An urban spatial atlas information platform of block form and its construction method

Abstract

The present invention discloses an urban spatial atlas information platform in the form of a block and a construction method thereof, belonging to the field of urban planning; the construction method of the information platform includes inputting block form data, building an urban spatial atlas information platform, fusing case multimodal data, training an automatic clustering model for block forms, training an intelligent link model cluster, and demonstrating and interacting with a holographic sand table; this method uses three-dimensional fusion entity linking and artificial intelligence technology to cluster block forms, and then forms an updated working model of the target block by linking the information of similar case blocks; the present invention can meet the construction requirements of old city renewal, generate a block renewal working model based on multimodal data, expand the collection scope of old city renewal data, and improve the working efficiency of old city renewal.

Description

Technical Field

[0001] The present invention belongs to the field of urban planning, and particularly relates to an urban spatial atlas information platform for block forms and a construction method thereof. Background Technique

[0002] Blocks are one of the basic elements of the physical spatial form of a city, the basic unit for urban planning compilation and management, and also the most direct environmental support for urban buildings. Urban blocks present diverse and complex forms. How to efficiently and scientifically classify and carry out block renewal and construction is a difficult task currently faced. The existing block renewal methods are mainly divided into comprehensive renovation types, reconstruction and addition types, demolition and reconstruction types, etc. For different renewal methods, due to different renovation intensities, scales, and dominant methods, their renovation effects are very different. Choosing the wrong block renewal method will cause a series of problems such as renewal failure, economic losses, and urban development imbalance. However, the existing technology is still difficult to accurately predict the applicable block renewal and development methods.

[0003] Currently, for the implementation of block renewal work, it mainly relies on experienced designers to subjectively classify and judge the block forms. This judgment method has great randomness, and the classification conclusions obtained through different classification criteria are also different, with low classification efficiency. Summary of the Invention

[0004] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide an urban spatial atlas information platform for block forms and a construction method thereof.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A construction method of an urban spatial atlas information platform for block forms includes the following steps:

[0007] S1, obtaining the three-dimensional vector data of the spatial forms of the roads, buildings, and natural elements of the blocks within the target area, and performing unit splitting, block grouping, and element coding;

[0008] S2, constructing an algorithm rule for spatial form attributes, building an urban spatial atlas information platform, inputting the processed data of the target blocks in S1 into the atlas information platform, and calculating the spatial unit attributes and spatial association attributes of the target blocks;

[0009] S3, obtaining the image, text, and model materials of block renewal design cases, forming three-dimensional fusion entity links and inputting them into the platform case library, and calculating the spatial unit attributes and spatial association attributes of the case blocks;

[0010] S4. Use the spatial unit attributes and spatial association attributes of the block as machine learning labels, and adopt a supervised clustering learning algorithm to perform clustering and association machine learning training on the target block and case blocks, form an automatic clustering model of block morphology, and optimize the model.

[0011] S5. Use the spatial unit keywords and spatial unit attributes of the block as machine learning labels, and adopt a supervised clustering learning algorithm to perform data transfer machine learning training on the three-dimensional fusion entity links of the case blocks and the three-dimensional model of the target block, form a cluster of intelligent link models for various entity data, and construct a multi-modal model for updating the target block.

[0012] S6. Demonstrate the multi-modal model for updating the target block on the holographic sand table. Through the interactive selection and operation of the user, feedback and optimize the automatic clustering model of block morphology and the cluster of intelligent link models for various entity data, form a working model for updating the target block, output an update work manual, connect to a 3D printer, and output a physical model.

[0013] Further, in S1, the road information data refers to the center line, width, and intersection morphology data of the block roads after rasterization processing; the building information data refers to the coordinate position, building height, and three-dimensional shape data after rasterization processing, and the natural element data refers to the natural element unit data distinguished by using infrared remote sensing sub-band technology and vectorized.

[0014] Further, in S1, the block grouping refers to taking the road data presented as a closed polygon as a block contour and grouping it with the internal building data and natural element data into a block data.

[0015] The element coding refers to performing a ten-digit coding on each entity unit within a block. The coding basis is that the first six digits are the block serial number, the seventh digit is the entity type to which it belongs, and the last three digits are the entity unit serial number.

[0016] Further, in S2, the steps for constructing the urban spatial atlas information platform are as follows:

[0017] 1) Structurally process the urban spatial data, perform data deduplication, feature calculation, and supplementation operations to generate spatial unit attributes and spatial association attributes.

[0018] 2) Digitally encode the read structured data, encapsulate the data into entities according to the ontology, and establish relationships between entities through algorithms to construct the urban spatial atlas information platform.

[0019] Further, in S3, the steps for forming three-dimensional fusion entity links and entering them into the platform case library are as follows:

[0020] 1) Extract keywords from the case vocabulary library as entities, extract keywords from the case block update regulations and the image library as candidate entities, and use a supervised method to calculate the matching degree between the entities and the candidate entities. The links with the highest matching degree form two-dimensional fusion entity links;

[0021] 2) Take the keywords in the case vocabulary library included in the obtained two-dimensional fusion entity links as entities, take the keywords extracted from the case three-dimensional model library as candidate entities, use a supervised method to calculate the matching degree between the entities and the candidate entities, and the links with the highest matching degree form three-dimensional fusion entity links. Input them into the urban spatial atlas information platform to form a multi-modal database of block update design cases.

[0022] Furthermore, in S4, the steps of machine learning training are as follows:

[0023] 1) Divide the target block and the case blocks into a training set, a validation set, and a test set according to a ratio of 6:2:2; use a holographic sand table with a platform size of not less than 200 cm × 200 cm to demonstrate the three-dimensional model of the block. The operator wears data gloves with a static accuracy of attitude solution Roll / pitch ≤ 1.0 deg to select valuable case blocks for the demonstrated block, and uses an eye tracker with a line-of-sight tracking accuracy of 0.4° to assist in obtaining the operator's selection tendency;

[0024] 2) Conduct machine learning training on block association through a deep learning system with 512GB video memory, and select a machine learning model with strong generalization performance as the automatic clustering model for block morphology through cross-validation and generalization testing.

[0025] Furthermore, the model optimization refers to determining whether there are n necessary association relationships between the case block and the target block; if the number of necessary association relationships > n, output the obtained case block; if the number of necessary association relationships < n, return to adjust and optimize the automatic clustering model for block morphology.

[0026] Furthermore, in S5, the steps of machine learning training are as follows:

[0027] 1) Divide the target block and the case blocks into a training set, a validation set, and a test set according to a ratio of 6:2:2; use a holographic sand table with a platform size of not less than 200 cm × 200 cm to demonstrate the three-dimensional model of the block. The operator wears data gloves with a static accuracy of attitude solution Roll / pitch ≤ 1.0 deg (RMS) to select entity materials from the multi-modal data of the case block links output in step S4-3, transfer them to the spatial units of the demonstrated block, and use an eye tracker with a line-of-sight tracking accuracy of 0.4° to assist in obtaining the user's selection tendency;

[0028] 2) Transfer the data for machine learning training through a deep learning system with 512GB video memory. Through cross-validation and generalization tests, select the data transfer machine learning model of updated regulations, effect diagrams, analysis diagrams, and post-construction photos with strong generalization performance as the entity data intelligent link model, and further combine them to form a cluster of intelligent link models for entity data.

[0029] Furthermore, the steps to establish the multi-modal model for block renewal are as follows: According to the three case blocks obtained by intelligent matching, use the cluster of entity data intelligent link models that have been trained to fuse and link the updated regulations, effect diagrams, analysis diagrams, and post-construction photos of the case blocks with the corresponding spatial units of the target block.

[0030] An urban spatial atlas information platform for block form is constructed using the above method.

[0031] Advantages of the present invention:

[0032] The automatic clustering model of block form obtained by machine learning in the present invention, compared with the existing classification methods, takes into account the complex and diverse forms of urban blocks and can obtain block clustering that is not only similar in shape but also similar in many aspects such as spatial relationship, morphological characteristics, and functional location; the application of this model in the field of block classification allows for multi-line parallel calculation of block classification, doubling the block classification efficiency, and at the same time considering multiple internal factors other than form, improving the accuracy of classification, and facilitating the subsequent classification to carry out block renewal work;

[0033] The present invention realizes the full automation of the whole process by obtaining the image, text, and model data of block renewal design cases, linking to obtain a multi-modal database of block renewal design cases, and intelligently matching to the blocks to be updated;

[0034] The multi-modal model for block renewal obtained in the present invention breaks through the traditional single block renewal results such as pictures and videos. By aggregating multi-source information, in addition to the original traditional text and model display methods, it also provides three updated display forms: holographic sand table interaction, MR mixed reality display glasses interaction, and 3D model printing, enabling the public to participate in the block renewal work process;

[0035] The present invention uses a holographic sand table and MR mixed reality display glasses to be able to correspond the block form and the updated effect in real time, shortening the decision-making time and realizing the immersive experience and real-time operation and modification of the block renewal work model. Description of the Drawings

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0037] Figure 1 Flow chart for constructing the urban spatial atlas information platform of the block form of the present invention;

[0038] Figure 2 Typical block clustering table in the embodiments of the present invention;

[0039] Figure 3 Updated work manual content table in the embodiments of the present invention. Detailed implementation manners

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0041] As Figure 1 shown, a method for constructing an urban spatial atlas information platform of a block form includes the following steps:

[0042] S1: Obtain the spatial form three-dimensional vector data of the road information, building information, and natural elements of the blocks in the target area, and perform unit splitting, block grouping, and element coding;

[0043] Among them, the road information data refers to the center line, width, and intersection form data of the block roads after rasterization processing; the building information data refers to the coordinate position, building height, and three-dimensional shape data after rasterization processing, and the natural element data refers to the natural element unit data distinguished by using the infrared remote sensing sub-band technology and vectorized;

[0044] Block grouping means taking the road data presented as a closed polygon as a block contour and grouping it with the internal building data and natural element data into a block data;

[0045] Element coding means performing a ten-digit coding on each entity unit in a block. The coding basis is that the first six digits are the block serial number, the seventh digit is the entity type to which it belongs, and the last three digits are the entity unit serial number.

[0046] S2: Construct the algorithm rules for spatial form attributes, build the urban spatial atlas information platform, input the processed data of the target block in S1 into the atlas information platform, and calculate the spatial unit attributes and spatial association attributes of the target block;

[0047] Constructing the algorithm rules for spatial form attributes and building the urban spatial atlas information platform means performing structured processing on urban spatial data, such as data deduplication, feature calculation, and supplementation, to generate spatial unit attributes and spatial association attributes; digitally encoding the read structured data, encapsulating the data into entities according to the ontology, and establishing relationships between entities through algorithms to build the urban spatial atlas information platform;

[0048] Among them, the spatial unit attributes are composed of the spatial attributes of six types of units: water systems, mountains, roads, blocks, land uses, and buildings; the spatial association attributes consist of three main types: inclusion, adjacency, and similarity.

[0049] S3: Obtain the image, text, and model materials of the block renewal design cases, form three-dimensional fusion entity links and input them into the platform case library, and calculate the spatial unit attributes and spatial association attributes of the case blocks;

[0050] Among them, for the text materials, a workstation with more than 64-core processors is used. The semantic analysis method is used to count the characteristic vocabulary in the text, construct a syntax tree of block characteristic indicators, sort the priorities of block characteristic parameters, classify the keywords, and extract the block case vocabulary library and block renewal regulations;

[0051] The image materials are composed of case effect diagrams, case analysis diagrams, and photos after completion. A workstation with more than 64-core processors is used. According to the extraction results of the key vocabulary in the text entity materials, keyword annotation of the case images is performed through full-pixel semantic segmentation machine learning, and the block case image library is extracted;

[0052] The model materials are composed of case three-dimensional models and site elevation models. A workstation with more than 64-core processors is used. According to the block grouping rules and element coding rules, the three-dimensional models are grouped and coded; according to the extraction results of the key vocabulary in the text materials, keyword annotation of the case three-dimensional models is performed through three-dimensional scene online semantic segmentation, and the three-dimensional model library of the renewal design cases is extracted;

[0053] Among them, the steps for forming three-dimensional fusion entity links and inputting them into the platform case library are as follows:

[0054] 1) Use the keywords extracted from the case vocabulary library as entities, use the keywords extracted from the block renewal regulations and the image library of the case as candidate entities, and use the supervised method to calculate the matching degree between the entity and the candidate entity. The link with the highest matching degree forms a two-dimensional fusion entity link;

[0055] 2) Use the keywords in the case vocabulary library included in the obtained 2D fusion entity link as entities, and use the keywords extracted from the case 3D model library as candidate entities. Use a supervised method to calculate the matching degree between the entities and the candidate entities. The link with the highest matching degree forms a 3D fusion entity link, which is input into the urban spatial atlas information platform to form a multimodal database of block renewal design cases.

[0056] S4: Use the spatial unit attributes and spatial association attributes of the block as machine learning labels, and adopt a supervised clustering learning algorithm to perform clustering association machine learning training on the target block 00 and the case blocks, forming an automatic block morphology clustering model. Recommend the three case blocks with the highest spatial morphology association degree with the target block in the platform case library and optimize the model;

[0057] The machine learning training mentioned above means: divide the target block and the case blocks into a training set, a validation set, and a test set according to the ratio of 6:2:2; use a holographic sand table with a platform size of not less than 200cm×200cm to demonstrate the 3D model of the block. The operator wears data gloves with a static accuracy of attitude solution Roll / pitch≤1.0deg (RMS) to select valuable case blocks for the demonstrated block, and uses an eye tracker with a line-of-sight tracking accuracy of 0.4° to assist in obtaining the operator's selection tendency; perform block association machine learning training through a deep learning system with 512GB video memory, and select a machine learning model with strong generalization performance as the automatic block morphology clustering model through cross-validation and generalization tests.

[0058] Model optimization means determining whether there are n necessary association relationships between the case block and the target block; if (the number of necessary association relationships>n) is satisfied, output the obtained case block; if (the number of necessary association relationships<n) is not satisfied, return to adjust and optimize the automatic block morphology clustering model.

[0059] S5: Use the spatial unit keywords and spatial unit attributes of the block as machine learning labels, and adopt a supervised clustering learning algorithm to perform data transfer machine learning training on the 3D fusion entity link of the case block and the 3D model of the target block, forming a cluster of intelligent link models of various entity data, and constructing a multimodal model for the renewal of the target block;

[0060] The machine learning training mentioned above refers to: dividing the target block and the case blocks into a training set, a validation set, and a test set according to the ratio of 6:2:2; using a holographic sand table with a platform size of not less than 200 cm × 200 cm to demonstrate the 3D model of the block. The operator wears data gloves with a static accuracy of attitude solution Roll / pitch ≤ 1.0 deg (RMS) and selects entity materials from the multi-modal data linked to the case blocks output in step S4-3, transfers them to the spatial units of the demonstrated block, and uses an eye tracker with a line-of-sight tracking accuracy of 0.4° to assist in obtaining the operator's selection tendency; conducts data transfer machine learning training through a deep learning system with 512GB video memory, and selects a data transfer machine learning model with strong generalization performance for updated regulations, renderings, analysis diagrams, and photos after completion of construction as the intelligent link model for entity materials through cross-validation and generalization tests. Further combine them to form a cluster of intelligent link models for entity materials;

[0061] Among them, the multi-modal model for block update refers to: based on the three case blocks obtained through intelligent matching, using the cluster of intelligent link models for entity materials that have been trained, fusing and linking the updated regulations, renderings, analysis diagrams, and photos after completion of construction of the case blocks with the corresponding spatial units of the target block, so as to obtain the multi-modal model for block update.

[0062] S6: Demonstrate the multi-modal model for the update of the target block on the holographic sand table. Through the interactive selection and operation of the user, feedback and optimize the automatic clustering model of the block form and the cluster of intelligent link models for various entity materials to form a working model for the update of the target block, output an update work manual, connect to a 3D printer, and output a physical model;

[0063] Among them, the interactive selection and operation of the user refer to: the user wears data gloves with a static accuracy of attitude solution Roll / pitch ≤ 1.0 deg (RMS) to connect the 3D model processing software and the holographic sand table for interaction, and modifies the model according to the case entity materials linked to the spatial units;

[0064] The update work manual refers to an update work manual that includes update guidelines, update material tables, schematic diagrams of landscape features models, and update intention diagrams, output by combining the entity materials in the case blocks;

[0065] Output entity model refers to: a 3D printer with a connection forming size greater than 255mm * 300mm * 300mm and supporting the printing of color entity updated models by PolyJet 3D printing technology; among them, the unupdated content is printed with white consumables, and the updated content is divided into three categories, including deletion (demolishing buildings), modification (facade modification, material modification), and addition (adding greenery, adding landscape features), which are printed with colored consumables of different colors; users can wear MR mixed display glasses to observe the entity updated model, and updated regulations, renderings, analysis diagrams, and photos after completion can be linked to the entity updated model for enhanced display.

[0066] Embodiment:

[0067] Taking a certain central urban area as an example, the technical solution of the present invention will be described in detail below.

[0068] S1. Obtain the three-dimensional vector data of the city by using a multi-rotor drone with a maximum payload greater than 3 kg and equipped with an oblique photography camera with more than 40 million pixels and the Remote Sensing Satellite Resource-3 with a flight altitude below 800 km, and obtain urban design plan data and urban design standard specification data from the urban planning department of the city; the specific steps are as follows:

[0069] S11. Use a multi-rotor drone with a maximum payload greater than 3 kg to carry an oblique photography camera with more than 40 million pixels to obtain aerial images within the target area, and use the oblique photography data to obtain the three-dimensional vector data of the spatial forms of roads and buildings in the target area of the central urban area of the city. Use the Remote Sensing Satellite Resource-3 (multi-spectral resolution of 5.8 meters) with a flight altitude below 800 km to obtain the three-dimensional vector data of the spatial forms of urban natural elements; among them, the road information data refers to the data of the center line, width, and intersection form of the streets in the target area of the central urban area of the city after rasterization processing, the building information data refers to the building coordinate position, building height (in the case of no height, the building height is deduced by the number of building floors, and the building height is the number of building floors * 3 meters), and three-dimensional shape data in the target area after rasterization processing, and the natural element data refers to the data of natural element units in the target area that are distinguished and vectorized by using the infrared remote sensing sub-band technology. The above vector data can be integrated into DWG or SHP format, and geographic coordinate data needs to be added;

[0070] S12. Take the road data of the central urban area of the city that currently appears as a complete closed polygon as a block contour, obtain the block contour range of independent plots, use row index in DataFrame to perform interval data segmentation on the three-dimensional vector data within the block, group the three-dimensional vector data within the block, import the CAD file that currently appears as a closed block, the internal building data, and natural element data into the geographic information system software, and export the data within the closed polyline as SHP format;

[0071] S13. Encode each entity unit within each block in the central urban area with a ten-digit code. The encoding basis is that the first six digits are the block serial number, the seventh digit is the entity type to which it belongs, and the last three digits are the entity unit serial number.

[0072] S2. Construct the urban spatial atlas information platform, construct the algorithm rules for spatial form attributes, build the urban spatial atlas information platform, input the processed data of the target blocks in the central urban area into the platform, and calculate the spatial unit attributes and spatial association attributes of the target blocks. The specific steps are as follows:

[0073] S21. Perform structured processing on the obtained urban spatial data of the target blocks in the central urban area. Use the data processing methods of duplicated and drop_duplicates provided by the python platform to perform operations such as duplicate removal, feature calculation, and supplementation on the urban spatial data, and generate spatial unit attributes and spatial association attributes.

[0074] S22. Digitally encode the read structured data of the central urban area using ProtoBuf, pack the data according to the ontology, perform error correction and encryption processing on the data, encapsulate it into entities, establish associations between the encapsulated entities through the Visual Studio algorithm, and construct the urban spatial atlas information platform based on the association attribute mapping relationship. Among them, the spatial unit attributes are composed of the spatial attributes of six types of units: water systems, mountains, roads, blocks, land uses, and buildings, and the spatial association attributes are composed of three main types: inclusion, adjacency, and similarity.

[0075] S23. Establish a data folder, use the data addition function of the geographic information platform to establish a connection with the folder, and input the data within each closed block contour shown in the central urban area obtained into the urban spatial atlas information platform.

[0076] S24. Use a workstation with more than 64-core processors to calculate the spatial unit attributes of the target blocks in the central urban area in the information platform; based on NEO4J, associate the target block form with an available knowledge graph network, establish the corresponding relationships in the ontology. For example, blockId in the land use is a foreign key that connects to the id field of the block, and then establish the ownership relationship between the corresponding block and the land use. According to the longitude and latitude information of the entities, calculate the two attributes of r and theta for the relationship between them, and then calculate the semantic similarity between the pictures through the deep learning BM25 model, establish the picture similarity relationship between the block entities, establish the association between the smallest spatial units in the same block according to the same, adjacent, and similar relationships between the smallest spatial unit attributes, and calculate the attributes of the spatial association in the information platform.

[0077] S3. Obtain the image, text, and model data of the block renewal design cases, form three-dimensional fusion entity links and input them into the platform case library, and calculate the spatial unit attributes and spatial association attributes of the case blocks. The specific steps are as follows:

[0078] S31. Obtain the text entity data of the existing block renewal design cases. Use a workstation with more than 64-core processors, utilize natural language processing (NLP), count the feature words in the text through semantic analysis, construct a syntax tree of the block feature index shell grammar, and use PowerShell operators to sort the priorities of the block feature parameters, classify the keywords, and extract the block case vocabulary library and block renewal regulations;

[0079] S32. Obtain the image entity data of the existing block renewal design cases. The entity data consists of case renderings, case analysis diagrams, and photos after completion. Use a workstation with more than 64-core processors, and according to the key word extraction results sorted by the syntax tree described in S31, perform keyword annotation on the case images through full-pixel semantic segmentation machine learning to extract the block case image library;

[0080] S33. Obtain the three-dimensional model entity data of the existing block renewal design cases. The entity data consists of case three-dimensional models and site elevation models. Use a workstation with more than 64-core processors to group and encode the three-dimensional models; according to the key word extraction results sorted by the syntax tree in S31, use ACM TOG based on supervoxel convolution to perform online semantic segmentation on the three-dimensional scene, and perform keyword annotation on the case three-dimensional models to extract the three-dimensional model library of the renewal design cases;

[0081] S34. Use the keywords extracted from the case vocabulary library as entities, and use the keywords extracted from the block renewal regulations and image library of the cases as candidate entities. Use a supervised method to calculate the matching degree between the entity and the candidate entity, and the link with the highest matching degree forms a two-dimensional fusion entity link;

[0082] S35. Use the keywords in the case vocabulary library included in the obtained two-dimensional fusion entity link as entities, and use the keywords extracted from the case three-dimensional model library as candidate entities. Use a supervised method to calculate the matching degree between the entity and the candidate entity, and the link with the highest matching degree forms a three-dimensional fusion entity link, and input it into the urban spatial atlas information platform to form a multi-modal database of block renewal design cases;

[0083] S36. Use a workstation with more than 64-core processors. Based on the knowledge graph platform provided by the Neo4j Graph Platform, associate the case block forms into a linkable knowledge graph network, establish ontology correspondence relationships. For example, connect the land use foreign key to the id field of the case block to establish the ownership relationship between the case block and the land use. According to the geographical location information of the entities, calculate the graph relationship attributes. Then, through the Cosin similarity research method, represent the pictures as vectors, and characterize the semantic similarity of two pictures by calculating the cosine distance between the vectors, calculate the similarity between pictures, establish the picture similarity relationship between case block entities. According to the same, adjacent, and similar relationships between the attributes of the smallest spatial units, establish the association between the smallest spatial units in the same block, and calculate the unit attributes and spatial association attributes of the case library space in the information platform.

[0084] S4. Conduct clustering association machine learning training on the target block and the case blocks to form an automatic block form clustering model, and recommend the three case blocks with the highest spatial form association degree with the target block in the platform case library. The specific steps are as follows:

[0085] S41. Take the calculated spatial unit attributes and spatial association attributes of the target block and the case blocks as machine learning labels, divide the target block and the case blocks into a training set, a validation set, and a test set according to the ratio of 6:2:2. Use a holographic sand table to demonstrate the 3D model of the block, and the operator wears data gloves to select valuable case blocks for the demonstrated block, and assist in obtaining the operator's selection tendency through an eye tracker.

[0086] S42. Conduct block association machine learning training with a deep learning system, select a machine learning model with strong generalization performance as the automatic block form clustering model through the K-fold cross-validation method, obtain 6 types of typical blocks, as Figure 2 shown, and output n necessary association relationships in the judgment clustering division.

[0087] S43. Use the automatic block form clustering model. Taking type A as an example, solve and screen out the three case blocks with the highest association degree with the target block, determine whether there are 5 necessary association relationships between the case block and the target block, and output the obtained case blocks; if not satisfied, return to adjust and optimize the automatic block form clustering model until the output conditions are met.

[0088] S5. Conduct data transfer machine learning training on the 3D fusion entity link of the case block and the 3D model of the target block to form an intelligent link model cluster of various entity data, and construct a target block update multi-modal model. The specific steps are as follows:

[0089] S51. Take the spatial unit keywords and spatial unit attributes of the block as machine learning labels. Divide the target block and case blocks into a training set, a validation set, and a test set according to the ratio of 6:2:2. Use a holographic sand table to demonstrate the 3D model of the block. The operator wears data gloves and selects entity materials from the multi-modal data linked to the output case blocks, transfers them to the spatial units of the demonstrated block, and obtains the operator's selection tendency with the assistance of an eye tracker.

[0090] S52. Use a deep learning system to perform data transfer machine learning training. Select a data transfer machine learning model with strong generalization performance for updated regulations, renderings, analysis diagrams, and post-construction photos as an entity data intelligent link model through the K-fold cross-validation method, and further combine them to form a cluster of intelligent link models for entity data.

[0091] S53. According to the three case blocks obtained by intelligent matching, use the trained cluster of intelligent link models to fuse and link the updated regulations, renderings, analysis diagrams, and post-construction photos of the case blocks with the corresponding spatial units of the target block, so as to construct an updated multi-modal model of the case blocks.

[0092] S6. Demonstrate the updated multi-modal model of the target block on the holographic sand table. Through the interactive selection and operation of the user, feedback and optimize the block form automatic clustering model and the cluster of intelligent link models for various entity materials to form an updated working model of the target block, output an updated work manual, connect a 3D printer, and output a physical model. The specific steps are as follows:

[0093] S61. Use the holographic sand table to demonstrate the constructed updated multi-modal model of the case blocks. The user wears data gloves to connect the 3D model processing software and the holographic sand table for interaction, and modifies the model according to the case entity materials linked to the spatial units to form an updated working model of the target block.

[0094] S62. The system further connects to the network to feedback and optimize the block form automatic clustering model and the cluster of intelligent link models according to the user's selection tendency and modification behavior.

[0095] S63. According to the modified block update model, combined with the entity materials in the case blocks, output an updated work manual. The specific content of the manual is as Figure 3 shown; connect a 3D printer to print a colored physical update model. The user can wear MR mixed display glasses to observe the physical update model, and the updated regulations, renderings, analysis diagrams, and post-construction photos in the case materials can be linked to the physical update model for enhanced display.

[0096] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0097] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and what is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. A construction method of an urban spatial atlas information platform with a block form, characterized in that, Including the following steps: S1. Obtain the three-dimensional vector data of the road information, building information, and spatial forms of natural elements within the target area, and perform unit splitting, block grouping, and element coding; S2. Construct the algorithm rules for spatial form attributes, build the urban spatial atlas information platform, and input the processed data of the target block in S1 into the atlas information Information platform to calculate the spatial unit attributes and spatial association attributes of the target block; S3. Obtain the image, text, and model materials of the block update design cases, form three-dimensional fusion entity links and input them into the platform case library, and calculate the spatial unit attributes and spatial association attributes of the case blocks; S4. Use the spatial unit attributes and spatial association attributes of the blocks as machine learning labels, and adopt a supervised clustering learning algorithm to perform clustering and association machine learning training on the target blocks and case blocks to form an automatic block form clustering model and optimize the model; S5. Use the spatial unit keywords and spatial unit attributes of the blocks as machine learning labels, and adopt a supervised clustering learning algorithm to perform data transfer machine learning training on the three-dimensional fusion entity links of the case blocks and the three-dimensional models of the target blocks to form an intelligent link model cluster of various entity materials and construct a multi-modal model for the update of the target blocks; S6. Demonstrate the multi-modal model for the update of the target blocks on the holographic sand table, and through the interactive selection and operation of the user, feedback and optimize the automatic block form clustering model and the intelligent link model cluster of various entity materials to form a working model for the update of the target blocks, output an update work manual, connect a 3D printer, and output a physical model.

2. The construction method of an urban spatial atlas information platform with a block form according to claim 1, characterized in that, In S1, the road information data refers to the data of the center line, width, and intersection form of the block roads after rasterization processing; the building information data refers to the coordinate position, building height, and three-dimensional shape data after rasterization processing, and the natural element data refers to the natural element unit data distinguished by using the infrared remote sensing sub-band technology and vectorized.

3. The construction method of an urban spatial atlas information platform with a block form according to claim 2, characterized in that, In S1, block grouping refers to taking the road data presented as a closed polygon as a block outline and grouping it with the internal building data and natural element data into a block data; The element coding refers to performing a ten-digit coding on each entity unit within a block. The coding basis is that the first six digits are the block serial number, the seventh digit is the entity type to which it belongs, and the last three digits are the entity unit serial number.

4. The construction method of an urban spatial atlas information platform with a block form according to claim 1, characterized in that, In S2, the steps for constructing the urban spatial atlas information platform are as follows: 1) Structurally process the urban spatial data, perform data deduplication, feature calculation, and supplementation operations to generate spatial unit attributes and spatial association attributes; 2) Digitally encode the read structured data, encapsulate the data into entities according to the ontology, and establish relationships between the entities through algorithms to construct the urban spatial atlas information platform.

5. The construction method of an urban spatial atlas information platform with a block form according to claim 1, characterized in that, In S3, the steps for forming three-dimensional fusion entity links and inputting them into the platform case library are as follows: 1) Extract keywords from the case vocabulary library as entities, extract keywords from the case block update regulations and image library as candidate entities, use a supervised method to calculate the matching degree between the entities and the candidate entities, and the link with the highest matching degree forms a two-dimensional fusion entity link; 2) Use the keywords in the case vocabulary library included in the two-dimensional fusion entity link as entities, and use the keywords extracted from the case three-dimensional model library as candidate entities. Use a supervised method to calculate the matching degree between the entities and the candidate entities. The link with the highest matching degree forms a three-dimensional fusion entity link, which is input into the urban spatial atlas information platform to form a multi-modal database of block renewal design cases.

6. The construction method of an urban spatial atlas information platform with a block form as claimed in claim 1, wherein, The steps of the machine learning training in S4 are as follows: 1) Divide the target block and the case blocks into a training set, a validation set, and a test set according to a ratio of 6:2:2; use a holographic sand table with a platform size of not less than 200 cm × 200 cm to demonstrate the three-dimensional model of the block. The operator wears data gloves with a static accuracy of attitude solution Roll / pitch ≤ 1.0 deg to select valuable case blocks for the demonstrated block, and uses an eye tracker with a line-of-sight tracking accuracy of 0.4° to assist in obtaining the selection tendency of the operator. 2) Conduct machine learning training on block association through a deep learning system with 512 GB video memory. Select a machine learning model with strong generalization performance as the block form automatic clustering model through cross-validation and generalization tests.

7. The construction method of an urban spatial atlas information platform for a block form according to claim 6, characterized in that, The model optimization in S4 refers to determining whether there are n necessary association relationships between the case block and the target block; if the number of necessary association relationships > n, output the obtained case block; if the number of necessary association relationships < n, return to adjust and optimize the block form automatic clustering model.

8. A method for constructing an urban spatial atlas information platform with a block form, characterized in that, The steps of the machine learning training in S5 are as follows: 1) Divide the target block and the case blocks into a training set, a validation set, and a test set according to a ratio of 6:2:2; use a holographic sand table with a platform size of not less than 200 cm × 200 cm to demonstrate the three-dimensional model of the block. The operator wears data gloves with a static accuracy of attitude solution Roll / pitch ≤ 1.0 deg to select entity materials from the three-dimensional fusion entity link of the case block output in step S3, transfer them to the spatial unit of the demonstrated block, and use an eye tracker with a line-of-sight tracking accuracy of 0.4° to assist in obtaining the selection tendency of the operator. 2) Conduct machine learning training on data transfer through a deep learning system with 512 GB video memory. Select a data transfer machine learning model with strong generalization performance for the updated regulations, renderings, analysis diagrams, and post-construction photos as the intelligent link model of entity materials through cross-validation and generalization tests, and further combine them to form a cluster of intelligent link models of entity materials.

9. The construction method of an urban spatial atlas information platform with a block form according to claim 8, characterized in that, The steps of establishing the block renewal multi-modal model in S5 are as follows: According to the three case blocks obtained by intelligent matching, use the trained cluster of intelligent link models of entity materials to fuse and link the updated regulations, renderings, analysis diagrams, and post-construction photos of the case blocks with the corresponding spatial units of the target block.

10. An urban spatial atlas information platform with a block form, characterized in that, Constructed using the method according to any one of claims 1-9.

Images