A method and system for detecting structural performance of tunnel lining construction

Abstract

The application provides a tunnel lining construction structure performance detection method and system, and relates to the technical field of tunnel structure detection. The method comprises the following steps: obtaining a preliminary optical image before tunnel lining construction; determining a plurality of landmark data point information by using a landmark data point determination model based on the preliminary optical image before tunnel lining construction; determining a plurality of preliminary reinforcement points based on a laser scanning image of a reference lining area; determining an estimated preliminary scanning reinforcement area of each remaining lining segmentation area based on the laser scanning image of the reference lining area, the preliminary optical image before tunnel lining construction and the plurality of preliminary reinforcement points; and determining a reinforcement point of each remaining lining segmentation area based on a laser scanning image of the estimated preliminary scanning reinforcement area of each remaining lining segmentation area. The method can efficiently and accurately realize intelligent zoning and reinforcement point positioning of the tunnel lining structure performance.

Description

Technical Field

[0001] This invention relates to the field of tunnel structure testing technology, and in particular to a method and system for testing the structural performance of tunnel lining construction. Background Technology

[0002] In tunnel construction and structural health monitoring, accurate and efficient testing of tunnel lining structure performance is crucial for ensuring its long-term safe and stable operation. Traditional testing methods mainly rely on manual inspections or fixed-point instrument measurements. These methods suffer from drawbacks such as low efficiency, limited coverage, insensitivity to hidden defects, and significant susceptibility to subjective experience. With the development of laser scanning and image recognition technologies, automated testing methods based on 3D point cloud data have begun to be applied. However, existing technologies typically require high-density, high-precision scanning of the entire tunnel wall, generating massive amounts of redundant data. This leads to high data processing costs and lengthy analysis cycles. Furthermore, this method lacks pre-emptive intelligent identification and guidance of lining structural features, failing to focus on critical areas with potential structural weaknesses from the outset. This results in a lack of specificity and economy in the testing process, making it difficult to conduct rapid and accurate surveys and assessments of large-scale tunnel networks in engineering practice.

[0003] Therefore, how to efficiently and accurately achieve intelligent zoning of tunnel lining structure performance and positioning of reinforcement points is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem solved by this invention is how to efficiently and accurately achieve intelligent zoning of tunnel lining structure performance and positioning of reinforcement points.

[0005] According to a first aspect, the present invention provides a method for structural performance testing of tunnel lining construction, comprising: acquiring a preliminary optical image before tunnel lining construction; determining multiple landmark data point information using a landmark data point determination model based on the preliminary optical image before tunnel lining construction; determining multiple lining segmentation regions based on the multiple landmark data point information, the multiple lining segmentation regions including a reference lining region and multiple remaining lining segmentation regions; acquiring a laser scanning image of the reference lining region; determining multiple preliminary reinforcement points based on the laser scanning image of the reference lining region; determining an estimated preliminary scanning reinforcement region for each remaining lining segmentation region based on the laser scanning image of the reference lining region, the preliminary optical image before tunnel lining construction, and the multiple preliminary reinforcement points; acquiring a laser scanning image of the estimated preliminary scanning reinforcement region for each remaining lining segmentation region; and determining reinforcement points for each remaining lining segmentation region based on the laser scanning image of the estimated preliminary scanning reinforcement region for each remaining lining segmentation region.

[0006] In one possible implementation, determining multiple lining segmentation regions based on the multiple marker data point information, wherein the multiple lining segmentation regions include a reference lining region and multiple remaining lining segmentation regions, comprises: clustering the multiple marker data point information to obtain K marker data point clusters; determining a reference lining region based on the K marker data point clusters; and determining multiple remaining lining segmentation regions based on the reference lining region and the K marker data point clusters.

[0007] In one possible implementation, determining the reinforcement points of each remaining lining segment based on the laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment includes: determining multiple reinforcement point information of the estimated preliminary scan reinforcement area of ​​each remaining lining segment based on the laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment; constructing a reinforcement knowledge graph, the reinforcement knowledge graph including multiple reinforcement points and multiple edges between the multiple reinforcement points, the node features of each reinforcement point being reinforcement point information and a preliminary optical image of each remaining lining segment; processing the reinforcement knowledge graph based on a graph neural network to determine multiple similar reinforcement areas of each remaining lining segment; acquiring laser scan images of the multiple similar reinforcement areas of each remaining lining segment; and determining the reinforcement points of the multiple similar reinforcement areas of each remaining lining segment based on the laser scan images of the multiple similar reinforcement areas of each remaining lining segment.

[0008] In one possible implementation, the marker data point determination model is a convolutional neural network.

[0009] According to a second aspect, the present invention provides a structural performance testing system for tunnel lining construction, comprising: an image acquisition module for acquiring a preliminary optical image before tunnel lining construction; a marker point determination module for determining multiple marker data point information based on the preliminary optical image before tunnel lining construction using a marker data point determination model; a region segmentation module for determining multiple lining segmentation regions based on the multiple marker data point information, the multiple lining segmentation regions including a reference lining region and multiple remaining lining segmentation regions; a reference scanning module for acquiring a laser scanning image of the reference lining region; and a preliminary reinforcement point determination module for... The system comprises: a laser scanning image of the reference lining area to determine multiple preliminary reinforcement points; a predicted area determination module to determine the predicted preliminary scan reinforcement area of ​​each remaining lining segment based on the laser scanning image of the reference lining area, the preliminary optical image before tunnel lining construction, and the multiple preliminary reinforcement points; a reinforcement area scanning module to acquire the laser scanning image of the predicted preliminary scan reinforcement area of ​​each remaining lining segment; and a final reinforcement point determination module to determine the reinforcement point of each remaining lining segment based on the laser scanning image of the predicted preliminary scan reinforcement area of ​​each remaining lining segment.

[0010] In one possible implementation, the region segmentation module is further configured to: cluster the multiple landmark data point information to obtain K landmark data point clusters; determine a reference lining region based on the K landmark data point clusters; and determine multiple remaining lining segmentation regions based on the reference lining region and the K landmark data point clusters.

[0011] In one possible implementation, the final reinforcement point determination module is further configured to: determine multiple reinforcement point information of the estimated preliminary scan reinforcement area of ​​each remaining lining segment based on the laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment; construct a reinforcement knowledge graph, the reinforcement knowledge graph including multiple reinforcement points and multiple edges between the multiple reinforcement points, the node features of each reinforcement point being reinforcement point information and a preliminary optical image of each remaining lining segment; process the reinforcement knowledge graph based on a graph neural network to determine multiple similar reinforcement areas of each remaining lining segment; acquire laser scan images of multiple similar reinforcement areas of each remaining lining segment; and determine the reinforcement points of the multiple similar reinforcement areas of each remaining lining segment based on the laser scan images of the multiple similar reinforcement areas of each remaining lining segment.

[0012] In one possible implementation, the marker data point determination model is a convolutional neural network.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method comprising: acquiring a preliminary optical image before tunnel lining construction; determining multiple landmark data point information using a landmark data point determination model based on the preliminary optical image before tunnel lining construction; determining multiple lining segmentation regions based on the multiple landmark data point information, the multiple lining segmentation regions including a reference lining region and multiple remaining lining segmentation regions; acquiring a laser scan image of the reference lining region; determining multiple preliminary reinforcement points based on the laser scan image of the reference lining region; determining an estimated preliminary scan reinforcement region for each remaining lining segmentation region based on the laser scan image of the reference lining region, the preliminary optical image before tunnel lining construction, and the multiple preliminary reinforcement points; acquiring a laser scan image of the estimated preliminary scan reinforcement region for each remaining lining segmentation region; and determining a reinforcement point for each remaining lining segmentation region based on the laser scan image of the estimated preliminary scan reinforcement region for each remaining lining segmentation region.

[0014] According to a fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for detecting the structural performance of tunnel lining construction. The method includes: acquiring a preliminary optical image before tunnel lining construction; determining multiple landmark data point information using a landmark data point determination model based on the preliminary optical image before tunnel lining construction; determining multiple lining segmentation regions based on the multiple landmark data point information, the multiple lining segmentation regions including a reference lining region and multiple remaining lining segmentation regions; acquiring a laser scanning image of the reference lining region; determining multiple preliminary reinforcement points based on the laser scanning image of the reference lining region; determining an estimated preliminary scanning reinforcement region for each remaining lining segmentation region based on the laser scanning image of the reference lining region, the preliminary optical image before tunnel lining construction, and the multiple preliminary reinforcement points; acquiring a laser scanning image of the estimated preliminary scanning reinforcement region for each remaining lining segmentation region; and determining reinforcement points for each remaining lining segmentation region based on the laser scanning image of the estimated preliminary scanning reinforcement region for each remaining lining segmentation region.

[0015] This invention provides a method and system for detecting the structural performance of tunnel lining construction. The method includes acquiring a preliminary optical image before tunnel lining construction; determining multiple landmark data point information using a landmark data point determination model based on the preliminary optical image before tunnel lining construction; determining multiple lining segmentation regions based on the multiple landmark data point information, the multiple lining segmentation regions including a reference lining region and multiple remaining lining segmentation regions; acquiring a laser scanning image of the reference lining region; determining multiple preliminary reinforcement points based on the laser scanning image of the reference lining region; determining an estimated preliminary scanning reinforcement region for each remaining lining segmentation region based on the laser scanning image of the reference lining region, the preliminary optical image before tunnel lining construction, and the multiple preliminary reinforcement points; acquiring a laser scanning image of the estimated preliminary scanning reinforcement region for each remaining lining segmentation region; and determining reinforcement points for each remaining lining segmentation region based on the laser scanning image of the estimated preliminary scanning reinforcement region for each remaining lining segmentation region. This method can efficiently and accurately achieve intelligent zoning and reinforcement point positioning of tunnel lining structural performance. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic flowchart illustrating a method for testing the structural performance of tunnel lining construction, provided in an embodiment of the present invention;

[0018] Figure 2 A schematic diagram of a tunnel before lining construction, provided as an embodiment of the present invention;

[0019] Figure 3 This is a schematic diagram of a process for determining multiple lining segmentation regions provided in an embodiment of the present invention;

[0020] Figure 4 A flowchart illustrating the process of determining reinforcement points for each remaining lining segmentation region, provided in an embodiment of the present invention.

[0021] Figure 5 This is a schematic diagram of a structural performance testing system for tunnel lining construction provided in an embodiment of the present invention.

[0022] Figure 6 This is a schematic diagram of the structure of an electronic device provided in one embodiment of this specification. Detailed Implementation

[0023] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0024] As used herein, the term "comprising" and its variations are open terms meaning "including but not limited to". The term "based on" means "at least partially based on". The terms "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other definitions, whether explicit or implicit, may be included below. Unless explicitly indicated by the context, the definition of a term shall remain consistent throughout the specification.

[0025] In this embodiment of the invention, the following are provided: Figure 1 The method for testing the structural performance of tunnel lining construction, as shown, includes steps S1 to S8:

[0026] Step S1: Obtain preliminary optical images before tunnel lining construction.

[0027] Preliminary optical images before tunnel lining construction are high-resolution images of the tunnel interior wall taken by industrial cameras installed at the tunnel construction site after the main excavation is completed and before the lining structure is constructed. Figure 2 This is a schematic diagram of a tunnel before lining construction, provided as an embodiment of the present invention.

[0028] Preliminary optical images before tunnel lining construction can record visual information such as the surface morphology, fissure distribution, and cross-sectional contour of the surrounding rock of the tunnel wall after the main tunnel body has been excavated and formed but before the lining structure is constructed.

[0029] Optionally, the camera continuously captures images of the tunnel interior wall along the tunnel axis at preset intervals and overlap rates (typically 30%-50%). Preliminary optical images before tunnel lining construction can record the surface morphology, fissure distribution, and cross-sectional profile of the surrounding rock of the tunnel wall after the main tunnel excavation and before the lining structure construction.

[0030] Step S2: Based on the preliminary optical image before the tunnel lining construction, use the landmark data point determination model to determine multiple landmark data point information.

[0031] The landmark data point determination model is a convolutional neural network. The input of the landmark data point determination model is the preliminary optical image before the tunnel lining construction, and the output of the landmark data point determination model is information on multiple landmark data points.

[0032] Convolutional Neural Networks (CNNs) can be applied to image feature extraction and object recognition tasks. A CNN consists of a multi-layered network structure, including convolutional layers, rectified linear unit layers, pooling layers, and fully connected layers. Convolutional layers extract local features from images through sliding kernel computation; pooling layers reduce feature dimensionality while preserving key information; and fully connected layers map the extracted features to the target output. CNNs can accurately capture discernible feature point information from optical images.

[0033] Multiple landmark data points are identified using a landmark data point determination model to pinpoint locations on the tunnel's inner wall surface that exhibit significant geometric or textural features. Each landmark data point includes its pixel coordinates, geometric distribution characteristics, and component attribute data.

[0034] Geometric distribution feature information represents the spatial distribution pattern and morphological attributes of landmark data points. Geometric distribution feature information includes the straight-line distance between landmark data points and adjacent landmark data points, the distribution density of landmark data points within a preset surrounding range, the contour curvature value of the local area where the landmark data point is located, and the undulation difference of the surface surrounding the landmark data point.

[0035] Component attribute data is the core attribute data of the tunnel structural components corresponding to the quantified marker data points. Component attribute data includes the crack width value at the location of the marker data point, the vertical distance to the nearest surrounding rock layer interface, the quantified value of the density of the surrounding rock material at the location, the pixel grayscale value, and the edge change intensity.

[0036] Preliminary optical images taken before tunnel lining construction contain rich visual textures and geometric shapes of the tunnel interior, implicitly containing feature points that can serve as positioning benchmarks. The model can identify highly recognizable anchor bolts, joints, and protruding structures, thereby accurately locating multiple landmark data points.

[0037] Convolutional neural networks (CNNs) can extract features from preliminary optical images before tunnel lining construction using convolutional layers. They can capture areas in the image with significant pixel grayscale variations and spatial distinctiveness, which correspond to key locations in the tunnel structure. CNNs use pooling layers to simplify feature data while preserving core information, and then fully connected layers classify and locate the features, ultimately identifying multiple landmark data points with clear coordinates and feature attributes.

[0038] Step S3: Based on the multiple landmark data point information, determine multiple lining segmentation regions, including a reference lining region and multiple remaining lining segmentation regions.

[0039] In some embodiments, Figure 3 This is a schematic flowchart illustrating the process of determining multiple lining segmentation regions according to an embodiment of the present invention. The determination of multiple lining segmentation regions includes steps S31 to S33:

[0040] Step S31: Cluster the multiple iconic data points to obtain K iconic data point clusters.

[0041] The clustering method used is K-means clustering. K-means clustering is an iterative clustering analysis algorithm that measures the similarity of samples by calculating the distance between them and grouping samples with high similarity into the same cluster. K-means clustering can divide data points into K clusters to minimize the sum of the squared distances from each data point to the centroid of its cluster.

[0042] The K-marker data point clusters are K sets of data formed by dividing multiple marker data points on the inner wall surface of the tunnel according to their spatial proximity and feature similarity. Each marker data point cluster contains several marker data points with similar pixel location coordinates, geometric distribution features, and component attribute data. Data points within the same cluster have strong correlations in spatial distribution and structural attributes, while data points between different clusters have obvious distinguishing features.

[0043] The pixel location coordinates, geometric distribution features, and component attribute data contained in multiple landmark data points are all quantifiable parameters. These parameters can clearly reflect the spatial location, distribution pattern, and structural attributes of each data point, and provide a clear basis for calculating the similarity between data points. The K-means clustering algorithm can achieve accurate grouping of data points based on these quantifiable indicators.

[0044] In some embodiments, the value of K can be determined by a preset relationship table between the value of K and the total length, cross-sectional perimeter, and total number of landmark data points of the tunnel lining. The longer the total length of the tunnel lining, the larger the cross-sectional perimeter, or the greater the total number of landmark data points, the larger the value of K. The preset relationship table between the value of K and the total length, cross-sectional perimeter, and total number of landmark data points of the tunnel lining is artificially constructed in advance based on tunnel engineering construction specifications and regional segmentation accuracy requirements.

[0045] The process of clustering multiple landmark data points using the K-means clustering algorithm is as follows: First, the K-means clustering algorithm randomly selects K points as initial cluster centers in the high-dimensional feature space of the multiple landmark data points. Then, it calculates the Euclidean distance or weighted feature distance between each landmark data point and these K initial cluster centers, and assigns each landmark data point to the cluster represented by the corresponding cluster center according to the nearest distance principle. After the initial assignment, the feature mean of all data points in each cluster is recalculated, and this mean is used as the new cluster center. Subsequently, the distance from all data points to the new cluster centers is recalculated, and the data points are reassigned. This assignment and update process is repeated iteratively. Each iteration reduces the intra-cluster variability while increasing the inter-cluster variability. As the number of iterations increases, the positions of the cluster centers gradually stabilize until the change in the cluster centers is less than a preset convergence threshold or the maximum number of iterations is reached, at which point the iteration stops. Finally, the converged clustering results are output as K landmark data point clusters.

[0046] K-means clustering is used to obtain K clusters of landmark data points. This allows landmark data points with similar pixel coordinates, geometric distribution features, and component attribute data to be grouped into the same cluster, thus achieving structured classification and management of scattered landmark data points. This operation can highlight the core feature differences between landmark data points in different clusters, while weakening the local feature fluctuations of landmark data points within a cluster. As a result, the distribution pattern of landmark data points on the tunnel inner wall is presented in cluster units, which facilitates the subsequent accurate determination of the benchmark lining area and multiple remaining lining segmentation areas based on cluster features.

[0047] Step S32: Determine the reference lining area based on the K clusters of landmark data points.

[0048] In some embodiments, a reference region determination model can be used to determine the reference lining region. The reference region determination model is a deep neural network. The input to the reference region determination model is the K clusters of landmark data points, and the output of the reference region determination model is the reference lining region.

[0049] A deep neural network is a neural network model that includes an input layer, multiple hidden layers, and an output layer. Deep neural networks can perform multi-level abstraction and feature extraction of input data through nonlinear transformations. They can handle complex, high-dimensional, nonlinear data relationships and can automatically learn latent patterns and regularities in the input data through forward propagation computation and backpropagation optimization, thereby achieving classification, regression, or recognition tasks.

[0050] The reference lining area is a sub-region selected from the tunnel space regions corresponding to K clusters of landmark data points by the reference area determination model as a reference standard.

[0051] The cluster of landmark data points corresponding to the benchmark lining area has the characteristics of uniform data distribution, regular geometric distribution features, and stable component attribute data. The benchmark lining area can comprehensively reflect the typical structural features of tunnel lining.

[0052] The baseline lining area can serve as a reference standard for subsequent comparative analysis and can be used to determine the relative deviation of other areas and to find defect patterns.

[0053] Each of the K clusters of landmark data points contains a set of data points with specific spatial ranges and characteristic attributes. These data aggregate information on the overall morphological quality, texture uniformity, and geometric regularity of the local area. These cluster-level features can reflect the differences in construction quality and changes in the surrounding rock condition in different sections of the tunnel.

[0054] Deep neural networks can extract and deeply analyze the features of K iconic data point clusters layer by layer. The model can mine deep information such as the spatial range, geometric distribution uniformity, and component property stability of each iconic data point cluster. Then, the deep neural network can comprehensively evaluate these features to select clusters with moderate spatial range, regular geometric distribution, and component property data that can represent the general characteristics of tunnel lining. The tunnel spatial region corresponding to this cluster is the benchmark lining region.

[0055] Step S33: Determine multiple remaining lining segmentation regions based on the reference lining region and the K clusters of landmark data points.

[0056] In some embodiments, a region segmentation model can be used to determine multiple remaining lining segmentation regions. The region segmentation model is a deep neural network. The input to the region segmentation model is the baseline lining region and the K clusters of landmark data points, and the output of the region segmentation model is the remaining lining segmentation regions.

[0057] Multiple remaining lining segmentation regions are determined by the region segmentation model, which are other independent sub-regions in the overall tunnel lining region other than the reference lining region.

[0058] Multiple remaining lining segments, together with the baseline lining segment, constitute the complete tunnel lining area. Each remaining lining segment includes a tunnel spatial region corresponding to one or more clusters of landmark data points.

[0059] The benchmark lining area clarifies the typical structural features and reference standards of the tunnel lining, and the K clusters of landmark data points provide information on the spatial distribution and structural attributes of different locations in the tunnel, thus fully reflecting the overall spatial structure and characteristic differences of the tunnel lining.

[0060] Deep neural networks can utilize the structural features and boundary information of the baseline lining region to determine its correspondence among K clusters of landmark data points, while also defining the cluster range corresponding to the baseline lining region. Subsequently, the deep neural network performs spatial correlation analysis and feature comparison on the remaining landmark data point clusters. The model first calculates the spatial distance between landmark data points in each cluster, the matching degree of geometric distribution features, and the similarity of component attribute data. Then, based on the accuracy requirements for tunnel lining region segmentation, a unified clustering merging threshold is set. When the spatial distance between two or more clusters is less than the threshold and the feature matching degree is higher than the threshold standard, the model merges them into an independent remaining lining segmentation region. When the spatial range and feature integrity of a single cluster meet the requirements for independent region division, and the feature difference with other clusters exceeds the merging threshold, the model separately classifies it as a remaining lining segmentation region.

[0061] Optionally, in some embodiments, a specific example of classifying the landmark data points and dividing the lining segmentation area can be illustrated as follows: 2000 landmark data points (e.g., including 1500 anchor points and the start, inflection, and end points of 200 cracks) are clustered to determine 50 landmark data point clusters (each cluster can correspond to a section of the tunnel approximately 10 meters long). The 50 clusters can be categorized as follows: 15 clusters corresponding to sections with neatly arranged anchors (with regular spacing and uniform distribution density within the cluster); 20 clusters corresponding to crack-developed areas (the points within the cluster are mainly crack feature points, and the crack width attributes are similar); 10 clusters corresponding to fractured surrounding rock zones (the points within the cluster include both anchors and cracks, and the undulation difference is large); and the remaining 5 clusters corresponding to local anomaly areas of over-excavation or under-excavation. Then, the 50 clusters were input into a deep neural network model for in-depth analysis of each cluster. When analyzing the 8th cluster, the anchor bolts were evenly arranged, the surrounding rock surface was smooth, there were no cracks, and the various attribute data were stable, representing the typical normal state of the tunnel. This cluster was marked as the baseline lining area. Then, the spatial distance, feature matching degree, and attribute similarity between each pair of the remaining 49 clusters were calculated, and combined with a preset merging threshold, the remaining lining segmentation areas were determined. These included: Clusters 1-7 (located to the left of the baseline area) and Clusters 9-15 (located to the right of the baseline area), although spatially continuous, had features (regular anchor bolt arrangement, no cracks) highly similar to the baseline area. The model merged them into two large remaining areas (left healthy area and right healthy area) based on spatial continuity; Clusters 16-35 (cracked area), although spatially discontinuous (scattered in multiple locations within the tunnel), had highly similar features (all crack characteristics). Based on the merging threshold, the model divides them into multiple independent remaining regions, each region consisting of one or more spatially continuous fracture clusters. Clusters 36-45 (surrounding rock fracture zone) are also divided into several independent remaining regions. Clusters 46-50 (over-excavation / under-excavation anomaly zone) have unique characteristics and differ significantly from other clusters, so each cluster is divided into an independent remaining region. This results in 15 remaining lining segmentation regions.

[0062] Step S4: Obtain a laser scan image of the reference lining area.

[0063] The laser scan image of the reference lining area is a point cloud projection image containing depth and reflection intensity information generated by a high-precision 3D laser scanner that performs a full-coverage scan of the defined reference lining area.

[0064] Laser scanning images of the reference lining area can accurately capture key information such as the three-dimensional spatial structure, surface defect details, and structural thickness distribution of the reference lining area.

[0065] Step S5: Determine multiple preliminary reinforcement points based on the laser scan image of the reference lining area.

[0066] In some embodiments, a first reinforcement point determination model can be used to determine a plurality of preliminary reinforcement points. The first reinforcement point determination model is a convolutional neural network. The input of the first reinforcement point determination model is a laser scan image of the reference lining area, and the output of the first reinforcement point determination model is a plurality of preliminary reinforcement points.

[0067] Multiple preliminary reinforcement points are determined by identifying specific locations within the reference lining area where structural defects, crack development, and surface anomalies exist, requiring structural reinforcement or special attention, through the first reinforcement point.

[0068] The laser scan image of the reference lining area contains depth field information and reflectivity variation information of the tunnel surface. Changes in these physical signals directly correspond to the physical state of the concrete surface. For example, cracks will appear as abrupt changes in pixel values ​​on the depth map, and water seepage areas will cause a decrease in laser reflection intensity.

[0069] Convolutional neural networks (CNNs) can extract multi-layer features from laser-scanned images of the baseline lining area. Through convolutional layers, CNNs can capture local features reflecting structural defects in the image, such as grayscale variations in cracks and the spatial distribution of areas with insufficient thickness. After simplifying feature data and retaining core information through pooling layers, the model then uses fully connected layers to classify and locate the features. This, combined with pre-defined structural safety standards, identifies specific locations where the structure is weak and requires reinforcement, ultimately determining multiple preliminary reinforcement points.

[0070] In some embodiments, determining multiple preliminary reinforcement points based on the laser scan image of the reference lining area includes steps S51-S53:

[0071] Step S51: Based on the laser scan image of the reference lining area, determine the height difference distribution map of the reference lining area, the distribution map of the contour change points of the reference lining area, and the reflection intensity fluctuation sequence of the reference lining area.

[0072] In some embodiments, a convolutional neural network can be used to determine the elevation difference distribution map of the reference lining area, the distribution map of abrupt change points in the contour of the reference lining area, and the reflection intensity fluctuation sequence of the reference lining area.

[0073] The elevation difference distribution map of the reference lining area is a two-dimensional image matrix output by a convolutional neural network, reflecting the changes in the smoothness of the tunnel inner wall surface. The value of each pixel in the elevation difference distribution map of the reference lining area represents the relative height deviation between its corresponding physical location and its preset range of neighboring surfaces.

[0074] The distribution map of abrupt changes in the profile of the reference lining area is a two-dimensional marker matrix output by a convolutional neural network, reflecting the distribution of points where the structural profile of the tunnel inner wall undergoes a sharp turn. The coordinate value of each marker point in the distribution map of abrupt changes in the profile of the reference lining area represents the three-dimensional spatial position where the turning angle of the lining profile exceeds a preset threshold at the corresponding physical location.

[0075] The reflection intensity fluctuation sequence of the reference lining area is a data sequence extracted by a convolutional neural network based on the reflection intensity channel data in the laser scanning image, which can reflect the change law of the optical properties of the material surface along the scanning path.

[0076] The reflection intensity fluctuation sequence of the reference lining area can record the echo intensity oscillation caused by the laser beam encountering different materials and different surface roughness during the scanning process, in the order of the scanning path.

[0077] Convolutional neural networks (CNNs) can process laser-scanned images of the reference lining area in parallel through a multi-branch convolutional architecture. The model can calculate the variance or gradient magnitude of the depth values ​​corresponding to physical points within a local area. When the local depth change exceeds a preset normal fluctuation threshold for a smooth surface, the corresponding neurons are activated, thereby generating a height difference distribution map of the reference lining area. Simultaneously, the model can use edge detection convolutional layers to identify discontinuous boundaries of the corresponding lining structure in the depth image. By analyzing the geometric connectivity between pixels, the CNN can extract isolated points or line features where depth values ​​undergo step changes, thus generating a distribution map of abrupt change points in the contour of the reference lining area. For the intensity channel data, the model can perform temporal and spatial analysis of the reflection intensity values ​​along the scan line direction using one-dimensional convolutional layers to capture the frequency of intensity abrupt changes and peak / trough characteristics. Furthermore, the model can identify abnormal fluctuation segments that differ from the normal reflective characteristics of concrete, ultimately encoding a reflection intensity fluctuation sequence for the reference lining area.

[0078] Step S52: Based on the height difference distribution map of the reference lining area, the abrupt change point distribution map of the reference lining area, and the reflection intensity fluctuation sequence of the reference lining area, determine the overall weakness heat map of the reference lining area, the set of abnormal points, and the degree of abnormality of each abnormal point.

[0079] In some embodiments, a convolutional neural network can be used to determine a global weakness heatmap of the baseline lining region, a set of anomalous points, and the degree of anomalousness of each anomalous point.

[0080] The global weakness heatmap of the reference lining area is a two-dimensional quantitative evaluation image matrix determined by a convolutional neural network, covering the structural weakness at various locations throughout the entire reference lining area. Each pixel in the global weakness heatmap corresponds to a physical location within the lining, and the pixel's value and color gradient directly represent the structural weakness at that location.

[0081] The set of anomaly points is a list of coordinates of suspected defect centers identified by a convolutional neural network. Each element in the set corresponds to a specific defect point on the tunnel wall that needs attention, such as the start or end point of a crack or the center point of a cavity.

[0082] The abnormality level data corresponding to the abnormal points is determined by a convolutional neural network and is used as a numerical indicator to quantify the severity of structural weakness of each abnormal point and its surrounding area in the abnormal point set.

[0083] Convolutional neural networks employ a multi-channel fusion strategy to concatenate elevation difference distribution maps, contour abrupt change point distribution maps, and dimensionally reshaped reflection intensity fluctuation sequences along the channel dimension, constructing a composite feature map rich in semantic information. The model extracts features from this composite feature map using a deep convolutional network, learning the nonlinear combination relationships between different features. For example, the model identifies regions with "huge elevation differences accompanied by contour abrupt changes" as typically corresponding to spalling or holes, while regions with "drastic fluctuations in reflection intensity and small elevation differences" may correspond to water seepage or micro-cracks. Based on these learned patterns, the model can output a pixel-level quantitative evaluation map, namely a heatmap of the overall weakness of the baseline lining area. Each pixel in the heatmap corresponds to a physical location of the lining, and the pixel value and color gradient directly represent the structural weakness at that location. Subsequently, the model can apply a region growing algorithm to analyze the heatmap, identifying regions with clustered high weakness values ​​as independent anomalous targets and extracting their geometric centers as a set of anomalous points. Meanwhile, the regression branch of the model calculates a comprehensive score based on the magnitude of the elevation difference at the location of the outlier, the severity of the contour change, and the attenuation of the reflection intensity, and through a fully connected layer, as the degree of anomalousness of each outlier, thus achieving a comprehensive evaluation from qualitative localization to quantitative classification.

[0084] Step S53: Based on the thermal map of the overall weakness of the reference lining area, the set of abnormal points, and the degree of abnormality of each abnormal point, determine multiple preliminary reinforcement points.

[0085] In some embodiments, a convolutional neural network can be used to determine multiple initial reinforcement points.

[0086] Convolutional neural networks extract high-risk areas from a heatmap of the overall weakness of the baseline lining area, identifying the highest-risk clusters as key reinforcement zones. The model then uses the 3D coordinates of anomaly points to spatially calibrate the points within the key reinforcement zone, ensuring coordinate accuracy. Subsequently, based on the risk level and quantified values ​​in the anomaly severity data, the model selects high-risk and extremely-risk points as multiple initial reinforcement points.

[0087] Step S6: Based on the laser scan image of the reference lining area, the preliminary optical image before tunnel lining construction, and the multiple preliminary reinforcement points, determine the estimated preliminary scan reinforcement area for each remaining lining segment.

[0088] In some embodiments, a reinforcement region determination model can be used to determine the estimated preliminary scan reinforcement region for each remaining lining segment. The reinforcement region determination model is a convolutional neural network. The inputs to the reinforcement region determination model are the laser scan image of the reference lining region, the preliminary optical image before tunnel lining construction, and the plurality of preliminary reinforcement points. The output of the reinforcement region determination model is the estimated preliminary scan reinforcement region for each remaining lining segment.

[0089] The estimated preliminary scan reinforcement area for each remaining lining segment is determined by the reinforcement area determination model. Within the remaining lining segment, there may be structural defects that require focused laser scanning verification of local areas.

[0090] Each remaining lining segment contains a pre-estimated preliminary scan reinforcement area.

[0091] The laser scan images of the baseline lining area and multiple preliminary reinforcement points constitute a precise correspondence between defects and features, and clearly define the characteristic manifestations of structural defects under the laser scanning dimension. Preliminary optical images before tunnel lining construction provide visual texture information across the entire tunnel, enabling the model to establish a correlation between structural defects identified by laser scanning and optical texture. Based on this correlation, the model can transfer and apply the defect features of the baseline area to the remaining lining segmentation areas where only optical images are available.

[0092] Convolutional neural networks can fuse features from laser scan images of the baseline lining area and multiple preliminary reinforcement points to extract typical features of structurally weak areas, including geometric features and component attribute features. Subsequently, the model can perform layer-by-layer comparison and analysis of these features with the structural features of the remaining lining segmented areas in the preliminary optical images before tunnel lining construction. Through the synergistic effect of convolutional layers, pooling layers, and fully connected layers, it can filter out the areas in the remaining lining segmented areas with the highest feature similarity to the baseline weak area, ultimately determining the estimated boundaries and extent of the preliminary scan reinforcement area for each remaining lining segmented area.

[0093] Step S7: Obtain the laser scan image of the estimated preliminary scan reinforcement area for each remaining lining segment.

[0094] The laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment is a high-precision image data obtained by scanning the estimated preliminary scan reinforcement area of ​​each remaining lining segment using a laser scanner.

[0095] The laser scanning images of the estimated preliminary scan reinforcement area of ​​the remaining lining segment can accurately capture key information such as the three-dimensional spatial structure, surface defect details, and structural thickness distribution of the estimated preliminary scan reinforcement area.

[0096] Step S8: Determine the reinforcement points for each remaining lining segment based on the laser scan image of the estimated preliminary scan reinforcement area for each remaining lining segment.

[0097] In some embodiments, Figure 4 This is a flowchart illustrating the process of determining the reinforcement points for each remaining lining segment region according to an embodiment of the present invention. The determination of the reinforcement points for each remaining lining segment region includes steps S81 to S85:

[0098] Step S81: Based on the laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment, determine multiple reinforcement point information of the estimated preliminary scan reinforcement area of ​​each remaining lining segment.

[0099] In some embodiments, a second reinforcement point determination model can be used to determine multiple reinforcement point information of the estimated preliminary scan reinforcement area for each remaining lining segment. The second reinforcement point determination model is a convolutional neural network. The input of the second reinforcement point determination model is the laser scan image of the estimated preliminary scan reinforcement area for each remaining lining segment, and the output of the second reinforcement point determination model is the multiple reinforcement point information of the estimated preliminary scan reinforcement area for each remaining lining segment.

[0100] The information on multiple reinforcement points in the estimated preliminary scan reinforcement area of ​​each remaining lining segment is determined by the second reinforcement point determination model. The information on the details of multiple reinforcement points in the estimated preliminary scan reinforcement area of ​​each remaining lining segment is also provided.

[0101] The reinforcement point information includes the three-dimensional spatial coordinates of each reinforcement point, the defect type, the quantified value of the defect size, the degree of surface deformation caused by the defect, and the reinforcement priority.

[0102] Defect types include surrounding rock fissures, surrounding rock fractures and cavities, surrounding rock layering and peeling, and sudden changes in the surface of surrounding rock.

[0103] Defect size quantification is a specific numerical representation of the actual size of a defect, such as the length and width of a crack, the diameter of a hole, and the area of ​​delamination.

[0104] The degree of surface deformation caused by defects is a quantitative indicator of the deviation of the tunnel inner wall surrounding rock surface from the normal flat shape caused by defects.

[0105] The reinforcement priority is a classification of reinforcement order levels based on the degree of impact of defects on the structural load-bearing stability after tunnel lining construction.

[0106] The laser scan images of the estimated preliminary scan reinforcement area for each remaining lining segment provide accurate physical surface data within the estimated preliminary scan reinforcement area, and can eliminate artifact interference factors such as stains and uneven lighting in the optical images. The laser scan images of the estimated preliminary scan reinforcement area can directly reflect the physical integrity of the lining structure and are the final physical basis for determining the reinforcement points.

[0107] The convolutional neural network (CNN) can extract features layer by layer from the laser scan images of the estimated preliminary scan reinforcement area for each remaining lining segment. Through convolutional layers, the CNN can accurately capture the characteristics of surrounding rock defects corresponding to the reinforcement points, such as the length, width, and depth of rock fissures, the diameter and depth of broken holes in the surrounding rock, the area and thickness of rock layering, and the undulation differences of abrupt changes in the rock surface. Fully connected layers can classify and quantify the extracted defect features, and clearly define the three-dimensional spatial coordinates and various defect quantification parameters for each reinforcement point. By combining the surrounding rock structure safety assessment standards for the bare tunnel stage and prioritizing reinforcement based on the impact of defects on subsequent lining construction and structural load-bearing stability, the model can ultimately generate multiple reinforcement point information for each remaining lining segment's estimated preliminary scan reinforcement area.

[0108] Step S82: Construct a reinforcement knowledge graph, which includes multiple reinforcement points and multiple edges between the reinforcement points. The node features of each reinforcement point are reinforcement point information and a preliminary optical image of each remaining lining segmentation area.

[0109] A knowledge graph for reinforcement is a data structure that presents information and relationships related to reinforcement points in a graphical form. It uses a combination of nodes (vertices) and edges (edges) to intuitively reflect the attribute characteristics of reinforcement points and the relationship patterns between them.

[0110] In the reinforcement knowledge graph, each reinforcement point represents a node. Each reinforcement point node has node features, which include reinforcement point information and a preliminary optical image of each remaining lining segmentation area.

[0111] The edges between reinforced nodes represent the positional relationships between different reinforced points. Each edge has edge characteristics, including relative direction and straight-line distance. For example, the edge between reinforced point M and reinforced point N has edge characteristics including the spatial pointing direction of reinforced point M relative to reinforced point N and the numerical straight-line distance between them.

[0112] Step S83: Based on the graph neural network, the reinforcement knowledge graph is processed to determine multiple similar reinforcement areas for each remaining lining segmentation area.

[0113] Graph Neural Networks (GNNs) are neural network models capable of processing knowledge graphs, a type of data with a graphical structure. GNNs can fully utilize node features and edge associations for feature learning and reasoning. By aggregating the features of a node itself with those of its neighboring nodes, GNNs can capture complex relationships within the knowledge graph, thereby enabling tasks such as node or graph classification and prediction. GNNs possess powerful structured data processing capabilities. The input to the GNN is the reinforced knowledge graph, and the output is multiple similar reinforced regions for each remaining lining segmentation region.

[0114] Multiple similar reinforcement regions for each remaining lining segment are determined by processing the reinforcement knowledge graph through a graph neural network. These regions are located outside the estimated preliminary scan reinforcement region of the corresponding remaining lining segment and are multiple independent reinforcement regions with similar reinforcement point information features to the estimated preliminary scan reinforcement region.

[0115] Each remaining lining segment contains the estimated preliminary scan reinforcement area of ​​the remaining lining segment and multiple similar reinforcement areas of the remaining lining segment.

[0116] By constructing a reinforcement knowledge graph, the network of connections between reinforcement points and surrounding areas, and between defect features and optical image features, can be clearly mapped. The distribution patterns of reinforcement points and the evolutionary characteristics of defects are often deeply correlated with the structural features of the surrounding environment. Therefore, using the location attributes and optical image features of reinforcement point nodes as node features, and the feature similarity between regions as edge features, can more fully unlock the intrinsic value of the data. This helps graph neural networks better understand the association logic and influence mechanism between reinforcement points and defect areas, and improves the recognition accuracy of hidden reinforcement areas. Processing reinforcement knowledge graph data based on graph neural network models can effectively learn the complex associations and feature transmission patterns between nodes, thereby more accurately mining the distribution patterns of reinforcement points and the common features of defects. Compared to traditional visual matching algorithms, graph neural networks have stronger feature representation and association learning capabilities when processing structured reinforcement knowledge graphs.

[0117] Graph neural networks (Graph Neural Networks) can perform multi-layer graph convolution operations on a hardened knowledge graph. In each layer, each hardened node aggregates feature information from its connected neighbors and updates its state by combining this with its own preliminary optical image features. This process enables the effective propagation of defect feature patterns within the hardened knowledge graph. By learning strong correlation patterns between known hardened points and their optical features, Graph Neural Networks can identify regions in the hardened knowledge graph that are not marked as hardened points but whose optical feature vectors are highly similar to those of high-severity hardened nodes. Graph Neural Networks can calculate the feature distance or connection strength between unscanned regions and core defect clusters in the graph. When the feature representation of an unscanned region highly matches the feature pattern of the corresponding defect of a known hardened point, the Graph Neural Network classifies this region as a similar hardened region. In this process, Graph Neural Networks can leverage the feature correlation transitivity of the hardened knowledge graph to uncover hidden regions that are easily overlooked by simple visual matching but are highly homologous to defects in the structured feature space.

[0118] Step S84: Obtain laser scan images of multiple similar reinforced areas in each remaining lining segmentation area.

[0119] The laser scan images of multiple similar reinforced areas within each remaining lining segment are high-precision image data obtained by scanning multiple similar reinforced areas within each remaining lining segment using a laser scanner.

[0120] Laser scanning images of similar reinforced areas can accurately present key information such as the three-dimensional spatial structure, surface defect details, and structural thickness distribution of similar reinforced areas.

[0121] Step S85: Based on the laser scan images of multiple similar reinforcement areas in each remaining lining segment, determine the reinforcement points of multiple similar reinforcement areas in each remaining lining segment.

[0122] In some embodiments, a third reinforcement point determination model can be used to determine the reinforcement points of multiple similar reinforcement areas in each remaining lining segment. The third reinforcement point determination model is a convolutional neural network. The input to the third reinforcement point determination model is a laser scan image of multiple similar reinforcement areas in each remaining lining segment, and the output of the third reinforcement point determination model is the reinforcement points of the multiple similar reinforcement areas in each remaining lining segment.

[0123] The reinforcement points of multiple similar reinforcement areas in each remaining lining segment are the specific locations that need reinforcement treatment, determined by the third reinforcement point determination model within the multiple similar reinforcement areas of each remaining lining segment.

[0124] The reinforcement points of multiple similar reinforcement zones in the remaining lining segment area correspond to locations in the similar reinforcement zones that are structurally weak and pose safety hazards. The reinforcement points of multiple similar reinforcement zones, together with the reinforcement points of the estimated preliminary scan reinforcement zone, constitute the complete set of reinforcement points for the remaining lining segment area.

[0125] Convolutional neural networks (CNNs) can extract and analyze features from laser scan images of multiple similar reinforcement areas within each remaining lining segment. Convolutional layers capture local features of structural defects in the image, such as grayscale variation areas corresponding to cracks and the spatial extent of areas with insufficient thickness. After the extracted feature data is simplified by pooling layers while retaining core information, fully connected layers classify and locate the features. Combined with pre-defined structural safety standards and the characteristic patterns of similar reinforcement areas, the model can then select specific locations requiring reinforcement, ultimately determining the reinforcement points for multiple similar reinforcement areas within each remaining lining segment, ensuring comprehensive and accurate reinforcement of the remaining lining segment.

[0126] Please refer to the following. Figure 5 , Figure 5 A structural schematic diagram of a structural performance testing system for tunnel lining construction provided in an embodiment of this specification is shown. It should be noted that... Figure 5 The structural performance testing system for tunnel lining construction shown is used to perform the functions described in this manual. Figure 1 The methods shown in the embodiments are illustrated for ease of explanation, showing only the parts related to the embodiments of this specification. For specific technical details not disclosed, please refer to this specification. Figure 1 The example shown.

[0127] Based on the same inventive concept Figure 5This is a schematic diagram of a structural performance testing system for tunnel lining construction provided by an embodiment of the present invention. The structural performance testing system for tunnel lining construction includes:

[0128] Image acquisition module 91 is used to acquire preliminary optical images before tunnel lining construction;

[0129] The marker point determination module 92 is used to determine multiple marker data point information based on the preliminary optical image before the tunnel lining construction using a marker data point determination model.

[0130] The region segmentation module 93 is used to determine multiple lining segmentation regions based on the multiple landmark data point information, the multiple lining segmentation regions including a reference lining region and multiple remaining lining segmentation regions;

[0131] The reference scanning module 94 is used to acquire laser scanning images of the reference lining area;

[0132] The preliminary reinforcement point determination module 95 is used to determine multiple preliminary reinforcement points based on the laser scanning image of the reference lining area;

[0133] The estimated area determination module 96 is used to determine the estimated preliminary scan reinforcement area of ​​each remaining lining segment based on the laser scan image of the reference lining area, the preliminary optical image before the tunnel lining construction, and the multiple preliminary reinforcement points.

[0134] The reinforcement area scanning module 97 is used to acquire the laser scanning image of the estimated preliminary scan reinforcement area for each remaining lining segmentation area;

[0135] The final reinforcement point determination module 98 is used to determine the reinforcement point of each remaining lining segment based on the laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment.

[0136] See Figure 6 It shows a schematic diagram of the structure of an electronic device according to an embodiment of this specification, which can be used to implement... Figure 1 The method in the illustrated embodiment. (As shown) Figure 6 As shown, the electronic device 600 may include: at least one central processing unit 601, at least one network interface 604, user interface 603, memory 605, and at least one communication bus 602.

[0137] The communication bus 602 is used to enable communication between these components.

[0138] The user interface 603 may include a display screen and a camera. Optionally, the user interface 603 may also include a standard wired interface and a wireless interface.

[0139] The network interface 604 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0140] The processor 601 may include one or more processing cores. The processor 601 connects to various parts within the electronic device 600 using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 605, and by calling data stored in the memory 605. Optionally, the processor 601 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 601 may integrate one or a combination of several of the following: a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), and a modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content to be displayed on the screen; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 601 and may be implemented as a separate chip.

[0141] The memory 605 may include random access memory (RAM) or read-only memory. Optionally, the memory 605 may include a non-transitory computer-readable storage medium. The memory 605 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 605 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 605 may also be at least one storage system located remotely from the aforementioned processor 601. Figure 6 As shown, the memory 605, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and program instructions.

[0142] exist Figure 6 In the illustrated electronic device 600, the user interface 603 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 601 can be used to call the image-based interactive application stored in the memory 605.

[0143] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0144] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for testing the structural performance of tunnel lining construction, characterized in that, include: Acquire preliminary optical images before tunnel lining construction; Based on the preliminary optical images before tunnel lining construction, multiple landmark data point information are determined using a landmark data point determination model. Based on the information of the multiple landmark data points, multiple lining segmentation regions are determined, including a reference lining region and multiple remaining lining segmentation regions. Acquire laser scan images of the reference lining area; Multiple preliminary reinforcement points were determined based on the laser scanning image of the reference lining area; Based on the laser scan image of the reference lining area, the preliminary optical image before tunnel lining construction, and the multiple preliminary reinforcement points, the estimated preliminary scan reinforcement area of ​​each remaining lining segment is determined. Acquire laser scan images of the estimated preliminary scan reinforcement area for each remaining lining segment; The reinforcement points for each remaining lining segment are determined based on the laser scan image of the estimated preliminary scan reinforcement area for each remaining lining segment.

2. The structural performance testing method for tunnel lining construction as described in claim 1, characterized in that, The determination of multiple lining segmentation regions based on the multiple landmark data point information, wherein the multiple lining segmentation regions include a reference lining region and multiple remaining lining segmentation regions, including: K clusters of iconic data points are obtained by clustering based on the information of the multiple iconic data points; The reference lining area is determined based on the K clusters of landmark data points; Based on the baseline lining area and the K clusters of landmark data points, multiple remaining lining segmentation areas are determined.

3. The structural performance testing method for tunnel lining construction as described in claim 1, characterized in that, The determination of reinforcement points for each remaining lining segment based on the laser scanning image of the estimated preliminary scan reinforcement area for each remaining lining segment includes: Based on the laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment, multiple reinforcement point information of the estimated preliminary scan reinforcement area of ​​each remaining lining segment is determined; Construct a reinforcement knowledge graph, which includes multiple reinforcement points and multiple edges between the reinforcement points. The node features of each reinforcement point are reinforcement point information and a preliminary optical image of each remaining lining segmentation region. The reinforcement knowledge graph is processed using a graph neural network to determine multiple similar reinforcement regions for each remaining lining segmentation region. Acquire laser scan images of multiple similar reinforced areas in each remaining lining segmentation region; The reinforcement points of the multiple similar reinforcement areas in each remaining lining segment are determined based on the laser scanning images of the multiple similar reinforcement areas in each remaining lining segment.

4. The structural performance testing method for tunnel lining construction as described in claim 1, characterized in that, The model for determining the key data points is a convolutional neural network.

5. A structural performance testing system for tunnel lining construction, characterized in that, include: The image acquisition module is used to acquire preliminary optical images before tunnel lining construction. The marker point determination module is used to determine multiple marker data point information based on the preliminary optical image before the tunnel lining construction using a marker data point determination model. The region segmentation module is used to determine multiple lining segmentation regions based on the multiple landmark data point information. The multiple lining segmentation regions include a reference lining region and multiple remaining lining segmentation regions. The reference scanning module is used to acquire laser scanning images of the reference lining area; The preliminary reinforcement point determination module is used to determine multiple preliminary reinforcement points based on the laser scanning image of the reference lining area; The estimated area determination module is used to determine the estimated preliminary scan reinforcement area of ​​each remaining lining segment based on the laser scan image of the reference lining area, the preliminary optical image before the tunnel lining construction, and the multiple preliminary reinforcement points. The reinforcement area scanning module is used to acquire the laser scan image of the estimated preliminary scan reinforcement area for each remaining lining segment; The final reinforcement point determination module is used to determine the reinforcement point of each remaining lining segment based on the laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment.

6. The structural performance testing system for tunnel lining construction as described in claim 5, characterized in that, The region segmentation module is also used for: K clusters of iconic data points are obtained by clustering based on the information of the multiple iconic data points; The reference lining area is determined based on the K clusters of landmark data points; Based on the baseline lining area and the K clusters of landmark data points, multiple remaining lining segmentation areas are determined.

7. The structural performance testing system for tunnel lining construction as described in claim 5, characterized in that, The final reinforcement point determination module is also used for: Based on the laser scan image of the estimated preliminary scan reinforcement area of ​​each remaining lining segment, multiple reinforcement point information of the estimated preliminary scan reinforcement area of ​​each remaining lining segment is determined; Construct a reinforcement knowledge graph, which includes multiple reinforcement points and multiple edges between the reinforcement points. The node features of each reinforcement point are reinforcement point information and a preliminary optical image of each remaining lining segmentation region. The reinforcement knowledge graph is processed using a graph neural network to determine multiple similar reinforcement regions for each remaining lining segmentation region. Acquire laser scan images of multiple similar reinforced areas in each remaining lining segmentation region; The reinforcement points of the multiple similar reinforcement areas in each remaining lining segment are determined based on the laser scanning images of the multiple similar reinforcement areas in each remaining lining segment.

8. The structural performance testing system for tunnel lining construction as described in claim 5, characterized in that, The model for determining the key data points is a convolutional neural network.

9. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the structural performance testing method for tunnel lining construction as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the structural performance testing method for tunnel lining construction as described in any one of claims 1 to 4.

Images