CN-121997413-A - Digital twinning-based harbor area navigation capacity assessment method and system

Abstract

The invention provides a digital twin-based port area navigation capability assessment method and system, which are used for integrating unmanned aerial vehicle oblique photography, laser point cloud data and a BIM parameterization assembly library to construct a self-adaptive modeling framework, acquiring global topography and building layout through laser point cloud scanning port facility details and an unmanned aerial vehicle, inputting two types of data into a pre-trained port scene recognition model, automatically matching a standardized module in the BIM parameterization assembly library to generate an initial three-dimensional model, reserving a parameterization adjustment interface of attributes of the initial three-dimensional model, establishing a physical and virtual dynamic calibration and standard adaptation dual mechanism, deploying a multidimensional sensing network at an edge node to acquire physical entity operation data in real time, eliminating data noise through a Kalman filtering algorithm, comparing the cleaned data with a virtual model simulation result, dynamically revising physical parameters of the model, synchronously formulating a port digital twin-modeling unified interface protocol compatible with a multi-system data format, and defining a model collaborative interaction rule.

Inventors

• YUE XIAOHAN
• HAO WEI
• SUN ZHICHAO
• LIU MING
• HE JIANFEI

Assignees

• 青岛市交通科学研究院

Dates

Publication Date

20260508

Application Date

20251226

Claims (10)

1. 1. The digital twinning-based harbor area navigation capacity assessment method is characterized by comprising the following steps of: The unmanned aerial vehicle oblique photography, the laser point cloud data and the BIM parameterization component library are fused to construct a self-adaptive modeling framework, the laser point cloud is used for scanning port facility details, the unmanned aerial vehicle obtains global topography and building layout, two types of data are input into a pre-trained port scene recognition model, a standardized module in the BIM parameterization component library is automatically matched to generate an initial three-dimensional model, and a parameterization adjustment interface of attributes is reserved for the initial three-dimensional model; Establishing a physical and virtual dynamic calibration and standard adaptation dual mechanism, deploying a multi-dimensional sensing network at an edge node to acquire physical entity operation data in real time, removing data noise through a Kalman filtering algorithm, comparing the cleaned data with a virtual model simulation result, dynamically correcting model physical parameters, synchronously formulating a port digital twin modeling unified interface protocol compatible with a multi-system data format, and defining a model collaborative interaction rule; Dividing a harbor model into a microscopic equipment component layer, a mesoscopic operation unit layer and a macroscopic universe system layer by adopting a trans-scale hierarchical heterogeneous modeling and intelligent linkage technology, wherein the microscopic layer is stored in a cloud by adopting a high-fidelity finite element model and is only called when a scene is analyzed with high precision, the mesoscopic layer is deployed at an edge node by adopting a lightweight model for retaining operation characteristics, the macroscopic layer is a topology model for focusing global indexes, and the hierarchical precision is automatically switched according to evaluation requirements by an intelligent linkage algorithm; Constructing a lightweight model, an edge and cloud cooperative computing architecture, adopting a model self-adaptive simplifying algorithm based on an attention mechanism, dynamically adjusting the precision level according to scene priority, enabling edge nodes to be responsible for real-time sensing data receiving, mesoscopic and macroscopic model simulation and simple decision computation, guaranteeing quick response by utilizing low-delay characteristics, enabling a cloud focusing microscopic model to perform high-fidelity simulation, historical data mining and model training optimization, and transmitting parameters and model updating fragments through an edge and cloud data increment synchronization mechanism; Introducing reinforcement learning-driven full life cycle closed-loop optimization, inputting a virtual model simulation result and harbor operation data as feedback signals into a reinforcement learning model, continuously optimizing a physical rule mapping and statistical prediction algorithm of the model, visually recognizing harbor facility changes through a computer and automatically triggering model structure updating, establishing a model precision evaluation index system, automatically checking model performance at regular intervals, and applying history modeling experience to a new scene through transfer learning; A scene self-adaptive calculation power scheduling algorithm is constructed, calculation resources are dynamically allocated based on harbor work load, when complex scenes trigger mass data processing, edge node cluster cooperative calculation is started, simulation tasks are disassembled into parallel subtasks and distributed through a load balancing algorithm, a model caching mechanism is adopted to store a high-frequency calling model, repeated calculation and data transmission are reduced, and on the premise that high accuracy of a core scene is ensured, simulation delay is controlled within a threshold value required by real-time evaluation.
2. 2. The method for estimating the navigation capacity of a harbor area based on digital twin according to claim 1, wherein the method for constructing a self-adaptive modeling framework by fusing unmanned aerial vehicle oblique photography, laser point cloud data and BIM parameterization component library, acquiring global topography and building layout by laser point cloud scanning harbor facility details and unmanned aerial vehicles, inputting the two types of data into a pre-trained harbor scene recognition model, automatically matching a standardized module in the BIM parameterization component library to generate an initial three-dimensional model, wherein the initial three-dimensional model reserves a parameterization adjustment interface of attributes, and comprises the following steps: Combing port area modeling requirements, defining fusion rules of laser point clouds, unmanned aerial vehicle oblique photography data and BIM parameterization component libraries, defining data interaction standards and format conversion protocols, constructing a BIM parameterization component library, classifying and constructing standardized modules according to port facility types, presetting editable attribute parameters for each module, and simultaneously establishing a module feature index library of related facility identification features; carrying out omnibearing scanning on harbor facility details by adopting ground laser scanning equipment to acquire data, planning an unmanned aerial vehicle oblique photography route and setting shooting parameters to acquire image data of topography and building layout covering the whole harbor area, and synchronously recording shooting auxiliary information; Denoising and registering the laser point cloud data, preprocessing the unmanned aerial vehicle oblique photographic image, generating a dense point cloud and digital orthophoto map, extracting terrain and building information, and carrying out coordinate conversion and fusion on the processed laser point cloud data and unmanned aerial vehicle related data to form a three-dimensional data body with a uniform format; After the preprocessed three-dimensional data body is processed according to a preset format, inputting a pre-trained port scene recognition model, extracting facility features through the model, comparing the facility features with a built-in feature library, and completing facility type recognition and positioning; Calling a feature index library of the BIM parameterized component library based on a scene recognition result, automatically matching a standardized module through a similarity matching algorithm, carrying out parameter self-adaptive adjustment on the matched module by combining size information in a three-dimensional data body, and automatically assembling each module according to a facility space position relation to generate an initial three-dimensional model covering the whole domain of a harbor area; And (3) reserving an attribute parameterization adjustment interface on the initial three-dimensional model component by adopting a parameterization programming technology, associating the interface with a BIM component library to preset attribute parameters, performing accuracy verification on the initial model, reversely fine-adjusting module parameters if the deviation exceeds a threshold value, and finally outputting the flexibly adjustable harbor initial three-dimensional model.
3. 3. The method for estimating the navigation ability of the harbor area based on the digital twin according to claim 2, wherein the steps of processing the preprocessed three-dimensional data volume according to a preset format, inputting a pre-trained harbor scene recognition model, extracting facility features through the model and comparing with a built-in feature library to complete the recognition and positioning of the facility types include: Analyzing the original format of the preprocessed three-dimensional data body, converting the original format into a unified format according to the input requirement of a pre-training model, reserving information, removing redundant attributes, resampling the data to be regular into a data set with preset dimensions to ensure the unification of the input dimensions, and normalizing the resampled data to eliminate characteristic extraction deviation caused by the difference of different facility sizes; Performing lightweight enhancement operation on the standardized point cloud data, improving robustness of the model to port facility posture change and environmental interference, and if the pre-training model is a two-dimensional convolution architecture, converting the three-dimensional point cloud data into a multi-view two-dimensional depth map, reserving mapping relation between depth information of each view and coordinates, and ensuring that the subsequent mapping to a three-dimensional space can be reversed; Loading a deep learning model which is pre-trained based on a port scene data set, importing a pre-training weight file, extracting parameters of the first half part of a network by fixed features, activating an inference mode of an output layer and a feature matching layer, initializing a built-in feature library of the model, wherein the feature library is a standard feature vector set of various port facilities, classifying and storing according to the types of the facilities, and constructing an index structure to improve the feature comparison efficiency; the input data after format processing is transmitted into a pre-training model, spatial alignment, local feature extraction and global feature fusion are sequentially carried out through a feature extraction network, a global feature vector capable of representing the overall structure and morphological features of the facility is generated, and then normalization processing is carried out on the global feature vector so as to eliminate scale difference; Inputting the global feature vector into a feature matching module of a model, searching similar feature templates in a built-in feature library through an index structure, calculating the similarity between the features to be matched and the features of each template, setting a similarity threshold value to screen candidate feature templates, counting the number of facility categories corresponding to the candidate templates to determine a primary recognition result, and if a plurality of category similarities are close, extracting local detail features of a region to be recognized for secondary comparison to determine the type of final facility; The method comprises the steps of positioning key feature points of a facility based on a feature point thermodynamic diagram output by a model, obtaining coordinates of the key feature points in a model input coordinate system, converting the feature point coordinates into a harbor region global unified coordinate system through a predefined coordinate mapping matrix, fitting a boundary frame of the facility according to the feature point coordinates, determining accurate positions and spatial postures of the facility in a harbor region three-dimensional space, and recording the positioning confidence of the facility; And carrying out consistency verification on the preliminary identification positioning result by combining with priori knowledge of harbor scenes, removing results of logic conflict with surrounding environments, marking facilities with positioning confidence coefficient lower than a preset threshold as suspicious areas, carrying out feature extraction and comparison again through supplementary data, finally outputting a structured identification positioning result comprising facility types, accurate coordinates, boundary frame parameters and identification confidence coefficient, and associating the result with a three-dimensional data body.
4. 4. The method for estimating the harbor navigation ability based on digital twin according to claim 1, wherein the steps of establishing a physical and virtual dynamic calibration and standard adaptation dual mechanism, deploying a multi-dimensional sensing network at an edge node to collect physical entity operation data in real time, removing data noise by a kalman filtering algorithm, comparing the cleaned data with a virtual model simulation result, dynamically modifying model physical parameters, synchronously formulating a harbor digital twin modeling unified interface protocol compatible with a multi-system data format, and defining model cooperative interaction rules comprise: Dividing a monitoring area according to port area physical entity distribution, arranging multi-dimensional sensing equipment in an edge node coverage area, determining each sensor installation point to comprehensively reflect an entity running state, configuring sensor acquisition parameters, transmitting acquired physical entity running data to an edge node local data buffer area in real time by adopting a mixed transmission mode, and setting a data transmission checking mechanism; Deploying a Kalman filtering algorithm module at an edge node, initializing a filtering parameter, setting a corresponding equation according to the type of the sensing data, inputting the original sensing data into the filtering module, removing random noise through a prediction and update iteration process, outputting an optimal estimated value, performing secondary cleaning on the filtered data, detecting and removing abnormal values, supplementing missing data, and performing standardization processing to form a physical entity operation data set; constructing a data comparison engine, setting a comparison period linked with data acquisition frequency, carrying out association matching on the cleaned physical entity operation data and corresponding simulation data output by a virtual model according to a preset mapping relation, calculating the difference degree of the two types of data by adopting a deviation calculation model, setting a deviation threshold value, triggering a model parameter correction process when the deviation exceeds the threshold value, and recording comparison information to form a comparison log; Starting a parameter correction engine based on a deviation calculation result, positioning virtual model physical parameters corresponding to the deviation, establishing a mapping model of the deviation and the parameter correction, calculating the correction according to the mapping model, carrying out iterative correction on the physical parameters corresponding to the virtual model according to a preset correction step length, driving the virtual model to simulate again and compare again until the deviation is lower than a threshold value, forming a closed loop calibration flow, and recording parameter correction information; Various systems related to digital twinning of a port area are combed, data source information, data formats, transmission protocols and interaction requirements of the systems are collected through interface debugging and document analysis modes, characteristics and difference points of data of the systems are analyzed, compatibility problems are summarized, requirements that an interface protocol needs to be covered are clearly unified, and a protocol requirement specification is formed; a unified interface protocol framework is designed based on a demand analysis result, a layered architecture is adopted, an application layer is taken as a core, a data coding rule, an interface calling specification and a message format are defined, a unified data field naming rule and a data type are established, a protocol communication specification is formulated, a synchronous/asynchronous transmission mode is supported, each system adaptation interface is designed, interface parameters, calling modes and error code definitions are defined explicitly, and a data encryption transmission and identity authentication mechanism is added to ensure transmission safety; Defining a model collaborative interaction rule by combining modeling requirements, wherein the model collaborative interaction rule comprises a data interaction time sequence rule, a data sharing authority rule and a conflict resolution rule, formulating a multi-model collaborative linkage flow, defining data interaction triggering conditions, constructing a verification environment to access simulation data of each system, testing protocol compatibility and rule validity, and optimizing the protocol and rule according to test results; integrating the dynamic calibration mechanism and the standard adaptation mechanism into an edge node unified management platform to realize data flow closed loop, deploying a monitoring module on the platform, monitoring the operation state of the double mechanisms in real time, triggering an alarm when the double mechanisms are abnormal, periodically collecting operation data and analyzing indexes, and iteratively optimizing filtering parameters, deviation thresholds, protocol details and collaborative rules by combining the port operation demand change.
5. 5. The method for estimating the harbor navigation ability based on digital twin according to claim 1, wherein the harbor model is divided into a micro device component layer, a mesoscopic operation unit layer and a macroscopic global system layer by using a trans-scale hierarchical heterogeneous modeling and intelligent linkage technology, the micro layer is stored in the cloud by using a high-fidelity finite element model, the mesoscopic layer is deployed at an edge node by using a lightweight model for retaining operation characteristics only when a scene is analyzed with high precision, the macroscopic layer is a topology model for focusing global indexes, and the hierarchical precision is automatically switched according to the estimation requirement by using an intelligent linkage algorithm, and the method comprises the following steps: Defining definition standards and targets of each level of a micro-device component layer, a mesoscopic operation unit layer and a macroscopic universe system layer, carding association mapping relations among each level, establishing an entity and parameter mapping table among the levels, and defining elements and precision requirements of modeling of each level; The microscopic equipment component layer builds a high-fidelity finite element model based on the three-dimensional geometric data and design parameters of the component, performs fine processing on key parts and moderately simplifies non-key structures, the mesoscopic operation unit layer eliminates microscopic details based on outline features and operation flows of the microscopic layer model by adopting a model light-weight technology, retains the operation features to generate a light-weight model, the macroscopic universe system layer adopts a topology abstraction method, abstracts mesoscopic operation units into topology nodes and association among operation units into topology links, and builds node and link topology models and associates parameters and indexes; Building a cloud and edge node collaborative storage architecture, storing a microscopic layer high-fidelity finite element model in a cloud distributed database in a classified manner, building a model index library, only calling when a scene is analyzed with high precision, deploying a mesoscopic layer lightweight model in a corresponding edge node local storage according to an operation unit type, adopting a redundancy backup mechanism to ensure real-time calling, and deploying a macroscopic layer topology model in the cloud and the edge node simultaneously; Constructing a linkage decision engine, defining a hierarchical switching trigger condition matrix comprising an evaluation requirement type, a data precision requirement, a response delay requirement and an abnormal event type, constructing an intelligent linkage algorithm based on a rule engine and a machine learning algorithm, processing clear trigger conditions, optimizing a hierarchical switching strategy and designing a hierarchical switching smooth transition mechanism; Designing a standardized linkage interface and a corresponding calling mode for the three-layer model, adapting to hierarchical switching requests with different real-time requirements, defining a unified data interaction protocol, defining a data transmission format and a data field, establishing a checking mechanism and a fault-tolerant mechanism of data interaction, and deploying a data gateway at an edge node; The three-layer heterogeneous model, the intelligent linkage algorithm and the linkage interface are integrated to a harbor digital twin unified management platform, a simulation test environment is built, the switching trigger accuracy, the response delay and the data synchronization accuracy of test levels of different evaluation requirement scenes are simulated, test indexes are collected, the optimization direction is analyzed, test point application is carried out on actual harbor scenes, and the hierarchy division standard, the model accuracy and the linkage logic are iteratively optimized by combining real operation data.
6. 6. The digital twinning-based harbor area navigation ability assessment method according to claim 5, wherein the integrating the three-layer heterogeneous model, the intelligent linkage algorithm and the linkage interface into the harbor area digital twinning unified management platform, constructing the simulation test environment, simulating the switching trigger accuracy, the response delay and the data synchronization accuracy of different assessment requirement scene test levels, collecting the test indexes, analyzing the optimization direction, and performing test point application in the actual harbor area scene comprises: Technical architecture characteristics of a three-layer heterogeneous model, an intelligent linkage algorithm and a linkage interface are combed, input and output parameters and operation dependent environments of each module are defined, each module is packaged based on a micro-service architecture, independent micro-services are formed through splitting, communication standards and registration mechanisms of each micro-service are defined, a harbor digital twin unified management platform framework is built, a data access layer, a model management layer, an algorithm engine layer, an interface gateway layer and a visual display layer are integrated, and functions of each layer are defined; The method comprises the steps of deploying the packaged three-layer heterogeneous model micro-service according to corresponding nodes, establishing model indexes and calling links through a model management layer to achieve unified scheduling, integrating an intelligent linkage algorithm into an algorithm engine layer, configuring interactive links of the algorithm and other levels, integrating a standardized linkage interface into an interface gateway layer, configuring an interface routing rule and an authority control strategy, developing a module joint scheduling test, verifying the cooperativity of model calling, algorithm decision and interface communication, and repairing various synergic problems in the joint scheduling process; Building a cloud and edge cooperative hardware architecture consistent with a real harbor area, building a mixed transmission network to simulate a real data transmission environment, building a software environment corresponding to the cloud and edge nodes, installing a test tool and a log collection system, building a test data set based on real harbor operation data, generating simulation data covering different assessment scenes, and marking expected results and core requirements corresponding to the data; Combing port area typical evaluation requirement scenes and defining test targets of each scene, designing test cases aiming at each scene, and defining test steps, input data and expected output; According to the test case sequence, executing single-scene independent test and multi-scene concurrent test, in the test process, recording the level switching response time and the data synchronization deviation through a test tool, recording the trigger result and the running state of each module through a log collecting system, marking the test abnormal scene and recording related information; Sorting test data, counting test indexes, setting an index qualification threshold, comparing test results with the threshold, screening indexes which are not up to standard and corresponding scenes, analyzing reasons for the indexes which are not up to standard, positioning an optimization direction and formulating a targeted optimization scheme; Carrying out optimization work according to an optimization scheme, carrying out regression testing by multiplexing the original test cases after optimization is finished, verifying whether indexes reach standards and the influence of the optimization scheme on other modules, and repeating the optimization and regression testing flow until the indexes meet the requirements; Selecting a representative test point area which covers corresponding entities of the three-layer model and has complete operation flow, carrying out field investigation on the test point area, adjusting platform parameters and interface configuration to adapt to the actual environment, deploying a unified management platform in the test point area, butting a field real sensor network with an operation system, completing cloud and edge node configuration and building a visual monitoring interface; Starting test point area platform test operation, monitoring the platform operation state in real time, recording actual operation data, collecting feedback comments of on-site staff on the platform, and recording actual problems in the test operation process; and comprehensively analyzing test run data and feedback comments, evaluating the suitability and practicality of the platform, optimizing the platform aiming at the problems found in test run by combining with the actual conditions of the site, and verifying again in a test point area after optimization is completed.
7. 7. The method for estimating the harbor navigation ability based on digital twin according to claim 1, wherein the constructing a lightweight model and edge-cloud collaborative computing architecture, adopting a model self-adaptive simplified algorithm based on an attention mechanism, dynamically adjusting the precision level according to scene priority, enabling edge nodes to be responsible for real-time sensing data receiving, mesoscopic and macroscopic model simulation and simple decision computation, guaranteeing quick response by using low-delay characteristics, enabling high-fidelity simulation of a cloud focusing microscopic model, historical data mining and model training optimization, and transmitting parameters and model updating fragments by an edge-cloud data increment synchronization mechanism, comprises: Combing the digital twin modeling and evaluation requirements of the harbor area, defining the functional boundary of edge and cloud cooperation, dividing the scene priority level, matching the corresponding model precision strategy, designing the whole framework of the edge real-time processing and cloud depth computing cooperation framework, and defining the interaction relation between the data flow and the modules; Constructing an algorithm framework to determine input and output elements, designing a scene feature attention extraction module, calculating feature attention weights, developing dynamic precision adjustment logic, adopting a differential simplification strategy according to scene priorities and the attention weights, and embedding a model complexity monitoring module; Aiming at the mesoscopic and macro-level models, carrying out light weight processing based on a self-adaptive simplification algorithm, generating a multi-precision level model version, establishing a precision and scene priority mapping table, retaining core features according to level characteristics, eliminating redundant details, carrying out performance test on the light weight model, and backtracking to optimize simplification parameters; selecting adaptive edge node hardware, configuring related components, deploying a software framework and middleware required by a lightweight operating system and real-time processing, configuring a communication link between an edge node and a sensor, starting a local data caching mechanism, deploying a model management module, preloading a corresponding precision model and establishing a quick calling index; deploying a cloud server cluster, configuring a storage and calculation component, installing an operating system, a distributed database, high-fidelity simulation software and a machine learning framework, constructing a cloud model training and optimizing module, deploying a cloud and edge collaborative management platform and definitely integrating functions; Defining task allocation rules of the edge nodes and the cloud end, designing a dynamic scheduling engine, monitoring the computing force load of the edge nodes and the task queue state in real time, and formulating a task migration and update instruction issuing strategy; Defining the increment synchronous data types of the edge and the cloud, adopting a synchronous trigger mechanism of a timing synchronization and event triggering synchronization dual-mode design, adopting a data compression and differential coding technology to process synchronous data, deploying a synchronous check and fault tolerance module and formulating a data conflict processing principle; Integrating the modules into a unified management platform to develop joint debugging test, verifying smoothness of a business process, a cloud high-fidelity simulation process and an edge and cloud increment synchronization process, testing response and synchronization performance in different scenes, and repairing various problems in the joint debugging process; deploying a collaborative architecture in a harbor area test point area, accessing real sensing data and an operation scene to perform test operation, monitoring the operation states of edge nodes, cloud end and data synchronization in real time, collecting test operation data and feedback, analyzing the balance effect of model precision and calculation efficiency, and iterating and optimizing algorithm strategies, calculation force configuration and synchronization mechanisms.
8. 8. The method for estimating the harbor navigation ability based on digital twin according to claim 7, wherein the method for combing the harbor digital twin modeling and estimating requirements, defining the functional boundaries of edge and cloud cooperation, dividing the scene priority level and matching the corresponding model precision strategies, designing the whole framework of the collaborative architecture of edge real-time processing and cloud depth computing, defining the interaction relation between the data flow and the modules comprises the following steps: Acquiring the requirements of each relevant party of the harbor district by adopting a combined investigation mode, defining the accuracy and consistency requirements of a modeling layer and the scene requirements of an evaluation layer, synchronously collecting real-time response, calculation power, bandwidth and compatibility constraint conditions, carding to form a requirement list, completing priority sorting, and outputting a requirement specification; Disassembling the combed requirements into modeling and evaluation specific indexes, and defining quantization standards and definitions of the indexes; defining task ranges of edge nodes and cloud ends based on the instantaneity and calculation force requirements of the requirement indexes, determining real-time processing and local caching tasks of the edge nodes, high calculation force of the cloud ends and non-real-time processing tasks, and determining cooperative connection points and data interaction contents of the edge nodes and the cloud ends; Dividing the scenes into three levels of priority by combining the harbor district operation characteristics and the demand priority, and defining the coverage range and the concerned object of each level of scenes; establishing differential model precision strategies aiming at scenes with different priorities, establishing mapping relations between scene priorities and model precision and simplifying strategies, and determining calculation amount adaptation requirements of each scene model; Adopting a layered thought to construct a four-level collaborative architecture of a perception layer, an edge layer, a cloud layer and an application layer, and defining the requirements of constituent modules, functions, software and hardware deployment of each level; Data links among all levels and modules of the framework are combed, and transmission content and modes of all data flows are defined to form a closed-loop data link; defining the interactive links of the modules in the architecture, defining task allocation, data synchronization, fault switching and authority management rules, and writing module interactive specifications; and verifying the functional coverage and index satisfaction of the collaborative architecture against the requirement specification, and adjusting the architecture design aiming at the suitability problem until the architecture completely adapts to the requirement.
9. 9. The method for estimating the harbour area navigation ability based on digital twinning according to claim 7, wherein the construction algorithm framework determines input and output elements, designs a scene feature attention extraction module, calculates feature attention weights, develops dynamic accuracy adjustment logic, adopts a differential simplification strategy according to scene priority and attention weights, and embeds a model complexity monitoring module, comprising: The explicit algorithm aims to realize scene self-adaptive precision adjustment and light weight of the harbor digital twin model, adapt the computing capacity of edge nodes, build a modularized algorithm frame based on logic, the system comprises an input analysis module, a scene feature attention extraction module, a weight calculation module, a dynamic accuracy adjustment module, a differential simplification execution module, a complexity monitoring feedback module and an output encapsulation module, wherein the input analysis module, the scene feature attention extraction module, the weight calculation module, the dynamic accuracy adjustment module, the differential simplification execution module, the complexity monitoring feedback module and the output encapsulation module define functional boundaries and data interaction specifications of the modules, and define data flow forms, transmission protocols and triggering conditions among the modules; The input elements comprise original three-dimensional model data, scene priority labels, real-time evaluation demand parameters and edge node calculation force state data, analysis rules, coding specifications and transmission interfaces of various input data are formulated, the output elements comprise self-adaptive simplified lightweight model files, model precision configuration reports and complexity monitoring result reports, and packaging formats of the output data and interface specifications of a downstream module are formulated; Constructing a harbor scene feature system, classifying and carding geometric features, operation features and physical features, building a feature classification dictionary and an attribute description system, building a feature extraction network architecture, developing a feature post-processing module, carrying out structural integer on extracted feature vectors to generate a feature map, and training an optimization network through a labeling data set to ensure that feature recognition and positioning accuracy reach the standard; Determining the inherent importance of the features, the priority of the scene and the real-time evaluation requirement as the weight calculation core dimension, designing a weight calculation model of multi-dimension weighting fusion, carrying out normalization processing on the initial weight, developing a weight dynamic calibration mechanism, updating weight parameters based on historical feedback data, and building a weight visualization module; Constructing an accuracy adjustment rule base based on scene priority and attention weight, establishing a three-dimensional mapping relation, defining technical parameters corresponding to each accuracy level, developing a dynamic accuracy adjustment decision engine, combining an edge node calculation force state fine adjustment accuracy level, writing an accuracy adjustment logic code, reserving a rule updating interface, and adding a logic verification module; Aiming at different scene priorities, attention weights and precision grade combinations, a differential simplification strategy system is formulated, a simplified algorithm component library is developed, a simplified execution scheduling module is constructed, corresponding algorithm components are called according to precision adjustment instructions to execute simplified operations, and a simplified log is recorded; defining a model complexity evaluation index system, defining each index calculation method and acquisition frequency, developing a complexity monitoring data acquisition module, setting index thresholds in different scenes, designing monitoring and feedback closed-loop logic, linking a monitoring module with a dynamic accuracy adjustment module, and developing a complexity monitoring visualization module; Completing the integrated butt joint of each module to open a data interface, developing single-module function test and end-to-end full-flow joint debugging test, verifying the module function and full-flow operation stability, and optimizing module parameters and interaction logic; and constructing a test data set covering the full scene type of the harbor district, carrying out algorithm performance batch verification, counting performance indexes, analyzing the root for the test cases which do not reach the standard, optimizing algorithm parameters, and carrying out iterative verification until all the performance indexes meet the requirements to form an algorithm performance verification report.
10. 10. A digital twinning-based port area capability assessment system, comprising: A processor; A machine-readable storage medium storing machine-executable instructions for the processor; Wherein the processor is configured to perform the digital twin based estuary navigability assessment method of any of claims 1 to 9 via execution of the machine executable instructions.

Description

Digital twinning-based harbor area navigation capacity assessment method and system Technical Field The invention relates to the technical field of port area capability assessment, in particular to a digital twin-based port area navigation capability assessment method and system. Background The current port area capacity assessment faces the core problem that modeling precision and global coverage are difficult to consider, the traditional modeling method relies on a single data source or manual modeling, the fusion of laser point clouds, unmanned aerial vehicle images and BIM component libraries lacks unified standards, the problems of incompatibility of data formats, characteristic extraction deviation and the like lead to insufficient initial model precision, and model attribute adjustment flexibility is poor, so that the method is difficult to adapt to complex and diverse facility types and layout changes of the port area. Meanwhile, an effective dynamic calibration mechanism is lacking between a physical entity and a virtual model, noise interference of sensing data is large, deviation between a simulation result and actual operation data is obvious, data interaction protocols among different systems are not uniform, and a barrier exists in model cooperative interaction, so that reliability of an evaluation result is seriously affected. In terms of computing architecture and optimization mechanism, the traditional harbor district evaluation system mostly adopts a single computational effort deployment mode, so that the real-time response and high-precision simulation requirements are difficult to balance, the delay is too high in a core operation scene, and the high-precision analysis faces the dilemma of insufficient computational effort. In addition, the model lacks the adaptive capacity of a cross-scale level, the precision level cannot be dynamically switched according to the evaluation requirement, a full life cycle optimization mechanism is not established, the performance of the model is difficult to continuously iterate through operation data feedback, the model has poor adaptability when facing the scenes such as port facility updating, operation flow adjustment and the like, the repeated modeling cost is high, and the long-term and dynamic capacity evaluation requirement of a port area cannot be met. Disclosure of Invention In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a digital twin-based port area navigability assessment method, which includes: The unmanned aerial vehicle oblique photography, the laser point cloud data and the BIM parameterization component library are fused to construct a self-adaptive modeling framework, the laser point cloud is used for scanning port facility details, the unmanned aerial vehicle obtains global topography and building layout, two types of data are input into a pre-trained port scene recognition model, a standardized module in the BIM parameterization component library is automatically matched to generate an initial three-dimensional model, and a parameterization adjustment interface of attributes is reserved for the initial three-dimensional model; Establishing a physical and virtual dynamic calibration and standard adaptation dual mechanism, deploying a multi-dimensional sensing network at an edge node to acquire physical entity operation data in real time, removing data noise through a Kalman filtering algorithm, comparing the cleaned data with a virtual model simulation result, dynamically correcting model physical parameters, synchronously formulating a port digital twin modeling unified interface protocol compatible with a multi-system data format, and defining a model collaborative interaction rule; Dividing a harbor model into a microscopic equipment component layer, a mesoscopic operation unit layer and a macroscopic universe system layer by adopting a trans-scale hierarchical heterogeneous modeling and intelligent linkage technology, wherein the microscopic layer is stored in a cloud by adopting a high-fidelity finite element model and is only called when a scene is analyzed with high precision, the mesoscopic layer is deployed at an edge node by adopting a lightweight model for retaining operation characteristics, the macroscopic layer is a topology model for focusing global indexes, and the hierarchical precision is automatically switched according to evaluation requirements by an intelligent linkage algorithm; Constructing a lightweight model, an edge and cloud cooperative computing architecture, adopting a model self-adaptive simplifying algorithm based on an attention mechanism, dynamically adjusting the precision level according to scene priority, enabling edge nodes to be responsible for real-time sensing data receiving, mesoscopic and macroscopic model simulation and simple decision computation, guaranteeing quick response by utilizin