Bridge construction process ship passing identification early warning system and method based on digital twinning

Abstract

The invention provides a bridge construction process ship passing identification early warning system and method based on digital twinning, and belongs to the field of traffic engineering. The problem that in a complex water area environment of bridge construction, sailing paths of ships of different types and sizes possibly intersect with a construction area, and the collision risk is increased is solved. A water level prediction module is used for obtaining real-time data according to an on-site water level monitor and a meteorological platform, and processing the real-time data through a multivariable LSTM time sequence prediction model to obtain a water level prediction result; the temporary structure and ship monitoring module is used for obtaining data according to an RGB camera and a laser radar, and obtaining three-dimensional space information of a temporary structure and a ship through data processing; and the digital twin platform and collision risk prediction module is used for integrating the water level prediction result and the three-dimensional space information of the temporary structure and the ship, and carrying out risk analysis and early warning. The method is applied to the field of ship traffic management and safety.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic engineering, and more specifically, to a ship passing recognition and early warning system and method during the bridge construction process based on digital twin. Background Art

[0002] In the complex water area environment during bridge construction, safe passage of ships is a highly challenging task. During the construction process, the construction area often changes, including temporarily erected platforms, floating equipment, etc. These changes not only alter the terrain of the water area but also affect the passage routes of ships and the predictability of safe passage. At the same time, the continuous changes in hydrological conditions, such as tides, water level fluctuations, etc., further increase the complexity of the ship navigation routes and safety risks.

[0003] In busy water areas, ships of different types and sizes may pass through simultaneously, and their navigation routes often intersect with the bridge construction area, which significantly increases the risk of collision. The lack of effective traffic management and conflict resolution mechanisms will make these risks more difficult to control, leading to serious safety accidents and property losses.

[0004] To effectively address these challenges, there is an urgent need to introduce a real-time collision risk identification and early warning system for the bridge construction site to ensure safety during construction, avoid property losses and casualties, especially in large infrastructure projects such as bridge construction. Summary of the Invention

[0005] The technical problem to be solved by the present invention is:

[0006] To solve the problem that in the complex water area environment during bridge construction, the navigation routes of ships of different types and sizes may intersect with the construction area, increasing the collision risk and causing safety accidents and property losses.

[0007] The technical solution adopted by the present invention to solve the above technical problem:

[0008] The present invention provides a ship passing recognition and early warning system during the bridge construction process based on digital twin, including a water level prediction module, a temporary structure and ship monitoring module, and a digital twin platform and collision risk prediction module;

[0009] The water level prediction module is used to obtain historical data according to the on-site water level monitor, water temperature monitor, and meteorological platform, train the model using a multivariate LSTM time series prediction model, and predict the water level information for the next day by combining the trained multivariate LSTM time series prediction model with real-time data;

[0010] The temporary structure and vessel monitoring module is used to obtain data from RGB cameras and lidar, and after calibration alignment, semantic segmentation, and three-dimensional conversion, the three-dimensional spatial information of the temporary structure and vessels during the bridge construction process is obtained;

[0011] The digital twin platform and collision risk prediction module are used to integrate the water level prediction results and the three-dimensional spatial information of the temporary structure and vessels, and conduct risk analysis and early warning.

[0012] Furthermore, in the water level prediction module, specifically,

[0013] (1) Data acquisition: Pre-collect the daily temperature, water temperature, weather conditions, and water level height data at the target location in the past five years, and process missing values or outliers;

[0014] (2) Data merging: Align and merge the data obtained in step (1) in chronological order, and perform timestamp standardization processing on each data source to ensure that all data points can correspond on the same time dimension; The merged dataset contains N variables of daily temperature, water temperature, weather conditions, and water level height, forming an N-dimensional time series dataset. The merged dataset M can be expressed as:

[0015] M = [x1, x2, …, x N

[0016] where x N represents the N variables collected;

[0017] (3) Data preprocessing: Normalize the data into the interval [0, 1]:

[0018]

[0019] where x is the original data, x' is the normalized data, and min(x) and max(x) are the minimum and maximum values of this variable respectively;

[0020] (4) Dataset division: Select 80% of the data as the training set for model training, and the remaining 20% of the data as the test set for evaluating the model performance;

[0021] (5) Model training: In the model training stage, first input the multi-variable time series data in the training set into the multi-variable LSTM time series prediction model; including the daily temperature T air , water temperature T water , weather conditions W weather , and water level height H water information in the past five years; The input at each time step is represented as a vector:

[0022] X​t = [T air (t), T water (t), W weather (t), H water (t)]

[0023] The model captures long-term dependencies in the time series through the LSTM layer. The forget gate f t determines what information to discard, the input gate i t determines what new information to add, and the output gate o t determines what information to output; the calculation formulas are as follows:

[0024] f t = σ(W f · [h t-1 , X t + b f )

[0025] i t = σ(W i · [h t-1 , X t + b i )

[0026] C' t = tanh(W C · [h t-1 , X t + b C )

[0027] C t = f t · C t-1 + i t · C t '

[0028] o t = σ(W o · [h t-1 , X t + b o )

[0029] h t = o t · tanh(C t )

[0030] Among them, h t-1 is the hidden state of the previous time step, C t-1 is the state of the previous time step, W f , W i , W C , W o are weight matrices, b f , b i , bC ,b o where \(b\) is the bias term, \(\sigma\) is the sigmoid activation function, and \(\tanh\) is the hyperbolic tangent activation function;

[0031] The fully connected layer converts the output of the LSTM layer into the final predicted value, i.e., the water level height of the next day. The calculation formula of the fully connected layer is:

[0032] y t = W·h t + b

[0033] where \(y\) t is the predicted water level height, \(W\) is the weight matrix, and \(b\) is the bias term;

[0034] During the training process, the mean squared error is used as the loss function, and the difference between the predicted value and the actual value is minimized through the Adam optimization algorithm, thereby continuously adjusting the weights and biases of the model. The calculation formula of the mean squared error is:

[0035]

[0036] where \(n\) is the number of samples, \(y\) i is the actual value, is the predicted value;

[0037] After multiple rounds of iterative training, the model training is completed;

[0038] (6) Water level prediction In the prediction stage, the trained multi-variable LSTM time series prediction model is used to predict the water level height of the next day according to the seven-day multi-variable time series data. The specific steps are as follows:

[0039] Normalize the seven-day data collected to meet the requirements of the model input;

[0040] Construct the normalized seven-day data into a sequence sample and input it into the LSTM model trained in step (5). The model processes the input sequence through its internal LSTM layer, utilizes the collaborative effects of the forget gate, input gate, and output gate to capture the long-term dependencies in the time series, and finally the fully connected layer outputs the predicted value of the water level height of the next day. The predicted value is the normalized value and needs to be converted back to the actual water level height through the inverse normalization formula:

[0041] H predicted (t + 7) = y′·(max(H water ) - min(H water )) + min(H water )

[0042] Among them, y′ is the normalized predicted value output by the model, and max(Hwater) and min(Hwater) are the maximum and minimum values of the water level height respectively.

[0043] Furthermore, in the temporary structure and vessel monitoring module, the data obtained from the RGB camera and lidar includes the height, speed, direction of the vessel, and the current state of the bridge.

[0044] Calibrate the dataset using the RGB camera and lidar to obtain the aligned RGB images and lidar data.

[0045] Furthermore, use the YOLOV8 model to perform semantic segmentation on the RGB image to identify the regions of the temporary structure and the vessel in the image, and map the regions of the temporary structure and the vessel in the image to the three-dimensional space using the point cloud data collected by the lidar, so as to obtain the positions and sizes of the temporary structure and the passing vessel in the three-dimensional space.

[0046] Furthermore, use Unity or UE5 to establish a digital twin platform. In the digital twin platform, evaluate the risk of the vessel passing through the bridge construction area based on real-time data. When the analysis result shows potential risks, the system automatically issues a warning to notify the construction team and relevant management personnel.

[0047] Furthermore, the warning information includes the estimated passing time of the vessel, the risk that the vessel cannot pass due to its excessive size, and the collision risk with the temporary structure.

[0048] A method for identifying and warning the passing of vessels during the bridge construction process based on digital twin, using the system for identifying and warning the passing of vessels during the bridge construction process based on digital twin, includes the following steps:

[0049] Obtain real-time data from the on-site water level monitor and meteorological platform, and obtain the water level prediction result through the processing of the multivariate LSTM time series prediction model;

[0050] Obtain data from the RGB camera and lidar, and obtain the three-dimensional space information of the temporary structure and the vessel through data processing;

[0051] Integrate the water level prediction result and the three-dimensional space information of the temporary structure and the vessel using the digital twin platform and the collision risk prediction module to conduct risk analysis and warning.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] The present invention overcomes the problem of the lack of a real-time collision risk identification and warning system at the traditional bridge construction site. Through the application of sensor technology and data processing capabilities, the system can monitor and predict the changes at the construction site in real time, including water level changes and vessel position changes, greatly improving the system's response speed to risks and the accuracy of warnings. The application of digital twin technology enables the system to more comprehensively simulate and analyze possible collision risks, take measures in advance to avoid the cross paths of different vessel types and temporary structures, and effectively reduce the accident risk. The data integration between different modules enables the system to have a high level of intelligence, be able to dynamically adjust the warning strategy according to the actual situation, and further improve the efficiency and level of safety management at the construction site. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a schematic structural diagram of a method for identifying and warning the passage of vessels during bridge construction based on digital twins in an embodiment of the present invention;

[0055] Figure 2 It is a schematic diagram of the water level prediction module in an embodiment of the present invention;

[0056] Figure 3 It is a schematic diagram of the temporary structure and vessel monitoring module in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be given with reference to the accompanying drawings.

[0058] Specific Embodiment 1: In combination with Figures 1 to 3 As shown, the present invention provides a vessel passage identification and warning system during bridge construction based on digital twins, including a water level prediction module, a temporary structure and vessel monitoring module, and a digital twin platform and collision risk prediction module;

[0059] The water level prediction module is used to obtain real-time data according to the on-site water level monitor and the meteorological platform, and obtain the water level prediction result after being processed by a multi-variable LSTM time series prediction model;

[0060] The water level prediction module obtains data through the on-site water level monitor and the meteorological platform, including air temperature, water temperature, weather conditions, and water level height, and processes these data using a multi-variable LSTM time series prediction model to predict future water level changes; the existence of this module solves the uncertainty problem of the prediction of vessel navigation paths due to changes in hydrological conditions and provides a reliable prediction basis for the safety of vessel passage;

[0061] The analysis of the data relies on a multivariate LSTM time series prediction model. By using the historical water level monitoring data and precipitation data from the meteorological platform as inputs, the training set and test set are divided, and the neural network is trained to obtain a high-precision water level change prediction model for this water area, which specifically includes:

[0062] 1. Data acquisition

[0063] To construct the dataset for the multivariate LSTM time series prediction model of water level prediction, it is necessary to pre-collect the daily temperature, water temperature, weather conditions, and water level height information at the target location in the past 5 years; these data can obtain temperature and weather data from the meteorological platform, water level height data from the water level monitor, and water temperature data from the water temperature monitor; during the collection process, it is necessary to ensure the integrity and accuracy of the data and handle missing values or outliers for subsequent data merging and model training;

[0064] 2. Data merging

[0065] During the process of constructing the dataset for the multivariate LSTM time series prediction model of water level prediction, data merging is a key step; specifically, it is necessary to precisely align and merge the daily temperature and weather condition data in the past 5 years obtained from the meteorological platform, the water level height data obtained from the water level monitor, and the water temperature data obtained from the water temperature monitor in chronological order; this process requires standardizing the timestamps of each data source to ensure that all data points can correspond on the same time dimension; the merged dataset will contain N variables including daily temperature, water temperature, weather conditions, and water level height, forming an N-dimensional time series dataset, providing comprehensive and accurate inputs for subsequent model training; the merged dataset M can be expressed as:

[0066] M = [x1, x2, …, x N

[0067] where, x N represents the N variables collected;

[0068] 3. Data preprocessing

[0069] Normalization processing: Normalize the data into the interval [0, 1] to eliminate the influence of dimension and order of magnitude, enhance the comparability between data

[0070] indicators, and accelerate the convergence speed of model training; the normalization formula is as follows:

[0071]

[0072] where, x is the original data, x’ is the normalized data, and min(x) and max(x) are the minimum and maximum values of this variable respectively; ​

[0073] 4. Dataset Division

[0074] In the process of constructing a multivariate LSTM time series prediction model for water level prediction, dataset division is to divide the collected and merged dataset into a training set and a test set according to a certain ratio. Usually, 80% of the data is selected as the training set for model training, and the remaining 20% of the data is used as the test set to evaluate the model performance. This can ensure that the model learns sufficient patterns during training and effectively evaluates the generalization ability of the model during testing, providing a basis for model training and evaluation;

[0075] 5. Model Training

[0076] In the model training stage, first, the multivariate time series data in the training set is input into the multivariate LSTM time series prediction model; this data includes the daily air temperature (T air ), water temperature (T water ), weather condition (W weather ), and water level height (H water ) information within the past 5 years; the input at each time step can be represented as a vector:

[0077] X t =[T air (t), T water (t), W weather (t), H water (t)]. The model captures the long-term dependencies in the time series through the LSTM layer; the forget gate, input gate, and output gate in the LSTM layer work together to determine which information needs to be retained, updated, or output; specifically, the forget gate (f t ) determines which information to discard, the input gate (i t ) determines which new information to add, and the output gate (o t ) determines which information to output; the calculation formulas for these gates are as follows:

[0078] f t =σ(W f ·[h t-1 , X t +b f )

[0079] i t =σ(W i ·[h t-1 , X t +b i )

[0080] C′ t =tanh(W C ·[h t-1 , X t+b C )

[0081] C t =f t ·C t-1 +i t ·C t ′

[0082] o t =σ(W o ·[h t-1 ,X t +b o )

[0083] h t =o t ·tanh(C t )

[0084] where h t-1 is the hidden state at the previous time step, C t-1 is the state at the previous time step, W f ,W i ,W C ,W o are weight matrices, b f ,b i ,b C ,b o are bias terms, σ is the sigmoid activation function, and tanh is the hyperbolic tangent activation function;

[0085] Next, the fully connected layer converts the output of the LSTM layer into the final predicted value, i.e., the water level height of the next day; the calculation formula of the fully connected layer is:

[0086] y t =W·h t +b

[0087] where y t is the predicted water level height, W is the weight matrix, and b is the bias term;

[0088] During the training process, the mean squared error is used as the loss function, and the difference between the predicted value and the actual value is minimized through the Adam optimization algorithm, thereby continuously adjusting the weights and biases of the model; the calculation formula of the mean squared error is:

[0089]

[0090] where n is the number of samples, y i is the actual value, is the predicted value;

[0091] After multiple rounds of iterative training, the model gradually learns the patterns and regularities in the data, improving the accuracy of predictions. Finally, the trained model will be used to predict the test set data to ensure that the model has good generalization ability.

[0092] 6. Water level prediction

[0093] In the prediction stage, the trained multivariate LSTM time series prediction model is used to predict the water level height of the next day based on the multivariate time series data of the past week (including air temperature, water temperature, weather conditions, and water level height). The specific steps are as follows:

[0094] First, normalize the data of these 7 days to meet the requirements of model input.

[0095] Next, construct the normalized 7-day data into a sequence sample and input it into the LSTM model. Through its internal LSTM layer, the model uses the collaborative effect of the forget gate, input gate, and output gate to process the input sequence, capture the long-term dependencies in the time series, and finally the fully connected layer outputs the predicted value of the water level height of the next day. The predicted value is the normalized value and needs to be converted back to the actual water level height through the inverse normalization formula:

[0096] H predicted (t + 7) = y'·(max(H water ) - min(H water )) + min(H water )

[0097] where y' is the normalized predicted value output by the model, and max(Hwater) and min(Hwater) are the maximum and minimum values of the water level height respectively.

[0098] Through this process, the predicted result of the water level height of the next day is obtained, providing a scientific basis for water level monitoring and related decision-making.

[0099] The temporary structure and vessel monitoring module is used to obtain data based on RGB cameras and lidar, and through data processing, obtain the three-dimensional spatial information of the temporary structure and vessels during the bridge construction process. It overcomes the limitations of the accuracy of vessel positions and construction temporary structures in traditional monitoring methods, enabling the system to accurately and in real-time grasp the actual situation of the construction site.

[0100] The digital twin platform and collision risk prediction module are used to integrate the water level prediction results and the three-dimensional spatial information of the temporary structure and vessels for risk analysis and early warning.

[0101] Build a digital twin platform using Unity or UE5. In the digital twin platform, evaluate the risk of ships passing through the bridge construction area based on real-time data. The real-time data includes the height, speed, direction of the ship, and the current state of the bridge. When the analysis results show potential risks, the system will automatically issue a warning to notify the construction team and relevant management personnel. The warning information should include specific details of the risk, such as the estimated passing time of the ship, the risk of the ship being unable to pass due to its large size, and the risk of collision with the temporary structure.

[0102] This module not only provides a real-time assessment of the risk of collision between the ship and the temporary structure, but also can issue a warning to the relevant responsible personnel in a timely manner to help them take necessary measures to avoid accidents.

[0103] Preferably, in the temporary structure and ship monitoring module, use an RGB camera and a lidar for calibration to obtain aligned RGB images and lidar data, specifically including,

[0104] Use the YOLOV8 model to perform semantic segmentation on the RGB image to identify the areas of the temporary structure and the ship in the image, and map the areas of the temporary structure and the ship in the image to the three-dimensional space using the point cloud data collected by the lidar to obtain the positions and sizes of the temporary structure and the ship in the three-dimensional space.

[0105] In this embodiment, the data of the RGB camera and the lidar are combined. Through calibration and alignment processing, visual images and high-precision three-dimensional point cloud data can be obtained simultaneously; this comprehensive perception system can more comprehensively capture the positions, shapes of the temporary structure and the ship and their relationships in the environment; using the YOLOV8 model for semantic segmentation and object recognition of the RGB image can quickly and accurately identify the temporary structure and the ship in the image; the YOLOV8 model has high accuracy and speed in object detection and is suitable for the application requirements in real-time scenarios; by mapping the positions of the identified temporary structure and the ship in the RGB image to the point cloud data collected by the lidar, precise positioning and size measurement of these objects in the three-dimensional space can be achieved; this three-dimensional reconstruction can provide the necessary spatial information basis for subsequent behavior prediction, path planning, and decision support; combining the above advantages, the system can monitor the dynamic positions and state changes of the ship and the temporary structure in real time, discover potential collision risks in a timely manner and give warnings; this real-time and accuracy are crucial for improving the safety and efficiency in the bridge construction process.

[0106] Specific implementation method two: Combine Figures 1 to 3 As shown, the present invention provides a method for identifying and warning ships passing through during the bridge construction process based on digital twin, including the following steps:

[0107] Obtain real-time data according to the on-site water level monitor and the meteorological platform, and obtain the water level prediction result through the processing of the multi-variable LSTM time series prediction model;

[0108] Obtain data according to the RGB camera and the lidar, and obtain the three-dimensional spatial information of the temporary structure and the ship through data processing;

[0109] Integrate the water level prediction result and the three-dimensional spatial information of the temporary structure and the ship by using the digital twin platform and the collision risk prediction module to conduct risk analysis and early warning.

[0110] Other combinations and connection relationships of this implementation plan are the same as those of the specific implementation plan one.

[0111] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art of the present invention can make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will all fall within the protection scope of the present invention.

Claims

1. A ship passing identification and warning system for bridge construction based on digital twin, characterized by: It includes water level prediction module, temporary structure and vessel monitoring module, and digital twin platform and collision risk prediction module; The water level prediction module is used to obtain data from previous years based on the on-site water level monitor, water temperature monitor and meteorological platform, use the multivariate LSTM time series prediction model to perform model training, and use the trained multivariate LSTM time series prediction model combined with real-time data to predict the water level information of the next day; The temporary structure and vessel monitoring module is used to obtain data from the RGB camera and the laser radar, and obtain the three-dimensional spatial information of the temporary structure and the vessel during the bridge construction process after calibration alignment, meaning segmentation and three-dimensional conversion; The digital twin platform and collision risk prediction module are used to integrate water level prediction results and three-dimensional spatial information of temporary structures and ships to conduct risk analysis and early warning.

2. The digital twin-based bridge construction process ship passing identification and warning system according to claim 1 is characterized by: In the water level prediction module, it specifically includes: (1) Data acquisition: collect daily air temperature, water temperature, weather conditions, and water level data at the target location over the past five years in advance, and handle missing values ​​or outliers; (2) Data merging: align and merge the data obtained in step (1) in chronological order, and perform time stamp standardization on each data source to ensure that all data points correspond to the same time dimension. The merged data set contains N variables, including daily air temperature, water temperature, weather conditions, and water level, forming an N-dimensional time series data set. The merged data set M can be expressed as: M=[x1,x2,…,x N ] Among them, x N represents the N variables collected; (3) Data preprocessing: normalize the data to the interval [0,1]: Among them, x is the original data, x' is the normalized data, min(x) and max(x) are the minimum and maximum values ​​of the variable respectively; (4) Data set division: 80% of the data is selected as the training set for model training, and the remaining 20% ​​of the data is used as the test set to evaluate model performance; (5) Model training. In the model training stage, the multivariate time series data in the training set are first input into the multivariate LSTM time series prediction model; including the daily temperature T in the past five years. air 、Water temperature T water 、Weather conditions weather and water level H water information; the input at each time step is represented as a vector: X t =[T air (t),T water (t),W weather (t),H water (t)] The model captures long-term dependencies in time series through the LSTM layer and the forget gate f t Decide which information to discard, input gate i t Decide what new information to add, output gate o t Determines what information to output; the calculation formula is as follows: f t =σ(W f ·[h t-1 ,X t ]+b f ) i t =σ(W i ·[h t-1 ,X t ]+b i ) C' t =tanh(W C ·[h t-1 ,X t ]+b C ) C t =f t ·C t-1 +i t ·C′ t the t =σ(W o ·[h t-1 ,X t ]+b o ) h t =o t ·tanh(C t ) Among them, h t-1 is the hidden state of the previous time step, C t-1 is the state of the previous time step, W f ,W i ,W C ,W o is the weight matrix, b f ,b i ,b C ,b o is the bias term, σ is the sigmoid activation function, and tanh is the hyperbolic tangent activation function; The fully connected layer converts the output of the LSTM layer into the final predicted value, which is the water level of the next day. The calculation formula of the fully connected layer is: y t =W·h t +b Among them, y t is the predicted water level, W is the weight matrix, and b is the bias term; During the training process, the mean square error is used as the loss function, and the difference between the predicted value and the actual value is minimized through the Adam optimization algorithm, so as to continuously adjust the weight and bias of the model; the calculation formula of the mean square error is: Where n is the number of samples, y i is the actual value, is the predicted value; After multiple rounds of iterative training, the model training is completed; (6) Water level prediction In the prediction stage, the trained multivariate LSTM time series prediction model is used to predict the water level of the next day based on the seven-day multivariate time series data. The specific steps are as follows: The seven-day collected data were normalized to meet the model input requirements; The normalized seven-day data is constructed into a sequence sample and input into the LSTM model trained in step (5). The model processes the input sequence through its internal LSTM layer, using the synergy of the forget gate, input gate, and output gate to capture the long-term dependencies in the time series. Finally, the fully connected layer outputs the predicted value of the water level for the next day. The predicted value is the normalized value and needs to be converted back to the actual water level through the denormalization formula: H predicted (t+7)=y′·(max(H water )-bright water ))+min(H water ) Among them, y′ is the normalized predicted value output by the model, max(Hwater) and min(Hwater) are the maximum and minimum values ​​of the water level, respectively.

3. The digital twin-based bridge construction process ship passing identification and warning system according to claim 2 is characterized by: In the temporary structure and vessel monitoring module, data acquired from RGB cameras and lidar include the height, speed, direction of the vessel, and the current status of the bridge.

4. The digital twin-based bridge construction process ship passing identification and warning system according to claim 3 is characterized by: The dataset is calibrated using an RGB camera and a LiDAR to obtain aligned RGB images and LiDAR data.

5. The digital twin-based bridge construction process ship passing identification and warning system according to claim 4 is characterized by: The YOLOV8 model is used to perform semantic segmentation on the RGB image to identify the areas of temporary structures and ships in the image, and the areas of temporary structures and ships in the image are mapped to three-dimensional space using point cloud data collected by lidar, thereby obtaining the position and size of the temporary structures and ships in the three-dimensional space.

6. The digital twin-based bridge construction process ship passing identification and warning system according to claim 5 is characterized by: Use Unity or UE5 to build a digital twin platform. In the digital twin platform, the risk of ships passing through the bridge construction area is assessed based on real-time data. When the analysis results show potential risks, the system automatically issues an early warning to notify the construction team and relevant managers.

7. The digital twin-based bridge construction process ship passing identification and warning system according to claim 6 is characterized by: Warning information includes the vessel's estimated time of passage, the risk of the vessel being too large to pass, and the risk of collision with temporary structures.

8. A ship passing identification and early warning method in the bridge construction process based on digital twin, characterized in that: The bridge construction process ship passing identification and warning system based on digital twin as described in any one of claims 1 to 7 comprises the following steps: Real-time data is obtained from on-site water level monitors and meteorological platforms, and the water level prediction results are obtained through processing with a multivariate LSTM time series prediction model; The data is acquired from the RGB camera and the LiDAR, and the three-dimensional spatial information of the temporary structure and the ship is obtained through data processing; The digital twin platform and collision risk prediction module are used to integrate water level prediction results and three-dimensional spatial information of temporary structures and ships for risk analysis and early warning.

Images