Intelligent engineering survey data analysis system

Abstract

The application discloses an intelligent engineering measurement data analysis system and relates to the technical field of engineering measurement data analysis.The system comprises a processing module, an analysis module and an interaction module.The original engineering measurement data is received, spatio-temporal registration and feature extraction are performed, the obtained feature information is subjected to information fusion, a fusion feature set is generated, the analysis result and the decision instruction are obtained by analyzing the fusion feature set and the second analysis source, and the analysis result and the decision instruction are displayed through the interaction module.The application significantly improves the accuracy and efficiency of engineering structure health monitoring by performing integrated analysis and intelligent decision on multi-source measurement data, perceives and predicts the system behavior by using a digital twin model and a graph neural network to construct a component correlation, adopts a synthetic rule to fuse multi-source conflict decision, improves the robustness of risk judgment in a complex environment, and provides a systematic solution for realizing intelligent operation and maintenance of infrastructure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of engineering measurement data analysis, and particularly relates to an intelligent engineering measurement data analysis system. BACKGROUND

[0002] Traditional engineering structure monitoring mainly relies on single sensor data and manual inspection, which is difficult to realize effective fusion and collaborative analysis of multi-source heterogeneous data, cannot accurately assess the impact of local anomalies on the overall structure, has low analysis efficiency and insufficient early warning accuracy, and cannot meet the urgent needs of intelligent management and maintenance of large-scale engineering infrastructure.

[0003] At present, the Chinese invention with the application number CN202411186584.6, an engineering data analysis method and system for engineering management. The method includes extracting personnel behavior images and building edge information from engineering monitoring information; dividing the dangerous area according to the edge information; obtaining key action information by analyzing the personnel behavior images in the dangerous area; judging whether there is dangerous behavior according to the key action information, and triggering high-altitude falling warning if there is. The invention improves the engineering safety warning ability by accurately dividing the dangerous area and behavior analysis, and ensures personnel safety. However, the related technology cannot perform spatio-temporal registration and deep fusion on multi-source measurement data, which is not conducive to forming unified digital perception of the engineering object, cannot perform component correlation analysis based on physical connection, lacks comprehensiveness, cannot solve the conflict problem between multi-source decision-making evidence through synthesis rules, and is not conducive to making reliable decisions under contradictory conditions. SUMMARY

[0004] The technical problem solved by the present application is that related technology cannot perform spatio-temporal registration and deep fusion on multi-source measurement data, which is not conducive to forming unified digital perception of the engineering object, cannot perform component correlation analysis based on physical connection, lacks comprehensiveness, cannot solve the conflict problem between multi-source decision-making evidence through synthesis rules, and is not conducive to making reliable decisions under contradictory conditions.

[0005] To solve the above technical problems, the present application provides the following technical solutions:

[0006] The intelligent engineering measurement data analysis system comprises a processing module, an analysis module and an interaction module.

[0007] The processing module is used for receiving original engineering measurement data, performing spatio-temporal registration on the original engineering measurement data to obtain engineering measurement data, and performing feature extraction on the engineering measurement data to obtain feature information.

[0008] The analysis module is used for receiving the feature information, performing information fusion on the feature information to generate a fusion feature set, and obtaining a global decision instruction by analyzing the fusion feature set and a second analysis source.

[0009] The interaction module is configured to display the global decision instruction and receive a feedback instruction.

[0010] As a preferred scheme of the intelligent engineering measurement data analysis system, the original engineering measurement data comprises point cloud data, image data and time series data.

[0011] The point cloud data comprises ground laser scanning point cloud, airborne laser radar point cloud and mobile measurement system point cloud.

[0012] The image data comprises orthophoto data, oblique photography image and infrared thermal imaging data.

[0013] The time series data comprises deformation time series data, structure response time series data and environmental factor time series data.

[0014] As a preferred scheme of the intelligent engineering measurement data analysis system, the processing module comprises a space-time registration unit, a data standardization unit and a feature extraction unit.

[0015] The space-time registration unit is configured to register the point cloud data and register the image data and the three-dimensional model, and use a GPS satellite timing signal as a time reference to stamp a unified time stamp on each data unit in the original engineering measurement data.

[0016] The data standardization unit is configured to perform normalization processing on the data after space-time registration in different resolutions and different dimensions to obtain engineering measurement data.

[0017] The feature extraction unit is configured to receive the engineering measurement data, extract feature information of the engineering measurement data, and the feature information comprises geometric features, texture features, apparent disease features, deformation trends and statistical features of the components, and the geometric features comprise normal vectors, curvatures and contour lines.

[0018] As a preferred scheme of the intelligent engineering measurement data analysis system, the point cloud data registration specifically comprises:

[0019] For two pieces of point cloud data, the nearest neighbor points in the target point cloud are found for the points in the source point cloud to establish a corresponding relationship, and the optimal rotation matrix and translation vector are obtained by calculation to minimize the Euclidean distance square sum between all corresponding point pairs.

[0020] The image data and the three-dimensional model registration specifically comprises:

[0021] A virtual orthographic image or perspective image is generated according to a viewpoint parameter of the three-dimensional model, feature points are extracted from the real image and the virtual image respectively and feature descriptors are generated, feature point matching is performed and false matching points are removed, and a homography matrix or a camera projection matrix for mapping the real image to the surface of the three-dimensional model is calculated.

[0022] As a preferred scheme of the intelligent engineering survey data analysis system, the analysis module comprises a digital twin unit, a feature association unit, a special analysis unit and a decision fusion unit.

[0023] The digital twin unit is configured to establish a digital twin model.

[0024] The feature association unit is configured to receive the feature information, bind and map the feature information to corresponding components in the digital twin unit, construct a graph neural network structure relationship graph with components as nodes, physical connection relationships between components as edges and bound component features as node attributes, perform cross-component information fusion through a message passing mechanism to generate a fusion feature set, and perform first processing based on the fusion feature set to obtain a first preliminary decision and a first confidence degree of each component.

[0025] The first processing comprises inputting the fusion feature set into a preset classifier for forward propagation calculation, outputting probability values of each component belonging to each predefined state category, the predefined state categories comprising normal, slight abnormality and serious abnormality, selecting the predefined state category with the highest probability value as the first preliminary decision of the component, and taking the probability value as the first confidence degree.

[0026] The feature association unit is a first analysis source.

[0027] The special analysis unit is configured to perform specific target analysis on the engineering survey data to output a second preliminary decision and a second confidence degree, the specific target analysis comprising one of crack analysis, settlement analysis and earthwork volume analysis.

[0028] The crack analysis is configured to analyze image data and output a second preliminary decision and a second confidence degree of whether there is a crack.

[0029] The settlement analysis is configured to analyze time series data and output a second preliminary decision and a second confidence degree of whether the settlement is out of limit.

[0030] The earthwork volume analysis is configured to analyze point cloud data and output a second preliminary decision and a second confidence degree of whether the earthwork volume change meets the standard.

[0031] The special analysis unit is a second analysis source.

[0032] The decision fusion unit is configured to receive a first preliminary decision and a first confidence thereof from the first analysis source, a second preliminary decision and a second confidence thereof from the second analysis source, perform a second processing on the first preliminary decision and the first confidence thereof and the second preliminary decision and the second confidence thereof, and generate the global decision instruction, which includes a decision conclusion, a decision confidence, a severity level, and a corresponding component identifier.

[0033] As a preferred scheme of the intelligent engineering survey data analysis system, the establishing a digital twin model comprises:

[0034] Inputting a building information model, a computer-aided design drawing, and geographic information system data to generate a digital twin model, the digital twin model comprising geometric attribute, static semantic attribute, and dynamic state attribute;

[0035] Receiving the engineering survey data and updating the digital twin model according to the engineering survey data, the updating comprising geometric updating, spatio-temporal state updating, and physical attribute updating;

[0036] The geometric updating is configured to compare and analyze the point cloud data and a corresponding component in the digital twin model, and calculate a deviation amount of the point cloud and a model surface;

[0037] The spatio-temporal state updating is configured to associate the deformation time series data and the corresponding component in the digital twin model, and drive the position and shape of the component in the three-dimensional space to change accordingly with time series;

[0038] The physical attribute updating is configured to write the physical attribute monitoring data into an attribute field of a corresponding component in the digital twin model in real time.

[0039] As a preferred scheme of the intelligent engineering survey data analysis system, the information fusion across components through the message passing mechanism comprises:

[0040] For each edge in the graph neural network structure relationship graph, a message is generated from a source node, the message being generated by linearly transforming a feature attribute of the source node to generate a message sent to a target node;

[0041] The target node receives messages from all adjacent nodes of the target node, aggregates the messages of the adjacent nodes into an aggregated message through a summation function, combines the aggregated message with a feature attribute of the target node, and generates an updated state feature of the target node through a nonlinear transformation function;

[0042] The nonlinear transformation function has a form of:

[0043] ;

[0044] wherein, is the state vector of node i at layer l+1, σ is a nonlinear activation function, W is a weight matrix, is the state vector of node i at layer l, m i is the message vector converged at node i;

[0045] After Q rounds of iteration of the message generation, the message aggregation and the node update, the state feature of each node is fused with the fused feature of all the neighboring nodes in the Q-hop neighborhood, to obtain a fused feature set.

[0046] As a preferred scheme of the intelligent engineering measurement data analysis system, the second processing includes:

[0047] assigning a basic support degree of each received preliminary decision and its confidence to all possible propositions, the propositions including all the preliminary decisions to be selected;

[0048] calculating the relationship between all the preliminary decisions by a synthetic rule expression, and updating the comprehensive support degree and uncertainty of each proposition according to the relationship between all the preliminary decisions, the synthetic rule expression being:

[0049] ;

[0050] wherein, m1 and m2 are two basic probability distribution functions to be fused, representing preliminary decisions and confidences from two different analysis sources, m1(B) represents the support degree of the first analysis source to proposition B, m2(C) represents the support degree of the second analysis source to proposition C, represents the joint support degree of the new analysis source to proposition A after fusion, K is a conflict coefficient, and the calculation expression of the conflict coefficient is:

[0051] ;

[0052] wherein, the value range of k is [0, 1), based on the joint support degree and uncertainty, the proposition with the highest comprehensive support degree and the lowest uncertainty is selected as the final global decision instruction output.

[0053] As a preferred scheme of the intelligent engineering measurement data analysis system, the interaction module includes a display unit, a warning unit and an interaction unit.

[0054] The display unit maps the decision conclusion and the severity level to a color gradient for display on the digital twin model, and provides a two-dimensional graphical user interface to display the decision conclusion and the decision confidence;

[0055] The early warning unit is configured to trigger different levels of early warning response according to the severity level of the global decision instruction.

[0056] The interaction unit is configured to receive feedback instructions, system control instructions, and feedback instructions on the decision conclusion, and send the feedback instructions to the analysis module.

[0057] As a preferred scheme of the intelligent engineering measurement data analysis system, the severity level of the global decision instruction includes high risk, medium risk and low risk.

[0058] When the severity level of the global decision instruction is high risk, short message and email push are automatically triggered.

[0059] When the severity level of the global decision instruction is medium risk, interface pop-up and sound-light alarm are triggered.

[0060] When the severity level of the global decision instruction is low risk, information is prompted in the two-dimensional graphical user interface.

[0061] The beneficial effects of the present application are as follows: the present application processes point cloud data, image data and time series data through space-time registration and feature extraction, simulates the physical association between components using a graph neural network structure relationship graph, realizes cross-component information fusion and state evaluation, further adopts a synthesis rule to fuse multiple source preliminary decisions, effectively solves the evidence conflict and uncertainty problem, generates a reliable global decision, and an interaction module performs three-dimensional visualization display, forms a closed loop optimization through hierarchical early warning and user feedback, and comprehensively improves the accuracy, systematicness and self-adaptive ability of engineering structure monitoring analysis. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1 The basic flowchart of the intelligent engineering measurement data analysis system provided by an embodiment of the present application is shown.

[0063] Figure 2 The signal transmission diagram of the intelligent engineering measurement data analysis system provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0064] In order to make the above-mentioned objects, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments.

[0065] Embodiment 1, refer to Figure 1 and Figure 2 For an embodiment of the present application, an intelligent engineering survey data analysis system is provided, comprising a processing module, an analysis module, and an interaction module:

[0066] The processing module is configured to receive raw engineering survey data, perform spatio-temporal registration on the raw engineering survey data to obtain engineering survey data, and perform feature extraction on the engineering survey data to obtain feature information.

[0067] The analysis module is configured to receive the feature information and perform information fusion on the feature information to generate a fusion feature set, and obtain a global decision instruction by analyzing the fusion feature set and a second analysis source.

[0068] The interaction module is configured to display the global decision instruction and receive a feedback instruction.

[0069] The raw engineering survey data includes point cloud data, image data, and time series data.

[0070] The point cloud data includes ground laser scanning point cloud, airborne laser radar point cloud, and mobile measurement system point cloud.

[0071] The image data includes orthophoto data, oblique photography image, and infrared thermal imaging data.

[0072] The time series data includes deformation time series data, structural response time series data, and environmental factor time series data.

[0073] In specific embodiments, the raw engineering survey data serves as the input basis for system analysis, including multi-dimensional information obtained through different sensing technologies. The point cloud data is obtained through laser scanning technology to obtain high-precision three-dimensional spatial coordinates, which is used to construct the geometric skeleton of the engineering entity. The image data is captured by various photography devices to capture the optical and radiation characteristics of the object surface, which is used to provide texture, color, and thermal state information. The time series data is obtained through a sensor network for long-term monitoring to obtain dynamic change information, which is used to reveal the structural behavior trend and environmental correlation. These multi-source data collectively provide comprehensive raw information support for subsequent registration, feature extraction, and fusion analysis.

[0074] The processing module includes a spatio-temporal registration unit, a data standardization unit, and a feature extraction unit.

[0075] The spatio-temporal registration unit is configured to register the point cloud data and register the image data and the three-dimensional model, with the GPS satellite timing signal as the time reference, to stamp a unified time stamp on each data unit in the raw engineering survey data.

[0076] The data standardization unit is configured to perform normalization processing on the data after spatio-temporal registration in different resolutions and different dimensions to obtain engineering survey data.

[0077] The feature extraction unit is configured to receive the engineering measurement data and extract feature information of the engineering measurement data, the feature information including geometric features, texture features, apparent disease features, deformation trends and statistical features of the component, the geometric features including normal vectors, curvatures and contour lines.

[0078] In specific embodiments, the processing module realizes the preprocessing and feature mining of the multi-source heterogeneous original engineering measurement data through the collaborative work of the space-time registration unit, the data standardization unit and the feature extraction unit, and provides a standardized data basis for subsequent analysis. Specifically, the space-time registration unit first solves the inconsistency of data space and time, and uses the GPS satellite time signal to stamp a unified timestamp on all data units in time to ensure the comparability of time series data. The data standardization unit then normalizes the registered data to eliminate the differences in resolution and dimension of different sensors, and generates engineering measurement data that can be directly used for analysis. The feature extraction unit finally extracts multi-dimensional feature information from the standardized data, such as calculating the normal vector and curvature of the point cloud to identify the geometric abnormalities of the component surface, analyzing the texture features of the image to detect crack diseases, or extracting the settlement trend features from the time series data, thereby converting the original data into feature information with clear engineering significance.

[0079] The point cloud data registration is specifically:

[0080] For two pieces of point cloud data, find the nearest neighbor point in the target point cloud for the point in the source point cloud to establish a corresponding relationship, and calculate the optimal rotation matrix and translation vector to minimize the sum of the squared Euclidean distances between all corresponding point pairs.

[0081] The image data and three-dimensional model registration is specifically:

[0082] Generate a virtual orthographic image or perspective image according to the viewpoint parameters of the three-dimensional model, extract feature points from the real image and virtual image respectively and generate feature descriptors, perform feature point matching and eliminate false matching points, and calculate the homography matrix or camera projection matrix that maps the real image to the surface of the three-dimensional model.

[0083] In specific embodiments, the point cloud data registration accurately aligns multiple pieces of point cloud collected at different angles or time points into a unified coordinate system, which is achieved by an iterative closest point algorithm. The specific process is to find the closest neighbor point in the target point cloud for each point in the source point cloud to establish a corresponding relationship, and to calculate the optimal rotation matrix and translation vector by mathematical methods such as singular value decomposition, to minimize the sum of squared Euclidean distances between all corresponding points. For example, two pieces of scanning point cloud on the left and right sides of a bridge pier are fused to generate a complete three-dimensional model. The image data and three-dimensional model registration is to achieve the spatial alignment of the real image and the virtual model through visual feature matching. First, a virtual image is generated according to the three-dimensional model view parameters. SIFT feature points and descriptors are extracted from real images and virtual images, respectively. After feature matching and false match elimination, the homography matrix or camera projection matrix is calculated, so as to accurately map the real image texture to the model surface. For example, the crack photos taken on site are accurately pasted to the corresponding position of the bridge digital twin model, realizing the visual positioning of the disease.

[0084] The analysis module includes a digital twin unit, a feature association unit, a special analysis unit, and a decision fusion unit.

[0085] The digital twin unit is configured to establish a digital twin model.

[0086] The feature association unit is configured to receive feature information, bind and map the feature information to corresponding components in the digital twin unit, construct a graph neural network structure relationship graph with components as nodes, physical connection relationships between components as edges, and bound construction features as node attributes, perform information fusion across components through a message passing mechanism, generate a fused feature set, and perform a first processing based on the fused feature set to obtain a first preliminary decision and a first confidence degree of each component.

[0087] The first processing includes inputting the fused feature set into a preset classifier for forward propagation calculation, outputting probability values of each component belonging to each predefined state category, selecting the predefined state category with the highest probability value as the first preliminary decision of the component, and taking the probability value as the first confidence degree.

[0088] In specific embodiments, the preset classifier can be selected from any one of a neural network classifier based on a fully connected layer, a support vector machine model, a decision tree model, and a random forest model. The classifier is trained to map the fused feature set to the predefined state category.

[0089] The forward propagation calculation refers to inputting the fused feature vector representing the state of the component into the trained classifier, performing calculation through a series of fixed mathematical layers in the classifier, and finally outputting a probability distribution representing various health states of the component.

[0090] The feature association unit is a first analysis source;

[0091] The special analysis unit is configured to perform a specific target analysis on the engineering survey data, and output a second preliminary decision and a second confidence thereof, the specific target analysis including one of crack analysis, settlement analysis, and earthwork analysis;

[0092] The crack analysis is configured to analyze the image data and output a second preliminary decision and a second confidence thereof as to whether there is a crack;

[0093] The settlement analysis is configured to analyze the time-series data and output a second preliminary decision and a second confidence thereof as to whether the settlement is out of limit;

[0094] The earthwork analysis is configured to analyze the point cloud data and output a second preliminary decision and a second confidence thereof as to whether the earthwork change meets the standard;

[0095] The special analysis unit is a second analysis source;

[0096] The decision fusion unit is configured to receive the first preliminary decision and the first confidence thereof from the first analysis source and the second preliminary decision and the second confidence thereof from the second analysis source, perform a second processing on the first preliminary decision and the first confidence thereof and the second preliminary decision and the second confidence thereof, and generate a global decision instruction, the global decision instruction including a decision conclusion, a decision confidence, a severity level, and a corresponding component identification.

[0097] In specific embodiments, the decision fusion unit integrates the component state judgment from the first analysis source and the analysis from the second analysis source, fuses these multi-source heterogeneous decisions through the second processing, and finally outputs a comprehensive global decision instruction. For example, if the feature association unit judges that a certain pier has a "severe abnormality" feature with a probability of 0.9, and the settlement analysis also shows that the settlement at this location is out of limit with a probability of 0.85, the decision fusion unit may generate a global decision instruction that the pier has a structural instability risk, thereby providing a comprehensive and reliable analysis conclusion for engineering maintenance.

[0098] The digital twin model is established by:

[0099] Inputting the building information model, the computer-aided design drawing, and the geographic information system data to generate the digital twin model, the digital twin model including geometric attribute, static semantic attribute, and dynamic state attribute;

[0100] Receiving the engineering survey data, and updating the digital twin model according to the engineering survey data, the updating including geometric updating, time-space state updating, and physical attribute updating;

[0101] The geometric updating is configured to compare and analyze the point cloud data and the corresponding component in the digital twin model, and calculate the deviation amount of the point cloud and the model surface;

[0102] The spatio-temporal state update is used to associate the deformation time series data with the corresponding component in the digital twin model, and drive the position and shape of the component in the three-dimensional space to change over time.

[0103] The physical property update is used to write the physical property monitoring data into the attribute field of the corresponding component in the digital twin model in real time.

[0104] In specific embodiments, the digital twin model provides a digital foundation for analysis, simulation and decision-making of the entire system. By continuously receiving measured engineering measurement data and driving the digital twin model to update in multiple dimensions, the geometric shape update compares the latest scanned point cloud data with the original component in the model, calculates the deviation to reflect the actual shape change, the spatio-temporal state update associates the deformation monitoring time series data with the model component, and drives the dynamic evolution of the component in the three-dimensional space, and the physical property update writes the real-time monitoring data into the attribute field of the corresponding component, so that the digital twin model can fully and real-time map the real state of the physical entity, and provide accurate data basis for subsequent analysis.

[0105] The information fusion across components through the message passing mechanism specifically includes:

[0106] For each edge in the structure relationship graph of the graph neural network, the edge connects a source node and a target node, and a message is generated through the source node. The message generation is to generate a message sent to the target node by linearly transforming the feature attributes of the source node;

[0107] The target node receives messages from all adjacent nodes of the target node, and aggregates the messages of the adjacent nodes into an aggregated message through a summation function. The target node combines the aggregated message with the feature attributes of the target node, and generates the updated state feature of the node through a nonlinear transformation function;

[0108] The form of the nonlinear transformation function is:

[0109] ;

[0110] Wherein, is the state vector of node i at layer l+1, σ is a nonlinear activation function, W is a weight matrix, is the state vector of node i at layer l, m i is the message vector converged at node i;

[0111] After Q rounds of message generation, message aggregation and node update iterations, the state feature of each node is fused with the fused features of all adjacent nodes in the Q-hop neighborhood, and a fused feature set is obtained.

[0112] In specific embodiments, information fusion is performed across components through a message passing mechanism, and each component can receive and integrate information from its neighboring components based on the physical connection relationship between components, so as to generate more comprehensive and accurate fusion features to reflect the real state of the component in the overall structure. For example, in a digital twin model of a bridge, when a small deformation occurs in a beam component, this change can be gradually spread through the message passing mechanism and affect the state features of the adjacent components such as the pier and the deck slab connected to the beam component, so that the system can determine whether the deformation is an isolated phenomenon or a sign that may trigger continuous damage based on more abundant context information, thereby significantly improving the reliability of state evaluation.

[0113] The second processing includes:

[0114] For each received preliminary decision and its confidence, a basic support degree of the preliminary decision and its confidence to all possible propositions is assigned, the propositions including all the preliminary decisions to be selected;

[0115] The relationship between all the preliminary decisions is calculated by a synthetic rule expression, and the comprehensive support degree and uncertainty of each proposition are updated according to the relationship between all the preliminary decisions, the synthetic rule expression being:

[0116] ;

[0117] wherein m1 and m2 are two basic probability distribution functions to be fused, representing preliminary decisions and confidences from two different analysis sources, m1(B) represents the support degree of the first analysis source to proposition B, and m2(C) represents the support degree of the second analysis source to proposition C, represents the joint support degree of the new analysis source after fusion to proposition A, K is a conflict coefficient, and the calculation expression of the conflict coefficient is:

[0118] ;

[0119] wherein the value range of k is [0, 1), based on the joint support degree and the uncertainty, the proposition with the highest comprehensive support degree and the lowest uncertainty is selected as the final global decision instruction output.

[0120] In specific embodiments, the second processing assigns a basic probability assignment to each preliminary decision and its confidence, for example, if the crack analysis output exists a crack decision and the confidence is 0.8, the support degree of the result to the proposition of existing cracks can be assigned as 0.8, and the remaining 0.2 is assigned to the unknown uncertain item; then the common support degree of different evidences is calculated through the synthesis rule, wherein the conflict coefficient K is used to measure the size of the conflict between evidences, when K approaches 1, it indicates that the evidences are highly conflicting; for example, when the first analysis source of the beam component judges that the support degree of serious abnormality is 0.7 and the settlement analysis output judges that the support degree of settlement not exceeding the limit is 0.6, the second processing will automatically calculate the conflict part and reassign the credibility, finally select the proposition with the highest comprehensive support degree and the lowest uncertainty as the global decision instruction output, so that a reliable conclusion can still be formed under the environment of contradictory evidences.

[0121] The interaction module includes a display unit, a warning unit and an interaction unit;

[0122] The display unit maps the decision conclusion and the severity level to a color gradient to be displayed on the digital twin model, and provides a two-dimensional graphical user interface to display the decision conclusion and the decision confidence;

[0123] The warning unit is used to trigger different levels of warning responses according to the severity level of the global decision instruction;

[0124] The interaction unit is used to receive feedback instructions, system control instructions and feedback instructions on the decision conclusion, and send the feedback instructions to the analysis module.

[0125] In specific embodiments, the interaction module converts the output of the analysis module into intuitive visual information, graded warning actions and a two-way human-computer interaction channel through the cooperative action of the display unit, the warning unit and the interaction unit, and its core role is to break the closed loop between system intelligent decision and user cognition and operation. Specifically, the display unit maps the decision conclusion and the severity level to a color gradient, green represents normal, yellow represents slight abnormality, and red represents serious abnormality, which is superimposed on the corresponding component of the digital twin model, and at the same time, the decision conclusion and the decision confidence are displayed through a two-dimensional graphical interface list, thereby providing multi-dimensional and visualized situation awareness. The warning unit automatically triggers differentiated responses based on the severity level of the global decision instruction, for example, when the system determines that there is a high risk of spalling of the tunnel lining, it immediately activates the short message and email push to the maintenance department responsible person. The interaction unit receives the feedback instructions of the user and the feedback on the system decision, and sends these feedback instructions back to the analysis module for optimizing the subsequent decision process, for example, when the user confirms that a certain type of crack is a false alarm several times, the system can adaptively reduce the confidence calculation weight of similar features based on this data, thereby continuously improving the accuracy of system judgment.

[0126] The severity level of the global decision instruction includes high risk, medium risk and low risk:

[0127] When the severity level of the global decision instruction is high risk, then the short message and email push are triggered automatically;

[0128] When the severity level of the global decision instruction is medium risk, then the interface pop-up window and sound-light alarm are triggered;

[0129] When the severity level of the global decision instruction is low risk, then the information is prompted in the two-dimensional graphical user interface.

[0130] The application obtains engineering measurement data and feature information thereof by receiving original engineering measurement data, performing space-time registration and feature extraction processing on the original engineering measurement data, performing feature correlation and multi-source decision fusion through a digital twin model, generating a global decision instruction, and displaying in an interaction module. The application realizes integrated processing, intelligent analysis and interactive control of multi-source heterogeneous engineering measurement data, and improves the accuracy and efficiency of engineering structure health monitoring and safety evaluation.

[0131] Those skilled in the art will understand that embodiments of the application can be provided as methods, systems or computer program products. Therefore, the application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, the application can take the form of a computer program product implemented on one or more computer-usable storage media having computer-usable program code embodied in the medium. The storage media can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk. These computer program instructions can also be stored in a computer readable storage medium that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction means, which realizes the processes specified in the flowcharts Figure 1 The functions specified in one or more flows and / or blocks Figure 1 The functions specified in one or more flows and / or blocks

[0132] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced, without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. An intelligent engineering survey data analysis system, characterized in that, The processing module, the analysis module and the interaction module are included. The processing module is configured to receive original engineering survey data, perform space-time registration on the original engineering survey data to obtain engineering survey data, perform feature extraction on the engineering survey data to obtain feature information. The analysis module is configured to receive the feature information, perform information fusion on the feature information to generate a fusion feature set, and obtain a global decision instruction by analyzing the fusion feature set and a second analysis source. The interaction module is configured to display the global decision instruction and receive a feedback instruction. The analysis module includes a digital twin unit, a feature association unit, a special analysis unit and a decision fusion unit. The digital twin unit is configured to establish a digital twin model. The feature association unit is configured to receive the feature information, bind and map the feature information with corresponding components in the digital twin unit, construct a graph neural network structure relationship graph with components as nodes, physical connection relationships between components as edges and bound feature information as node attributes, perform cross-component information fusion through a message passing mechanism to generate a fusion feature set, and perform a first processing based on the fusion feature set to obtain a first preliminary decision of each component and a first confidence thereof. The first processing includes inputting the fusion feature set into a preset classifier to perform forward propagation calculation, outputting probability values of each component belonging to each predefined state category, selecting the predefined state category with the highest probability value as the first preliminary decision of the component, and taking the probability value as the first confidence. The feature association unit is a first analysis source. The special analysis unit is configured to perform specific target analysis on the engineering survey data to output a second preliminary decision and a second confidence thereof, and the specific target analysis includes one of crack analysis, settlement analysis and earthwork volume analysis. The crack analysis is configured to analyze image data and output a second preliminary decision of whether there is a crack and a second confidence thereof. The settlement analysis is configured to analyze time series data and output a second preliminary decision of whether the settlement is out of limit and a second confidence thereof. The earthwork volume analysis is configured to analyze point cloud data and output a second preliminary decision of whether the earthwork volume change meets the standard and a second confidence thereof. The special analysis unit is a second analysis source. The decision fusion unit is configured to receive the first preliminary decision and the first confidence thereof from the first analysis source and the second preliminary decision and the second confidence thereof from the second analysis source, perform a second processing on the first preliminary decision and the first confidence thereof and the second preliminary decision and the second confidence thereof, and generate the global decision instruction, which includes a decision conclusion, a decision confidence, a severity level and a corresponding component identifier.

2. The intelligent engineering survey data analysis system of claim 1, wherein, The original engineering survey data includes point cloud data, image data and time series data. The point cloud data includes ground laser scanning point cloud, airborne laser radar point cloud and mobile measurement system point cloud. The image data includes orthophoto image data, oblique photography image and infrared thermal imaging data. The time series data comprises deformation time series data, structure response time series data and environmental factor time series data.

3. The intelligent engineering survey data analysis system of claim 2, wherein, The processing module comprises a space-time registration unit, a data standardization unit and a feature extraction unit. The space-time registration unit is configured to register point cloud data and image data and a three-dimensional model, and to stamp a uniform timestamp on each data unit in the original engineering survey data by taking a GPS satellite timing signal as a time reference. The data standardization unit is configured to normalize the data after space-time registration to obtain engineering survey data. The feature extraction unit is configured to receive the engineering survey data and extract feature information of the engineering survey data, the feature information comprising geometric features, texture features, apparent disease features, deformation trends and statistical features of a component, the geometric features comprising normal vectors, curvatures and contour lines.

4. The intelligent engineering survey data analysis system of claim 3, wherein, The point cloud data registration specifically comprises: For two pieces of point cloud data, a nearest neighbor point in a target point cloud is found for a point in a source point cloud to establish a corresponding relationship, and an optimal rotation matrix and translation vector are obtained by calculation to minimize the sum of squared Euclidean distances between all corresponding point pairs. The image data and three-dimensional model registration specifically comprises: A virtual orthographic image or perspective image is generated according to view point parameters of the three-dimensional model, feature points are extracted from the real image and the virtual image respectively and feature descriptors are generated, feature point matching is performed and false matching points are removed, and a homography matrix or camera projection matrix that maps the real image to the surface of the three-dimensional model is calculated.

5. The intelligent engineering survey data analysis system of claim 4, wherein, The establishment of the digital twin model comprises: An input building information model, computer-aided design drawing and geographic information system data are used to generate a digital twin model, the digital twin model comprising geometric shape attributes, static semantic attributes and dynamic state attributes; The engineering survey data is received, and the digital twin model is updated according to the engineering survey data, the update comprising geometric shape update, space-time state update and physical property update; The geometric shape update is configured to compare and analyze the point cloud data and a corresponding component in the digital twin model, and to calculate a deviation amount of the point cloud and the model surface; The space-time state update is configured to associate the deformation time series data with the corresponding component in the digital twin model, and to drive the position and shape of the component in the three-dimensional space to change accordingly with the time sequence; The physical property update is configured to write physical property monitoring data into an attribute field of a corresponding component in the digital twin model in real time.

6. The intelligent engineering survey data analysis system of claim 5, wherein, The information fusion across components through the message passing mechanism specifically comprises: For each edge between a source node and a target node in the graph neural network structure relationship graph, a message is generated by the source node, the message generation comprising generating a message sent to the target node by linearly transforming the feature attributes of the source node; The target node receives messages from all adjacent nodes of the target node, aggregates the messages of the adjacent nodes into an aggregated message by a summation function, combines the aggregated message with the feature attributes of the target node, and generates an updated state feature of the node by a nonlinear transformation function. The form of the nonlinear transformation function is: ； wherein, is the state vector of node i at layer l + 1, σ is a non-linear activation function, W is a weight matrix, is the state vector of node i at layer l, m i is the message vector converging at node i; After the iteration of the message generation, the message aggregation and the node update for Q rounds, the state feature of each node is fused with the fusion feature of all the neighboring nodes in the Q-hop neighborhood, to obtain a fusion feature set.

7. The intelligent engineering survey data analysis system of claim 6, wherein, The second processing includes: Assigning a basic support degree of each received preliminary decision and its confidence to all possible propositions, including all the preliminary decisions to be selected; Calculating the relationship between all the preliminary decisions by a synthetic rule expression, and updating the comprehensive support degree and uncertainty of each proposition according to the relationship between all the preliminary decisions, wherein the synthetic rule expression is: ； wherein m1 and m2 are two basic probability assignment functions to be fused, representing preliminary decisions and confidences from two different analysis sources, m1(B) represents the support degree of the first analysis source to proposition B, m2(C) represents the support degree of the second analysis source to proposition C, represents the joint support degree of the new analysis source to proposition A after fusion, K is a conflict coefficient, and the calculation expression of the conflict coefficient is: ； Wherein the value range of k is [0, 1), and based on the joint support degree and uncertainty, the proposition with the highest comprehensive support degree and the lowest uncertainty is selected as the global decision instruction output.

8. The intelligent engineering survey data analysis system of claim 7, wherein, The interaction module includes a display unit, a warning unit and an interaction unit; The display unit maps the decision conclusion and the severity level to a color gradient for display on the digital twin model, and provides a two-dimensional graphical user interface to display the decision conclusion and the decision confidence; The warning unit is configured to trigger different levels of warning responses according to the severity level of the global decision instruction; The interaction unit is configured to receive feedback instructions, system control instructions and feedback instructions for the decision conclusion, and send the feedback instructions to the analysis module.

9. The intelligent engineering survey data analysis system of claim 8, wherein, The severity level of the global decision instruction includes high risk, medium risk and low risk: When the severity level of the global decision instruction is high risk, then automatically trigger SMS and email push; When the severity level of the global decision instruction is medium risk, then trigger interface pop-up and sound and light alarm; When the severity level of the global decision instruction is low risk, then information is prompted in the two-dimensional graphical user interface.

Images