Complex scene three-dimensional object matching method and device based on first-order distance-topological compatibility index

Abstract

This invention discloses a method and device for matching three-dimensional objects in complex scenes based on a first-order distance-topological compatibility index. The method first designs a first-order distance-topological compatibility index for three-dimensional objects, aiming to express the positioning characteristics of three-dimensional objects, thereby alleviating the geometric matching problem caused by random position changes, which is difficult to handle with traditional first-order or second-order spatial compatibility indices. Secondly, a three-dimensional object similarity calculation technique based on the first-order distance-topological compatibility index is proposed to extract exact matching pairs and potential matching pairs. Finally, a multi-line-of-sight constraint equation system is constructed using the relative positional relationship between exact matching pairs and potential matching pairs to achieve iterative self-correction of the spatial position of BIM models in potential matching pairs. This invention achieves high-precision matching between real-world point cloud objects and BIM three-dimensional models in complex scenes, providing a basis for automatic correction of BIM models.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional object matching, and in particular to a method and device for matching three-dimensional objects in complex scenes based on a first-order distance-topological compatibility index. Background Art

[0002] As the digital economy enters its full-scale expansion phase, traditional substations are gradually evolving towards digitalization and intelligentization. Against this backdrop, the construction of a substation digital twin system can visualize complex substation power equipment using three-dimensional digital models, reflecting the operating status of power equipment in real time. This creates a virtual scene that seamlessly integrates the real-world substation scene, enabling comprehensive, full-process intelligent perception, analysis, prediction, and control of the substation, thereby empowering digital substation construction. However, discrepancies exist between the actual cloud model of the substation during construction and operation and the three-dimensional BIM equipment models from the design phase. This leads to inconsistencies between the digital twin scene and the real-world power equipment in terms of quantity, location, and distribution. This makes it impossible to directly map the BIM model to the corresponding real-world power equipment, hindering the practical application of BIM models in substation digital twins. Currently, BIM models in digital twin systems rely primarily on manual matching and correction, which is not only time-consuming and labor-intensive but also prone to human error. Therefore, it is urgent to carry out research on automatic matching methods between three-dimensional objects in different scenes, realize the one-to-one mapping and matching of BIM equipment three-dimensional models to real-world point cloud objects, and provide a basis for the automatic correction of BIM equipment three-dimensional models, which has important research significance and application value.

[0003] Currently, existing techniques for automatically matching 3D objects across different scenes mostly focus on matching based on geometric consistency. These methods use geometrically consistent overlapping regions between the scenes to be matched as references and capture key points to achieve efficient and accurate 3D scene alignment. For example, some methods assume geometric consistency between BIM scenes and real-world scenes, thereby solving the problems of matching real-world point clouds to BIM space and relocating moving measurement targets. However, assumptions based on geometric consistency transformations such as rotation, scaling, and translation are generally not applicable to object matching and spatial alignment between two different data sources. Recent research has attempted to incorporate the dynamic changes of the 3D objects to be matched into the object matching task, focusing primarily on the movement, matching, and relocation of objects within 3D scenes. Furthermore, some publicly available datasets use RGBD cameras or laser point clouds to generate 3D matching tasks that present dynamic changes. While these datasets and methods consider geometric inconsistencies between different scene data, they primarily employ matching criteria based on the geometric similarity of 3D objects. However, in complex substations, there are numerous electrical equipment with similar shapes but different types. This requires fully leveraging information about equipment types and their spatial neighbors to improve matching accuracy, making it difficult to directly apply these techniques to achieve high-precision matching.

[0004] Traditional graph matching aims to establish a one-to-one correspondence between nodes in two graphs. By constructing a graph network between 3D objects, graph-based matching techniques can be used to achieve 3D object matching. However, in complex substation scenarios, matching BIM equipment models with real-world point cloud objects is a partial graph matching problem, and traditional graph matching methods struggle to achieve high matching accuracy. Recent research has attempted to incorporate deep learning methods into partial graph matching tasks and proposed deep graph matching techniques for 3D object matching in changing scenarios. For example, some classic graph network node encoding methods can encode topological features of graph nodes through methods such as random walks. Deep learning-based methods such as SuperGlue and LightGlue significantly enhance the topological localization features of graph nodes through internal self-attention and cross-attention, achieving high-precision matching between BIM equipment models and real-world point cloud objects. Despite this, deep graph-based matching techniques still face a major challenge: training deep graph networks requires a large number of sample matching pairs. This, in part, requires manual annotation of complex scenes, significantly increasing training costs. Therefore, in order to reduce mismatches between device objects of different categories and lower the cost of 3D object matching tasks, it is necessary to utilize device objects with semantic labels, expand the spatial compatibility index of device graph network matching, and further improve the matching accuracy of 3D objects. Summary of the Invention

[0005] In response to the above defects or improvement needs of the existing technology, the present invention provides a three-dimensional object matching method based on first-order distance-topological similarity and adaptive correction of spatial position, which achieves high-precision matching between real-world point cloud objects and BIM three-dimensional models in complex scenes, and assists in the automatic correction of the three-dimensional model of substation digital twin BIM equipment. The main contributions and innovations include: (1) a first-order distance-topological compatibility index for quantitative expression of three-dimensional object positioning features is designed, which can effectively alleviate the geometric matching problem caused by random position changes that traditional first-order or second-order spatial compatibility indexes are difficult to handle; (2) a three-dimensional object similarity calculation technology based on the first-order distance-topological compatibility index is proposed, which aims to identify and extract accurate matching pairs and potential matching pairs, providing a basis for the correction of the spatial position of three-dimensional objects; (3) the idea of ​​photogrammetry forward reciprocation is innovatively used to realize the automatic update of the spatial position of three-dimensional objects in potential matching pairs. The method proposed in the present invention is an iterative automatic matching and position update algorithm for real-world point cloud objects and BIM three-dimensional models, that is, the spatial position of the BIM model is continuously updated through iterative matching until the number of accurate matching pairs remains relatively stable. The method of the present invention was experimented on a three-dimensional dataset of a complex substation. By comparing it with the existing benchmark method, the experimental results showed the effectiveness and reliability of the method of the present invention.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A first aspect provides a method for matching three-dimensional objects in complex scenes based on a first-order distance-topological compatibility index, comprising:

[0008] According to the relative distance and spatial topological positioning characteristics between three-dimensional objects, a first-order distance-topological compatibility index of three-dimensional objects is constructed;

[0009] According to the spatial layout of complex 3D objects, the first-order distance-topological compatibility index between the real-world point cloud objects and the BIM objects is calculated respectively;

[0010] Based on the first-order distance-topological compatibility index of the real-world point cloud objects and the BIM objects, the first-order distance similarity and spatial topological similarity between the BIM objects and the real-world point cloud objects are calculated;

[0011] Generate initial matching pairs by maximizing the first-order distance similarity;

[0012] The exact matching pairs are established by maximizing the first-order topological neighbor similarity;

[0013] Potential matching pairs are searched based on the first-order neighbor objects of the exact matching pairs, and the spatial positions of the BIM objects are adaptively updated.

[0014] In one embodiment, a first-order distance-topological compatibility index of three-dimensional objects is constructed based on the relative distance and spatial topological positioning features between the three-dimensional objects, including:

[0015] Calculate the geometric distance between any two three-dimensional objects and obtain the distance matrix;

[0016] Traverse each row of the distance matrix and express the relationship between the corresponding 3D object and the other 3D objects in the scene using the key-value pair of "type-distance", thus completing the generation of the first-order geometric distance positioning feature vector of the 3D object in the global space;

[0017] The topological relationship of 3D objects in the scene is established based on the 3D object Voronoi diagram. The type labels of the neighboring objects of the 3D objects are counted to generate a type label histogram. The first-order topological feature vector is obtained as the first-order topological positioning feature expression.

[0018] The first-order geometric distance positioning feature vector and the first-order topological feature vector are used as the first-order distance-topological compatibility index of three-dimensional objects.

[0019] In one embodiment, the first-order geometric distance positioning feature vector is calculated as follows:

[0020]

[0021] Among them, i and j represent two three-dimensional objects, D i→j Indicates the geometric distance between 3D objects i and j in the scene, C M-1 Indicates the device type of the M-1th 3D object, where M represents the number of 3D objects in the scene. Represents the first-order distance feature vector of three-dimensional object i;

[0022] The first-order topological eigenvector is calculated as:

[0023]

[0024] Among them, U i represents the first-order topological neighbor set of three-dimensional object i, q represents the first-order topological neighbor of three-dimensional object i, C q represents the type of the first-order topological neighbors of a three-dimensional object i, Indicates that the neighbor type of the three-dimensional object i is C j the number of represents the first-order topological eigenvector of three-dimensional object i, and k represents a three-dimensional object.

[0025] In one embodiment, the first-order distance similarity between the BIM object and the real-world point cloud object is calculated based on the first-order distance-topology compatibility index between the real-world point cloud object and the BIM object, including:

[0026] For the real-world point cloud object s, use its equipment type as the filtering tag, obtain BIM objects with the same equipment type, and add them to the matching object set;

[0027] For the current real-world point cloud object s and BIM object t, the minimum distance difference is calculated dimension by dimension. Specifically, each key-value pair in the first-order distance feature vector of the real-world point cloud object s is selected in turn, and the set of key-value pairs with the same device type in the first-order distance feature vector of the BIM object t is filtered based on the device type of the key-value pair. The difference between the distance in the current key-value pair and each distance in the filtered key-value pair set is calculated, and the minimum value of the distance difference is used as the distance between the real-world point cloud object s and the BIM object t in the current dimension.

[0028] Sum the minimum distance differences between the real-world point cloud object s and the BIM object t in all dimensions to obtain the similarity between them in the first-order distance feature, which is used as the first-order distance similarity between the BIM object and the real-world point cloud object;

[0029] For the real-world point cloud object s and the BIM object t, the spatial topological similarity between the BIM object and the real-world point cloud object is obtained by calculating the quantitative differences in the corresponding dimensions of the first-order topological feature vectors of the two objects and summing up the differences in all dimensions.

[0030] In one embodiment, the calculation formula for the first-order distance similarity between the BIM object and the field point cloud object is:

[0031]

[0032] in, It represents the distance difference between the mth dimension of the real-world point cloud object s and the nth dimension of the BIM object t after filtering the same type of key-value pairs, D s→m Indicates the distance value on the mth dimension in the real-time point cloud object s, D t→n Represents the distance value in the nth dimension of BIM object t, represents the minimum distance difference between the field point cloud object s and the BIM object t in the mth dimension, Δd s→t It represents the first-order distance similarity between the field point cloud object s and the BIM object t, indicating the similarity between the two objects in the first-order geometric distance feature;

[0033] The calculation formula for the spatial topological similarity between BIM objects and field point cloud objects is:

[0034]

[0035] and Specifically, the topological neighbors of the real-world point cloud object s and the BIM object t belong to type C q The number of neighbors, Indicates that in type C q The difference in the number of topological neighbors between the two objects, Δn s→t Represents the spatial topological similarity between the field point cloud object s and the BIM object t.

[0036] In one embodiment, establishing an exact matching pair by maximizing the first-order topological neighbor similarity includes:

[0037] By maximizing the first-order topological neighbor similarity, the pairs with the largest first-order topological neighbor similarity and the smallest first-order geometric distance are selected from the initial matching pairs as the exact matching pairs.

[0038] In one embodiment, searching for potential matching pairs based on first-order neighbor objects of exact matching pairs and adaptively updating the spatial positions of BIM objects includes:

[0039] Select potential matching pairs from the first-order neighbors of the exact matching pairs, requiring that the neighbors of the potential matching pairs contain exact matching objects and are exact matching pairs of each other;

[0040] Taking the distance relationship between all accurate matching pairs and potential matching pairs as the constraint condition, a multi-view constraint equation group is established to achieve adaptive correction of the spatial position of BIM objects in potential matching pairs.

[0041] Based on the same inventive concept, the second aspect of the present invention provides a complex scene three-dimensional object matching device based on a first-order distance-topological compatibility index, comprising:

[0042] An index construction module is used to construct a first-order distance-topological compatibility index of three-dimensional objects based on the relative distance and spatial topological positioning features between three-dimensional objects;

[0043] An index calculation module is used to calculate the first-order distance-topological compatibility index of the real-world point cloud object and the BIM object based on the spatial layout of the complex three-dimensional object;

[0044] A similarity calculation module is used to calculate the first-order distance similarity and spatial topological similarity between the BIM object and the real-world point cloud object based on the first-order distance-topological compatibility index of the real-world point cloud object and the BIM object;

[0045] An initial matching pair generation module is used to generate initial matching pairs by maximizing the first-order distance similarity;

[0046] The precise matching pair establishment module is used to establish precise matching pairs by maximizing the first-order topological neighbor similarity;

[0047] Adaptive correction module, used to search for potential matching pairs based on the first-order neighbor objects of the exact matching pairs and adaptively update the spatial position of the BIM objects

[0048] Based on the same inventive concept, the third aspect of the present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, it implements the complex scene three-dimensional object matching method based on the first-order distance-topological compatibility index described in the first aspect.

[0049] Based on the same inventive concept, the fourth aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the program, the complex scene three-dimensional object matching method based on the first-order distance-topological compatibility index described in the first aspect is implemented.

[0050] Compared with the prior art, the advantages and beneficial technical effects of the present invention are as follows:

[0051] The present invention proposes a method for matching three-dimensional objects in complex scenes based on the first-order distance-topological compatibility index. First, a first-order distance-topological compatibility index for three-dimensional objects is constructed to express the positioning characteristics of three-dimensional objects, thereby alleviating the problem of geometric matching caused by random position changes that traditional first-order or second-order spatial compatibility indices cannot handle. A calculation technique for the similarity of three-dimensional objects based on the first-order distance-topological compatibility index is proposed to extract precise matching pairs and potential matching pairs. Finally, a multi-view constraint equation set is constructed using the relative positional relationship between precise matching pairs and potential matching pairs to achieve iterative self-correction of the spatial position of the BIM model in the potential matching pairs, thereby achieving precise matching of three-dimensional objects. The method of the present invention was experimented on a complex scene three-dimensional dataset, and the overall matching accuracy reached 97.92%. Experimental results show that compared with the current benchmark method, the method of the present invention has higher accuracy and reliability in the task of matching three-dimensional objects in complex scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0053] Figure 1 This is an overall flow chart of a complex scene three-dimensional object matching method based on a first-order distance-topology compatibility index in an embodiment of the present invention;

[0054] Figure 2 This is a flowchart of calculating the first-order distance-topological similarity between a field point cloud object and a BIM object in an embodiment of the present invention;

[0055] Figure 3 Schematic diagram of adaptive correction of the spatial position of a three-dimensional object in an embodiment of the present invention;

[0056] Figure 4 This is the experimental dataset in the embodiment of the present invention, where (a) is the single device distribution of the real-time point cloud scene, and (b) is the single device distribution of the BIM scene;

[0057] Figure 5 Figures 1 and 2 show the BIM 3D object matching results during the iterative process of an embodiment of the present invention: (a) shows the matching result after 1 iteration; (b) shows the matching result after 3 iterations; (c) shows the matching result after 5 iterations; (d) shows the matching result after 7 iterations; and (e) shows the percentage of 3D object matching in the BIM scene as a function of the number of iterations. In the figure, green, red, and gray represent correctly matched, incorrectly matched, and unused BIM 3D objects, respectively.

[0058] Figure 6 This is the spatial distribution of BIM 3D object matching results in a complex substation scenario. In the figure, green, blue, red, and orange represent correctly matched, correctly deleted, incorrectly matched, and incorrectly deleted BIM 3D objects, respectively.

[0059] Figure 7 Comparison of the spatial distribution of BIM 3D object matching effects using different methods in a complex substation scenario: (a) FGM; (b) DGM; (c) DeepWalk-LightGlue; (d) GraRep-LightGlue; (e) NetMF-LightGlue; (f) Node2Vec-LightGlue; (g) NodeSketch-LightGlue; (h) the method of the present invention.

[0060] Figure 8 FIG. 4 is a graph showing how the overall accuracy of different methods varies with the size of the displacement noise in a specific implementation manner. DETAILED DESCRIPTION

[0061] In response to the differences between the actual point cloud model of complex construction scenes and the BIM three-dimensional model in the design stage, which leads to inconsistencies in the number, position, and distribution of three-dimensional objects in different scenes, hindering the practical application of BIM models in digital twins, this paper proposes a three-dimensional object matching method based on first-order distance-topological similarity and spatial position adaptive correction, which achieves high-precision matching between actual point cloud objects and BIM three-dimensional models in complex scenes and provides a basis for automatic correction of BIM models. First, a first-order distance-topological compatibility index for three-dimensional objects is designed to express the positioning characteristics of three-dimensional objects, thereby alleviating the problem of geometric matching caused by random position changes that traditional first-order or second-order spatial compatibility indexes are difficult to handle; secondly, a three-dimensional object similarity calculation technology based on the first-order distance-topological compatibility index is proposed to extract accurate matching pairs and potential matching pairs; finally, using the relative position relationship between accurate matching pairs and potential matching pairs, a multi-line of sight constraint equation set is constructed to achieve iterative self-correction of the spatial position of BIM models in potential matching pairs.

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0063] Example 1

[0064] The present invention discloses a method for matching three-dimensional objects in complex scenes based on a first-order distance-topological compatibility index, comprising:

[0065] S1: Construct a first-order distance-topological compatibility index of three-dimensional objects based on the relative distance and spatial topological positioning features between three-dimensional objects;

[0066] S2: Based on the spatial layout of complex 3D objects, the first-order distance-topological compatibility index between the real-world point cloud objects and the BIM objects is calculated respectively;

[0067] S3: Based on the first-order distance-topological compatibility index of the real-world point cloud object and the BIM object, the first-order distance similarity and spatial topological similarity between the BIM object and the real-world point cloud object are calculated;

[0068] S4: Generate initial matching pairs by maximizing the first-order distance similarity;

[0069] S5: Establish accurate matching pairs by maximizing the first-order topological neighbor similarity;

[0070] S6: Search for potential matching pairs based on the first-order neighbor objects of the exact matching pairs and adaptively update the spatial positions of the BIM objects.

[0071] Specifically, the present invention proposes a spatial position iterative correction three-dimensional object matching method based on the first-order distance-topological compatibility index. It uses the semantic labels and spatial relationships between the objects to be matched to generate positioning feature descriptors, and iteratively updates the spatial position of the three-dimensional objects to achieve high-precision matching between BIM objects and real-world point cloud objects in complex scenes.

[0072] Specifically, if Figure 1 As shown, the method of the present invention can be implemented through a calculation process, which mainly includes the following steps:

[0073] Step 1: Define the first-order distance-topological compatibility index of the three-dimensional object to describe the positioning characteristics of the three-dimensional object

[0074] Step 2: Based on the spatial layout of complex 3D objects, calculate the first-order distance-topological compatibility index of the real-world point cloud objects and BIM objects respectively;

[0075] Step 3: Based on the first-order distance-topological compatibility index of the 3D object, calculate the first-order distance similarity and spatial topological similarity between the BIM object and the real-world point cloud object;

[0076] Step 4: Generate initial matching pairs. That is, use the method of maximizing the first-order distance similarity to match the real-world point cloud objects to the BIM objects with the smallest first-order distance. If the BIM object with the smallest first-order distance is matched to another real-world point cloud object, then match the BIM object with the next smallest first-order distance until the initial matching is completed.

[0077] Step 5: Establish an exact matching pair, that is, use the method of maximizing the first-order topological neighbor similarity to select the pair with the largest first-order topological neighbor similarity and the smallest first-order geometric distance from the initial matching pairs as the exact matching pair;

[0078] Step 6: Search for potential matching pairs and adaptively update the spatial positions of BIM objects. That is, potential matching pairs are selected from the first-order neighbors of exact matching pairs. It is required that the neighbors of potential matching pairs contain exact matching objects and are exact matching pairs of each other. On this basis, the distance relationship between all exact matching pairs and potential matching pairs is used as a constraint condition to establish a multi-view constraint equation group to achieve adaptive correction of the spatial positions of BIM objects in potential matching pairs.

[0079] Repeat steps 2 to 6 above until the number of exact matching pairs reaches a stable level, and then terminate the iteration.

[0080] The key contents of the present invention include the first-order distance-topological compatibility index of three-dimensional objects, the calculation of three-dimensional object matching similarity, and the adaptive correction of three-dimensional object spatial position. The three aspects are described in detail below.

[0081] 1.1 First-order distance of three-dimensional objects - topological compatibility index

[0082] Based on the relative distance and spatial topological positioning features between 3D objects, the first-order distance-topological compatibility index of 3D objects can be constructed in the following ways:

[0083] Calculate the geometric distance between any two three-dimensional objects and obtain the distance matrix;

[0084] Traverse each row of the distance matrix and express the relationship between the corresponding 3D object and the other 3D objects in the scene using the key-value pair of "type-distance", thus completing the generation of the first-order geometric distance positioning feature vector of the 3D object in the global space;

[0085] The topological relationship of 3D objects in the scene is established based on the 3D object Voronoi diagram. The type labels of the neighboring objects of the 3D objects are counted to generate a type label histogram. The first-order topological feature vector is obtained as the first-order topological positioning feature expression.

[0086] The first-order geometric distance positioning feature vector and the first-order topological feature vector are used as the first-order distance-topological compatibility index of three-dimensional objects.

[0087] Specifically, the relative distance between a three-dimensional object and other objects in a spatial scene can effectively describe its orientation information, and is one of the geometric expression features for matching three-dimensional objects. This type of feature can usually be expressed using first-order or second-order spatial compatibility indices. Traditional methods can solve geometric matching problems caused by transformations such as translation and rotation, but cannot effectively deal with geometric matching problems caused by changes in relative spatial layout due to random position offsets. To this end, the present invention introduces spatial topological positioning features on the basis of the first-order spatial compatibility positioning feature expression, and proposes a first-order distance-topological compatibility index for three-dimensional objects, thereby effectively dealing with geometric transformation problems caused by random position changes, and facilitating high-precision geometric matching between real-world point cloud objects and BIM objects in complex substation scenarios.

[0088] During the specific implementation process, assuming that there are M three-dimensional objects in the scene, this embodiment method provides a calculation process for the first-order distance-topological compatibility index of the three-dimensional objects. First, the geometric distance between any two three-dimensional objects is calculated to obtain a distance matrix. Secondly, each row of the distance matrix is ​​traversed, and the relationship between the corresponding three-dimensional object L and the remaining M-1 three-dimensional objects in the scene is expressed in a key-value pair of "type-distance", thereby completing the generation of the first-order geometric distance positioning feature vector of the three-dimensional object L in the global space. As shown in formula (1), where i and j represent two three-dimensional objects respectively, D i→j Indicates the geometric distance between 3D objects i and j in the scene, C M-1 Indicates the device type of the M-1th 3D object, F i Geoi Represents the first-order distance feature vector of the three-dimensional object i. Third, construct the three-dimensional object Voronoi diagram

[20] , establish the topological relationship of the three-dimensional objects in the scene, and count the type labels of the neighboring objects of the three-dimensional objects to generate a type label histogram as the first-order topological positioning feature expression. As shown in formula (2), where U i represents the first-order topological neighbor set of three-dimensional object i, Indicates that the neighbor type of the three-dimensional object i is C j the number of Represents the first-order topological eigenvector of three-dimensional object i.

[0089]

[0090] 1.2 3D Object Matching Similarity Calculation

[0091] Feature similarity between three-dimensional objects is an important basis for object matching. Traditional feature similarity algorithms require the consistency of the feature dimensions of the two objects. However, in complex substation scenarios, it is difficult to ensure that the first-order distance features and topological features of two three-dimensional objects have the same feature dimensions. Therefore, this implementation uses the minimum distance difference sum method to estimate the similarity of first-order distance features and the topological neighbor type statistical histogram difference method to estimate the similarity of first-order topological features.

[0092] like Figure 2 As shown, since the number of real-world point cloud objects is inconsistent with the number of BIM objects, this embodiment assumes that a relatively small number of real-world point clouds are used to calculate the similarity between each real-world point cloud object and each BIM object.

[0093] First, for the real-world point cloud object s, use its device type as a filter label to quickly obtain BIM objects with the same device type and add them to the set of matching objects; secondly, for the current real-world point cloud object s and BIM object t, calculate the minimum distance difference dimension by dimension, that is, select each key-value pair in the first-order distance feature vector of the real-world point cloud object s in turn, and use the device type of the key-value pair to filter the key-value pair set with the same device type in the first-order distance feature vector of the BIM object t, and calculate the difference between the distance in the current key-value pair and each distance in the filtered key-value pair set, and use the minimum value of the distance difference as the distance between the real-world point cloud object s and the BIM object t in the current dimension; thirdly, sum the minimum distance differences between the real-world point cloud object s and the BIM object t in all dimensions to obtain the similarity between them in the first-order distance feature. As shown in formula (3), It represents the distance difference between the mth dimension of the real-world point cloud object s and the nth dimension of the BIM object t after filtering the same type of key-value pairs, D s→m Indicates the distance value on the mth dimension in the real-time point cloud object s, D t→n Represents the distance value in the nth dimension of BIM object t, represents the minimum distance difference between the field point cloud object s and the BIM object t in the mth dimension, Δd s→t Represents the similarity between the field point cloud object s and the BIM object t in the first-order geometric distance feature.

[0094] Similar to the calculation strategy of first-order distance feature similarity, the calculation of first-order topological feature similarity is also performed between objects with the same type of label. Specifically, for the field point cloud object s and the BIM object t, this embodiment calculates the quantitative difference in the corresponding dimensions of the first-order topological feature vectors of the two objects and sums the differences in all dimensions to obtain the first-order topological feature similarity between the two objects. Its essence is the difference in the statistical histogram of the topological neighbor types between the two objects. As shown in formula (4), and Specifically, the topological neighbors of the real-world point cloud object s and the BIM object t belong to type C q The number of neighbors, Indicates that in type C q The difference in the number of topological neighbors between the two objects, Δn s→t Represents the first-order topological feature similarity between the field point cloud object s and the BIM object t.

[0095]

[0096] 1.3 Adaptive correction of 3D object spatial position

[0097] Exact matching pairs are generated when the real-world point cloud objects and BIM objects strictly meet the conditions of maximum first-order topological neighbor similarity and minimum first-order geometric distance. They are the most likely matching pairs between the real-world point cloud and the BIM scene, and their relative positions and spatial distributions are highly stable. In addition, given that the topological neighbor type statistical histograms of the exact matching pairs are exactly the same, the first-order topological neighbors of the exact matching pairs are more likely to become potential matching pairs, that is, the potential matching pairs have exact matching objects in their respective neighbors and are exact matching pairs with each other. However, if Figure 3 As shown, the relative positions of potential matching pairs (blue five-pointed star symbols) may have random position deviations. Therefore, the accurate matching pairs can be used as a control for the position correction of potential matching pairs to generate more accurate matching pairs in the object matching iteration process.

[0098] After determining several potential matching pairs through precise matching within the scene, the spatial positions of the potential matching pairs can be recalculated and updated using the multi-point positioning method. The algorithm diagram for self-correction of the potential matching pair positions is shown in the figure below. Figure 3 As shown in the figure, the main idea is based on the idea of ​​forward intersection based on photogrammetry, that is, taking the spatial position of the precisely matched 3D object in the BIM scene as the center of the circle and the spatial distance between the precisely matched 3D object and its potential matched 3D object in the real-world cloud scene as the radius, the forward intersection technology is used to establish a multi-view constraint equation group, and combined with the least squares principle, the spatial position of the corresponding potential matched 3D object in the BIM scene is determined, thereby obtaining its position correction vector, and then realizing the automatic update of the position of the potential matched object in the matching iteration process. The specific calculation formula is shown in (5), which gives the equation group of multi-view constraints, where, and They represent the coordinates of the potential matching pair L in the real-world cloud scene and the BIM scene, n represents the number of accurate matching pairs in the current iteration process, and Respectively represent the coordinates of the first accurate matching pair in the real-time point cloud scene and the BIM scene. and Under known conditions, the least squares technique is used to solve Then we get the position correction vector (i.e. This enables automatic correction of the spatial position of potential three-dimensional objects.

[0099]

[0100] The effectiveness and reliability of the method proposed in the present invention are verified through specific experiments below.

[0101] 2.1 Experimental Dataset and Model Evaluation

[0102] In order to verify the effectiveness and reliability of the method proposed in this invention, a complex substation scene was selected to carry out experiments. On the one hand, the real-life point cloud data of the substation was obtained using a ground-based laser scanner to reflect the three-dimensional equipment spatial layout during the actual operation of the substation; on the other hand, based on the substation digital twin platform, the BIM three-dimensional model data of the equipment running on the platform was obtained. The real-life three-dimensional point cloud obtained by the ground-based laser scanner contains a large number of spatially continuous three-dimensional points. First, the ground three-dimensional points need to be removed through a cloth algorithm, and then divided into multiple independent individual devices. On this basis, a type label is assigned to each device. Figure 4 Part (a) shows the 3D point clouds of different types of equipment obtained by segmenting the real-world point cloud. The BIM scene contains 180 equipment 3D models, each of which corresponds to a single BIM 3D model with detailed equipment type information and is stored in the form of fbx file type. The present invention first performs spatial uniform sampling on the BIM 3D model to obtain the BIM point cloud of the single equipment, then calculates the geometric center and stores it in the form of "object ID-type label-geometric center coordinates". Figure 4 Part (b) shows the 3D point clouds of different types of equipment in the BIM scene.

[0103] In this embodiment, the overall accuracy index (OA) is used to quantitatively evaluate the matching accuracy of all devices in the BIM scene. As shown in formula (6), OA represents the overall accuracy of BIM three-dimensional object matching, Indicates the number of correctly matched BIM 3D objects, Indicates the number of BIM 3D objects that are correctly deleted. Indicates the number of incorrectly matched BIM 3D objects. Indicates the number of incorrectly deleted BIM 3D objects. It should be noted that, given the potential inconsistency in the number of equipment objects between the real-world cloud scene and the BIM scene, the calculation of BIM object matching accuracy requires considering correctly matched BIM objects, correctly deleted BIM objects, incorrectly matched BIM objects, and incorrectly deleted BIM objects.

[0104]

[0105] 2.2 Experimental Results Analysis

[0106] Figure 5(a)-(d) respectively show the specific matching process of the real-world point cloud 3D objects and the BIM 3D objects in the 1st, 3rd, 5th and 7th iterations of the method of the present invention. It can be clearly seen that with the increase in the number of iterations, more and more real-world point cloud 3D objects establish precise matching relationships with the corresponding BIM 3D objects. Finally, the matching is completed through 7 iterations, and the overall matching accuracy reaches 97.92%. In the result diagram of each iteration, the green point cloud object represents the correct matching pair in the current iteration process, the red point cloud object represents the incorrect matching pair in the current iteration process, and the gray point cloud object represents the 3D object that has not been matched in the current iteration process. Figure 5 As shown in (e), when the number of iterations increases from 1 to 7, the percentage of correctly matched BIM 3D objects in the total BIM population increases from 37.78% to 59.44%, while the percentage of incorrectly matched BIM 3D objects in the total BIM population does not exceed 1.11%. Ultimately, the percentages of correctly deleted and incorrectly deleted BIM 3D objects are 38.33% and 1.11%, respectively. This result demonstrates that through iterative matching and adaptive position updates, this method can accurately complete 3D object matching in complex substation scenarios, demonstrating its effectiveness and accuracy.

[0107] In order to demonstrate the matching effect of 3D objects in complex substation scenes, Figure 6 The spatial overlay of the real-world point cloud objects and BIM 3D objects is plotted using different colors. Green represents correctly matched BIM 3D objects, blue represents correctly deleted BIM 3D objects, red represents incorrectly deleted BIM 3D objects, and orange represents incorrectly matched BIM 3D objects. As can be seen, only four BIM 3D objects were incorrectly matched and deleted, further demonstrating the effectiveness of the proposed algorithm.

[0108] 2.3 Comparative analysis of experimental methods

[0109] To further verify the effectiveness of the method of the present invention, this embodiment conducted a comparative analysis experiment with a benchmark method, including two traditional graph network matching methods, Factorized Graph Matching (FGM) and Deformable Graph Matching (DGM), and five deep learning matching methods. Due to the differences in graph node feature encoders, this embodiment uses DeepWalk, GraRep, NetMF, Node2Vec, and NodeSketch as feature encoders for graph nodes, and uses the latest deep learning matching algorithm LightGlue to complete the matching of real-world point cloud objects and BIM three-dimensional objects. Therefore, the deep learning matching method includes five different types, including DeepWalk-LightGlue, GraRep-LightGlue, NetMF-LightGlue, Node2Vec-LightGlue, and NodeSketch-LightGlue.

[0110] Table 1 shows the comparative experimental results of different methods. It can be seen that the matching accuracy of the traditional graph network matching method is significantly lower than that of the proposed method and the deep learning-based method, demonstrating the limitations of the traditional graph network matching method in handling complex substation 3D object matching tasks. The matching accuracy of the proposed method and the deep learning-based method both exceeds 90%, demonstrating the effectiveness of these methods in complex substation 3D object matching tasks. Specifically, the proposed method achieved the highest matching accuracy, reaching 97.92%, while the DGM method had the lowest matching accuracy, at only 54.67%. Among the deep learning-based matching methods, NetMF-LightGlue achieved a relatively high matching accuracy of 95.81%, while NodeSketch-LightGlue achieved a relatively low matching accuracy of 92.39%. In particular, the proposed method improves the accuracy by 39.44%, 43.25%, 3.80%, 3.46%, 2.11%, 4.15% and 5.53% compared with FGM, DGM, DeepWalk-LightGlue, GraRep-LightGlue, NetMF-LightGlue, Node2Vec-LightGlue and NodeSketch-LightGlue, respectively. Figure 7 The matching effects of BIM 3D objects using different methods in a complex substation scenario are demonstrated. Green and blue represent correctly matched and correctly deleted BIM 3D objects, respectively, while red and orange represent incorrectly matched and incorrectly deleted BIM 3D objects, respectively. These matching effects intuitively demonstrate the effectiveness of this method compared to the comparison method.

[0111] Table 1 Comparison of the overall accuracy of the proposed method and existing methods in the literature on 3D object matching in complex substation scenes

[0112]

[0113] 2.4 Sensitivity analysis of experimental methods

[0114] To further verify the reliability of the method of the present invention, this embodiment constructs a three-dimensional object dataset with different noise levels and conducts sensitivity analysis of different methods. By adding different levels of random displacement noise to the center point coordinates of the real-world point cloud three-dimensional objects and BIM three-dimensional objects, a three-dimensional object dataset with random position offset is formed to verify the reliability of different methods on more complex three-dimensional object datasets. In particular, for each three-dimensional object, with its center point as the circle point and the north direction as the starting direction, a uniformly distributed offset angle between 0-360 degrees is randomly generated, and an offset length that conforms to the normal distribution Norm(r,1) is further randomly generated. Finally, the three-dimensional object is randomly translated according to the offset angle and offset length. In this experiment, the value of r is set to 0.2 meters, 0.4 meters, 0.6 meters, 0.8 meters, 1.0 meters and 1.2 meters, thereby obtaining 6 sets of real-world point cloud three-dimensional object datasets and BIM three-dimensional object datasets with different noise levels.

[0115] The experimental results under different noise levels are as follows Figure 8 As shown in the figure, for most noisy 3D datasets, the proposed method achieves the highest overall accuracy. Specifically, when r is 0.2m, 0.4m, 0.6m, 0.8m, 1.0m, and 1.2m, the proposed method achieves overall accuracies of 97.92%, 96.89%, 90.66%, 92.73%, 92.39%, and 93.08%, respectively. Among deep learning-based matching methods, NodeSketch-LightGlue consistently performs relatively poorly in overall accuracy. Among traditional graph network-based matching methods, the overall accuracy of FGM and DGM both significantly falls below 50% when r exceeds 0.4m, consistently performing the worst. Specifically, with the increase of r value, the overall accuracy of the method of the present invention and other comparison methods shows a roughly downward trend, but the overall accuracy of the method of the present invention is always higher than 90%; when the value of r is 1.2 meters, the overall accuracy of the method of the present invention is 4.84%, 2.08%, 3.11%, 4.15% and 10.38% higher than that of DeepWalk-LightGlue, GraRep-LightGlue, NetMF-LightGlue, Node2Vec-LightGlue and NodeSketch-LightGlue based on deep learning, respectively, demonstrating the reliability of the method of the present invention in matching three-dimensional objects in complex substations.

[0116] In general, the present invention proposes a high-precision matching method between real-world point cloud three-dimensional objects and BIM three-dimensional objects in complex scenes, designs a three-dimensional object positioning feature expression model based on the first-order distance-topological compatibility index, constructs a similarity measurement calculation technology based on three-dimensional object type labels, and uses the idea of ​​photogrammetric forward intersection to iteratively update the spatial position of BIM three-dimensional objects through adaptive calculation, thereby achieving effective and reliable matching between three-dimensional objects of different modalities in complex scenes, and solving the problems of inconsistency in quantity, difference in geometric position, and mismatch in spatial relationship between BIM three-dimensional objects and actual running objects in the digital twin platform. The method of the present invention was experimentally compared with existing three-dimensional object spatial matching methods, showing that the method of the present invention has a high overall object matching accuracy, proving its effectiveness in the task of matching three-dimensional objects in complex scenes; by constructing three-dimensional object datasets with different noise levels, the stability of the method of the present invention in the overall object matching accuracy was verified, further proving the reliability of the method of the present invention in the task of matching three-dimensional objects in complex scenes.

[0117] Example 2

[0118] Based on the same inventive concept, this embodiment discloses a complex scene three-dimensional object matching device based on a first-order distance-topological compatibility index, comprising:

[0119] An index construction module is used to construct a first-order distance-topological compatibility index of three-dimensional objects based on the relative distance and spatial topological positioning features between three-dimensional objects;

[0120] An index calculation module is used to calculate the first-order distance-topological compatibility index of the real-world point cloud object and the BIM object based on the spatial layout of the complex three-dimensional object;

[0121] A similarity calculation module is used to calculate the first-order distance similarity and spatial topological similarity between the BIM object and the real-world point cloud object based on the first-order distance-topological compatibility index of the real-world point cloud object and the BIM object;

[0122] An initial matching pair generation module is used to generate initial matching pairs by maximizing the first-order distance similarity;

[0123] The precise matching pair establishment module is used to establish precise matching pairs by maximizing the first-order topological neighbor similarity;

[0124] The adaptive correction module is used to search for potential matching pairs based on the first-order neighbor objects of the exact matching pairs and adaptively update the spatial positions of BIM objects.

[0125] Since the device described in the second embodiment of the present invention is a device used to implement the complex scene three-dimensional object matching method based on the first-order distance-topological compatibility index in the first embodiment of the present invention, based on the method described in the first embodiment of the present invention, those skilled in the art will be able to understand the specific structure and variations of the device, so they will not be described in detail here. All devices used in the method in the first embodiment of the present invention fall within the scope of protection of the present invention.

[0126] Example 3

[0127] Based on the same inventive concept, the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method described in the first embodiment is implemented.

[0128] Since the computer-readable storage medium introduced in the third embodiment of the present invention is the computer-readable storage medium used to implement the complex scene three-dimensional object matching method based on the first-order distance-topological compatibility index in the first embodiment of the present invention, based on the method introduced in the first embodiment of the present invention, those skilled in the art can understand the specific structure and deformation of the computer-readable storage medium, so they are not described in detail here. All computer-readable storage media used in the method of the first embodiment of the present invention fall within the scope of protection of the present invention.

[0129] Example 4

[0130] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first embodiment when executing the program.

[0131] Since the computer device introduced in the fourth embodiment of the present invention is a computer device used to implement the complex scene three-dimensional object matching method based on the first-order distance-topological compatibility index in the first embodiment of the present invention, based on the method introduced in the first embodiment of the present invention, those skilled in the art can understand the specific structure and deformation of the computer device, so they are not described in detail here. All computer devices used in the method of the first embodiment of the present invention fall within the scope of protection of the present invention.

[0132] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0133] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0134] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, the present invention is intended to include such changes and modifications to the embodiments of the present invention if they fall within the scope of the claims and their equivalents.

Claims

1. A method for matching three-dimensional objects in complex scenes based on first-order distance-topological compatibility index, characterized in that: include: According to the relative distance and spatial topological positioning characteristics between three-dimensional objects, a first-order distance-topological compatibility index of three-dimensional objects is constructed; According to the spatial layout of complex 3D objects, the first-order distance-topological compatibility index between the real-world point cloud objects and the BIM objects is calculated respectively; Based on the first-order distance-topological compatibility index of the real-world point cloud objects and the BIM objects, the first-order distance similarity and spatial topological similarity between the BIM objects and the real-world point cloud objects are calculated; Generate initial matching pairs by maximizing the first-order distance similarity; The exact matching pairs are established by maximizing the first-order topological neighbor similarity; Potential matching pairs are searched based on the first-order neighbor objects of the exact matching pairs, and the spatial positions of the BIM objects are adaptively updated.

2. The complex scene three-dimensional object matching method based on first-order distance-topological compatibility index according to claim 1 is characterized in that: Based on the relative distance and spatial topological positioning features between 3D objects, a first-order distance-topological compatibility index of 3D objects is constructed, including: Calculate the geometric distance between any two three-dimensional objects and obtain the distance matrix; Traverse each row of the distance matrix and express the relationship between the corresponding 3D object and the other 3D objects in the scene using "type-distance" key-value pairs to complete the generation of the first-order geometric distance positioning feature vector of the 3D object in the global space; The topological relationship of 3D objects in the scene is established based on the 3D object Voronoi diagram. The type labels of the neighboring objects of the 3D objects are counted to generate a type label histogram. The first-order topological feature vector is obtained as the first-order topological positioning feature expression. The first-order geometric distance positioning feature vector and the first-order topological feature vector are used as the first-order distance-topological compatibility index of three-dimensional objects.

3. The complex scene three-dimensional object matching method based on first-order distance-topological compatibility index according to claim 2 is characterized in that: The calculation method of the first-order geometric distance positioning feature vector is: Among them, i and j represent two three-dimensional objects, D i→j Indicates the geometric distance between 3D objects i and j in the scene, C M-1 Indicates the device type of the M-1th 3D object, where M represents the number of 3D objects in the scene. Represents the first-order distance feature vector of three-dimensional object i; The first-order topological eigenvector is calculated as: Among them, U i represents the first-order topological neighbor set of three-dimensional object i, q represents the first-order topological neighbor of three-dimensional object i, C q represents the type of the first-order topological neighbors of a three-dimensional object i, Indicates that the neighbor type of the three-dimensional object i is C j the number of represents the first-order topological eigenvector of three-dimensional object i, and k represents a three-dimensional object.

4. The complex scene three-dimensional object matching method based on first-order distance-topological compatibility index according to claim 1 is characterized in that: Based on the first-order distance-topological compatibility index between the real-world point cloud object and the BIM object, the first-order distance similarity between the BIM object and the real-world point cloud object is calculated, including: For the real-world point cloud object s, use its equipment type as the filtering tag, obtain BIM objects with the same equipment type, and add them to the matching object set; For the current real-world point cloud object s and BIM object t, the minimum distance difference is calculated dimension by dimension. Specifically, each key-value pair in the first-order distance feature vector of the real-world point cloud object s is selected in turn, and the set of key-value pairs with the same device type in the first-order distance feature vector of the BIM object t is filtered based on the device type of the key-value pair. The difference between the distance in the current key-value pair and each distance in the filtered key-value pair set is calculated, and the minimum value of the distance difference is used as the distance between the real-world point cloud object s and the BIM object t in the current dimension. Sum the minimum distance differences between the real-world point cloud object s and the BIM object t in all dimensions to obtain the similarity between them in the first-order distance feature, which is used as the first-order distance similarity between the BIM object and the real-world point cloud object; For the real-world point cloud object s and the BIM object t, the spatial topological similarity between the BIM object and the real-world point cloud object is obtained by calculating the quantitative differences in the corresponding dimensions of the first-order topological feature vectors of the two objects and summing up the differences in all dimensions.

5. The complex scene three-dimensional object matching method based on first-order distance-topological compatibility index according to claim 4 is characterized in that: The calculation formula for the first-order distance similarity between BIM objects and field point cloud objects is: in, It represents the distance difference between the mth dimension of the real-world point cloud object s and the nth dimension of the BIM object t after filtering the same type of key-value pairs, D s→m Indicates the distance value on the mth dimension in the real-time point cloud object s, D t→n Represents the distance value in the nth dimension of BIM object t, represents the minimum distance difference between the field point cloud object s and the BIM object t in the mth dimension, Δd s→t It represents the first-order distance similarity between the field point cloud object s and the BIM object t, indicating the similarity between the two objects in the first-order geometric distance feature; The calculation formula for the spatial topological similarity between BIM objects and field point cloud objects is: and Specifically, the topological neighbors of the field point cloud object s and the BIM object t belong to type C q The number of neighbors, Indicates that in type C q The difference in the number of topological neighbors between the two objects, Δn s→t Represents the spatial topological similarity between the field point cloud object s and the BIM object t.

6. The complex scene three-dimensional object matching method based on first-order distance-topological compatibility index according to claim 1 is characterized in that: The method of maximizing the first-order topological neighbor similarity is used to establish accurate matching pairs, including: By maximizing the first-order topological neighbor similarity, the pairs with the largest first-order topological neighbor similarity and the smallest first-order geometric distance are selected from the initial matching pairs as the exact matching pairs.

7. The complex scene three-dimensional object matching method based on first-order distance-topological compatibility index according to claim 1 is characterized in that: Search for potential matching pairs based on the first-order neighbor objects of exact matching pairs and adaptively update the spatial positions of BIM objects, including: Select potential matching pairs from the first-order neighbors of the exact matching pairs, requiring that the neighbors of the potential matching pairs contain exact matching objects and are exact matching pairs of each other; Taking the distance relationship between all accurate matching pairs and potential matching pairs as the constraint condition, a multi-view constraint equation group is established to achieve adaptive correction of the spatial position of BIM objects in potential matching pairs.

8. A complex scene three-dimensional object matching device based on first-order distance-topological compatibility index, characterized in that: include: An index construction module is used to construct a first-order distance-topological compatibility index of three-dimensional objects based on the relative distance and spatial topological positioning features between three-dimensional objects; An index calculation module is used to calculate the first-order distance-topological compatibility index of the real-world point cloud object and the BIM object based on the spatial layout of the complex three-dimensional object; A similarity calculation module is used to calculate the first-order distance similarity and spatial topological similarity between the BIM object and the real-world point cloud object based on the first-order distance-topological compatibility index of the real-world point cloud object and the BIM object; An initial matching pair generation module is used to generate initial matching pairs by maximizing the first-order distance similarity; The precise matching pair establishment module is used to establish precise matching pairs by maximizing the first-order topological neighbor similarity; The adaptive correction module is used to search for potential matching pairs based on the first-order neighbor objects of the exact matching pairs and adaptively update the spatial positions of BIM objects.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for matching three-dimensional objects in complex scenes based on the first-order distance-topology compatibility index as described in any one of claims 1 to 7 is implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for matching three-dimensional objects in complex scenes based on the first-order distance-topology compatibility index as described in any one of claims 1 to 7 is implemented.

Images