting the city management data to be analyzed into the smart city cloud early warning prediction model to obtain a city early warning prediction result;

[0074] A real-scene three-dimensional city information map is generated according to the city early warning prediction results, the city integration management characteristics, the city opinions and the real-scene three-dimensional city map; the real-scene three-dimensional city information map is composed of multiple management units.

[0075] In this embodiment, the method for generating a realistic three-dimensional city map includes:

[0076] The urban scene data includes satellite street view maps, building information, drone scanning data and city light data;

[0077] Adaptive edge detection is performed using a Canny operator to calculate a dynamic threshold to extract boundary features of urban functional groups in satellite street view maps. A U-Net deep learning model is used to perform pixel classification on the edge-detected satellite street view maps to output vector boundaries of urban functional groups for semantic segmentation. The satellite street view maps are gridded using the H3 geographic grid system to determine management units. Image feature extraction is performed on the satellite street view maps corresponding to the management units to obtain topographic features. The urban functional groups include roads, rivers, parks, and buildings.

[0078] The city light data map is converted into city coordinates, and radiation correction and light intensity are performed to identify the population density. Linear regression is performed based on the light intensity to obtain the population density. The population density is associated with the corresponding management unit based on the city coordinates. The population density expression is:

[0079]

[0080] Among them, Pd is the population density of the d grid, α1 and α2 are the linear regression weights of light intensity, τ1 and τ2 are the light intensity attenuation parameters, and L is the light intensity;

[0081] Importing building information into REVIT software to generate an initial BIM model, converting drone coordinates of drone point cloud data into city coordinates, aligning the initial BIM model with the drone point cloud data on the city coordinates, extracting SIFT key points from the point cloud data and the initial BIM model for feature matching, calculating the initial rigid body transformation matrix and optimizing it using an improved iterative closest point algorithm, calculating the Hausdorff distance between the corresponding surfaces of the point cloud data and the initial BIM model for difference detection and annotating the differences, and adjusting the initial BIM model based on the differences to obtain the city BIM model; the improved iterative closest point algorithm is combined with a kd tree algorithm and singular value decomposition;

[0082] A support vector machine classifier is used to classify buildings according to the city BIM model to obtain building uses. The building uses are then associated with the city BIM model. Satellite street view maps in different management units are aligned with the city BIM model according to city coordinates. The RGB information of the satellite street view maps is projected onto the surface of the city BIM model to generate a real-life 3D city map.

[0083] In the actual evaluation, for a 100km 2 The total area of ​​the city is digitally managed to obtain urban landscape data, where the satellite image resolution is 0.5m / pixel and the drone point cloud density is 1000 points / ㎡. The city is divided into 85 management units based on the satellite street view map (H3 grid resolution 8, each management unit is about 1.2km 2 , each management unit covers about 5 blocks). Taking the management unit H3-8-B1 / 5 as an example, the linear regression weights α1 and α2 of light intensity are set to 0.78 and 1.15, and the light intensity attenuation parameters τ1 and τ2 are set to 1.2 and 85. The urban light data map of this management unit corresponds to a light intensity of 120, and the population density of this management unit is obtained as 310 people / km 2 ;

[0084] Management unit H3-8-B1 / 5 includes one commercial area, three residential areas, and one park. The resolution of the drone point cloud data is 0.05m and the coverage area is 1.18km. 2 ,The initial error of the Hausdorff distance corresponding to the conversion of the drone ,point cloud data to city coordinates was 0.3m.,The differences were marked and the model was adjusted. Finally, the ,difference was reduced to 0.05m. The building classification results were ,60% residential, 20% commercial, 10% public facilities, and 10% industrial;

[0085] The satellite street view maps within 85 management units are aligned with the city BIM model according to the city coordinates, and the RGB information of the satellite street view maps is projected onto the surface of the city BIM model to generate a real-life three-dimensional city map.

[0086] In this embodiment, the method for locally storing the city management data and the city management features according to the corresponding path information includes:

[0087] Different types of urban management data are processed by corresponding functional departments to obtain different types of urban management characteristics; the urban management characteristics include perception network characteristics, demographic characteristics, resource demand characteristics, traffic characteristics, transaction characteristics, network characteristics and security characteristics;

[0088] Generate path information of different types of urban management features and urban management data according to path rules, and store the different types of urban management features and urban management data locally in the functional department according to the path information; the path rules specifically mean that the path information needs to cover the functional department, data category, coordinate area and collection time;

[0089] In actual evaluations, the perception network characteristics are related to data collected by urban pipe networks and meteorological systems; the demographic characteristics include spatial distribution, gender, age, cultural occupation, and family ethnicity; the resource demand characteristics are related to the resources necessary for the normal work and life of urban residents, specifically including water demand, electricity demand, gas demand, and food demand; the traffic characteristics specifically include traffic flow, traffic facilities, traffic modes, traffic land occupation, traffic greening, and traffic attraction points; the transaction characteristics are related to transaction events occurring in the region, specifically including transaction type, transaction volume, transaction price, transaction channel, and security characteristics; the network characteristics reflect the network operation status in the region, specifically including network infrastructure, network coverage, network operation indicators, and network security; the network operation indicators specifically include network speed and bandwidth, network stability, and network latency; the security characteristics reflect the living safety environment of residents in the region, specifically including the number of police calls, police call rate, and police call type;

[0090] Taking the generation of path information of traffic characteristics and resource demand characteristics within management unit H3-8-B1 / 5 as an example, the corresponding paths (functional department / data category / management unit center coordinate area / collection time) are obtained: 1. Department A / traffic flow / 116.40E-39.90N / 20231001T100000, 2. Department B / water resource demand / 116.40E-39.90N / 20231001T080000.

[0091] In this embodiment, the city management platform includes a management terminal, a resident terminal, and a functional terminal; the management terminal allows managers to authenticate their identities and obtain management permissions; the management permissions include viewing real-life three-dimensional city information maps, marking and storing historical city opinions, and linking the functional terminal to view corresponding city management data based on the path information of urban integrated management characteristics; the resident terminal allows city residents to view the opinions of nearby residents and upload resident opinions; the functional terminal allows different functional departments to transmit corresponding city management characteristics to the city management platform and store the corresponding city management characteristics and city management data locally;

[0092] In the actual assessment, administrator GL001-8-B1 / 5 can obtain management rights for the corresponding management unit H3-8-B1 / 5 by logging into the city management platform through the management terminal. Department A JT001 can upload urban traffic characteristics by logging into the management platform through the functional terminal. Residents (urban resident number 11010119900101XXXX) can view and upload nearby residents' opinions by logging into the management platform through the resident terminal (the management platform calls the C database to verify the validity of the urban resident number and obtains the current location 116.402° / 39.905° through GPS, which matches the coordinate range of the management unit H3-8-12345 and is determined to be valid).

[0093] In this embodiment, the method of setting encryption transmission rules to encrypt and transmit city management features of different functional terminals to the city management platform includes:

[0094] Encrypting urban management features using encryption transmission rules; the encryption transmission rules include a hybrid layered encryption algorithm and a dynamic fragmentation transmission protocol;

[0095] The specific steps of the hybrid layered encryption algorithm are: using the BLAKE3 hash function to process the plaintext data to obtain a hash value, using AES encryption and GCM mode to symmetric encrypt the hash value, and using the elliptic curve encryption algorithm to asymmetric encrypt the symmetric encrypted ciphertext to obtain the encryption result. The expressions of symmetric encryption and asymmetric encryption are:

[0096]

[0097] C final =ECC secp521r1 (K sym ∥Nonce)∥C1

[0098] Where C1 is the ciphertext after symmetric encryption, C final is the ciphertext after asymmetric encryption, It is a symmetric encryption algorithm that combines AES encryption with GCM mode, K symis a dynamically generated 32-bit session key, M is the plaintext data, that is, the urban management features to be encrypted and transmitted, Hash blake3 (·) is the BLAKE3 hash function, Timestamp is the corresponding timestamp, ECC secp521r1 (·) is the elliptic curve encryption algorithm using the secp521r1 curve, K sym is the same session key as in the symmetric encryption stage, Nonce is a one-time random number;

[0099] Generate public and private keys according to the hybrid hierarchical encryption algorithm, and the expression is:

[0100]

[0101] where K public is the public key, used to encrypt data or verify signatures, K private is the private key, used to decrypt data or generate signatures, G is the elliptic curve base point, d base is the dynamic reference private key, mod p is the modulo operation of parameter p in the elliptic curve equation, Hash sm3 (·) is the SM3 hash function, NodeID is the unique node identifier, mod n is the modulo operation of the n-th order of the elliptic curve;

[0102] The specific dynamic sharding transmission protocol is as follows: Divide the urban management features into N shards according to management units, encrypt each shard independently, and randomly route the shards to different edge nodes through the Kademlia distributed hash table. Only when M shards (M < N) are collected, decrypt and recover the data through the threshold. The expression for encrypting each shard independently is:

[0103]

[0104] K i = HKDF(K master , Salt i )

[0105] where S i is the i-th encrypted data shard, K i is the encryption key for the i-th shard, is the symmetric encryption of the i-th shard, M i is the i-th plaintext data shard, CRC32(M i ) is the CRC32 check of the plaintext data shard M i , HKDF(·) is the key derivation function, K master is the master key, Salt i is the salt value for the i-th shard;

[0106] A privacy protocol enhancement mechanism is introduced to achieve data availability, invisibility, and fine-grained access control. The privacy protocol enhancement mechanism includes zero-knowledge proof and dynamic attribute-based encryption. The zero-knowledge proof is generated by the functional end using the Groth16 protocol to prove its validity. The validity proof and the data hash value are written into the blockchain and quickly verified by the management platform. The dynamic attribute-based encryption defines data category permissions for each management unit, binds the category policy during encryption, and allows decryption when the management unit data category permissions are consistent with the category policy. At the same time, an overlay tree is used to assign permission time limits to category permissions.

[0107] The encrypted city management features are transmitted to the management platform, which verifies the validity, category, and timeliness of the encrypted city management features. After verification, the encrypted city management features are decrypted using the public key and private key to obtain the corresponding city management features. Different types of city management features are associated with corresponding management units according to the city coordinates.

[0108] In the actual evaluation, the encrypted transmission of the urban traffic flow of 120 vehicles per hour in the traffic characteristics of the management unit H3-8-B1 / 5 was taken as an example. A 16-bit master key 1A2B3C4D5E6F7E8H was randomly generated. The data was sharded to obtain the unique salt value 9X8Y7Z6W5V4U3T2S of the urban traffic flow. The traffic characteristics were divided into 6 shards. The encryption key of the urban traffic flow corresponding to the shard was f1rstP4tch4466. The corresponding check value crc32V1 was obtained by CRC32 check. The zero-knowledge proof z3r0Kn0wledge of the urban traffic flow was generated using the Groth16 protocol. It was written into the blockchain together with the hash value d4b0c1a2f3e4g5h6 for rapid verification by the management platform.

[0109] The BLAKE3 hash function generates a 16-bit hash value of d4b0c1a2f3e4g5h6. Symmetric encryption with AES-GCM yields the symmetric ciphertext e4c2a1f3b4d5g6h7 (corresponding to the dynamic session key s3kr3tK3y1697823). Asymmetric encryption yields the ciphertext n0nC3ph3rT3xt123 (corresponding to the one-time random number r4nd0m96485225). The public key used to encrypt the data is pUbL1cK3y6583219, and the private key used to decrypt the data is pr1v4t3K3y158743. Both public and private keys are updated every 30 days.

[0110] Define the data category permission of "Urban Traffic Flow" as "Traffic-Vehicle-Quantity", and bind the corresponding category policy during encryption. The data category permission of the management unit is consistent with the category policy and allows decryption. Transmission to the management platform is allowed within 7 days of the time limit permission.

[0111] In this embodiment, the method of setting display rules to filter resident opinions uploaded by residents to obtain city opinions includes:

[0112] City residents use the resident terminal to authenticate their identity and determine their management unit, then log in to the city management platform to obtain resident privileges. The authentication includes ID number verification and city coordinate verification. Resident privileges include viewing a real-life 3D city map, viewing the opinions of residents within their management unit, and liking and commenting on the opinions of residents within their management unit.

[0113] The display rules specifically include: setting an opinion timeliness threshold, displaying resident opinions within the opinion timeliness threshold on the resident side; setting a popularity threshold, selecting resident opinions with popularity greater than the popularity threshold as city opinions to be displayed on the management side, and updating the management side city opinions based on the opinion timeliness threshold; the popularity of the resident opinions is calculated by weighting the number of likes and comments on the corresponding resident opinions;

[0114] In the actual evaluation, management unit H3-8-B1 / 5 received 5 resident opinions. The data corresponding to the likes and comments on the resident side were (number of likes / number of comments / time): 100 / 20 / 23-10-01-09:00, 80 / 30 / 23-10-02-14:00, 150 / 40 / 23-10-03-08:00, 50 / 10 / 23-09-29-12:00, 200 / 50 / 23-10-03-21:00. Taking the time threshold as 3 days, opinion 4 was filtered out. According to the likes and comments thresholds of 0.7 and 0.3 respectively, the opinion heat was calculated to be 76, 65, 117, and 155 respectively. Resident opinions 3 and 5 (serious roadside waterlogging and travel congestion) with a heat threshold greater than 80 were screened as city opinions and displayed on the management side.

[0115] In this embodiment, the method for performing platform edge processing to obtain city integration management features includes:

[0116] The platform edge processing includes edge correction and construction of perception factors;

[0117] The specific steps of the edge correction are: using clustering to classify resource demand characteristics into a first demand characteristic, a second demand characteristic, a third demand characteristic and a fourth demand characteristic, performing linear regression prediction on the first demand characteristic, the second demand characteristic and the third demand characteristic to obtain a first population prediction, a second population prediction and a third population prediction respectively, performing feature fusion on the first population prediction, the second population prediction and the third population prediction to obtain a population prediction characteristic, inputting the fourth demand characteristic into the resident food database to obtain a first population structure prediction characteristic, extracting network features to obtain a second population structure prediction characteristic, intersecting the first population structure prediction characteristic with the second population structure prediction characteristic to obtain a population structure prediction characteristic, performing feature splicing on the population structure prediction characteristic and the population prediction characteristic to obtain a demographic prediction characteristic, and performing feature fusion on the demographic prediction characteristic and the demographic characteristic to obtain a corrected demographic characteristic;

[0118] The first, second and third demand characteristics are related to water resources, electricity and gas respectively; the fourth demand characteristic is related to food structure; the resident food database reflects the relationship between food type and population gender, age and ethnicity; the first population structure prediction characteristics include population gender, age and ethnicity; the second population structure prediction characteristics include population gender, age and occupation;

[0119] The specific steps of constructing the perception factor are: inputting the perception network feature into the perception function to obtain the perception factor; the perception function expression is:

[0120]

[0121] Where Q is the perception factor, R int R is the short-term rainfall intensity score, spa is the spatial heterogeneity score, W is the drainage efficiency dynamic index, β1 is the drainage efficiency dynamic index power coefficient, β2 is the topological robustness factor power coefficient, β1 and β2 are obtained by fitting historical data, l j is the length of the j-th segment of the pipeline, is the health of the j-segment pipeline, is the material aging rate, t is the service life, M is the number of pipeline sections, F acc is the weather forecast accuracy, F0 is the standard weather forecast accuracy, λ1 and λ2 are the attenuation coefficients, T resp is the response delay time, T target Target response delay time, N sensor is the number of time-sensitive nodes, N total is the total number of monitoring nodes, I t is the current rainfall intensity, I max is the maximum rainfall intensity in history, T is the duration of rainfall, and T critis the critical rainfall time, N is the number of management units, A i is the monitoring area of ​​management unit i, is the rainfall gradient of management unit i, A total is the total monitoring area of ​​the city, q real is the real-time flow rate of the pipe network, q design is the design flow rate of the pipeline network, γ is the blockage influence coefficient, obtained by fitting historical data, S clog is the pipe blockage rate, ΔH is the upstream and downstream water level difference, H crit is the critical safety water level;

[0122] The perception factors, modified demographic characteristics, topographical features, building uses, perception network characteristics, resource demand characteristics, traffic characteristics, transaction characteristics, network characteristics and security characteristics within the same management unit are combined into the urban integration management characteristics;

[0123] In the actual assessment, the resource demand characteristics of the management unit H3-8-B1 / 5 are obtained: the first demand characteristic is 1000 tons / day, the second demand characteristic is 5000 degrees / day, and the third demand characteristic is 2000m 3 / day, fourth demand characteristics (1,000 tons / day, rice and flour 60% / meat 20% / vegetables 15% / others 5%);

[0124] The linear regression prediction of population size obtained the first population prediction, the second population prediction and the third population prediction as 5500, 5100 and 6700 respectively, and the population prediction was 5766.

[0125] The resident food database rules are as follows: the proportion of rice and flour corresponds to the Han ethnic group and blue-collar workers, the proportion of meat corresponds to young and middle-aged people and white-collar workers, and the proportion of vegetables corresponds to women. Based on the resident food database, the first population structure prediction characteristics are obtained (80% Han ethnic group, 60% of the population aged 20-40 years, 55% of women). The network feature extraction results are 40% white-collar workers, 30% blue-collar workers, and 30% others (students, retirees, freelancers, etc.). Combined with the resident food database rules, the second population structure prediction characteristics are obtained (40% white-collar workers, 55% of the population aged 20-40 years, 48% of women). Cross-feature analysis is performed to obtain the population structure prediction characteristics (80% Han ethnic group, 40% white-collar workers, 60% of the population aged 20-40 years, 55% of women). Feature correction is performed to obtain the revised demographic characteristics (6,667 people, 80% Han ethnic group, 40% white-collar workers, 60% of the population aged 20-40 years, 55% of women).

[0126] In the management unit H3-8-B1 / 5, the drainage efficiency dynamic index power coefficient β1 is taken as 0.5, the topological robustness factor power coefficient β2 is taken as 0.3, the blockage influence coefficient γ is taken as 0.2, and the attenuation coefficients λ1 and λ2 are taken as 0.1 and 0.2 respectively;

[0127] According to the characteristics of the sensing network (pipeline service life of 5 years, material aging rate of 0.05, number of pipeline sections of 10, weather forecast accuracy of 0.8, standard weather forecast accuracy of 0.7, response delay time of 0.5h, target response delay time of 0.3h, number of time-sensitive nodes of 50, total number of monitoring nodes of 100, current rainfall intensity of 50mm / h, historical maximum rainfall intensity of 100mm / h, rainfall duration of 2h, critical rainfall time of 1h, N is the number of management units, A i is the monitoring area of ​​management unit i, is the rainfall gradient of management unit i, A total The total monitoring area of ​​the city, the real-time flow of the pipe network is 80m 3 / s, pipe network design flow rate 100m 3 / s, pipeline blockage rate 0.1, upstream and downstream water level difference 2m, critical safety water level 3m) the calculated perception factor is 0.242.

[0128] In this embodiment, the method for obtaining city early warning forecast results includes:

[0129] The urban integration management features are randomly divided into training set and test set according to the ratio of 7:3;

[0130] Constructing a smart city cloud early warning prediction model, comprising a data early warning module, a sudden meteorological event prediction module, and a BP variable prediction module;

[0131] The data warning module detects abnormal patterns of urban integration management characteristics other than intra-unit perception factors through high-dimensional spatial mapping to obtain early risk warnings of corresponding categories. Specifically, it includes a graph attention network and a hypersphere detection algorithm.

[0132] The sudden meteorological event prediction module uses a spatiotemporal graph convolutional network to learn the propagation laws of different meteorological elements and predict extreme meteorological events based on real-time data;

[0133] The BP variable prediction module uses a causal reasoning network to learn the causal relationship of urban integration management characteristics other than perception network characteristics, and predicts the evolution trend of the target city's integration management characteristics, including a causal discovery layer and an inference layer;

[0134] Use the training set to train the model, use the test set to evaluate the model performance, and output the smart city cloud early warning prediction model;

[0135] Inputting the city management data to be analyzed into the smart city cloud early warning prediction model to obtain the city early warning prediction results;

[0136] In the actual evaluation, in the data warning module: the graph attention network encodes the urban integration management features to obtain a unified feature vector; the hypersphere detection algorithm uses deep support vector data description to map normal data to the minimum hypersphere space to identify abnormal data; the data warning module uses SVDD reconstruction error and feature space sparse regularization as the loss function to evaluate the abnormal warning error, and uses Nesterov accelerated gradient descent to adjust the learning rate;

[0137] In the sudden meteorological event prediction module: the spatiotemporal graph convolutional network uses weighted cross entropy and prediction result spatial smoothness constraint as the loss function to evaluate the meteorological forecast error, and uses the AdamW optimizer to adjust the meteorological event prediction module parameters;

[0138] In the BP variable prediction module: the causal discovery layer uses conditional independence testing to construct a feature causal graph; the inference layer dynamically adjusts the input feature weights based on the gated recursive unit of causal relationships; the BP variable prediction module uses random forest importance ranking to screen the top-20 key features, uses quantile loss and causal consistency constraints to evaluate the trend prediction error of urban integration management features, and uses the Lookahead optimizer and gradient clipping to adjust optimization efficiency;

[0139] In the actual assessment, the data warning module monitored and found that the traffic volume in management units H3-8-B1 / 5, H3-9-B1 / 5 and H3-10-B1 / 5 dropped by 50% and the rainfall lasted too long, triggering the "traffic congestion risk" and "waterlogging" warnings;

[0140] In the sudden meteorological event prediction module, it is predicted that the entire city will experience traffic congestion and waterlogging caused by typhoon and heavy rain;

[0141] In the BP variable prediction module, the characteristic prediction results include a 20% increase in network traffic, an 8% average increase in demand for various resources, a 40% decrease in traffic flow, and a 20% increase in police response rate;

[0142] A real-life 3D city information map is generated based on the city's early warning forecast results, urban integrated management characteristics, city opinions, and the real-life 3D city map. The real-life 3D city information map simulates traffic congestion and heavy rainfall near management unit H3-8-B1 / 5.

[0143] The manager clicks on the management unit H3-8-B1 / 5 on the management platform to view the urban integration management characteristics of the current unit. At the same time, the manager pays attention to the city's opinions (serious roadside waterlogging, travel congestion) and warning prediction results ("traffic congestion risk", "inland waterlogging"), and accesses the functional end (Department A / Transportation, Department D / Municipality, Department E / Meteorology) based on the path information of the corresponding urban integration management characteristics (traffic characteristics, perception network characteristics) to view the corresponding urban management data and formulate a response strategy.

[0144] The second aspect is a smart city digital management system based on real-scene 3D, including:

[0145] City map module: used to obtain city scene data and city management data, divide the management units according to the city scene data, match different modal city scenes within the management units, and generate a real-life 3D city map;

[0146] Edge processing module: used to perform functional edge processing on the city management data to obtain city management features, generate path information according to path rules, store the city management data and the city management features locally according to the corresponding path information, and perform platform edge processing to obtain city integration management features;

[0147] Cloud early warning prediction model module: used to build a smart city cloud early warning prediction model based on the city integrated management characteristics, and input the city management data to be analyzed into the smart city cloud early warning prediction model to obtain the city early warning prediction results;

[0148] Management platform module: used to connect the management end with the resident end and the functional end, for managers to view real-life three-dimensional city information maps, mark and store historical city opinions, connect the functional end to view corresponding city management data based on the path information of urban integrated management characteristics, for city residents to view the opinions of nearby residents and upload residents' opinions, for different functional end departments to transmit corresponding city management characteristics to the city management platform, and store corresponding city management characteristics and city management data locally; used to notify managers and corresponding functional end departments based on the city early warning forecast results.

[0149] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A digital management method for smart cities based on real-world 3D, characterized in that: The following steps are involved: S1. Acquire urban scene data and urban management data, divide the urban scene data into management units, match different modal urban scenes within the management units, and generate a real-life three-dimensional city map; the urban management data is collected by the corresponding functional departments; S2. Performing functional edge processing on the city management data to obtain city management features, generating path information according to path rules, and locally storing the city management data and the city management features according to the corresponding path information; S3. Build a city management platform, directly input the real-life three-dimensional city map into the city management platform, set encryption transmission rules to encrypt and transmit city management features of different functional terminals to the city management platform, and set display rules to filter resident opinions uploaded by resident terminals to obtain city opinions; S4. Performing platform edge processing to obtain city integration management characteristics, constructing a smart city cloud early warning prediction model based on the city integration management characteristics, and inputting the city management data to be analyzed into the smart city cloud early warning prediction model to obtain a city early warning prediction result; S5. Generate a real 3D city information map based on the city early warning prediction results, the city integration management characteristics, the city opinions and the real 3D city map; the real 3D city information map is composed of multiple management units.

2. The method for digital management of smart cities based on real-scene 3D according to claim 1, characterized in that: The method for generating a real-life three-dimensional city map comprises: The urban scene data includes satellite street view maps, building information, drone scanning data and city light data; Adaptive edge detection is performed using a Canny operator to calculate a dynamic threshold to extract boundary features of urban functional groups in satellite street view maps. A U-Net deep learning model is used to perform pixel classification on the edge-detected satellite street view maps to output vector boundaries of urban functional groups for semantic segmentation. The satellite street view maps are gridded using the H3 geographic grid system to determine management units. Image feature extraction is performed on the satellite street view maps corresponding to the management units to obtain topographic features. The urban functional groups include roads, rivers, parks, and buildings. The city light data map is converted into city coordinates, and radiation correction and light intensity are performed to identify the population density. Linear regression is performed based on the light intensity to obtain the population density. The population density is associated with the corresponding management unit based on the city coordinates. The population density expression is: Among them, P d is the population density of the d grid, α1 and α2 are the linear regression weights of light intensity, τ1 and τ2 are the light intensity attenuation parameters, and L is the light intensity; Importing building information into REVIT software to generate an initial BIM model, converting drone coordinates of drone point cloud data into city coordinates, aligning the initial BIM model with the drone point cloud data on the city coordinates, extracting SIFT key points from the point cloud data and the initial BIM model for feature matching, calculating the initial rigid body transformation matrix and optimizing it using an improved iterative closest point algorithm, calculating the Hausdorff distance between the corresponding surfaces of the point cloud data and the initial BIM model for difference detection and annotating the differences, and adjusting the initial BIM model based on the differences to obtain the city BIM model; the improved iterative closest point algorithm is combined with a kd tree algorithm and singular value decomposition; Using a support vector machine classifier to classify buildings based on the urban BIM model to obtain the building uses, associating the building uses with the urban BIM model, aligning the satellite street view maps within different management units with the urban BIM model according to the urban coordinates, and projecting the RGB information of the satellite street view maps onto the surface of the urban BIM model to generate a real-scene three-dimensional urban map.

3. The method for digital management of smart cities based on real-scene 3D according to claim 1, characterized in that: A method for locally storing the urban management data and the urban management features according to the corresponding path information, including: Performing feature processing on different types of urban management data at the corresponding functional department ends to obtain different types of urban management features; the urban management features include perception network features, demographic features, resource demand features, traffic features, transaction features, network features, and security features; Generating path information for different types of urban management features and urban management data according to the path rules, and locally storing different types of urban management features and urban management data at the functional department ends according to the path information; the path rules specifically refer to that the path information needs to cover the functional department end, data category, coordinate area, and collection time.

4. The method for digital management of a smart city based on real-scene 3D according to claim 1, characterized in that: The urban management platform includes a management end, a resident end, and a functional end; the management end is for managers to perform identity verification to obtain management permissions; the management permissions include viewing the real-scene three-dimensional urban information map, marking and storing historical urban opinions, and linking to the functional end according to the path information of the urban integration management features to view the corresponding urban management data; the resident end is for urban residents to view the opinions of nearby residents and upload resident opinions; the functional end is for different functional department ends to transmit the corresponding urban management features to the urban management platform and locally store the corresponding urban management features and urban management data.

5. The method for digital management of smart cities based on real-scene 3D according to claim 1, characterized in that: A method for encrypting and transmitting different functional end urban management features to the urban management platform by setting encryption transmission rules, including: Encrypting the urban management features using encryption transmission rules; the encryption transmission rules include a hybrid hierarchical encryption algorithm and a dynamic sharding transmission protocol; The specific steps of the hybrid hierarchical encryption algorithm are as follows: using the BLAKE3 hash function to process the plaintext data to obtain a hash value, using AES encryption and the GCM mode to perform symmetric encryption on the hash value, and using the elliptic curve encryption algorithm to perform asymmetric encryption on the symmetric encryption ciphertext to obtain an encryption result. The expressions for symmetric encryption and asymmetric encryption are: C1=AES-GCM Ksym [M∥Hash blake3 (M)∥Timestamp] C final =ECC secp521r1 (K sym ∥Nonce)∥C1 Where C1 is the ciphertext after symmetric encryption, C final Asymmetric encryption ciphertext, AES-GCM Ksym (·) is the symmetric encryption algorithm of AES encryption combined with GCM mode, K sym is a dynamically generated 256-bit session key, M is plaintext data, i.e., the city management features that need to be encrypted for transmission, and Hash blake3 (·) is the BLAKE3 hash function, Timestamp is the corresponding timestamp, ECC secp521r1 (·) is the elliptic curve encryption algorithm using the secp521r1 curve, K sym The session key is the same as that in the symmetric encryption phase, and Nonce is a one-time random number; Generating a public key and a private key according to the hybrid hierarchical encryption algorithm, the expression is: where K public is the public key, used to encrypt data or verify signatures, K private is the private key used to decrypt data or generate signatures, G is the base point of the elliptic curve, d base is the dynamic base private key, mod p is the modular operation of the parameter p in the elliptic curve equation, Hash sm3 (·) is the SM3 hash function, NodeID is the unique identifier of the node, and mod n is the modular operation of the elliptic curve of order n; The dynamic sharding transmission protocol is specifically: dividing the urban management features into N shards according to the management unit, independently encrypting each shard, randomly routing the shards to different edge nodes through the Kademlia distributed hash table, and only decrypting and restoring the data through threshold when M shards (M < N) are collected. The expression for independently encrypting each shard is: K i =HKDF(K master ,Salt i ) Among them S i is the i-th encrypted data shard, K i is the encryption key of the i-th shard, To symmetric encrypt i shards, M i For the i-th plaintext data fragment, CRC32(M i ) is the fragmentation of plaintext data M i Perform CRC32 check, HKDF(·) is the key derivation function, K master Master key, Salt i is the salt value of the i-th shard; A privacy protocol enhancement mechanism is introduced to achieve data availability, invisibility, and fine-grained access control. The privacy protocol enhancement mechanism includes zero-knowledge proof and dynamic attribute-based encryption. The zero-knowledge proof is generated by the functional end using the Groth16 protocol to prove its validity. The validity proof and the data hash value are written into the blockchain and quickly verified by the management platform. The dynamic attribute-based encryption defines data category permissions for each management unit, binds the category policy during encryption, and allows decryption when the management unit data category permissions are consistent with the category policy. At the same time, an overlay tree is used to assign permission time limits to category permissions. The encrypted urban management features are transmitted to the management platform, which verifies the validity, category and timeliness of the encrypted urban management features. After the verification, the encrypted urban management features are decrypted using the public key and private key to obtain the corresponding urban management features. Different types of urban management features are associated with corresponding management units according to the city coordinates.

6. The method for digital management of smart cities based on real-scene 3D according to claim 1, characterized in that: The method of setting display rules to filter the resident opinions uploaded by the resident terminal to obtain the city opinions includes: City residents use the resident terminal to authenticate their identity and determine their management unit, then log in to the city management platform to obtain resident privileges. The authentication includes ID number verification and city coordinate verification. Resident privileges include viewing a real-life 3D city map, viewing the opinions of residents within their management unit, and liking and commenting on the opinions of residents within their management unit. The display rules specifically include: setting an opinion timeliness threshold, displaying resident opinions within the opinion timeliness threshold on the resident side, setting a heat threshold, selecting resident opinions with heat greater than the heat threshold as city opinions to display on the management side, and updating the management side city opinions based on the opinion timeliness threshold; the heat of the resident opinions is calculated by weighting the number of likes and comments on the corresponding resident opinions.

7. The method for digital management of smart cities based on real-scene 3D according to claim 1, characterized in that: The method for performing platform edge processing to obtain urban integration management features includes: The platform edge processing includes edge correction and construction of perception factors; The specific steps of the edge correction are: using clustering to classify resource demand characteristics into a first demand characteristic, a second demand characteristic, a third demand characteristic and a fourth demand characteristic, performing linear regression prediction on the first demand characteristic, the second demand characteristic and the third demand characteristic to obtain a first population prediction, a second population prediction and a third population prediction respectively, performing feature fusion on the first population prediction, the second population prediction and the third population prediction to obtain a population prediction characteristic, inputting the fourth demand characteristic into the resident food database to obtain a first population structure prediction characteristic, extracting network features to obtain a second population structure prediction characteristic, intersecting the first population structure prediction characteristic with the second population structure prediction characteristic to obtain a population structure prediction characteristic, performing feature splicing on the population structure prediction characteristic and the population prediction characteristic to obtain a demographic prediction characteristic, and performing feature fusion on the demographic prediction characteristic and the demographic characteristic to obtain a corrected demographic characteristic; The first, second and third demand characteristics are related to water resources, electricity and gas respectively; the fourth demand characteristic is related to food structure; the resident food database reflects the relationship between food type and population gender, age and ethnicity; the first population structure prediction characteristics include population gender, age and ethnicity; the second population structure prediction characteristics include population gender, age and occupation; The specific steps of constructing the perception factor are: inputting the perception network feature into the perception function to obtain the perception factor; the perception function expression is: Where Q is the perception factor, R int R is the short-term rainfall intensity score, spa is the spatial heterogeneity score, W is the drainage efficiency dynamic index, β1 is the drainage efficiency dynamic index power coefficient, β2 is the topological robustness factor power coefficient, β1 and β2 are obtained by fitting historical data, l j is the length of the j-th segment of the pipeline, is the health of the j-segment pipeline, is the material aging rate, t is the service life, M is the number of pipeline sections, F acc is the weather forecast accuracy, F0 is the standard weather forecast accuracy, λ1 and λ2 are the attenuation coefficients, T resp is the response delay time, T target Target response delay time, N sensor is the number of time-sensitive nodes, N total is the total number of monitoring nodes, I t is the current rainfall intensity, I max is the maximum rainfall intensity in history, T is the duration of rainfall, and T crit is the critical rainfall time, N is the number of management units, A i is the monitoring area of ​​management unit i, is the rainfall gradient of management unit i, A total is the total monitoring area of ​​the city, q real is the real-time flow rate of the pipe network, q design is the design flow rate of the pipeline network, γ is the blockage influence coefficient, obtained by fitting historical data, S clog is the pipe blockage rate, ΔH is the upstream and downstream water level difference, H crit is the critical safety water level; The perception factors, modified demographic characteristics, topographical features, building uses, perception network characteristics, resource demand characteristics, traffic characteristics, transaction characteristics, network characteristics and security characteristics within the same management unit are combined to form the urban integration management characteristics.

8. The method for digital management of smart cities based on real-scene 3D according to claim 1, characterized in that: The method for obtaining city early warning prediction results includes: The urban integration management features are randomly divided into training set and test set according to the ratio of 7:3; Constructing a smart city cloud early warning prediction model, comprising a data early warning module, a sudden meteorological event prediction module, and a BP variable prediction module; The data warning module detects abnormal patterns of urban integration management characteristics other than intra-unit perception factors through high-dimensional spatial mapping to obtain early risk warnings of corresponding categories. Specifically, it includes a graph attention network and a hypersphere detection algorithm. The sudden meteorological event prediction module uses a spatiotemporal graph convolutional network to learn the propagation laws of different meteorological elements and predict extreme meteorological events based on real-time data; The BP variable prediction module uses a causal reasoning network to learn the causal relationship of urban integration management characteristics other than perception network characteristics, and predicts the evolution trend of the target city's integration management characteristics, including a causal discovery layer and an inference layer; Use the training set to train the model, use the test set to evaluate the model performance, and output the smart city cloud early warning prediction model; The city management data to be analyzed is input into the smart city cloud early warning prediction model to obtain the city early warning prediction results.

9. A smart city digital management system based on real-scene 3D, used to execute the method according to any one of claims 1 to 8, characterized in that: include: City map module: used to obtain city scene data and city management data, divide the management units according to the city scene data, match different modal city scenes within the management units, and generate a real-life 3D city map; Edge processing module: used to perform functional edge processing on the city management data to obtain city management features, generate path information according to path rules, store the city management data and the city management features locally according to the corresponding path information, and perform platform edge processing to obtain city integration management features; Cloud early warning prediction model module: used to build a smart city cloud early warning prediction model based on the city integrated management characteristics, and input the city management data to be analyzed into the smart city cloud early warning prediction model to obtain the city early warning prediction results; Management platform module: used to connect the management end with the resident end and the functional end, for managers to view real-life three-dimensional city information maps, mark and store historical city opinions, connect the functional end to view corresponding city management data based on the path information of urban integrated management characteristics, for city residents to view the opinions of nearby residents and upload residents' opinions, for different functional end departments to transmit corresponding city management characteristics to the city management platform, and store corresponding city management characteristics and city management data locally; used to notify managers and corresponding functional end departments based on the city early warning forecast results.

Images