3D box girder point cloud data topological structure identification method based on continuous coherence

Abstract

The invention discloses a 3D box girder point cloud data topological structure identification method based on continuous coherence, and the method comprises the steps: S1, obtaining a point cloud of a box girder, carrying out the preprocessing of the point cloud, and generating a three-dimensional data set; s2, performing continuous homology analysis according to the three-dimensional data set to obtain topological characteristics of the box girder; s3, topological characteristics of the box girder are analyzed, and areas with structural problems are marked; and S4, according to the marked region, carrying out physical interpretation on the topological characteristics, and generating a health assessment report. Through a continuous coherence technology, in combination with topology analysis, noise robustness, geometric mechanics interpretation and multi-scale feature capture of point cloud data, multi-scale, robust and highly-interpretable topology structure analysis of the point cloud data of the box girder is realized, and complex topology features of the box girder can be comprehensively captured under different scales. The method overcomes the defects of a traditional geometric analysis method, and especially has higher robustness and reliability when processing the problems of noise, non-uniform point cloud data and the like.

Description

Technical Field

[0001] The invention belongs to the field of data structure analysis, and in particular relates to a 3D box girder point cloud data topological structure recognition method based on continuous coherence. Background Art

[0002] In bridge engineering, box girders are key load-bearing components in bridges, buildings and infrastructure. Their structural stability and deformation are directly related to the safety of the entire project. The geometric shape of the box girder has a decisive influence on its bearing capacity and mechanical response under load. Therefore, a comprehensive understanding and accurate analysis of the structural state, deformation and potential damage of the box girder is an important prerequisite for ensuring project safety.

[0003] With the continuous development of 3D scanning, remote sensing technology and computer vision technology, 3D point cloud data has been widely used in many fields such as bridge inspection, building modeling, medical imaging, etc. 3D point cloud data can accurately capture the geometric shape of objects and provide rich spatial information, which enables engineers to conduct detailed structural analysis based on point cloud data. These analyses can not only help evaluate the structural health of box girders, but also identify their deformation, damage, wear and other problems, thereby providing a basis for subsequent quality assessment, maintenance inspection and design optimization.

[0004] At present, the existing point cloud data analysis technologies are mainly divided into analysis based on geometric features and methods based on machine learning. Methods based on geometric features, such as KD tree, quadtree, etc., usually analyze the spatial distribution of point cloud data through local features and distance metrics. However, such methods mainly rely on local geometric information, and it is difficult to capture the changes in the overall structure of the box girder under complex stress environments. Especially when multi-scale topological features are involved, the existing geometric analysis techniques often seem to be inadequate. This means that when faced with complex structural deformations or potential damage, the results of geometric feature analysis may be limited and cannot fully reflect the actual situation of the box girder under various loads.

[0005] On the other hand, machine learning methods, especially deep learning techniques such as PointNet and PointCNN, have performed well in feature extraction and classification of point cloud data in recent years. These methods can automatically extract useful features from point cloud data through an end-to-end learning framework, reducing the burden of manual feature engineering. However, machine learning methods also face some challenges. First, deep learning models usually require a large amount of high-quality training data, and in bridge engineering, it may be relatively difficult to obtain comprehensive and accurately annotated point cloud datasets. Second, although these models perform well in feature extraction, they lack the ability to directly interpret key topological features in box girder structures, and it is difficult to reveal topological relationships in point cloud data, such as connectivity, holes, etc.

[0006] In summary, existing technologies have a common limitation in the analysis of point cloud data, that is, they mainly rely on geometric metrics and it is difficult to effectively capture the topological information in the box girder point cloud data. Especially in multi-scale analysis and complex structural scenarios, both geometric analysis methods and machine learning models have difficulty distinguishing between real structural features and noise. In addition, they lack the ability to identify important topological structures in box girders, which is particularly important when evaluating the overall structural stability of box girders. Therefore, it is necessary to introduce more comprehensive analysis methods, such as topological data analysis methods based on continuous homology, to overcome the limitations of these existing technologies, especially in the detection of multi-scale topological features of complex structures. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a 3D box girder point cloud data topological structure recognition method based on continuous coherence, which solves the problems of multi-scale topology being difficult to identify, sensitive to noise and poor interpretability in the existing 3D box girder point cloud data analysis.

[0008] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a method for topological structure recognition of 3D box girder point cloud data based on continuous coherence, comprising the following steps: S1, obtaining the point cloud of the box girder, preprocessing the point cloud, and generating a three-dimensional data set; S2, perform continuous homology analysis based on the three-dimensional data set to obtain the topological characteristics of the box girder; S3, analyze the topological characteristics of the box girder and mark the areas with structural problems; S4. Physically interpret the topological features based on the marked areas and generate a health assessment report.

[0009] Further: In S1, the method for preprocessing point cloud data includes the following steps: S11, calculating the neighborhood density of each point in the point cloud, and removing abnormal points in the point cloud whose neighborhood density is lower than the neighborhood density threshold; S12, registering the point cloud after removing abnormal points; S13. The method of coarsening the data granularity is used to reduce the volume of the registered point cloud and generate a three-dimensional data set.

[0010] The beneficial effect of the above further solution is: in order to improve the accuracy and efficiency of subsequent continuous coherence analysis, denoising, registration and simplified data processing are performed in sequence to ensure its integrity and accuracy. The present invention ensures that the simplification process does not cause the loss of key geometric information by measuring the distance error between the simplified point cloud and the original point cloud.

[0011] Further: In said S11, i Neighborhood density of a point The specific expression is:

[0012] In the formula, For the i Point and j The Euclidean distance between points, is the selected cutoff distance.

[0013] Further: S13 specifically includes: S131, taking the point cloud as a point set to be divided; S132, obtaining two points in the point set with the farthest interval, establishing a first subset and a second subset based on the obtained points, and allocating the remaining points in the point set to the nearest subset; S133, calculate the compactness of the point set, the first subset and the second subset, the compactness The specific expression is:

[0014] In the formula, m Point Set D The number of midpoints, p j Point Set D The point in C Point Set D The centroid of Determine whether the splitting conditions are met. The splitting conditions are as follows: and

[0015] In the formula, is the compactness of the first subset, is the compactness of the second subset, is the compactness of the point set, Point Set D The total number of points, n is the total number of points in the point cloud; If yes, both the first subset and the second subset are used as the point set to be divided, and the process returns to S132; if no, the current point set is used as the simplified point cloud, and the process goes to S134; S134. Establish a three-dimensional data set based on all simplified point clouds.

[0016] The beneficial effect of the above further scheme is: in order to improve the computational efficiency of subsequent continuous coherence analysis, the present invention introduces a method of coarsening data granularity to reduce the volume of point cloud data, and divides the overall data set into multiple point sets. The centroid of each point set represents these point sets. In this process, key geometric information is retained and the integrity and accuracy of the data are ensured.

[0017] Further: S2 comprises the following sub-steps: S21, based on each point in the point cloud of the three-dimensional data set, calculating the Euclidean distance between each point and generating a distance matrix; S22. Set the neighborhood radius, create a persistent bar chart to analyze the persistence of the topological features generated by each point cloud, and obtain the topological features of the box girder.

[0018] Further: S22 is specifically: Set the neighborhood radius, gradually increase the neighborhood radius under the framework of continuous homology, calculate the topological features under neighborhood radius of different scales, create a persistent bar chart based on the life cycle of the topological features, and retain the topological features whose life cycle exceeds 10% of the overall time length as the topological features of the box girder; Among them, the neighborhood radius is used to determine whether points are considered connected. In response to the Euclidean distance between two points being less than the neighborhood radius, the two points are connected in the topological space; The life cycle is specifically the time when a topological feature appears and disappears.

[0019] The beneficial effect of the above further scheme is: to ensure the accuracy of the analysis results, the present invention filters out noise by setting a threshold. Generally, only those topological features whose life cycle exceeds 10% of the overall time length are retained, which can effectively remove short-lived features caused by noise or local microstructures. The retained persistent topological features need to be physically interpreted in combination with the specific structure of the box girder in order to better guide the health status assessment and structural analysis of the box girder.

[0020] Further: the dimensions of the topological features include 0-dimension, 1-dimension and 2-dimension; Among them, 0-dimensional topological features represent connected branches in the point cloud, 1-dimensional topological features represent closed loops or holes, and 2-dimensional topological features represent cavities or closed three-dimensional spaces.

[0021] Further: In S3, the method for analyzing the topological characteristics of the box girder is specifically: (1) Perform stress analysis on the box girder through finite element analysis, calculate the stress distribution of different areas under the action of external forces, and identify the degree of overlap between the topological features and the stress concentration area by comparing the stress distribution with the topological features; (2) For the geometric structure area where the topological features are located, fatigue life assessment is carried out in combination with the mechanical model. The stress state and load history of the geometric structure area are analyzed to evaluate the possibility of fatigue failure under long-term cyclic loading.

[0022] Further: in said S4, the method of physically interpreting the topological features includes crack and damage detection and cavity and material defect identification; Among them, the methods for crack and damage detection are specifically: Detecting connectivity interruptions or new connectivity branches, which indicate that there is a break or damage in a local area of ​​the box girder, and analyzing the impact of the connectivity branch on the overall performance of the box girder through geometric structure and mechanics to guide maintenance work; The specific method of material defect identification is: The hollow structure in the 2D topological features is extracted, and the specific position and shape of the cavity structure are analyzed through continuous coherence analysis. Combined with mechanical analysis, the potential threat of the cavity structure to structural safety is determined.

[0023] Further: In S4, the health assessment report includes the overall connectivity of the box girder, local damage and areas that require key monitoring or maintenance.

[0024] The beneficial effects of the present invention are as follows: the present invention provides a 3D box girder point cloud data topological structure recognition method based on continuous coherence, aiming to achieve multi-scale, robust and highly interpretable topological structure analysis of box girder point cloud data. Through this technology, the present invention can fully capture the complex topological features of the box girder at different scales, overcome the shortcomings of traditional geometric analysis methods, especially when dealing with problems such as noise and uneven point cloud data, it has higher robustness and reliability, and compared with the existing technology, has the following advantages: (1) Capturing multi-scale topological features: The present invention uses continuous coherence technology to analyze 3D point cloud data at different scales and capture topological features such as connected branches, holes, and cavities in the box girder structure. This method can reveal the geometric and topological information of the structure from multiple scales, both globally and locally, and provides a more comprehensive topological structure analysis.

[0025] (2) High noise robustness: Through the persistence analysis of topological features, the present invention can effectively distinguish between noise features and persistent features that reflect the real structure.

[0026] (3) Enhanced interpretability: The present invention generates a persistent bar chart to intuitively display the topological features in the box girder point cloud data and evaluates the importance of the topological features by their life cycle length. Combined with the actual geometric morphology and mechanical analysis, engineers can clearly determine the physical meaning of the topological features, such as holes and cracks.

[0027] (4) Strong adaptability to unlabeled data: Continuous coherence analysis does not rely on large-scale labeled data, but is purely based on topological analysis to mine the intrinsic structure of the data. Therefore, the present invention can adapt to the scene of box girder 3D point cloud data that is unlabeled or has limited annotations.

[0028] (5) Combined analysis of topology and mechanical properties: The present invention can not only identify the topological features in the box girder point cloud data, but also combine the geometric structure and mechanical properties to conduct in-depth physical interpretation and structural health assessment of these topological features.

[0029] (6) Real-time monitoring and maintenance support: The systematic analysis process of the present invention provides structural health assessment and maintenance support. Through long-term monitoring, the system can dynamically track the changes in the box girder structure, detect potential damage or deformation in a timely manner, and generate a health assessment report to provide powerful decision-making support for engineers.

[0030] In summary, the present invention solves the problems of multi-scale topology being difficult to identify, sensitive to noise, and poor interpretability in the prior art by combining continuous coherence technology with topological analysis, noise robustness, geometric mechanics interpretation, and multi-scale feature capture of point cloud data, and provides a more robust, accurate, and interpretable box girder structure analysis method. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The figure is a flow chart of a method for topological structure recognition of 3D box girder point cloud data based on continuous homology. DETAILED DESCRIPTION

[0032] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0033] like Figure 1 As shown, in one embodiment of the present invention, a method for topological structure recognition of 3D box girder point cloud data based on continuous coherence includes the following steps: S1, obtaining the point cloud of the box girder, preprocessing the point cloud, and generating a three-dimensional data set; S2, perform continuous homology analysis based on the three-dimensional data set to obtain the topological characteristics of the box girder; S3, analyze the topological characteristics of the box girder and mark the areas with structural problems; S4. Physically interpret the topological features based on the marked areas and generate a health assessment report.

[0034] In this embodiment, the present invention uses a suitable laser radar device to scan the box girder structure, obtains its high-precision 3D point cloud data, and pre-processes these data. The pre-processing steps include denoising, data alignment and simplification to ensure the quality and accuracy of the analyzed data. On this basis, the processed point cloud data is subjected to continuous homology analysis to generate a persistent bar chart to reveal key topological features in the point cloud, such as connected branches, holes, cavities, etc. By analyzing the duration of these topological features, the importance of each feature can be effectively evaluated, and the key elements in the global structure can be distinguished from the transient features that may be caused by noise. Next, the identified topological features are combined with the geometric structure and mechanical properties of the box girder for in-depth physical interpretation. This process can help engineers evaluate the health of the box girder and detect potential structural problems, such as cracks, cavities or other damage. Through this systematic topological analysis method, a scientific basis can be provided for the maintenance, repair and monitoring of the box girder, and long-term tracking and risk management of the structural status can be supported.

[0035] In S1, the method for preprocessing point cloud data includes the following steps: S11, calculating the neighborhood density of each point in the point cloud, and removing abnormal points in the point cloud whose neighborhood density is lower than the neighborhood density threshold; S12, registering the point cloud after removing abnormal points; S13. The method of coarsening the data granularity is used to reduce the volume of the registered point cloud and generate a three-dimensional data set.

[0036] In this embodiment, after obtaining the 3D point cloud data of the box girder, in order to improve the accuracy and efficiency of the subsequent continuous coherence analysis, denoising, registration and simplified data processing are performed in sequence to ensure its integrity and accuracy. The present invention ensures that the simplification process does not cause the loss of key geometric information by measuring the distance error between the simplified point cloud and the original point cloud.

[0037] In the S11, i Neighborhood density of a point The specific expression is:

[0038] In the formula, For the i Point and j The Euclidean distance between points, The selected cutoff distance should ensure that the average number of samples contained in the neighborhood of each data point accounts for 2% of the total number of samples.

[0039] In this embodiment, the present invention sets a neighborhood density threshold. If the number of other points in the local neighborhood of a point is too small, it is generally considered that it may be a noise point or a mismeasured point. The present invention will determine the threshold with the goal of eliminating 10% of low-density points. Through this density filtering method, noise points caused by surface reflection, occlusion or sensor error during the scanning process can be effectively removed, thereby improving the overall quality of the point cloud data.

[0040] In S12, the present invention uses the built-in GPS and IMU data of the laser radar to perform preliminary registration of the point cloud data. The purpose of registration is to align the point cloud data collected from multiple angles into a unified spatial coordinate system to construct a complete box girder model. In this step, by matching the position information of each scan segment, a rough alignment can be achieved, laying the foundation for subsequent fine registration and analysis.

[0041] The S13 is specifically: S131, taking the point cloud as a point set to be divided; S132, obtaining two points in the point set with the farthest interval, establishing a first subset and a second subset based on the obtained points, and allocating the remaining points in the point set to the nearest subset; S133, calculate the compactness of the point set, the first subset and the second subset, the compactness The specific expression is:

[0042] In the formula, m Point Set D The number of midpoints, p j Point Set D The point in C Point Set D The centroid of Determine whether the splitting conditions are met. The splitting conditions are as follows: and

[0043] In the formula, is the compactness of the first subset, is the compactness of the second subset, is the compactness of the point set, Point Set D The total number of points, n is the total number of points in the point cloud; If yes, both the first subset and the second subset are used as the point set to be divided, and the process returns to S132; if no, the current point set is used as the simplified point cloud, and the process goes to S134; S134. Establish a three-dimensional data set based on all simplified point clouds.

[0044] In this embodiment, since the point cloud data generated by high-precision scanning is large in volume and complex to process, in order to improve the computational efficiency of subsequent continuous coherence analysis, the present invention introduces a method of coarsening data granularity to reduce the volume of point cloud data, and divides the overall data set into multiple point sets. The centroid of each point set represents these point sets. In this process, key geometric information is retained, and the integrity and accuracy of the data are ensured.

[0045] The S2 comprises the following sub-steps: S21, based on each point in the point cloud of the three-dimensional data set, calculating the Euclidean distance between each point and generating a distance matrix; S22. Set the neighborhood radius, create a persistent bar chart to analyze the persistence of the topological features generated by each point cloud, and obtain the topological features of the box girder.

[0046] In this embodiment, the distance matrix will provide a basis for subsequent topological calculations. On this basis, the present invention sets a neighborhood radius , which determines whether points are considered connected. If the distance between two points is less than this radius, the two points are considered connected in the topological space, thus forming an edge between the two points. As the neighborhood radius gradually increases, a complex body composed of points and edges begins to be constructed. In this process, the change in the neighborhood radius corresponds to the change in connectivity and structure in the topological space. Different topological structures will be generated under different radii, including connected branches, ring structures, and closed cavities.

[0047] The S22 is specifically: Set the neighborhood radius, gradually increase the neighborhood radius under the framework of continuous homology, calculate the topological features under neighborhood radius of different scales, create a persistent bar chart based on the life cycle of the topological features, and retain the topological features whose life cycle exceeds 10% of the overall time length as the topological features of the box girder; Among them, the neighborhood radius is used to determine whether points are considered connected. In response to the Euclidean distance between two points being less than the neighborhood radius, the two points are connected in the topological space; The life cycle is specifically the time when a topological feature appears and disappears.

[0048] In this embodiment, as the neighborhood radius increases, the topological features will gradually appear and disappear. This means that the life cycle of different features can be recorded to determine the importance of these features. Topological structures with longer persistence times usually represent more global or stable geometric structures in the data, while short-lived features may be noise or local microstructures. Based on this, a persistent bar graph can be created according to the time when the topological features appear and disappear, which represents the life cycle of the topological features with the length of the bar. The horizontal axis is the neighborhood radius, and each bar corresponds to the life cycle of a topological feature-the starting point of the bar represents the moment when the feature is "born", and the end point represents the moment of "death". The longer the bar, the more stable the topological feature remains over a larger scale range, which means that the feature is more important. For example, a persistent connected branch or hole may correspond to some key structure in the box girder, while a bar with a shorter duration may be just local noise or unimportant details.

[0049] After generating the persistence bar graph, it is necessary to evaluate the importance of the topological features by analyzing the persistence of each topological feature (i.e., the length of its life cycle). To ensure the accuracy of the analysis results, the present invention filters out noise by setting a threshold. Usually, only those topological features whose life cycle exceeds 10% of the overall time length are retained, which can effectively remove short-lived features caused by noise or local microstructures. The retained persistent topological features need to be physically interpreted in combination with the specific structure of the box girder in order to better guide the health status assessment and structural analysis of the box girder.

[0050] The dimensions of the topological features include 0 dimension, 1 dimension and 2 dimension; Among them, 0-dimensional topological features represent connected branches in the point cloud, 1-dimensional topological features represent closed loops or holes, and 2-dimensional topological features represent cavities or closed three-dimensional spaces.

[0051] In this embodiment, the dimensions of topological features can be used to classify and identify these features: 0D topological features mainly represent the connectivity of point cloud data. For box girder structures, 0D topological features can reflect the overall connectivity of the box girder. If there are breaks or discontinuities in certain areas, the changes in 0D features can keenly capture these break points or structural defects.

[0052] 1D topological features represent ring structures or holes in point cloud data. In box girder applications, 1D features can reflect holes or cracks that may exist on the surface or inside the beam. These structures may be formed due to manufacturing defects or long-term stress. By analyzing the persistence time of 1D features, the scope of these holes or cracks and their importance in the box girder structure can be determined.

[0053] 2D topological features describe closed 3D spaces or cavities. For example, there may be areas inside a box girder that are not fully filled, or the structure itself forms some kind of closed space. By analyzing the persistence of these cavities, it can be determined whether they are critical features in the box girder structure, and thus an effective assessment of the health of the box girder can be made.

[0054] In S3, the method for analyzing the topological characteristics of the box girder is specifically as follows: (1) Perform stress analysis on the box girder through finite element analysis, calculate the stress distribution of different areas under the action of external forces, and identify the degree of overlap between the topological features and the stress concentration area by comparing the stress distribution with the topological features; (2) For the geometric structure area where the topological features are located, fatigue life assessment is carried out in combination with the mechanical model. The stress state and load history of the geometric structure area are analyzed to evaluate the possibility of fatigue failure under long-term cyclic loading.

[0055] In this embodiment, stress analysis is performed on the box girder through finite element analysis to calculate the stress distribution of different areas under external forces. By comparing the results of the stress analysis with the topological features, the degree of overlap between the topological features and the stress concentration areas is identified. For example, around the holes revealed by the 1D topological features, if the stress concentration phenomenon is severe, this may indicate that the material in this area is at risk of fatigue or crack propagation. This type of information is of great value for structural health monitoring and predictive maintenance.

[0056] In addition, for the geometric structure area where the topological features are located, the present invention combines the mechanical model to perform fatigue life assessment. By analyzing the stress state and load history of these areas, the possibility of fatigue damage under long-term cyclic load is assessed. For example, the cavity structure identified by the 2D topological features may lead to local stress concentration under long-term dynamic load. This area of ​​concentrated stress is often the starting point for fatigue crack initiation. By analyzing these areas, an important basis can be provided for the fatigue life prediction of the box girder.

[0057] In said S4, the method for physically interpreting the topological features includes crack and breakage detection and cavity and material defect identification; Among them, the methods for crack and damage detection are specifically: Detecting connectivity interruptions or new connectivity branches, which indicate that there is a break or damage in a local area of ​​the box girder, and analyzing the impact of the connectivity branch on the overall performance of the box girder through geometric structure and mechanics to guide maintenance work; The specific method of material defect identification is: The hollow structure in the 2D topological features is extracted, and the specific position and shape of the cavity structure are analyzed through continuous coherence analysis. Combined with mechanical analysis, the potential threat of the cavity structure to structural safety is determined.

[0058] In this embodiment, the present invention physically interprets the topological features and provides a basis for the health assessment of the box girder, for example: Crack and damage detection: The interruption of connectivity or the appearance of new connected branches may indicate that there is a fracture or damage in a local area of ​​the box girder. In most cases, these fractures initially appear as small cracks, which may gradually expand over time and with the accumulation of stress. Through the continuous coherence method, cracks can be detected in time before they expand to the extent that they threaten the safety of the structure, and combined with geometric structure and mechanical analysis, their impact on the overall performance of the box girder can be judged, thereby guiding the maintenance work.

[0059] Identification of cavities and material defects: Cavity structures extracted by 2D topological features may be caused by material defects, uneven pouring or omissions during construction in actual box girders. Such problems may lead to local failures during the long-term service of the box girder, especially when stress concentration occurs around these areas. Through continuous coherence analysis, engineers can determine the specific location and shape of these cavities, and combine mechanical analysis to determine their potential threat to structural safety.

[0060] In S4, the health assessment report includes the overall connectivity of the box girder, local damage, and areas that require key monitoring or maintenance.

[0061] The beneficial effects of the present invention are as follows: the present invention provides a 3D box girder point cloud data topological structure recognition method based on continuous coherence, aiming to achieve multi-scale, robust and highly interpretable topological structure analysis of box girder point cloud data. Through this technology, the present invention can fully capture the complex topological features of the box girder at different scales, overcome the shortcomings of traditional geometric analysis methods, especially when dealing with problems such as noise and uneven point cloud data, it has higher robustness and reliability, and compared with the existing technology, has the following advantages: (1) Capturing multi-scale topological features: The present invention uses continuous coherence technology to analyze 3D point cloud data at different scales and capture topological features such as connected branches, holes, and cavities in the box girder structure. This method can reveal the geometric and topological information of the structure from multiple scales, both globally and locally, and provides a more comprehensive topological structure analysis.

[0062] (2) High noise robustness: Through the persistence analysis of topological features, the present invention can effectively distinguish between noise features and persistent features that reflect the real structure.

[0063] (3) Enhanced interpretability: The present invention generates a persistent bar chart to intuitively display the topological features in the box girder point cloud data and evaluates the importance of the topological features by their life cycle length. Combined with the actual geometric morphology and mechanical analysis, engineers can clearly determine the physical meaning of the topological features, such as holes and cracks.

[0064] (4) Strong adaptability to unlabeled data: Continuous coherence analysis does not rely on large-scale labeled data, but is purely based on topological analysis to mine the intrinsic structure of the data. Therefore, the present invention can adapt to the scene of box girder 3D point cloud data that is unlabeled or has limited annotations.

[0065] (5) Combined analysis of topology and mechanical properties: The present invention can not only identify the topological features in the box girder point cloud data, but also combine the geometric structure and mechanical properties to conduct in-depth physical interpretation and structural health assessment of these topological features.

[0066] (6) Real-time monitoring and maintenance support: The systematic analysis process of the present invention provides structural health assessment and maintenance support. Through long-term monitoring, the system can dynamically track the changes in the box girder structure, detect potential damage or deformation in a timely manner, and generate a health assessment report to provide powerful decision-making support for engineers.

[0067] In summary, the present invention solves the problems of multi-scale topology being difficult to identify, sensitive to noise, and poor interpretability in the prior art by combining continuous coherence technology with topological analysis, noise robustness, geometric mechanics interpretation, and multi-scale feature capture of point cloud data, and provides a more robust, accurate, and interpretable box girder structure analysis method.

[0068] In the description of the present invention, it is necessary to understand that the orientation or positional relationship indicated by the terms "center", "thickness", "upper", "lower", "horizontal", "top", "bottom", "inner", "outer", "radial", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", and "third" are used only for descriptive purposes, and cannot be understood as indicating or implying the relative importance or the number of implicitly specified technical features. Therefore, the features defined by "first", "second", and "third" may explicitly or implicitly include one or more of the features.

Claims

1. A method for topological structure recognition of 3D box girder point cloud data based on continuous homology, characterized in that: The following steps are involved: S1, obtaining the point cloud of the box girder, preprocessing the point cloud, and generating a three-dimensional data set; S2, perform continuous homology analysis based on the three-dimensional data set to obtain the topological characteristics of the box girder; S3, analyze the topological characteristics of the box girder and mark the areas with structural problems; S4. Physically interpret the topological features based on the marked areas and generate a health assessment report.

2. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 1, characterized in that: In S1, the method for preprocessing point cloud data includes the following steps: S11, calculating the neighborhood density of each point in the point cloud, and removing abnormal points in the point cloud whose neighborhood density is lower than the neighborhood density threshold; S12, registering the point cloud after removing abnormal points; S13. The method of coarsening the data granularity is used to reduce the volume of the registered point cloud and generate a three-dimensional data set.

3. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 2, characterized in that: In the S11, i Neighborhood density of a point The specific expression is: In the formula, For the i Point and j The Euclidean distance between points, is the selected cutoff distance.

4. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 2, characterized in that: The S13 is specifically: S131, taking the point cloud as a point set to be divided; S132, obtaining two points in the point set with the farthest interval, establishing a first subset and a second subset based on the obtained points, and allocating the remaining points in the point set to the nearest subset; S133, calculate the compactness of the point set, the first subset and the second subset, the compactness The specific expression is: In the formula, m Point Set D The number of midpoints, p j Point Set D The point in C Point Set D The centroid of Determine whether the splitting conditions are met. The splitting conditions are as follows: and In the formula, is the compactness of the first subset, is the compactness of the second subset, is the compactness of the point set, Point Set D The total number of points, n is the total number of points in the point cloud; If yes, both the first subset and the second subset are used as the point set to be divided, and the process returns to S132; if no, the current point set is used as the simplified point cloud, and the process goes to S134; S134. Establish a three-dimensional data set based on all simplified point clouds.

5. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 1, characterized in that: The S2 comprises the following sub-steps: S21, based on each point in the point cloud of the three-dimensional data set, calculating the Euclidean distance between each point and generating a distance matrix; S22. Set the neighborhood radius, create a persistent bar chart to analyze the persistence of the topological features generated by each point cloud, and obtain the topological features of the box girder.

6. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 5, characterized in that: The S22 is specifically: Set the neighborhood radius, gradually increase the neighborhood radius under the framework of continuous homology, calculate the topological features under neighborhood radius of different scales, create a persistent bar chart based on the life cycle of the topological features, and retain the topological features whose life cycle exceeds 10% of the overall time length as the topological features of the box girder; Among them, the neighborhood radius is used to determine whether points are considered connected. In response to the Euclidean distance between two points being less than the neighborhood radius, the two points are connected in the topological space; The life cycle is specifically the time when a topological feature appears and disappears.

7. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 6, characterized in that: The dimensions of the topological features include 0 dimension, 1 dimension and 2 dimension; Among them, 0-dimensional topological features represent connected branches in the point cloud, 1-dimensional topological features represent closed loops or holes, and 2-dimensional topological features represent cavities or closed three-dimensional spaces.

8. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 1, characterized in that: In S3, the method for analyzing the topological characteristics of the box girder is specifically as follows: (1) Perform stress analysis on the box girder through finite element analysis, calculate the stress distribution of different areas under the action of external forces, and identify the degree of overlap between the topological features and the stress concentration area by comparing the stress distribution with the topological features; (2) For the geometric structure area where the topological features are located, fatigue life assessment is carried out in combination with the mechanical model. The stress state and load history of the geometric structure area are analyzed to evaluate the possibility of fatigue failure under long-term cyclic loading.

9. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 1, characterized in that: In said S4, the method for physically interpreting the topological features includes crack and breakage detection and cavity and material defect identification; Among them, the methods for crack and damage detection are specifically: Detecting connectivity interruptions or new connectivity branches, which indicate that there is a break or damage in a local area of ​​the box girder, and analyzing the impact of the connectivity branch on the overall performance of the box girder through geometric structure and mechanics to guide maintenance work; The specific method of material defect identification is: The hollow structure in the 2D topological features is extracted, and the specific position and shape of the cavity structure are analyzed through continuous coherence analysis. Combined with mechanical analysis, the potential threat of the cavity structure to structural safety is determined.

10. The method for topological structure recognition of 3D box girder point cloud data based on continuous coherence according to claim 1, characterized in that: In S4, the health assessment report includes the overall connectivity of the box girder, local damage, and areas that require key monitoring or maintenance.

Images