Tunnel portal light-dark boundary mileage prediction method based on deep learning

Abstract

The application relates to the technical field of tunnel design, and discloses a tunnel portal light-dark boundary mileage prediction method based on deep learning, which comprises the following steps: generating sequence features and relative positions of light-dark boundary mileages of each historical tunnel based on a tunnel portal data set of the historical tunnels; determining a model training data set according to the sequence features and the relative positions of the light-dark boundary mileages of the historical tunnels; constructing a regression prediction model according to the model training data set; obtaining portal data of a tunnel to be predicted, and predicting the portal data of the tunnel to be predicted by the regression prediction model to obtain light-dark boundary mileages of the tunnel to be predicted. The application can effectively improve the design efficiency and precision of the light-dark boundary mileages of the tunnel portal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel design, in particular to a tunnel portal light-dark boundary mileage prediction method based on deep learning. BACKGROUND

[0002] In the design of a tunnel portal, the selection of a light-dark boundary mileage will directly affect the selection of design parameters such as a tunnel portal type, a portal length, and a light tunnel length, and further affect the construction safety and efficiency of a tunnel entrance and exit. In order to determine a suitable tunnel portal light-dark boundary mileage, a designer needs to comprehensively measure factors such as a relative position of a terrain and a tunnel and an excavation difficulty of the terrain by experience. For an entrance or an exit, the main process is as follows: (1) selecting a certain mileage between a design starting mileage and a design ending mileage as a potential light-dark boundary mileage; (2) drawing a terrain and a tunnel cross-section and longitudinal section diagram corresponding to the mileage in step (1); (3) observing a relative position of the terrain and the tunnel and a terrain change trend on the cross-section and longitudinal section in step (2); and (4) reselecting a new mileage as a potential light-dark boundary mileage and repeating steps (2) and (3) until a suitable tunnel portal light-dark boundary mileage is found. As can be seen, the design of the tunnel portal light-dark boundary mileage needs to be iteratively tried, the design process is time-consuming and laborious, the entrance and the exit need to be independently designed without a unified method, and the design is highly dependent on the experience of the designer, resulting in low design efficiency and precision of the tunnel portal light-dark boundary mileage.

[0003] Therefore, how to effectively improve the design efficiency and precision of the tunnel portal light-dark boundary mileage is a problem to be solved at present.

[0004] The above content is only used to assist in understanding the technical solutions of the present application and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0005] The main purpose of the present application is to provide a tunnel portal light-dark boundary mileage prediction method based on deep learning, aiming to solve the technical problem of how to effectively improve the design efficiency and precision of the tunnel portal light-dark boundary mileage.

[0006] To achieve the above purpose, the present application provides a tunnel portal light-dark boundary mileage prediction method based on deep learning, which comprises the following steps:

[0007] generating sequence features and relative positions of light-dark boundary mileages of each historical tunnel based on a tunnel portal data set of the historical tunnel;

[0008] determining a model training data set according to the sequence features and the relative positions of the light-dark boundary mileages of each historical tunnel;

[0009] constructing a regression prediction model according to the model training data set;

[0010] obtain the portal data of a tunnel to be predicted, and perform prediction on the portal data of the tunnel to be predicted by the regression prediction model to obtain the light-dark demarcation mileage of the tunnel to be predicted.

[0011] In an embodiment, the generating of the sequence features and the relative positions of the light-dark demarcation mileages of each historical tunnel based on the historical tunnel portal data set comprises:

[0012] determining the discrete terrain point cloud data of each historical tunnel, the light-dark demarcation mileages of each historical tunnel and the tunnel location information of each historical tunnel based on the historical tunnel portal data set;

[0013] constructing the sequence features of each historical tunnel according to the discrete terrain point cloud data of each historical tunnel and the tunnel location information of each historical tunnel;

[0014] encoding the light-dark demarcation mileages of each historical tunnel to obtain the relative positions of the light-dark demarcation mileages of each historical tunnel.

[0015] In an embodiment, the constructing of the sequence features of each historical tunnel according to the discrete terrain point cloud data of each historical tunnel and the tunnel location information of each historical tunnel comprises:

[0016] determining the design start mileage of the tunnel axis, the design end mileage of the tunnel axis, the tunnel radius and the tunnel center position of each historical tunnel according to the tunnel location information of each historical tunnel;

[0017] determining the tunnel design range according to the design start mileage and the design end mileage;

[0018] determining a preset number of mileage points on the tunnel axis based on the tunnel design range and a preset distance interval, wherein the preset number of mileage points are uniformly distributed;

[0019] obtaining the cross sections corresponding to the preset number of mileage points, and performing linear interpolation on the discrete terrain point cloud data of each historical tunnel according to the cross sections to obtain the uniform terrain point cloud data of each historical tunnel;

[0020] determining the features on the cross sections according to the uniform terrain point cloud data, the tunnel center position and the tunnel radius;

[0021] generating the sequence features of each historical tunnel based on the features on the cross sections.

[0022] In an embodiment, the encoding of the light-dark demarcation mileages of each historical tunnel to obtain the relative positions of the light-dark demarcation mileages of each historical tunnel comprises:

[0023] Based on the design starting mileage, design ending mileage, and light-dark boundary mileage of each historical tunnel, the location information of the design starting mileage, the location information of the design ending mileage, and the location information of the light-dark boundary mileage are obtained by converting the design starting mileage, the design ending mileage, and the location information of the light-dark boundary mileage of each historical tunnel.

[0024] The relative positions of the light and dark boundary mileages of each historical tunnel are obtained by encoding the location information of the design start mileage, the design end mileage, and the light and dark boundary mileage.

[0025] In one embodiment, constructing a regression prediction model based on the model training dataset includes:

[0026] The model training dataset is divided according to a preset ratio to generate a training set and a validation set;

[0027] The preset deep learning model is trained based on the training set to obtain the trained deep learning model.

[0028] The trained deep learning model is validated based on the validation set to obtain the mean squared error loss value of the model in the validation set.

[0029] When the mean squared error loss of the model on the validation set reaches a preset threshold, the trained deep learning model is used as a regression prediction model.

[0030] In one embodiment, training a preset deep learning model based on the training set to obtain a trained deep learning model includes:

[0031] The sequence features in the training set are input into the preset deep learning model to obtain the predicted relative position;

[0032] The mean squared error loss value of the training set is calculated based on the predicted relative position and the relative position corresponding to the sequence features in the training set.

[0033] The weight parameters and bias terms of the preset deep learning model are optimized based on the mean squared error loss value of the model in the training set to obtain the trained deep learning model.

[0034] In one embodiment, the step of predicting the light-dark boundary mileage of the tunnel to be predicted based on the tunnel entrance data using the regression prediction model includes:

[0035] Construct the predictable sequence features based on the tunnel entrance data;

[0036] The features of the sequence to be predicted are standardized to obtain standardized features of the sequence to be predicted.

[0037] The standardized features of the sequence to be predicted are input into the regression prediction model for prediction, and the relative position of the light-dark boundary mileage of the tunnel to be predicted is obtained.

[0038] The relative position of the light-dark boundary mileage of the tunnel to be predicted is used for reverse encoding to obtain the light-dark boundary mileage of the tunnel to be predicted.

[0039] Furthermore, to achieve the above objectives, this application also proposes a deep learning-based device for predicting the mileage of the tunnel entrance light-dark boundary, the deep learning-based device for predicting the mileage of the tunnel entrance light-dark boundary comprising:

[0040] The generation module is used to generate the sequence features and relative positions of the light and dark boundary mileages of each historical tunnel based on the historical tunnel entrance dataset.

[0041] The determination module is used to determine the model training dataset based on the sequence characteristics of each historical tunnel and the relative position of the light-dark boundary mileage.

[0042] The building module is used to construct a regression prediction model based on the model training dataset;

[0043] The prediction module is used to acquire the portal data of the tunnel to be predicted, and to make a prediction based on the portal data of the tunnel to be predicted using the regression prediction model, so as to obtain the mileage of the light-dark boundary of the tunnel to be predicted.

[0044] Furthermore, to achieve the above objectives, this application also proposes a deep learning-based tunnel entrance light-dark boundary mileage prediction device. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the deep learning-based tunnel entrance light-dark boundary mileage prediction method described above.

[0045] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the deep learning-based tunnel entrance light-dark boundary mileage prediction method described above.

[0046] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the deep learning-based tunnel entrance light-dark boundary mileage prediction method described above.

[0047] This application provides a deep learning-based method for predicting the mileage of the tunnel entrance light-dark boundary. First, it generates the sequence features and relative positions of the light-dark boundary mileage for each historical tunnel based on a historical tunnel entrance dataset, thereby improving the representativeness and accuracy of the data. Then, it determines a model training dataset based on the sequence features and relative positions of the light-dark boundary mileage for each historical tunnel, improving the accuracy of the data samples in the dataset. Next, it constructs a regression prediction model based on the model training dataset, improving the prediction accuracy of the regression prediction model. Finally, it obtains the tunnel entrance data to be predicted and uses the regression prediction model to predict the light-dark boundary mileage of the tunnel, effectively improving the design efficiency and accuracy of the tunnel entrance light-dark boundary mileage.

[0048] In summary, this application extracts sequence features and encodes the relative positions of light and dark boundary mileage based on the portal dataset of designed and constructed historical tunnels, and establishes a deep learning regression prediction model through the model training dataset, thereby predicting the light and dark boundary mileage of the tunnel to be predicted. This overcomes the technical defects of low design efficiency and accuracy of the light and dark boundary mileage of the tunnel portal, and effectively improves the design efficiency and accuracy of the light and dark boundary mileage of the tunnel portal. Attached Figure Description

[0049] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating an embodiment of the deep learning-based method for predicting the mileage of the tunnel entrance light-dark boundary in this application.

[0052] Figure 2 This is a flowchart illustrating Embodiment 2 of the tunnel entrance light-dark boundary mileage prediction method based on deep learning in this application.

[0053] Figure 3 The contour map of the designed and constructed tunnel for the deep learning-based tunnel portal light-dark boundary mileage prediction method provided in Embodiment 1 of this application;

[0054] Figure 4 The contour map of the tunnel to be predicted is provided by the deep learning-based tunnel entrance light-dark boundary mileage prediction method in Embodiment 1 of this application.

[0055] Figure 5 A diagram illustrating the sequence feature extraction process of the deep learning-based tunnel entrance light-dark boundary mileage prediction method provided in Embodiment 1 of this application;

[0056] Figure 6 A schematic diagram of cross-sectional features of the tunnel entrance light-dark boundary mileage prediction method based on deep learning provided in Embodiment 2 of this application;

[0057] Figure 7 This is a flowchart illustrating Embodiment 3 of the deep learning-based method for predicting the mileage of the tunnel entrance light-dark boundary in this application.

[0058] Figure 8 A trend chart showing the variation of the model loss value of the deep learning-based tunnel entrance light-dark boundary mileage prediction method provided in Embodiment 3 of this application;

[0059] Figure 9 This is a schematic diagram of the module structure of the tunnel entrance light-dark boundary mileage prediction device based on deep learning, as described in an embodiment of this application.

[0060] Figure 10 This is a schematic diagram of the hardware operating environment involved in the deep learning-based method for predicting the light and dark boundary mileage of tunnel entrances in this application embodiment.

[0061] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0062] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0063] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0064] The main solution of this application embodiment is as follows: generating the sequence features and relative positions of the light-dark boundary mileage of each historical tunnel based on the tunnel entrance dataset; determining the model training dataset based on the sequence features and relative positions of the light-dark boundary mileage of each historical tunnel; constructing a regression prediction model based on the model training dataset; obtaining the tunnel entrance data of the tunnel to be predicted, and using the regression prediction model to predict based on the tunnel entrance data of the tunnel to be predicted, to obtain the light-dark boundary mileage of the tunnel to be predicted.

[0065] In the design of tunnel entrances, the selection of the boundary mileage between open and closed sections directly affects the selection of design parameters such as tunnel entrance type, entrance length, and open section length, thereby affecting the construction safety and efficiency of tunnel entrances and exits. Existing technical means require designers to comprehensively consider factors such as the relative position of the terrain and the tunnel, and the excavation difficulty of the terrain based on experience in order to determine a suitable boundary mileage between open and closed sections of the tunnel entrance. For entrances or exits, the main process is as follows: (1) Select a certain mileage between the design start mileage and the end mileage as a potential boundary mileage between open and closed sections; (2) Draw the terrain and tunnel cross-section diagrams corresponding to the mileage in step (1); (3) Observe the relative position of the terrain and the tunnel on the cross-section diagrams in step (2), and the trend diagram of terrain change; (4) Select a new mileage as a potential boundary mileage between open and closed sections, and repeat steps (2) and (3) until a suitable boundary mileage between open and closed sections of the tunnel entrance is found. It is evident that the design of the boundary mileage between open and closed tunnel entrances requires continuous iterative attempts, a time-consuming and labor-intensive process. The lack of a unified method for both entrances and exits necessitates independent design, and the process heavily relies on the experience of the designers, resulting in low efficiency and accuracy in the design of the boundary mileage. Therefore, effectively improving the design efficiency and accuracy of the boundary mileage between open and closed tunnel entrances is a pressing issue that needs to be addressed.

[0066] This application extracts sequence features and encodes the relative positions of light and dark boundary mileage based on the portal dataset of historically designed and constructed tunnels. It then establishes a deep learning regression prediction model using the model training dataset to predict the light and dark boundary mileage of the tunnel to be predicted. This overcomes the technical defects of low design efficiency and accuracy of the light and dark boundary mileage of the tunnel portal, and effectively improves the design efficiency and accuracy of the light and dark boundary mileage of the tunnel portal.

[0067] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a tunnel entrance light-dark boundary mileage prediction device based on deep learning. The following description uses a tunnel entrance light-dark boundary mileage prediction device based on deep learning as an example to illustrate this embodiment and the subsequent embodiments.

[0068] Based on this, this application provides a method for predicting the mileage of the light and dark boundary at the tunnel entrance based on deep learning, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the tunnel entrance light-dark boundary mileage prediction method based on deep learning in this application.

[0069] In this embodiment, the deep learning-based method for predicting the mileage of the tunnel entrance light-dark boundary includes steps S10 to S40:

[0070] Step S10: Generate the sequence features and relative positions of the light and dark boundary mileages of each historical tunnel based on the historical tunnel entrance dataset.

[0071] It should be noted that historical tunnels refer to tunnels that have been designed and constructed. The historical tunnel portal dataset is the same as the portal dataset of tunnels that have been designed and constructed. This dataset includes at least the topographic point cloud data surveyed near the tunnel portal entrance and exit, the design starting mileage and corresponding starting coordinates of the tunnel axis, the design ending mileage and corresponding ending coordinates of the tunnel axis, the tunnel radius, and the mileage of the light-dark boundary, etc. This embodiment does not impose specific limitations on these.

[0072] Understandably, sequence features refer to the sequence features of relative elevation and slope of the terrain, generated based on terrain point cloud data and tunnel location information, in order to fully reflect the relative positions of the terrain and tunnel, the trend of terrain changes, and the three-dimensional features within the survey area. The open-cut / cut-cover boundary mileage refers to the boundary mileage between the open-cut and cut-cover sections at the tunnel entrance; it is an absolute distance represented by characters and numbers.

[0073] Step S20: Determine the model training dataset based on the sequence characteristics of each historical tunnel and the relative position of the light-dark boundary mileage.

[0074] It should be noted that the sequential features of each historical tunnel are paired with the relative positions of the light-dark boundary mileage to form a new dataset, which will be used for subsequent model training, i.e., the model training dataset. Each sample in the model training dataset includes a sequential feature (i.e., relative terrain height and slope sequential features) and a corresponding relative position of the light-dark boundary mileage.

[0075] In the specific implementation, the sequence features of each historical tunnel are standardized and combined with the relative position labels of the corresponding light and dark boundary mileage to form a learnable model training dataset. The standardization of sequence features is as follows: Equation 1:

[0076]

[0077] In Equation 1, i represents the value of the feature index, such as 1, 2, etc. These are the sequence feature values ​​before and after standardization, u i σ i These are the mean and standard deviation of the sequence characteristics, respectively.

[0078] Step S30: Construct a regression prediction model based on the model training dataset.

[0079] It should be noted that deep learning algorithms (such as neural networks) are used to train the dataset to learn the relationship between sequence features and the relative positions of light and dark boundary mileage. By continuously iterating and optimizing the model parameters, a regression prediction model that can accurately predict the relative positions of light and dark boundary mileage is finally obtained.

[0080] Step S40: Obtain the tunnel entrance data of the tunnel to be predicted, and use the regression prediction model to predict the light-dark boundary mileage of the tunnel to be predicted based on the tunnel entrance data.

[0081] It should be noted that after obtaining the regression prediction model, the sequence features of the tunnel to be predicted are input into the model, and the model is used to make predictions to obtain the relative position of the light-dark boundary mileage of the tunnel to be predicted. Then, the light-dark boundary mileage of the predicted tunnel is obtained based on the relative position of the light-dark boundary mileage of the predicted tunnel, which greatly improves the design efficiency and accuracy of the light-dark boundary mileage of the tunnel entrance.

[0082] In one feasible implementation, step S40 may include: constructing a predictable sequence feature based on the tunnel entrance data of the tunnel to be predicted; standardizing the predictable sequence feature to obtain a standardized predictable sequence feature; inputting the standardized predictable sequence feature into the regression prediction model for prediction to obtain the relative position of the light-dark boundary mileage of the tunnel to be predicted; and performing reverse encoding based on the relative position of the light-dark boundary mileage of the tunnel to be predicted to obtain the light-dark boundary mileage of the tunnel to be predicted.

[0083] It should be noted that for the tunnel to be predicted, the sequence features corresponding to the three-dimensional features of the tunnel topography and the relative position of the tunnel, the trend of topographic change, and the survey range are extracted. These are the sequence features to be predicted. The extracted sequence features are standardized and input into the trained deep learning model to obtain the prediction result. The prediction result of the deep learning is a representation of the relative position of the light and dark boundary mileage of the tunnel entrance. Therefore, it needs to be transformed into the light and dark boundary mileage that represents the real distance, which is the final predicted light and dark boundary mileage.

[0084] Understandably, the predicted value of the deep learning model is a value between 0 and 1, representing the relative position of the light-dark boundary mileage at the tunnel entrance. Further reverse encoding is needed to obtain the corresponding light-dark boundary mileage, as shown in Equations 2 and 3 below:

[0085]

[0086]

[0087] In equations 2 and 3, These are the reverse-encoded entry and exit distances, representing the light and dark boundary mileage, y. a yb S represents the predicted relative position of the light and dark boundary mileage at the entrance and exit of the cave. s S e These are the design starting mileage and design ending mileage of the tunnel entrance to be predicted, respectively.

[0088] The final representation of the light-dark boundary mileage is composed of the numerical values ​​corresponding to the characters "km" and "m". Therefore, it is necessary to transform the numerical values ​​of the light-dark boundary mileage for entering and exiting the cave after reverse encoding to obtain the numerical values ​​required for character representation, as shown in Equations 4 and 5 below:

[0089] a = y / / 1000 (Equation 4)

[0090] b = y%1000 (Equation 5)

[0091] In Equations 4 and 5, y is the mileage value of the light-dark boundary after reverse encoding, a and b are the values ​​of km and m in the final mileage representation, / / is the floor operator, and % is the modulo operator.

[0092] In the specific implementation, the sequence feature values ​​of the entrance and exit of the tunnel to be predicted are standardized and input into the trained deep learning model. The predicted values ​​of the entrance and exit are 0.539 and 0.489, respectively. After inverse normalization of the predicted values, the light-dark boundary mileage values ​​corresponding to the entrance and exit of the tunnel to be predicted are 388333m and 389355m, respectively. Further conversion into character representations of the light-dark boundary mileages DK388+333 and DK389+355 are the final predicted light-dark boundary mileages.

[0093] This embodiment provides a deep learning-based method for predicting the mileage of the tunnel entrance light-dark boundary. First, it generates the sequence features and relative positions of the light-dark boundary mileage for each historical tunnel based on a historical tunnel entrance dataset, improving the representativeness and accuracy of the data. Then, it determines a model training dataset based on the sequence features and relative positions of the light-dark boundary mileage for each historical tunnel, improving the accuracy of the data samples in the dataset. Next, it constructs a regression prediction model based on the model training dataset, improving the prediction accuracy of the regression prediction model. Finally, it obtains the tunnel entrance data to be predicted and uses the regression prediction model to predict the light-dark boundary mileage of the tunnel, effectively improving the design efficiency and accuracy of the tunnel entrance light-dark boundary mileage.

[0094] In summary, this embodiment extracts sequence features and encodes the relative positions of light and dark boundary mileage based on the portal dataset of historically designed and constructed tunnels. It then establishes a deep learning regression prediction model using the model training dataset to predict the light and dark boundary mileage of the tunnel to be predicted. This overcomes the technical defects of low design efficiency and accuracy of the light and dark boundary mileage of tunnel portals and effectively improves the design efficiency and accuracy of the light and dark boundary mileage of tunnel portals.

[0095] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 Step S10 further includes steps S101-S103:

[0096] Step S101: Based on the tunnel entrance dataset of the historical tunnels, determine the discrete terrain point cloud data of each historical tunnel, the light-dark boundary mileage of each historical tunnel, and the tunnel location information of each historical tunnel.

[0097] It should be noted that the collected dataset of designed and constructed tunnels includes topographic point cloud data and design-related parameter information. Specifically, the collected discrete topographic point cloud data describes the x, y, and z coordinates of discrete topographic points near the tunnel entrance, such as... Figure 3 As shown, Figure 3 The diagram shows the contour lines of the designed and constructed tunnels, which include at least Tunnel 1 and Tunnel 2. The diagram includes (a) the contour lines of the entrance to Tunnel 1; (b) the contour lines of the exit to Tunnel 1; (c) the contour lines of the entrance to Tunnel 2; and (d) the contour lines of the exit to Tunnel 2.

[0098] As shown in Table 1, Table 1 contains relevant parameters for the designed and constructed tunnel entrances. The table includes the tunnel name, entrance / exit, design starting mileage, design starting mileage coordinates, design ending mileage, design ending mileage coordinates, the boundary between open and closed sections, and the tunnel radius. For example, the design starting mileage of the entrance of Tunnel 1 is DK84+88, the design starting mileage coordinates are (504046.26, 3119485.00, 222.12), the design ending mileage is DK84+999, the design ending mileage coordinates are (504162.92, 3119502.75, 223.25); the boundary between open and closed sections is DK84+959, and the tunnel radius is 7.5m.

[0099] Table 1

[0100]

[0101] The contour map drawn from the topographic point cloud data of the tunnel to be predicted, such as... Figure 4 As shown, Figure 4The image shows the contour map of the tunnel to be predicted, including (a) the entrance contour and (b) the exit contour.

[0102] As shown in Table 2, Table 2 contains relevant parameters for the tunnel entrance to be predicted. The table includes the entrance / exit, design starting mileage, design starting mileage coordinates, design ending mileage, design ending mileage coordinates, and tunnel radius. For example, the design starting mileage of the entrance of the tunnel to be predicted is DK388+282, the design starting mileage coordinates are (505532.04, 2880791.01, 205.93), the design ending mileage is DK388+377, the design ending mileage coordinates are (505583.42, 2880711.11, 206.22), and the tunnel radius is 7.5m.

[0103] Table 2

[0104]

[0105]

[0106] Step S102: Construct sequence features of each historical tunnel based on the discrete terrain point cloud data and the tunnel location information of each historical tunnel.

[0107] It should be noted that the sequence features are constructed based on the three-dimensional topographic point cloud data and tunnel location information within the range from the design start mileage to the end mileage. They can not only reflect the relative position of the terrain and tunnel on a single cross section and the trend of terrain change, but also reflect the three-dimensional features of the entire survey range of the tunnel entrance, which is more conducive to obtaining the optimal solution of the light and dark boundary mileage.

[0108] Understandably, this involves extracting topographic point cloud data from historical tunnel entrance datasets and using this data, along with tunnel location information, to generate a sequence of relative elevation and slope characteristics—essentially, a sequence feature. This sequence is generated based on the three-dimensional coordinates (including x, y, and y coordinates) of each topographic point in the point cloud data and the tunnel's location information. Specifically, it can be achieved by arranging the two feature values ​​of each topographic point in the point cloud data sequentially along the tunnel axis, forming a two-dimensional feature sequence. This two-dimensional sequence feature reflects the elevation and slope changes of the terrain near the tunnel entrance, providing crucial information for subsequent predictions.

[0109] In one feasible implementation, step S102 may include: determining the design start mileage, design end mileage, tunnel radius, and tunnel center position of each historical tunnel based on the tunnel location information of each historical tunnel; determining the tunnel design range based on the design start mileage and the design end mileage; determining a preset number of mileage points on the tunnel axis based on the tunnel design range and a preset distance interval, wherein the preset number of mileage points are uniformly distributed; obtaining the cross-sections corresponding to the preset number of mileage points; performing linear interpolation on the discrete terrain point cloud data of each historical tunnel based on the cross-sections to obtain uniform terrain point cloud data of each historical tunnel; determining the features on the cross-sections based on the uniform terrain point cloud data, the tunnel center position, and the tunnel radius; and generating sequence features of each historical tunnel based on the features on the cross-sections.

[0110] It should be noted that a fixed number of mileage points (N, where N is a positive integer) are selected at equal intervals along the tunnel design axis. The value of N is a crucial parameter. If the value is too small, the three-dimensional features of the entire survey area around the tunnel entrance cannot be represented; if the value is too large, the feature values ​​will be too biased due to excessive density when calculating terrain points based on interpolation. The value of N is determined by the average distance between the design end mileage and the starting mileage of all collected tunnel entrances and the selected mileage intervals. N also represents the length of the constructed sequence features, as shown in Equation 6 below:

[0111]

[0112] In Equation 6, S s S e denoted as the design starting mileage and design ending mileage of the tunnel entrance, respectively; i and n are the i-th tunnel and the total number of tunnels collected, respectively; and δ is the distance interval between uniform mileages, with a value of 1m.

[0113] The discrete topographic point cloud data from the survey is subjected to Delaunay triangulation. Further linear interpolation is performed on the cross-sections corresponding to N mileages uniformly selected based on the design start and end points to generate uniform topographic point cloud data. Then, various features on the cross-sections are calculated, such as feature 1 and feature 2, generating a feature sequence, such as... Figure 5 As shown, Figure 5 This is a diagram of the sequence feature extraction process.

[0114] To fully reflect the relative position of the terrain and the tunnel, and the changing trends of the terrain, various features on the cross-section are calculated, such as feature 1 and feature 2, etc. Figure 6 As shown, Figure 6This is a schematic diagram of the features on the cross section. The cross section includes feature 1 and feature 2. Feature 1 is the difference in elevation between the terrain at the centerline of the tunnel and the top of the tunnel entrance on the cross section. Feature 2 is the slope of the terrain on the cross section. Based on the left and right reference lines at a distance d from the centerline of the tunnel, the coordinates of terrain points A and B are calculated using uniform terrain point data, and then feature 2 is determined.

[0115] Feature 1 is the difference in elevation between the terrain at the tunnel centerline and the tunnel entrance top on the cross-section, representing the relative position of the terrain and the tunnel on that cross-section, as shown in Equation 7 below:

[0116] x1 = gz - pz - R (Equation 7)

[0117] In Equation 7, p z g z These represent the z-coordinate of the tunnel center location in the cross-section, the corresponding terrain point elevation, R, the tunnel radius, and f1, the value of feature 1. Where g... z It can be calculated from uniform terrain point data.

[0118] Feature 2 represents the slope of the terrain on the cross section. First, left and right reference lines with a distance d from the tunnel centerline are selected. The coordinates of terrain points A and B are calculated using uniform terrain point data. The value of Feature 2 is then calculated. Feature 2 effectively reflects the slope of the terrain, as shown in Equation 8 below:

[0119]

[0120] In Equation 8, p A p B Let p be the coordinates of points A and B respectively. n f1 is the unit normal vector of the longitudinal section where the tunnel axis is located, and f2 is the value of feature 2.

[0121] For the tunnel entrance, along the tunnel design axis from the design start mileage to the design end mileage, the two features corresponding to the selected mileages on the cross section are calculated sequentially and formed into a sequence feature. For the tunnel exit, along the tunnel design axis from the design end mileage to the design start mileage, the two features corresponding to the selected mileages on the cross section are calculated sequentially and formed into a sequence feature. Therefore, the form of the sequence features constructed for both the entrance and exit is {(x...} 11 ,x 21 ),...,(x 1t ,x 2t ),...,(x 1N ,x 2N )}, where x 1t x 2t These represent two features on the t-th cross-section, and N represents the number of selected mileages, i.e., the length of the sequence features.

[0122] It is worth noting that when the designed and constructed tunnels include tunnel 1 and tunnel 2, and N is 100, that is, the sequence feature shape is 100×2, the constructed sequence features are shown in Table 3. Table 3 is the inlet and outlet sequence feature table of the designed and constructed tunnels. The table includes the values ​​of feature 1 and feature 2 at the inlet of tunnel 1, the values ​​of feature 1 and feature 2 at the outlet of tunnel 1, the values ​​of feature 1 and feature 2 at the inlet of tunnel 2, and the values ​​of feature 1 and feature 2 at the outlet of tunnel 2.

[0123] Table 3

[0124]

[0125]

[0126] For the structural sequence characteristics of the tunnel to be predicted, the sequence feature values ​​obtained for the tunnel entrance and exit are shown in Tables 4 and 5. Table 4 is the sequence feature value table for the tunnel entrance, and Table 5 is the sequence feature value table for the tunnel exit. Table 4 includes the values ​​of feature 1 and feature 2 for the tunnel entrance, and Table 5 includes the values ​​of feature 1 and feature 2 for the tunnel exit.

[0127] Table 4

[0128]

[0129]

[0130] Table 5

[0131]

[0132]

[0133] Step S103: Encode the light-dark boundary mileage of each historical tunnel to obtain the relative position of the light-dark boundary mileage of each historical tunnel.

[0134] It should be noted that determining the relative position of the historical tunnel's light-dark boundary mileage within the starting and ending mileage ranges along the tunnel axis is crucial. This relative position can be represented by a relative distance. For the entrance, this is the ratio of the difference between the light-dark boundary mileage and the starting mileage to the difference between the ending mileage and the starting mileage. For the exit, it is the ratio of the difference between the ending mileage and the light-dark boundary mileage to the ratio between the ending mileage and the starting mileage. This relative position reflects the specific location of the light-dark boundary mileage on the tunnel axis and has significant reference value for predicting the light-dark boundary mileage of new tunnels.

[0135] Understandably, the light-dark boundary mileage is an absolute distance represented by characters and numbers. On the one hand, this is to quantify the light-dark boundary mileage; on the other hand, it is to unify the light-dark boundary mileage of different tunnels within the same value range. Therefore, the light-dark boundary mileage is encoded as a relative position.

[0136] In one feasible implementation, step S103 may include: converting the design starting mileage, design ending mileage, and light-dark boundary mileage of each historical tunnel according to the design starting mileage, design ending mileage, and light-dark boundary mileage of each historical tunnel respectively to obtain the location information of the design starting mileage, the location information of the design ending mileage, and the location information of the light-dark boundary mileage; encoding the location information of the design starting mileage, the location information of the design ending mileage, and the location information of the light-dark boundary mileage to obtain the relative position of the light-dark boundary mileage of each historical tunnel.

[0137] It should be noted that the description of tunnel mileage is based on characters, km and numbers, m and numbers, and needs to be quantified to convert it into absolute distance in meters, i.e., location information, as shown in Equation 9 below:

[0138] S = a × 1000 + b (Equation 9)

[0139] In Equation 9, a and b are the numerical descriptions of km and m in the character mileage, respectively, and S is the absolute distance of the mileage.

[0140] To unify the light-dark boundary mileage of different tunnels within the same value range, they are coded as their relative positions within the design start and end mileage ranges. The coding for the entrance and exit is as follows: Equations 10 and 11:

[0141]

[0142]

[0143] In Equations 10 and 11, S b S s S e These are the light / dark boundary mileage, the design start mileage, and the design end mileage, respectively. a y b These represent the relative positions of the light and dark boundary mileage for import and export, respectively. The coded light and dark boundary mileage has a value range between 0 and 1.

[0144] When the designed and constructed tunnels include Tunnel 1 and Tunnel 2, the light-dark boundary mileage at the entrance and exit of Tunnel 1 and Tunnel 2 is coded as shown in Table 6. Table 6 is the light-dark boundary mileage coding table, which includes the tunnel name, entrance / exit, light-dark boundary mileage and the coded light-dark boundary mileage. For example, the light-dark boundary mileage at the entrance of Tunnel 1 is DK84+959, and the coded light-dark boundary mileage is 0.661.

[0145] Table 6

[0146]

[0147] In this embodiment, by constructing sequence features based on discrete topographic point cloud data of each historical tunnel and encoding the relative position of the light-dark boundary mileage, the three-dimensional features of the entire survey range of the tunnel entrance can be reflected, which is more conducive to obtaining the optimal solution of the light-dark boundary mileage and effectively improving the prediction accuracy of the light-dark boundary mileage of the tunnel entrance.

[0148] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the tunnel entrance light-dark boundary mileage prediction method based on deep learning. Any simple modifications based on this technical concept are within the protection scope of this application.

[0149] Based on the first embodiment of this application, in the third embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 7 Step S30 further includes steps S301-S304:

[0150] Step S301: Divide the model training dataset according to a preset ratio to generate a training set and a validation set.

[0151] It should be noted that the preset ratio can be 8:2, and this embodiment does not impose a specific limitation on it. For example, the light and dark boundary mileage after standardizing and encoding the sequence features is used to form a dataset. The order of the dataset is shuffled, and 80% of the dataset is randomly selected as the training set and 20% as the validation set. The purpose of the training set is to train the model, and the purpose of the validation set is to judge the training effect of the model.

[0152] Step S302: Train the preset deep learning model according to the training set to obtain the trained deep learning model.

[0153] It should be noted that, in order to fully explore the temporal evolution characteristics of sequence data, the deep learning model consists of an input layer, a memory unit layer, a dropout layer, a fully connected layer, and an output layer. The memory unit layer is used to capture the long-term dependencies of the sequence data, and Dropout is used to suppress overfitting of the model.Figure 8 As shown, Figure 8 This is a schematic diagram of a deep learning model.

[0154] In one feasible implementation, step S302 may include: inputting the sequence features in the training set into the preset deep learning model to obtain the predicted relative position; calculating the mean squared error loss value of the training set model based on the predicted relative position and the relative position corresponding to the sequence features in the training set; optimizing the weight parameters and bias terms of the preset deep learning model based on the mean squared error loss value of the training set model to obtain the trained deep learning model.

[0155] It should be noted that the memory units and weight parameters in the fully connected layers of the deep learning model are initialized based on the normal distribution, and the bias term parameters are initialized to 0. The sequence feature values ​​are input into the deep learning model, and the output results of the first memory unit layer, the second memory unit layer, and the third memory unit layer are calculated in sequence. The output result of the last memory unit of the third memory unit layer is further input into the fully connected layer to obtain the final output result.

[0156] The output of the memory unit is shown in Equations 12 to 16 below:

[0157]

[0158]

[0159]

[0160]

[0161]

[0162] In equations 12 to 16, These represent the outputs of the forget gate, input gate, and output gate at time t in the memory unit, respectively. and These are the bias terms for the forget gate, output gate, and input gate, respectively. f W o W i and W c These are the weight parameters for the forget gate, output gate, and input gate, respectively. Let be the input value at time t. These are the state values ​​of the memory cells at times t-1 and t. These are the output values ​​of the memory cells at times t-1 and t, respectively.

[0163] The mean squared error loss of the training set is calculated based on the output of the memory unit and the relative position labels of the dataset. Then, the weight parameters and bias terms in the deep learning model are further optimized based on the Adam algorithm according to the mean squared error loss value, as shown in Equation 17 below:

[0164]

[0165] In Equation 17, y i , These represent the true value of the relative position and the output value of the deep learning model, respectively, where n is the number of samples.

[0166] Step S303: Validate the trained deep learning model based on the validation set to obtain the mean squared error loss value of the model in the validation set.

[0167] It should be noted that the validation set samples are input into the trained deep learning model, and the mean squared error loss value of the validation set is obtained by calculating according to Equation 17.

[0168] Step S304: When the mean squared error loss of the model in the validation set reaches a preset threshold, the trained deep learning model is used as a regression prediction model.

[0169] It should be noted that training the model will be terminated when the maximum number of iterations is reached or the loss value of the model on the test set no longer changes. At this point, the model is saved, which is the trained deep learning model. Otherwise, the training of the model will be repeated.

[0170] Understandably, if each layer in a deep learning model has 32 memory units and the neuron inactivation rate in the Dropout layer is set to 0.3, then the trend of the model's loss value on the training and validation sets will be as follows. Figure 8 As shown, Figure 8 The graph shows the trend of the model's loss value. The loss value of the deep learning model on both the training and test sets decreases and tends to stabilize as the number of training rounds increases, indicating that the model has achieved good convergence. The model's accuracy in predicting the light-dark boundary mileage on the training and test sets is 87.8% and 87.5%, respectively, showing that the model has a high prediction accuracy.

[0171] In this embodiment, a preset deep learning model is trained and validated using a training set and a validation set to obtain a regression prediction model, thereby improving the accuracy of the regression prediction model and thus improving the accuracy of the prediction of the mileage of the tunnel entrance light and dark boundary.

[0172] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the tunnel entrance light-dark boundary mileage prediction method based on deep learning. Any simple modifications based on this technical concept are within the protection scope of this application.

[0173] This application also provides a deep learning-based device for predicting the mileage of the light and dark boundary at the tunnel entrance. Please refer to [reference needed]. Figure 9 The deep learning-based tunnel entrance light-dark boundary mileage prediction device includes:

[0174] The generation module 10 is used to generate the sequence features and relative positions of the light and dark boundary mileage of each historical tunnel based on the historical tunnel entrance dataset.

[0175] The determination module 20 is used to determine the model training dataset based on the sequence characteristics of each historical tunnel and the relative position of the light-dark boundary mileage.

[0176] Module 30 is used to construct a regression prediction model based on the model training dataset.

[0177] The prediction module 40 is used to acquire the portal data of the tunnel to be predicted, and to make a prediction based on the portal data of the tunnel to be predicted using the regression prediction model, so as to obtain the light-dark boundary mileage of the tunnel to be predicted.

[0178] This embodiment provides a deep learning-based device for predicting the mileage of the tunnel entrance light-dark boundary. First, it generates the sequence features and relative positions of the light-dark boundary mileage for each historical tunnel based on a historical tunnel entrance dataset, thereby improving the representativeness and accuracy of the data. Then, it determines a model training dataset based on the sequence features and relative positions of the light-dark boundary mileage for each historical tunnel, improving the accuracy of the data samples in the dataset. Next, it constructs a regression prediction model based on the model training dataset, improving the prediction accuracy of the regression prediction model. Finally, it acquires the tunnel entrance data to be predicted and uses the regression prediction model to predict the light-dark boundary mileage of the tunnel to be predicted, effectively improving the design efficiency and accuracy of the tunnel entrance light-dark boundary mileage.

[0179] In summary, this embodiment extracts sequence features and encodes the relative positions of light and dark boundary mileage based on the portal dataset of historically designed and constructed tunnels. It then establishes a deep learning regression prediction model using the model training dataset to predict the light and dark boundary mileage of the tunnel to be predicted. This overcomes the technical defects of low design efficiency and accuracy of the light and dark boundary mileage of tunnel portals and effectively improves the design efficiency and accuracy of the light and dark boundary mileage of tunnel portals.

[0180] Optionally, the generation module 10 is further configured to determine the discrete terrain point cloud data of each historical tunnel, the light-dark boundary mileage of each historical tunnel, and the tunnel location information of each historical tunnel based on the tunnel entrance dataset of the historical tunnels; construct the sequence features of each historical tunnel based on the discrete terrain point cloud data of each historical tunnel and the tunnel location information of each historical tunnel; and encode the light-dark boundary mileage of each historical tunnel to obtain the relative position of the light-dark boundary mileage of each historical tunnel.

[0181] Optionally, the generation module 10 is further configured to: determine the design start mileage, design end mileage, tunnel radius, and tunnel center position of each historical tunnel based on the tunnel location information of each historical tunnel; determine the tunnel design range based on the design start mileage and the design end mileage; determine a preset number of mileage points on the tunnel axis based on the tunnel design range and a preset distance interval, wherein the preset number of mileage points are uniformly distributed; obtain the cross-sections corresponding to the preset number of mileage points; perform linear interpolation on the discrete terrain point cloud data of each historical tunnel based on the cross-sections to obtain uniform terrain point cloud data of each historical tunnel; determine the features on the cross-sections based on the uniform terrain point cloud data, the tunnel center position, and the tunnel radius; and generate sequence features of each historical tunnel based on the features on the cross-sections.

[0182] Optionally, the generation module 10 is further configured to convert the design starting mileage, the design ending mileage, and the light-dark boundary mileage of each historical tunnel according to the design starting mileage, the design ending mileage, and the light-dark boundary mileage of each historical tunnel respectively, to obtain the location information of the design starting mileage, the location information of the design ending mileage, and the location information of the light-dark boundary mileage; and to encode the location information of the design starting mileage, the location information of the design ending mileage, and the location information of the light-dark boundary mileage to obtain the relative position of the light-dark boundary mileage of each historical tunnel.

[0183] Optionally, the construction module 30 is further configured to divide the model training dataset according to a preset ratio to generate a training set and a validation set; train a preset deep learning model according to the training set to obtain a trained deep learning model; validate the trained deep learning model according to the validation set to obtain the model mean squared error loss value of the validation set; and when the model mean squared error loss value of the validation set reaches a preset threshold, use the trained deep learning model as a regression prediction model.

[0184] Optionally, the construction module 30 is further configured to input the sequence features in the training set into the preset deep learning model to obtain the predicted relative position; calculate the mean squared error loss value of the training set model based on the predicted relative position and the relative position corresponding to the sequence features in the training set; and optimize the weight parameters and bias terms of the preset deep learning model based on the mean squared error loss value of the training set model to obtain the trained deep learning model.

[0185] Optionally, the prediction module 40 is further configured to construct a predictable sequence feature based on the tunnel entrance data of the tunnel to be predicted; standardize the predictable sequence feature to obtain a standardized predictable sequence feature; input the standardized predictable sequence feature into the regression prediction model for prediction to obtain the relative position of the light-dark boundary mileage of the tunnel to be predicted; and perform reverse encoding based on the relative position of the light-dark boundary mileage of the tunnel to be predicted to obtain the light-dark boundary mileage of the tunnel to be predicted.

[0186] The tunnel entrance light-dark boundary mileage prediction device based on deep learning provided in this application adopts the tunnel entrance light-dark boundary mileage prediction method based on deep learning in the above embodiments, and can solve the technical problem of tunnel entrance light-dark boundary mileage prediction based on deep learning. Compared with the prior art, the beneficial effects of the tunnel entrance light-dark boundary mileage prediction device based on deep learning provided in this application are the same as the beneficial effects of the tunnel entrance light-dark boundary mileage prediction method based on deep learning provided in the above embodiments, and other technical features in the tunnel entrance light-dark boundary mileage prediction device based on deep learning are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0187] This application provides a deep learning-based tunnel entrance light-dark boundary mileage prediction device. The deep learning-based tunnel entrance light-dark boundary mileage prediction device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the deep learning-based tunnel entrance light-dark boundary mileage prediction method in the first embodiment described above.

[0188] The following is for reference. Figure 10This document illustrates a structural schematic diagram of a deep learning-based tunnel entrance light-dark boundary mileage prediction device suitable for implementing embodiments of this application. The deep learning-based tunnel entrance light-dark boundary mileage prediction device in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 10 The deep learning-based tunnel entrance light-dark boundary mileage prediction device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0189] like Figure 10 As shown, the deep learning-based tunnel entrance light-dark boundary mileage prediction device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the deep learning-based tunnel entrance light-dark boundary mileage prediction device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the deep learning-based tunnel entrance light-dark boundary mileage prediction device to exchange data with other devices wirelessly or via wired communication. Although the figure shows a deep learning-based tunnel entrance light-dark boundary mileage prediction device with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0190] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0191] The tunnel entrance light-dark boundary mileage prediction device provided in this application employs the deep learning-based tunnel entrance light-dark boundary mileage prediction method described in the above embodiments, and can solve the technical problem of deep learning-based tunnel entrance light-dark boundary mileage prediction. Compared with the prior art, the beneficial effects of the deep learning-based tunnel entrance light-dark boundary mileage prediction device provided in this application are the same as those of the deep learning-based tunnel entrance light-dark boundary mileage prediction method provided in the above embodiments, and other technical features in this deep learning-based tunnel entrance light-dark boundary mileage prediction device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0192] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0193] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0194] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the deep learning-based method for predicting the mileage of the tunnel entrance light and dark boundary in the above embodiments.

[0195] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0196] The aforementioned computer-readable storage medium may be included in a deep learning-based tunnel entrance light-dark boundary mileage prediction device; or it may exist independently and not be assembled into a deep learning-based tunnel entrance light-dark boundary mileage prediction device.

[0197] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a deep learning-based tunnel entrance light-dark boundary mileage prediction device, the deep learning-based tunnel entrance light-dark boundary mileage prediction device: generates sequence features and relative positions of light-dark boundary mileage for each historical tunnel based on a historical tunnel entrance dataset; determines a model training dataset based on the sequence features and relative positions of light-dark boundary mileage for each historical tunnel; constructs a regression prediction model based on the model training dataset; acquires entrance data for the tunnel to be predicted, and uses the regression prediction model to predict the light-dark boundary mileage of the tunnel to be predicted based on the entrance data of the tunnel to be predicted.

[0198] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0199] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0200] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0201] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described deep learning-based method for predicting the mileage of the tunnel entrance light-dark boundary, thereby solving the technical problem of predicting the mileage of the tunnel entrance light-dark boundary based on deep learning. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the deep learning-based method for predicting the mileage of the tunnel entrance light-dark boundary provided in the above embodiments, and will not be repeated here.

[0202] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the deep learning-based method for predicting the mileage of the tunnel entrance light and dark boundary as described above.

[0203] The computer program product provided in this application can solve the technical problem of predicting the mileage of the tunnel entrance light-dark boundary based on deep learning. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the deep learning-based tunnel entrance light-dark boundary mileage prediction method provided in the above embodiments, and will not be repeated here.

[0204] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for predicting the mileage of the light-dark boundary at a tunnel entrance based on deep learning, characterized in that, The method includes: Based on the historical tunnel entrance dataset, the sequence features of each historical tunnel and the relative positions of the light and dark boundary mileage are generated. The model training dataset is determined based on the sequence characteristics of each historical tunnel and the relative position of the light-dark boundary mileage. Construct a regression prediction model based on the training dataset of the model described above; The portal data of the tunnel to be predicted is obtained, and the regression prediction model is used to predict the mileage of the light-dark boundary of the tunnel to be predicted based on the portal data. The dataset of tunnel entrances based on historical tunnels generates sequential features and relative positions of light and dark boundary mileage for each historical tunnel, including: Based on the tunnel entrance dataset of the historical tunnels, the discrete terrain point cloud data of each historical tunnel, the light-dark boundary mileage of each historical tunnel, and the tunnel location information of each historical tunnel are determined. The sequence features of each historical tunnel are constructed based on the discrete terrain point cloud data and the tunnel location information of each historical tunnel. The relative positions of the light and dark boundary mileage of each historical tunnel are obtained by encoding the light and dark boundary mileage of each historical tunnel. The construction of sequence features for each historical tunnel based on discrete terrain point cloud data and tunnel location information includes: Based on the tunnel location information of each historical tunnel, the design starting mileage, design ending mileage, tunnel radius, and tunnel center location of each historical tunnel are determined. The tunnel design range is determined based on the design start mileage and the design end mileage; Based on the tunnel design range and preset distance intervals, a preset number of mileage points are determined on the tunnel axis, wherein the preset number of mileage points are evenly distributed; Obtain the cross-sections corresponding to the preset number of mileage points, and perform linear interpolation on the discrete terrain point cloud data of each historical tunnel based on the cross-sections to obtain the uniform terrain point cloud data of each historical tunnel. The features on the cross section are determined based on the uniform terrain point cloud data, the center position of the tunnel, and the radius of the tunnel. Based on the features on the cross-section, sequence features of each historical tunnel are generated; The step of encoding the light-dark boundary mileage of each historical tunnel to obtain the relative position of the light-dark boundary mileage of each historical tunnel includes: Based on the design starting mileage, design ending mileage, and light-dark boundary mileage of each historical tunnel, the location information of the design starting mileage, the location information of the design ending mileage, and the location information of the light-dark boundary mileage are obtained by converting the design starting mileage, the design ending mileage, and the location information of the light-dark boundary mileage of each historical tunnel. The relative positions of the light and dark boundary mileages of each historical tunnel are obtained by encoding the location information of the design start mileage, the design end mileage, and the light and dark boundary mileage.

2. The method as described in claim 1, characterized in that, The step of constructing a regression prediction model based on the model training dataset includes: The model training dataset is divided according to a preset ratio to generate a training set and a validation set; The preset deep learning model is trained based on the training set to obtain the trained deep learning model. The trained deep learning model is validated based on the validation set to obtain the mean squared error loss value of the model in the validation set. When the mean squared error loss of the model on the validation set reaches a preset threshold, the trained deep learning model is used as a regression prediction model.

3. The method as described in claim 2, characterized in that, The step of training a preset deep learning model based on the training set to obtain a trained deep learning model includes: The sequence features in the training set are input into the preset deep learning model to obtain the predicted relative position; The mean squared error loss value of the training set is calculated based on the predicted relative position and the relative position corresponding to the sequence features in the training set. The weight parameters and bias terms of the preset deep learning model are optimized based on the mean squared error loss value of the model in the training set to obtain the trained deep learning model.

4. The method as described in claim 1, characterized in that, The step of predicting the light-dark boundary mileage of the tunnel to be predicted by using the regression prediction model based on the tunnel entrance data includes: Construct the predictable sequence features based on the tunnel entrance data; The features of the sequence to be predicted are standardized to obtain standardized features of the sequence to be predicted. The standardized features of the sequence to be predicted are input into the regression prediction model for prediction, and the relative position of the light-dark boundary mileage of the tunnel to be predicted is obtained. The relative position of the light-dark boundary mileage of the tunnel to be predicted is used for reverse encoding to obtain the light-dark boundary mileage of the tunnel to be predicted.

5. A deep learning-based device for predicting the mileage of the light-dark boundary at a tunnel entrance, characterized in that, The deep learning-based tunnel entrance light-dark boundary mileage prediction device includes: The generation module is used to generate the sequence features and relative positions of the light and dark boundary mileages of each historical tunnel based on the historical tunnel entrance dataset. The determination module is used to determine the model training dataset based on the sequence characteristics of each historical tunnel and the relative position of the light-dark boundary mileage. The building module is used to construct a regression prediction model based on the model training dataset; The prediction module is used to acquire the portal data of the tunnel to be predicted, and to make a prediction based on the portal data of the tunnel to be predicted using the regression prediction model, so as to obtain the mileage of the light-dark boundary of the tunnel to be predicted. The generation module is further configured to determine the discrete terrain point cloud data, the light-dark boundary mileage, and the tunnel location information of each historical tunnel based on the tunnel entrance dataset of the historical tunnels; construct the sequence features of each historical tunnel based on the discrete terrain point cloud data and the tunnel location information of each historical tunnel; and encode the light-dark boundary mileage of each historical tunnel to obtain the relative position of the light-dark boundary mileage of each historical tunnel. The generation module is further configured to: determine the design start mileage, design end mileage, tunnel radius, and tunnel center position of each historical tunnel based on the tunnel location information of each historical tunnel; determine the tunnel design range based on the design start mileage and the design end mileage; determine a preset number of mileage points on the tunnel axis based on the tunnel design range and a preset distance interval, wherein the preset number of mileage points are uniformly distributed; obtain the cross-sections corresponding to the preset number of mileage points; perform linear interpolation on the discrete terrain point cloud data of each historical tunnel based on the cross-sections to obtain uniform terrain point cloud data of each historical tunnel; determine the features on the cross-sections based on the uniform terrain point cloud data, the tunnel center position, and the tunnel radius; and generate sequence features of each historical tunnel based on the features on the cross-sections. The generation module is further configured to convert the design starting mileage, design ending mileage, and light-dark boundary mileage of each historical tunnel according to the design starting mileage, design ending mileage, and light-dark boundary mileage of each historical tunnel respectively, to obtain the location information of the design starting mileage, the location information of the design ending mileage, and the location information of the light-dark boundary mileage; and to encode the location information of the design starting mileage, the location information of the design ending mileage, and the location information of the light-dark boundary mileage to obtain the relative position of the light-dark boundary mileage of each historical tunnel.

6. A deep learning-based device for predicting the mileage of the light-dark boundary at a tunnel entrance, characterized in that, The deep learning-based tunnel entrance light-dark boundary mileage prediction device includes: a memory, a processor, and a deep learning-based tunnel entrance light-dark boundary mileage prediction program stored in the memory and executable on the processor. The deep learning-based tunnel entrance light-dark boundary mileage prediction program is configured to implement the deep learning-based tunnel entrance light-dark boundary mileage prediction method as described in any one of claims 1 to 4.

7. A storage medium, characterized in that, The storage medium stores a deep learning-based tunnel entrance light-dark boundary mileage prediction program. When the deep learning-based tunnel entrance light-dark boundary mileage prediction program is executed by the processor, it implements the deep learning-based tunnel entrance light-dark boundary mileage prediction method as described in any one of claims 1 to 4.

Images