WO2022047960A1 - 一种基于人工智能的商业街区建筑体块生成的方法 - Google Patents

一种基于人工智能的商业街区建筑体块生成的方法 Download PDF

Info

Publication number
WO2022047960A1
WO2022047960A1 PCT/CN2020/124323 CN2020124323W WO2022047960A1 WO 2022047960 A1 WO2022047960 A1 WO 2022047960A1 CN 2020124323 W CN2020124323 W CN 2020124323W WO 2022047960 A1 WO2022047960 A1 WO 2022047960A1
Authority
WO
WIPO (PCT)
Prior art keywords
block
building
dimensional
blocks
tower
Prior art date
Application number
PCT/CN2020/124323
Other languages
English (en)
French (fr)
Inventor
杨俊宴
陈代俊
史宜
朱骁
盛华星
史北祥
Original Assignee
东南大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 东南大学 filed Critical 东南大学
Priority to US17/614,291 priority Critical patent/US20220318443A1/en
Publication of WO2022047960A1 publication Critical patent/WO2022047960A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/088—Non-supervised learning, e.g. competitive learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Definitions

  • the invention relates to the field of artificial intelligence urban design, in particular to a method for generating building blocks of commercial blocks based on artificial intelligence.
  • the purpose of the present invention is to provide a method for generating building blocks of commercial blocks based on artificial intelligence, which provides an automated and intelligent design method for commercial block space design.
  • the system realizes the generation of effective multi-scheme and parameter adjustment after space generation, which effectively improves the design efficiency of planners.
  • a method for generating building blocks of commercial blocks based on artificial intelligence comprising the following steps:
  • the block two-dimensional plane, the block three-dimensional space block, the block public space, the block pedestrian entrance and exit, the block tower building and the podium building are generated in sequence;
  • step S1 includes the following steps:
  • a surveying and mapping drone with a pixel of more than 20 million is used to collect the geospatial information of the target block and a block extending outward with the target block as the center.
  • the picture information in raster format is converted into vector data, and Enter the geographic information platform;
  • the geospatial vector data in step S11 is converted to a unified coordinate system, the spatial geographic coordinates and the projected coordinates are aligned, and a high-precision three-dimensional space sand table is made.
  • geospatial information includes block roads, block buildings, block public spaces and block topography.
  • step S2 includes the following steps:
  • the built-in data collection interface collect the control planning texts and relevant legal norm documents related to the target block, and extract the block design conditions by loading the database component.
  • the block design conditions are preset in the component. , building height limit, building setback, commercial building fire protection distance, pedestrian entrance and exit distance, parking sight distance of surrounding roads;
  • Standardize the block design conditions collected in S21 unify the data format, perform spatial matching with the three-dimensional space sand table, and link to the target block in the form of an attribute table.
  • the building setback includes a podium building setback and a high-rise building setback.
  • step S3 includes the following steps:
  • step S22 Identify the podium building setback index and the surrounding road parking sight distance index in the block design conditions in step S22, shrink the block area contour line inward by a corresponding distance to form the block plane contour line, and fill in the block plane contour line, Form a two-dimensional plane of the block;
  • step S22 Identifying the building height limit index in the block design conditions in step S22, and pulling the block two-dimensional plane in step S31 to the building height limit height to form a block three-dimensional space block;
  • the length and width of the horizontal contour line of the three-dimensional space body of the block in step S32 are determined according to the building level reference modulus unit. If the short side of the block horizontal outline is greater than two building horizontal datum modulus units, then the block horizontal outline is shrunk inward by two building horizontal datum modulus units to form Horizontal outline of public space;
  • the short side of the horizontal contour line of the public space is greater than 1.5 meters, the part enclosed by the horizontal contour line of the public space is cut off from the three-dimensional space body of the block to form the public space of the block; if it is less than 1.5 meters, no public space will be generated;
  • Identify the entrance and exit spacing index in step 22 set pedestrian entrances and exits every other spacing index, with a width of 4 meters, to form a horizontal outline of entrances and exits, in step 33 to generate a public space block three-dimensional space body to cut the part of the entrance and exit outlines to form pedestrians entrances and building blocks of the same height;
  • step S32 Judging the height of the block three-dimensional space body generated in step S32, according to the minimum number of floors of the tower building and the average height of each floor of the tower building, if the height is less than 48 meters, no tower building will be generated, and it is a podium building; if the height is greater than or equal to 48 meters. , take the apex of the block plane contour line in step 31 as the starting point to expand 30-40 meters to both sides to form the tower plane contour line, and identify the high-rise building setback line index in step 22, and retreat the tower plane contour line inward by the corresponding distance;
  • step S4 includes the following steps:
  • the remote sensing images and street view images of different urban blocks are collected, and the unified scale of the remote sensing images and street view images is 1:2000 and the size is 1920*1080, forming a block sample library and generating block morphological characteristic indicators. Then input the morphological characteristic index of the target block, select the blocks with a matching degree of more than 90%, and extract the three-dimensional contour lines of the blocks to form a block three-dimensional contour line training sample library, and the number of training samples is 10,000;
  • a convolutional neural network model is built to identify the concave and convex features of the three-dimensional contour lines of the blocks in the training sample library, and the three-dimensional contour lines of the target blocks are generated.
  • the three-dimensional contour line of the target block is confronted with training, so that the generated sample gradually approaches the training sample, and the block three-dimensional contour line scheme set is output, and then the block building height is generated and the podium building shape in step S35 is optimized;
  • Identify the block floor area ratio index in step 22 conduct interactive verification with the different height building block plans generated in step S36, and adjust the height of the plans that do not meet the floor area ratio requirements until they meet the floor area ratio requirements.
  • the formula for calculating the floor area ratio R is:
  • the H tower is the height of the tower building
  • the S tower is the bottom area of the tower building
  • the H podium is the height of the podium building
  • the S podium is the bottom area of the podium building
  • the S block is the block area
  • block shape characteristic index includes block shape index, block area, building density, plot ratio, and land use property
  • the concave-convex features of the three-dimensional contour line of the block include the position of elevation bumps, the degree of elevation, and the degree of plane concave and convex. building form.
  • step S5 includes the following steps:
  • step S43 For the multi-scheme of the block building block generated in step S43, embed it into a three-dimensional space sand table, and use a 360° holographic display device to perform scheme simulation display and scheme index display;
  • step S43 Print the block plan of the block in step S43 as a paper drawing through a color printer with a resolution of not less than 4800dpi, and export it to SketchUp and AutoCAD aided design software through the built-in file format conversion module for further design and development by planning staff. optimization.
  • the method designed by the present invention can generate multiple plans simultaneously in a short time by setting the design conditions of the commercial block building block plan, which can be reduced from the previous design time of at least two weeks to one day. It can be completed within 12 months, and from the need to invest at least ten designers to generate and compare multiple solutions to the need to invest only one designer to complete the generation of multiple solutions, and it can also be generated from a maximum of twenty effective solutions. At least 100 effective schemes can be generated from the scheme, which effectively reduces labor costs and improves design efficiency;
  • Validity of the scheme The method designed by the present invention learns, identifies and extracts the characteristics of the mature building block scheme of the commercial block by applying the confrontation generation network, and on this basis automatically generates the commercial block building block scheme, which can ensure the generation of the scheme. It avoids the uncontrollable generation of artificial intelligence urban design schemes in the past, and the process of selecting effective schemes from tens of millions of schemes is inefficient and time-consuming. Instead, it directly compares and selects effective multiple schemes to promote The reliability and efficiency of the scheme comparison are improved.
  • FIG. 1 is a schematic flow chart of a method for generating a commercial block building block of the present invention
  • FIG. 2 is a schematic diagram of the planning scope of the commercial block building block scheme of the present invention.
  • Figure 3 is a schematic diagram of the generation of the commercial block building block scheme of the present invention.
  • Fig. 4 is the generation schematic diagram of the two-dimensional plane of the commercial block building block block of the present invention.
  • FIG. 5 is a schematic diagram of the generation of a three-dimensional space block of a commercial block building block block according to the present invention.
  • Fig. 6 is the generation schematic diagram of the public space of the commercial block building block block of the present invention.
  • Fig. 7 is the generation schematic diagram of the pedestrian entrance and exit of the commercial block building block block of the present invention.
  • Fig. 8 is the generation schematic diagram of the commercial block building block block tower building and the podium building of the present invention.
  • FIG. 9 is a schematic diagram illustrating the generation of the building height of the commercial block building block block and the optimization of the podium building shape according to the present invention.
  • a method for generating building blocks of commercial blocks based on artificial intelligence includes the following steps:
  • a surveying and mapping drone with a pixel of more than 20 million is used to collect the geospatial information of relevant blocks (including the target block and a block extending outward with the target block as the center), as shown in Figure 2, through the built-in data acquisition module, the grid
  • the image information in grid format is converted into vector data and entered into the geographic information platform.
  • the geospatial information includes block roads, block buildings, block public spaces, and block topography.
  • a surveying and mapping drone with a pixel of more than 20 million is used, and a certain core business plot and an area of one block that expands outwards (as shown in Figure 2) are set as the flight surveying and mapping area of the surveying and mapping drones .
  • the collected block data is imported into the computer.
  • the Streamline vectorization system is used to convert the picture information in raster format into vector data, and input it into the geographic information platform built by the computer server.
  • the geospatial vector data in step S11 is converted to a unified coordinate system, the spatial geographic coordinates and the projected coordinates are aligned, and a high-precision three-dimensional space sand table is made.
  • the vector data obtained in S11 is converted to a unified coordinate system using a spatial adjustment tool, and the spatial geographic coordinates and the projected coordinates are aligned.
  • the translation extracts the design conditions in various planning documents and local legal codes, performs spatial information registration, and embeds the design conditions in the target block in the form of attribute tables.
  • the block design conditions are preset in the component to extract the content, including: plot ratio, building density , Building height limit, building setback (podium building setback, high-rise building setback), commercial building fire protection distance, pedestrian entrance and exit distance, parking sight distance on surrounding roads.
  • Relevant statutory regulations include "Code for Design of Store Buildings", “Code for Fire Protection in Architectural Design”, and “Code for Planning of Urban Road Intersections”.
  • the "Detailed Controlling Plan for a Commercial Plot in Sichuan”, "Code for Design of Store Buildings”, “Code for Fire Protection of Architectural Design”, and “Code for Planning of Urban Road Intersections” are imported into the computer, and extracted through content recognition tools 07-43
  • the plot ratio of 07-43 plot is 5.2
  • the building density is 35%
  • the building height limit is 150 meters
  • the setback line of high-rise buildings is 15 meters
  • the setback line of podium buildings is 10 meters
  • the fire separation distance of high-rise commercial buildings is 13 meters.
  • the entrance distance is 80 meters
  • the parking sight distance of the surrounding roads is 40 meters.
  • Standardize the block design conditions collected in S21 unify the data format, perform spatial matching with the three-dimensional space sand table, and link to the target block in the form of an attribute table.
  • the block design conditions collected in S21 are standardized, the data format is unified, spatial matching is performed with the three-dimensional space sand table, and the block is linked to the target block in the form of an attribute table.
  • the attribute table information includes the standardised plot ratio, building density, building height limit, building setback (podium building setback, high-rise building setback), commercial building fire protection distance, pedestrian entrance and exit distance, and surrounding roads. Parking sight distance.
  • the block two-dimensional plane, the block three-dimensional space block, the block public space, the block pedestrian entrance and exit, the block tower building and the podium building are sequentially generated.
  • step S22 Identify the building setback index and the parking sight distance index in the block design conditions in step S22, shrink the block area contour line inward by a corresponding distance to form the block plane contour line, and fill the block plane contour line to form a two-dimensional block. flat.
  • the setback line of the podium building in the attribute table in the extraction step S22 is 10 meters and the parking sight distance index is 40 meters, the outline of the block area is shrunk inward by 10 meters, and the street corners are set back according to the parking sight distance.
  • step S22 Identify the building height limit index in the block design conditions in step S22, and pull the block two-dimensional plane in step S31 to the building height limit height to form a block three-dimensional space block.
  • the building height limit in the design conditions of the block in step S22 is identified as 150 meters, and the two-dimensional plane of the block in step S31 is pulled to a height of 150 meters to form a three-dimensional block of the block.
  • the length and width of the horizontal contour line of the three-dimensional space body of the block in step S32 are determined according to the building horizontal reference modulus unit. Less than or equal to two building level datum modulo units (16.8 meters), no public space is generated.
  • the block horizontal contour line is contracted inward by two building horizontal reference modulus units (16.8 meters) to form the public space horizontal contour line. If the short side of the horizontal contour line of the public space is greater than 1.5 meters, the part enclosed by the horizontal contour line of the public space is cut off from the three-dimensional space body of the block to form the public space of the block; if it is less than 1.5 meters, no public space will be generated.
  • the column net modulus (8.4 meters) of the commercial building is used as the building level reference modulus unit, in order to meet the actual column net requirement for parking three vehicles in the underground garage of the commercial building.
  • the short side of the horizontal outline of this block is 110 meters, which is larger than two building horizontal reference modulus units (16.8 meters).
  • the horizontal outline of the block is contracted inward by 16.8 meters to form the horizontal outline of public space.
  • the short side of the horizontal outline of the public space is 76.4 meters and the minimum distance for crowd activities is greater than 1.5 meters. Therefore, the part enclosed by the horizontal outline of the public space is cut from the three-dimensional space body of the block to form the public space of the block.
  • step 22 Identify the entrance and exit spacing index in step 22, and set pedestrian entrances and exits at every spacing index, with a width of 4 meters to form a horizontal outline of the entrance and exit.
  • step 33 the block three-dimensional space volume of the public space is generated, and the outline of the entrance and exit is cut to form pedestrian entrances and building blocks.
  • the index of the distance between the entrances and exits in the identification step 22 is 80 meters
  • the side length of the two-dimensional plane of the local block is between 80 meters and 160 meters
  • an entrance and exit are set at the midpoint of each side with a width of 4 meters.
  • the block three-dimensional space volume of the public space is generated, and the outline of the entrance and exit is cut to form pedestrian entrances and building blocks.
  • step S32 Determine the height of the block three-dimensional space body generated in step S32. According to the minimum floor number requirements of the tower building and the average height of each floor of the tower building, if the height is less than 48 meters, no tower building will be generated, and all are podium buildings. If the height is greater than or equal to 48 meters, take the apex of the horizontal outline of the block in step 31 as the starting point to expand 30-40 meters to both sides to form the outline of the tower, and identify the high-rise building setback index in step 22, and turn the outline of the tower inward Back off the appropriate distance. Further, the number of towers generated is determined.
  • the long side of the horizontal contour line of the block is less than 53 meters, one tower will be generated; if the long side of the horizontal contour line of the block is greater than 73 meters, and the short side is less than 53 meters, 2 towers will be generated; if both sides of the block line are greater than 73 meters, 4 towers will be generated. Further, the generated plane outlines of the towers are screened, and the plane outlines of the towers not facing the road on both sides are removed.
  • 48 meters is the minimum number of floors of the tower, 12 times the average height of each floor of the tower, 4 meters.
  • the height of the three-dimensional space volume of the block generated in step S32 is 150 meters, which is higher than the height limit of the podium of a conventional building of 50 meters, and a tower should be generated.
  • the difference between the two indicators is 5 meters.
  • the fire protection distance of commercial buildings in the extraction step 22 is 13 meters.
  • the side length of this side should be greater than 73 meters (the sum of the minimum fire protection distance and the width of the two towers). All are greater than 73 meters and generate 4 towers. It is determined that the plane outlines of the four towers have both sides facing the road, and will not be removed.
  • the remote sensing images and street view images of different urban blocks are collected, and the unified scale of the remote sensing images and street view images is 1:2000 and the size is 1920*1080, forming a block sample database and generating block morphological characteristic indicators. Further, input the morphological characteristic index of the target block, select the block with a matching degree of more than 90%, and extract the three-dimensional contour line of the block to form a block three-dimensional contour line training sample library, and the number of training samples is 10,000.
  • the block morphological characteristic indicators include block shape index, block area, building density, floor area ratio, and land use properties.
  • remote sensing images and street view images of different urban blocks are collected through the built-in data collection module, and the unified scale is 1:2000 and the size is 1920*1080.
  • the Standard for the Compilation Results of Urban Controlled Detailed Planning five indexes, shape index, block area, building density, floor area ratio, and land use property, which need to be compared, are selected as the morphological characteristic indexes of the sample database. Further, extract the morphological characteristic indicators of the block in this example from the three-dimensional space sand table, select blocks with a matching degree of more than 90% of the five indicators, and extract the three-dimensional contour lines of the blocks to form a training sample library containing 10,000 block three-dimensional contour lines. .
  • a convolutional neural network (CNN) model is built to identify the concave and convex features of the three-dimensional contour lines of the blocks in the training sample library, and the three-dimensional contour lines of the target blocks are generated;
  • the network (GAN) model performs adversarial training on the generated three-dimensional contour lines of the target block, so that the generated samples gradually approach the training samples, and outputs the three-dimensional contour line scheme set of the block, and then generates the block building height and performs the podium building shape in step S35. optimization.
  • the concave-convex feature of the three-dimensional contour line of the block includes the position of the elevation convex point, the elevation concave-convex degree, and the plane concave-convex degree, wherein the elevation convex point position can determine the position of the tower, the elevation concave-convex degree can determine the building height, and the plane concave-convex degree can determine the height of the building.
  • the podium building form can be optimized.
  • a computer is used to build a convolutional neural network (CNN) model, and the block three-dimensional contour line features (elevation bump position, elevation concave-convexity) of the training sample library are analyzed. , plane concave-convex degree) to identify, and construct training sample feature clusters through K-means clustering algorithm.
  • CNN convolutional neural network
  • GAN confrontational generative network
  • the generated three-dimensional contour lines of the target blocks are subjected to adversarial training, so that the generated samples gradually approach the training samples, and a set of three-dimensional contour lines is output, and then the block building heights are generated and analyzed in step S35.
  • the podium building form is optimized. Among them, the position of the elevation bump can determine the location of the tower, the elevation can determine the height of the building, and the plane concave can optimize the shape of the podium.
  • step S22 Identify the block plot ratio index in step S22, conduct interactive verification with the different height building block schemes generated in step S36, and adjust the height of the scheme that does not meet the plot ratio requirements until the plot ratio requirements are met.
  • the block plot ratio index in the identification step S22 is 5.2, and the plot ratio of the generated solution is calculated by the plot ratio calculation formula and compared with the index.
  • the genetic algorithm is used to adjust the height of the scheme that exceeds the plot ratio of 5.2, and iteratively generates until the plot ratio requirement is met.
  • the formula for calculating the volume ratio R is:
  • the H tower is the height of the tower building
  • the S tower is the bottom area of the tower building
  • the H podium is the height of the podium building
  • the S podium is the bottom area of the podium building
  • the S block is the block area
  • step S43 For the multi-scheme of the block building block generated in step S43, it is embedded in the three-dimensional space sand table, and the 360° holographic display equipment is used for scheme simulation display and scheme index display.
  • a computer server is used to import the multi-scheme of the block building blocks generated in step S43 into the geographic information platform, and the spatial adjustment tool is used to correct the coordinates and elevation of the case model, so that the case model is accurately embedded in the block generated in step S12.
  • 3D space sand table Use a 360° holographic display device with a display area of more than 150 cm * 150 cm to display 3D models and indicators, and recognize the user's action instructions through the action recognition module to realize the display operations such as moving, zooming and rotating the sand table content.
  • step S43 Print the block plan of the block in step S43 as a paper drawing through a color printer with a resolution of not less than 4800dpi, and export it to the auxiliary design software such as SketchUp and AutoCAD through the built-in file format conversion module for the planning staff to further Design and optimize.
  • auxiliary design software such as SketchUp and AutoCAD

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Civil Engineering (AREA)
  • Mathematical Optimization (AREA)
  • Medical Informatics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Architecture (AREA)
  • Computational Mathematics (AREA)
  • Structural Engineering (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Graphics (AREA)
  • Multimedia (AREA)
  • Processing Or Creating Images (AREA)

Abstract

一种基于人工智能的商业街区建筑体块生成的方法，涉及人工智能城市设计领域。该方法首先通过获取目标街区及周围街区的地理信息数据，构建出三维空间沙盘；其次通过转译提取各类规划文件和当地法定规范中的设计条件，生成街区三维建筑体块；然后构建街区三维轮廓线训练样本库，通过加载机器学习模型生成街区三维建筑高度并对建筑形态进行优化，生成街区建筑体块多方案；最后使用全息展示设备进行方案模拟展示和方案指标显示，并输出方案。该方法针对上一代人工智能技术生成大量无效方案的问题，实现了基于人工智能有效多方案的生成，解决了上一代人工智能有效方案筛选过程中耗时长、人力投入大等难题，提高了规划师设计效率。

Description

一种基于人工智能的商业街区建筑体块生成的方法 技术领域
本发明涉及人工智能城市设计领域，具体的是一种基于人工智能的商业街区建筑体块生成的方法。
背景技术
在城市设计过程中，需要在开发指标既定的情况下，对街区三维空间形态进行多种可能性的推敲。不仅需要配合开发总量在多个街区之间的分布设定与调整，同时，也需要推敲各街区开发指标对应下三维建筑群体形态的多种可能性，实现街区所在区域内的城市形态控制。在以往的设计过程中，这项从开发指标到建筑三维形态的生成工作都是由规划设计人员手动完成，当需要推敲方案的多种可能性时，则需要投入大量的人员和时间来完成这项工作。
随着人工智能技术逐步应用于城市设计领域，实现了从开发指标到三维建筑群体形态的自动生成工作，但由于传统智能生成手段的不可控性，往往产生过多无效方案，无法满足当前城市设计的实际需求。
发明内容
为解决上述背景技术中提到的不足，本发明的目的在于提供一种基于人工智能的商业街区建筑体块生成的方法，该方法为商业街区空间设计提供一种自动化，智能化的设计方法与系统，实现了有效多方案的生成和空间生成后的参数调整，有效提高了规划师设计效率。
本发明的目的可以通过以下技术方案实现：
一种基于人工智能的商业街区建筑体块生成的方法，包括以下步骤：
S1、街区基础数据采集与三维空间沙盘构建
获取目标街区及其周围街区的地理信息数据，进行坐标系的统一，构建出三维空间沙盘；
S2、街区设计条件提取与空间匹配
转译提取各类规划文件和当地法定规范中的设计条件并进行空间信息配准，将设计条件以属性表形式嵌入目标街区；
S3、基于设计条件的街区三维建筑体块的生成
基于所提取的设计条件，依次生成街区二维平面、街区三维空间体块、街区公共空间、街区人行出入口以及街区塔楼建筑和裙房建筑；
S4、基于机器学习的街区三维建筑高度生成和建筑形态优化
构建街区三维轮廓线训练样本库，通过加载机器学习模型生成街区三维建筑高度并对建筑形态进行优化，生成街区建筑体块多方案；
S5、人机交互方案展示与方案输出
使用全息展示设备进行方案模拟展示和方案指标显示，并输出方案。
进一步地，所述步骤S1包括以下步骤：
S11、街区地理信息数据采集
采用搭载像素为2000万以上的测绘无人机采集目标街区及以目标街区为中心向外扩展一个街区的地理空间信息，通过内置数据采集模块，将栅格格式的图片信息转换为矢量数据，并录入地理信息平台；
S12、街区三维空间沙盘构建
在地理信息平台中，将步骤S11中的地理空间矢量数据转换至统一坐标系，进行空间地理坐标与投影坐标对位，并制成高精度的三维空间沙盘。
进一步地，所述地理空间信息包括街区道路，街区建筑，街区公共空间以 及街区地形地貌。
进一步地，所述步骤S2包括以下步骤：
S21、街区设计条件提取
通过内置数据采集接口，采集涉及目标街区的控规规划文本和相关法定规范文件，并通过加载数据库组件提取街区设计条件，组件中预先设定街区设计条件的提取内容，包括：容积率、建筑密度、建筑限高、建筑退线、商业建筑消防间距、人行出入口间距、周边道路的停车视距；
S22、街区设计条件的标准化处理和空间匹配
将S21中采集的街区设计条件进行标准化处理，统一数据格式，与三维空间沙盘进行空间匹配，并以属性表的形式链接至目标街区。
进一步地，所述建筑退线包括裙房建筑退线和高层建筑退线。
进一步地，所述相关法定规范包括《商店建筑设计规范》、《建筑设计防火规范》和《城市道路交叉口规划规范》。
进一步地，所述步骤S3包括以下步骤：
S31、街区二维平面的生成
识别步骤S22中街区设计条件中的裙房建筑退线指标和周边道路停车视距指标，将街区范围轮廓线向内收缩相应距离，形成街区平面轮廓线，通过对街区平面轮廓线内进行填充，形成街区二维平面；
S32、街区三维空间体块的生成
识别步骤S22中街区设计条件中建筑限高指标，将步骤S31中的街区二维平面拉升至建筑限高高度，形成街区三维空间体块；
S33、街区公共空间的生成
以商业建筑柱网模数为建筑水平基准模数单位，根据建筑水平基准模数单 位对步骤S32中街区三维空间体的水平轮廓线长宽进行判定，若街区水平轮廓线长边小于或等于两个建筑水平基准模数单位，则不生成公共空间；若街区水平轮廓线短边大于两个建筑水平基准模数单位，则将街区水平轮廓线向内收缩两个建筑水平基准模数单位，形成公共空间水平轮廓线；
若公共空间水平轮廓线短边大于1.5米，则从街区三维空间体中切掉公共空间水平轮廓线围合的部分，形成街区公共空间；若小于1.5米，则不生成公共空间；
S34、街区人行出入口的生成
识别步骤22中出入口间距指标，每隔一间距指标设置人行出入口，宽度为4米，形成出入口水平轮廓线，在步骤33中生成公共空间的街区三维空间体对出入口轮廓线部分进行切割，形成人行出入口和同一高度的建筑体块；
S35、街区塔楼建筑和裙房建筑的生成
对步骤S32生成的街区三维空间体高度进行判定，根据塔楼最低层数要求和塔楼每层平均高度，若高度小于48米，则不生成塔楼建筑，均为裙房建筑；若高度大于等于48米，以步骤31中街区平面轮廓线顶点为起点向两边扩展30-40米形成塔楼平面轮廓线，并识别步骤22中的高层建筑退线指标，将塔楼平面轮廓线向内退让相应距离；
对塔楼生成数量进行判定，根据最小消防间距和塔楼宽度，若街区水平轮廓线长边小于53米，则生成1个塔楼；若街区水平轮廓线长边大于73米，且短边小于53米，生成2个塔楼；若街区线两边均大于73米，则生成4个塔楼，同时，对生成的塔楼平面轮廓线进行筛选，去除两边不临路的塔楼平面轮廓线。
进一步地，所述步骤S4包括以下步骤：
S41、构建街区三维轮廓线训练样本库
通过内置数据采集模块，对不同城市街区遥感影像和街景影像进行采集，并将遥感影像和街景影像图片统一比例尺为1：2000，尺寸为1920*1080，形成街区样本库并生成街区形态特征指标，进而输入目标街区形态特征指标，选取匹配度达90％以上的街区并提取街区三维轮廓线构成街区三维轮廓线训练样本库，训练样本数量为10000个；
S42、街区建筑高度的生成和建筑形态的优化
通过步骤S41生成的街区三维轮廓线训练样本库，搭建卷积神经网络模型对训练样本库街区三维轮廓线的凹凸特征进行识别，生成目标街区三维轮廓线，然后，通过构建对抗生成网络模型对生成的目标街区三维轮廓线进行对抗训练，使生成样本逐渐逼近训练样本，并输出街区三维轮廓线方案集，进而生成街区建筑高度并对步骤S35中的裙房建筑形态进行优化；
S43、街区建筑体块方案生成
识别步骤22中街区容积率指标，与步骤S36生成的不同高度建筑体块方案进行交互验证，对不满足容积率要求的方案进行高度调整，直到满足容积率要求为止，容积率R计算公式为：
Figure PCTCN2020124323-appb-000001
其中，H 塔楼为塔楼建筑高度，S 塔楼为塔楼建筑底面积，H 裙房为裙房建筑高度，S 裙房为裙房建筑底面积，S 街区为街区面积；
进一步地，所述街区形态特征指标包括街区形状指数，街区面积，建筑密度，容积率，用地性质；
所述街区三维轮廓线的凹凸特征包括立面凸点位置、立面凹凸度、平面凹凸度，立面凸点位置可以确定塔楼位置，立面凹凸度可以确定建筑高度，平面凹凸度可以优化裙房建筑形态。
进一步地，所述步骤S5包括以下步骤：
S51、街区三维建筑体块设计方案可视化
对于步骤S43中生成的街区建筑体块多方案，嵌入至三维空间沙盘，使用360°全息展示设备进行方案模拟展示和方案指标显示；
S52、方案结果输出
将步骤S43中街区建筑体块方案，通过分辨率不小于4800dpi彩色打印机将其打印为纸质图纸，并通过内置文件格式转化模块，导出到SketchUp、AutoCAD辅助设计软件，供规划工作人员进一步设计和优化。
本发明的有益效果：
1、过程高效性：本发明设计的方法通过设定商业街区建筑体块方案的设计条件，能够在短时间内同时生成多个方案，可以从以往的需要至少两周的设计时间缩减至一天之内完成，并且从需要投入至少十名设计人员来进行多方案的生成和比选到只需要投入一名设计人员就可以完成多方案的生成工作，同时也可以从最多只能生成二十个有效方案到至少能生成一百个有效方案，有效地减少了人力成本，提高了设计效率；
2、方案有效性：本发明设计的方法通过应用对抗生成网络，对商业街区成熟建筑体块方案特征的学习、识别和提取，在此基础上自动生成商业街区建筑体块方案，可以保证方案生成的有效性，避免了以往人工智能城市设计方案生成的不可控性，出现从几千万个方案中筛选出有效方案过程低效、费时的情况，而是直接对有效多方案进行比选，促进了方案比选的可靠性和高效性。
附图说明
下面结合附图对本发明作进一步的说明。
图1是本发明商业街区建筑体块生成的方法流程示意图；
图2是本发明商业街区建筑体块方案规划范围示意图；
图3是本发明商业街区建筑体块方案生成示意图；
图4是本发明商业街区建筑体块街区二维平面的生成示意图；
图5是本发明商业街区建筑体块街区三维空间体块的生成示意图；
图6是本发明商业街区建筑体块街区公共空间的生成示意图；
图7是本发明商业街区建筑体块街区人行出入口的生成示意图；
图8是本发明商业街区建筑体块街区塔楼建筑和裙房建筑的生成示意图；
图9是本发明商业街区建筑体块街区建筑高度的生成和裙房建筑形态的优化示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其它实施例，都属于本发明保护的范围。
本发明将结合四川省某核心商务地块的街区建筑体块生成案例和附图来详细地说明本发明的技术方案。
一种基于人工智能的商业街区建筑体块生成的方法，如图1所示，包括以下步骤：
S1、街区基础数据采集与三维空间沙盘构建
获取目标街区及其周围街区的地理信息数据，进行坐标系的统一，构建出三维空间沙盘。
S11、街区地理信息数据采集
采用搭载像素为2000万以上的测绘无人机采集相关街区(包括目标街区及 以目标街区为中心向外扩展一个街区)的地理空间信息，如图2所示，通过内置数据采集模块，将栅格格式的图片信息转换为矢量数据，并录入地理信息平台。其中，地理空间信息包括街区道路，街区建筑，街区公共空间以及街区地形地貌。
本实施例中，使用搭载像素为2000万以上的测绘无人机，将某核心商务地块及向外扩展的一个街区的区域(如图2所示)设为测绘无人机的飞行测绘区域。操作测绘无人机在飞行测绘区域内飞行并采集控规调整相关地块的地理空间信息，包括街区道路，街区建筑，街区公共空间以及街区地形地貌信息。回收测绘无人机后，将其中采集的街区数据导入计算机。对于导入的数据利用Streamline矢量化系统，将栅格格式的图片信息转换为矢量数据，录入计算机服务器搭建的地理信息平台中。
S12、街区三维空间沙盘构建
在地理信息平台中，将步骤S11中的地理空间矢量数据转换至统一坐标系，进行空间地理坐标与投影坐标对位，并制成高精度的三维空间沙盘。
本实施例中，对S11中获得的矢量数据使用spatial adjustment(空间校正)工具，转换至统一坐标系，进行空间地理坐标与投影坐标对位。
S2、街区设计条件提取与空间匹配
转译提取各类规划文件和当地法定规范中的设计条件并进行空间信息配准，将设计条件以属性表形式嵌入目标街区。
S21、街区设计条件提取
通过内置数据采集接口，采集涉及目标街区的控规规划文本和相关法定规范文件，并通过加载数据库组件提取街区设计条件，组件中预先设定街区设计条件的提取内容，包括：容积率、建筑密度、建筑限高、建筑退线(裙房建筑 退线、高层建筑退线)、商业建筑消防间距、人行出入口间距、周边道路停车视距。相关法定规范包括《商店建筑设计规范》、《建筑设计防火规范》、《城市道路交叉口规划规范》。
本实施例中，向计算机导入《四川某商务地块控制性详细规划》《商店建筑设计规范》、《建筑设计防火规范》、《城市道路交叉口规划规范》，通过内容识别工具提取07-43地块的容积率、建筑密度、建筑限高、建筑退线(裙房建筑退线、高层建筑退线)、商业建筑消防间距、人行出入口间距、周边道路停车视距，并导入至地理信息平台中。其中，07-43地块的容积率为5.2、建筑密度为35％、建筑限高为150米、高层建筑退线为15米、裙房建筑退线为10米，高层商业建筑消防间距为13米，入口间距为80米，周边道路的停车视距为40米。
S22、街区设计条件的标准化处理和空间匹配
将S21中采集的街区设计条件进行标准化处理，统一数据格式，与三维空间沙盘进行空间匹配，并以属性表的形式链接至目标街区。
本实施例中，将S21中采集的街区设计条件进行标准化处理，统一数据格式，与三维空间沙盘进行空间匹配，并以属性表的形式链接至目标街区。其中属性表信息包括本实例地块标准化后的容积率、建筑密度、建筑限高、建筑退线(裙房建筑退线、高层建筑退线)、商业建筑消防间距、人行出入口间距、周边道路的停车视距。
S3、基于设计条件的街区三维建筑体块的生成
基于所提取的设计条件，依次生成街区二维平面、街区三维空间体块、街区公共空间、街区人行出入口以及街区塔楼建筑和裙房建筑。
S31、街区二维平面的生成
识别步骤S22中街区设计条件中的建筑退线指标和停车视距指标，将街区范围轮廓线向内收缩相应距离，形成街区平面轮廓线，通过对街区平面轮廓线内进行填充，形成街区二维平面。
本实施例中，提取步骤S22属性表中的裙房建筑退线为10米和停车视距指标为40米，将街区范围轮廓线向内收缩10米、街角根据停车视距向内退让视距三角形，形成街区平面轮廓线，通过线面工具对街区平面轮廓线内进行填充，形成街区二维平面。
S32、街区三维空间体块的生成
识别步骤S22中街区设计条件中建筑限高指标，将步骤S31中的街区二维平面拉升至建筑限高高度，形成街区三维空间体块。
本实施例中，识别步骤S22中街区设计条件中建筑限高为150米，将步骤S31中的街区二维平面拉升至150米高度，形成街区三维空间体块。
S33、街区公共空间的生成
以商业建筑柱网模数(8.4米)为建筑水平基准模数单位，根据建筑水平基准模数单位对步骤S32中街区三维空间体的水平轮廓线长宽进行判定，若街区水平轮廓线长边小于或等于两个建筑水平基准模数单位(16.8米)，则不生成公共空间。
若街区水平轮廓线短边大于两个建筑水平基准模数单位(16.8米)，则将街区水平轮廓线向内收缩两个建筑水平基准模数单位(16.8米)，形成公共空间水平轮廓线。若公共空间水平轮廓线短边大于1.5米，则从街区三维空间体中切掉公共空间水平轮廓线围合的部分，形成街区公共空间；若小于1.5米，则不生成公共空间。
本实施例中，以商业建筑柱网模数(8.4米)为建筑水平基准模数单位，是 为了满足商业建筑地下车库停放三辆车的实际柱网需求。本街区水平轮廓线短边为110米，大于两个建筑水平基准模数单位(16.8米)，将街区水平轮廓线向内收缩16.8米，形成公共空间水平轮廓线。公共空间水平轮廓线短边为76.4米大于1.5米的人群活动最小间距，故从街区三维空间体中切掉公共空间水平轮廓线围合的部分，形成街区公共空间。
S34、街区人行出入口的生成
识别步骤22中出入口间距指标，每隔间距指标设置人行出入口，宽度为4米，形成出入口水平轮廓线。在步骤33中生成公共空间的街区三维空间体对出入口轮廓线部分进行切割，形成人行出入口和建筑体块。
本实施例中，识别步骤22中出入口间距指标为80米，本地块街区二维平面边长在80米到160米之间，在每边中点处设置一个出入口，宽度为4米，形成出入口水平轮廓线。在步骤33中生成公共空间的街区三维空间体对出入口轮廓线部分进行切割，形成人行出入口和建筑体块。
S35、街区塔楼建筑和裙房建筑的生成
对步骤S32生成的街区三维空间体高度进行判定，根据塔楼最低层数要求和塔楼每层平均高度，若高度小于48米，则不生成塔楼建筑，均为裙房建筑。若高度大于等于48米，以步骤31中街区水平轮廓线顶点为起点向两边扩展30-40米形成塔楼平面轮廓线，并识别步骤22中的高层建筑退线指标，将塔楼平面轮廓线向内退让相应距离。进一步的，对塔楼生成数量进行判定，根据最小消防间距和塔楼宽度，若街区水平轮廓线长边小于53米，则生成1个塔楼；若街区水平轮廓线长边大于73米，且短边小于53米，生成2个塔楼；若街区线两边均大于73米，则生成4个塔楼。进一步的，对生成的塔楼平面轮廓线进行筛选，去除两边不临路的塔楼平面轮廓线。
其中，48米为塔楼的最低层数12乘以塔楼每层平均高度4米。
本实施例中，步骤S32生成的街区三维空间体高度为150米，高于50米的常规建筑裙房高度界限，应当生成塔楼。以街区水平轮廓线顶点为起点向两边扩展30米形成塔楼平面轮廓线，并识别步骤22中的群房退线指标为10米、高层建筑退线指标为15米，将塔楼平面轮廓线向内再退让两指标之差5米。进一步的，提取步骤22中的商业建筑消防间距为13米，若两塔楼同边则此边边长应大于73米(最小消防间距和两塔楼宽度之和)，本实施例街区水平轮廓线两边均大于73米，生成4个塔楼。判定4个塔楼平面轮廓线都有两边临路，不进行去除。
S4、基于机器学习的街区三维建筑高度生成和建筑形态优化
构建街区三维轮廓线训练样本库，通过加载机器学习模型生成街区三维建筑高度并对裙房建筑形态进行优化，生成街区建筑体块多方案。
S41、构建街区三维轮廓线训练样本库
通过内置数据采集模块，对不同城市街区遥感影像和街景影像进行采集，并将遥感影像和街景影像图片统一比例尺为1：2000，尺寸为1920*1080，形成街区样本库并生成街区形态特征指标。进一步的，输入目标街区形态特征指标，选取匹配度达90％以上的街区并提取街区三维轮廓线构成街区三维轮廓线训练样本库，训练样本数量为10000个。所述街区形态特征指标包括街区形状指数，街区面积，建筑密度，容积率，用地性质。
本实施例中，通过内置数据采集模块，对不同城市街区遥感影像和街景影像进行采集，并统一比例尺为1：2000，尺寸为1920*1080。根据《城市控制性详细规划编制成果规范标准》，选择需进行对照的形状指数，街区面积，建筑密度，容积率，用地性质5个指标作为样本库形态特征指标。进一步的，从三 维空间沙盘中提取本实例街区的形态特征指标，选取5项指标匹配度都达90％以上的街区，提取街区三维轮廓线，构成包含10000个街区三维轮廓线实例的训练样本库。
S42、街区建筑高度的生成和裙房建筑形态的优化
通过步骤S41生成的街区三维轮廓线训练样本库，搭建卷积神经网络(CNN)模型对训练样本库街区三维轮廓线的凹凸特征进行识别，生成目标街区三维轮廓线；进一步的，通过构建对抗生成网络(GAN)模型对生成的目标街区三维轮廓线进行对抗训练，使生成样本逐渐逼近训练样本，并输出街区三维轮廓线方案集，进而生成街区建筑高度并对步骤S35中的裙房建筑形态进行优化。其中，所述街区三维轮廓线凹凸特征包括立面凸点位置、立面凹凸度、平面凹凸度，其中，立面凸点位置可以确定塔楼位置，立面凹凸度可以确定建筑高度，平面凹凸度可以优化裙房建筑形态。
本实施例中，利用步骤S41生成的街区三维轮廓线训练样本库，使用计算机搭建卷积神经网络(CNN)模型，对训练样本库街区三维轮廓线特征(立面凸点位置、立面凹凸度、平面凹凸度)进行识别，通过K-means聚类算法构建训练样本特征聚类。进一步的，通过构建对抗生成网络(GAN)模型对生成的目标街区三维轮廓线进行对抗训练，使生成样本逐渐逼近训练样本，并输出三维轮廓线方案集，进而生成街区建筑高度并对步骤S35中的裙房建筑形态进行优化。其中，立面凸点位置可以确定塔楼位置，立面凹凸度可以确定建筑高度，平面凹凸度可以优化裙房建筑形态。
S43、街区建筑体块方案生成
识别步骤S22中街区容积率指标，与步骤S36生成的不同高度建筑体块方案进行交互验证，对不满足容积率要求的方案进行高度调整，直到满足容积率 要求为止。
本实施例中，识别步骤S22中街区容积率指标为5.2，通过容积率计算公式测算生成方案的容积率并与指标进行比对。对于超过容积率5.2指标的方案使用遗传算法进行高度调整，不断迭代生成直到满足容积率要求为止。容积率R计算公式为：
Figure PCTCN2020124323-appb-000002
其中，H 塔楼为塔楼建筑高度，S 塔楼为塔楼建筑底面积，H 裙房为裙房建筑高度，S 裙房为裙房建筑底面积，S 街区为街区面积；
S5、人机交互方案展示与方案输出
使用全息展示设备进行方案模拟展示和方案指标显示，并输出方案。
S51、街区三维建筑体块设计方案可视化
对于步骤S43中生成的街区建筑体块多方案，嵌入至三维空间沙盘，使用360°全息展示设备进行方案模拟展示和方案指标显示。
本实例中，使用计算机服务器，将步骤S43中生成的街区建筑体块多方案导入地理信息平台，使用spatialadjustment(空间校正)工具矫正案例模型的坐标和高程，使案例模型准确嵌入步骤S12中生成的三维空间沙盘。使用展示区域面积在150厘米*150厘米以上的360°全息展示设备进行三维模型和指标显示，通过动作识别模块识别使用者的动作指令，实现对沙盘内容的移动、缩放和旋转等展示操作。
S52、方案结果输出
将步骤S43中街区建筑体块方案，通过分辨率不小于4800dpi的彩色打印机将其打印为纸质图纸，并通过内置文件格式转化模块，导出到SketchUp、AutoCAD等辅助设计软件，供规划工作人员进一步设计和优化。
以上显示和描述了本发明的基本原理、主要特征和本发明的优点。本行业的技术人员应该了解，本发明不受上述实施例的限制，上述实施例和说明书中描述的只是说明本发明的原理，在不脱离本发明精神和范围的前提下，本发明还会有各种变化和改进，这些变化和改进都落入要求保护的本发明范围内。

Claims (10)

  1. 一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，包括以下步骤：
    S1、街区基础数据采集与三维空间沙盘构建
    获取目标街区及其周围街区的地理信息数据，进行坐标系的统一，构建出三维空间沙盘；
    S2、街区设计条件提取与空间匹配
    转译提取各类规划文件和当地法定规范中的设计条件并进行空间信息配准，将设计条件以属性表形式嵌入目标街区；
    S3、基于设计条件的街区三维建筑体块的生成
    基于所提取的设计条件，依次生成街区二维平面、街区三维空间体块、街区公共空间、街区人行出入口以及街区塔楼建筑和裙房建筑；
    S4、基于机器学习的街区三维建筑高度生成和建筑形态优化
    构建街区三维轮廓线训练样本库，通过加载机器学习模型生成街区三维建筑高度并对建筑形态进行优化，生成街区建筑体块多方案；
    S5、人机交互方案展示与方案输出
    使用全息展示设备进行方案模拟展示和方案指标显示，并输出方案。
  2. 根据权利要求1所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述步骤S1包括以下步骤：
    S11、街区地理信息数据采集
    采用搭载像素为2000万以上的测绘无人机采集目标街区及以目标街区为中心向外扩展一个街区的地理空间信息，通过内置数据采集模块，将栅格格式的图片信息转换为矢量数据，并录入地理信息平台；
    S12、街区三维空间沙盘构建
    在地理信息平台中，将步骤S11中的地理空间矢量数据转换至统一坐标系，进行空间地理坐标与投影坐标对位，并制成高精度的三维空间沙盘。
  3. 根据权利要求2所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述地理空间信息包括街区道路，街区建筑，街区公共空间以及街区地形地貌。
  4. 根据权利要求1所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述步骤S2包括以下步骤：
    S21、街区设计条件提取
    通过内置数据采集接口，采集涉及目标街区的控规规划文本和相关法定规范文件，并通过加载数据库组件提取街区设计条件，组件中预先设定街区设计条件的提取内容，包括：容积率、建筑密度、建筑限高、建筑退线、商业建筑消防间距、人行出入口间距、周边道路的停车视距；
    S22、街区设计条件的标准化处理和空间匹配
    将S21中采集的街区设计条件进行标准化处理，统一数据格式，与三维空间沙盘进行空间匹配，并以属性表的形式链接至目标街区。
  5. 根据权利要求4所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述建筑退线包括裙房建筑退线和高层建筑退线。
  6. 根据权利要求4所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述相关法定规范包括《商店建筑设计规范》、《建筑设计防火规范》和《城市道路交叉口规划规范》。
  7. 根据权利要求1所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述步骤S3包括以下步骤：
    S31、街区二维平面的生成
    识别步骤S22中街区设计条件中的裙房建筑退线指标和周边道路停车视距指标，将街区范围轮廓线向内收缩相应距离，形成街区平面轮廓线，通过对街区平面轮廓线内进行填充，形成街区二维平面；
    S32、街区三维空间体块的生成
    识别步骤S22中街区设计条件中建筑限高指标，将步骤S31中的街区二维平面拉升至建筑限高高度，形成街区三维空间体块；
    S33、街区公共空间的生成
    以商业建筑柱网模数为建筑水平基准模数单位，根据建筑水平基准模数单位对步骤S32中街区三维空间体的水平轮廓线长宽进行判定，若街区水平轮廓线长边小于或等于两个建筑水平基准模数单位，则不生成公共空间；若街区水平轮廓线短边大于两个建筑水平基准模数单位，则将街区水平轮廓线向内收缩两个建筑水平基准模数单位，形成公共空间水平轮廓线；
    若公共空间水平轮廓线短边大于1.5米，则从街区三维空间体中切掉公共空间水平轮廓线围合的部分，形成街区公共空间；若小于1.5米，则不生成公共空间；
    S34、街区人行出入口的生成
    识别步骤22中出入口间距指标，每隔一间距指标设置人行出入口，宽度为4米，形成出入口水平轮廓线，在步骤33中生成公共空间的街区三维空间体对出入口轮廓线部分进行切割，形成人行出入口和同一高度的建筑体块；
    S35、街区塔楼建筑和裙房建筑的生成
    对步骤S32生成的街区三维空间体高度进行判定，根据塔楼最低层数要求和塔楼每层平均高度，若高度小于48米，则不生成塔楼建筑，均为多层裙房建 筑；若高度大于等于48米，以步骤31中街区平面轮廓线顶点为起点向两边扩展30-40米形成塔楼平面轮廓线，并识别步骤22中的高层建筑退线指标，将塔楼平面轮廓线向内退让相应距离；
    对塔楼生成数量进行判定，根据最小消防间距和塔楼宽度，若街区水平轮廓线长边小于53米，则生成1个塔楼；若街区水平轮廓线长边大于73米，且短边小于53米，生成2个塔楼；若街区线两边均大于73米，则生成4个塔楼，同时，对生成的塔楼平面轮廓线进行筛选，去除两边不临路的塔楼平面轮廓线。
  8. 根据权利要求1所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述步骤S4包括以下步骤：
    S41、构建街区三维轮廓线训练样本库
    通过内置数据采集模块，对不同城市街区遥感影像和街景影像进行采集，并将遥感影像和街景影像图片统一比例尺为1：2000，尺寸为1920*1080，形成街区样本库并生成街区形态特征指标，进而输入目标街区形态特征指标，选取匹配度达90％以上的街区并提取街区三维轮廓线构成街区三维轮廓线训练样本库，训练样本数量为10000个；
    S42、街区建筑高度的生成和建筑形态的优化
    通过步骤S41生成的街区三维轮廓线训练样本库，搭建卷积神经网络模型对训练样本库街区三维轮廓线的凹凸特征进行识别，生成目标街区三维轮廓线，然后，通过构建对抗生成网络模型对生成的目标街区三维轮廓线进行对抗训练，使生成样本逐渐逼近训练样本，并输出街区三维轮廓线方案集，进而生成街区建筑高度并对步骤S35中的裙房建筑形态进行优化；
    S43、街区建筑体块方案生成
    识别步骤22中街区容积率指标，与步骤S36生成的不同高度建筑体块方案 进行交互验证，对不满足容积率要求的方案进行高度调整，直到满足容积率要求为止，容积率R计算公式为：
    Figure PCTCN2020124323-appb-100001
    其中，H 塔楼为塔楼建筑高度，S 塔楼为塔楼建筑底面积，H 裙房为裙房建筑高度，S 裙房为裙房建筑底面积，S 街区为街区面积。
  9. 根据权利要求8所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述街区形态特征指标包括街区形状指数，街区面积，建筑密度，容积率，用地性质；
    所述街区三维轮廓线的凹凸特征包括立面凸点位置、立面凹凸度、平面凹凸度，立面凸点位置用以确定塔楼位置，立面凹凸度用以确定建筑高度，平面凹凸度用以优化裙房建筑形态。
  10. 根据权利要求1所述的一种基于人工智能的商业街区建筑体块生成的方法，其特征在于，所述步骤S5包括以下步骤：
    S51、街区三维建筑体块设计方案可视化
    对于步骤S43中生成的街区建筑体块多方案，嵌入至三维空间沙盘，使用360°全息展示设备进行方案模拟展示和方案指标显示；
    S52、方案结果输出
    将步骤S43中街区建筑体块方案，通过分辨率不小于4800dpi彩色打印机将其打印为纸质图纸，并通过内置文件格式转化模块，导出到SketchUp、AutoCAD辅助设计软件，供规划工作人员进一步设计和优化。
PCT/CN2020/124323 2020-09-04 2020-10-28 一种基于人工智能的商业街区建筑体块生成的方法 WO2022047960A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/614,291 US20220318443A1 (en) 2020-09-04 2020-10-28 Ai-based method for generating building block in commercial district

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010923491.2 2020-09-04
CN202010923491.2A CN112052503B (zh) 2020-09-04 2020-09-04 一种基于人工智能的商业街区建筑体块生成的方法

Publications (1)

Publication Number Publication Date
WO2022047960A1 true WO2022047960A1 (zh) 2022-03-10

Family

ID=73608265

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/124323 WO2022047960A1 (zh) 2020-09-04 2020-10-28 一种基于人工智能的商业街区建筑体块生成的方法

Country Status (3)

Country Link
US (1) US20220318443A1 (zh)
CN (1) CN112052503B (zh)
WO (1) WO2022047960A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114898215A (zh) * 2022-06-09 2022-08-12 西南交通大学 一种声屏障自动布置方法

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019205069A1 (en) * 2018-04-27 2019-10-31 Beijing Didi Infinity Technology And Development Co., Ltd. Systems and methods for updating 3d model of building
CN112651059B (zh) * 2020-12-18 2022-11-08 东南大学 一种基于人工智能的控规地块城市设计多方案生成方法
CN113033484B (zh) * 2021-04-21 2022-11-22 河北工程大学 一种面向无人机应急网络部署的城市分类方法
CN113822526A (zh) * 2021-08-02 2021-12-21 东南大学建筑设计研究院有限公司 控制性详细规划阶段商办类用地地块指标适配性验核方法
CN113657669B (zh) * 2021-08-17 2023-11-14 东南大学 一种街区空间编码图的智能建构方法、系统及存储介质
CN113963104B (zh) * 2021-09-08 2024-04-26 华南理工大学 基于gan模型生成的布局图像的三维立体模型生成方法
CN113987637B (zh) * 2021-10-25 2022-07-01 清华大学 基于生成对抗网络的楼盖结构设计方法和装置
CN114218640B (zh) * 2021-12-15 2023-10-31 东南大学 一种基于人工智能的假山体量设计方法
CN115329691B (zh) * 2022-08-01 2024-03-12 上海师范大学 一种基于cfd与gis的超大城市风环境模拟方法
CN115292789B (zh) * 2022-08-12 2023-04-28 东南大学建筑设计研究院有限公司 城市设计中基于形态类型的建筑体量数字化生成方法
US11850005B1 (en) * 2022-10-27 2023-12-26 Mammen Thomas Use of immersive real-time metaverse and avatar and 3-D hologram for medical and veterinary applications using spatially coordinated multi-imager based 3-D imaging
CN115492413A (zh) * 2022-10-31 2022-12-20 温州理工学院 一种智能砌墙机器人的控制系统及方法
WO2024100866A1 (ja) * 2022-11-10 2024-05-16 日本電信電話株式会社 学習装置、建造物領域分類装置、学習方法、建造物領域分類方法、及びプログラム
CN117437364B (zh) * 2023-12-20 2024-04-26 深圳大学 基于残缺点云数据的建筑三维结构提取方法和装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104732591A (zh) * 2015-03-10 2015-06-24 中国科学院遥感与数字地球研究所 三维虚拟城市的自动生成方法
CN108265820A (zh) * 2017-12-28 2018-07-10 北京建筑大学 一种城市街道设计方法
CN109635511A (zh) * 2019-01-16 2019-04-16 哈尔滨工业大学 一种基于条件生成对抗网络的高层居住区强排方案生成设计方法
US20190228115A1 (en) * 2018-01-19 2019-07-25 Autodesk, Inc. Building information design synthesis (bids)
CN111383303A (zh) * 2020-03-12 2020-07-07 深圳小库科技有限公司 一种自动生成住宅建筑平面的方法及装置

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102663957B (zh) * 2012-03-08 2016-01-13 北京师范大学 一种交互式三维城市全景地图的自动生成方法
CN104392064B (zh) * 2014-12-09 2017-12-19 南京大学 一种中观尺度下缓解热岛效应的优化城市形态设计方法
CN105159951A (zh) * 2015-08-17 2015-12-16 成都中科大旗软件有限公司 一种开放式的旅游多源异构数据融合方法及系统
JP6778631B2 (ja) * 2017-02-16 2020-11-04 日立Ｇｅニュークリア・エナジー株式会社 建物設計情報修正支援装置、建物設計情報修正支援方法、及びプログラム
CN109241580B (zh) * 2018-08-15 2023-08-18 深圳大学 一种地块设计方法、装置、计算机设备和存储介质
EP3623967A1 (en) * 2018-09-17 2020-03-18 Bricsys NV Improved 3d wall drawing in computer-aided design
CN110427724A (zh) * 2019-08-12 2019-11-08 深圳市艾伯信息科技有限公司 基于WebGL三维消防建筑模型可视化方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104732591A (zh) * 2015-03-10 2015-06-24 中国科学院遥感与数字地球研究所 三维虚拟城市的自动生成方法
CN108265820A (zh) * 2017-12-28 2018-07-10 北京建筑大学 一种城市街道设计方法
US20190228115A1 (en) * 2018-01-19 2019-07-25 Autodesk, Inc. Building information design synthesis (bids)
CN109635511A (zh) * 2019-01-16 2019-04-16 哈尔滨工业大学 一种基于条件生成对抗网络的高层居住区强排方案生成设计方法
CN111383303A (zh) * 2020-03-12 2020-07-07 深圳小库科技有限公司 一种自动生成住宅建筑平面的方法及装置

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114898215A (zh) * 2022-06-09 2022-08-12 西南交通大学 一种声屏障自动布置方法

Also Published As

Publication number Publication date
US20220318443A1 (en) 2022-10-06
CN112052503A (zh) 2020-12-08
CN112052503B (zh) 2021-04-13

Similar Documents

Publication Publication Date Title
WO2022047960A1 (zh) 一种基于人工智能的商业街区建筑体块生成的方法
WO2022088407A1 (zh) 基于实体模型的城市设计无纸化智能交互审查方法
Lin et al. Semantic decomposition and reconstruction of residential scenes from LiDAR data
CN102004963B (zh) 一种数字城市生成方法
US20220309201A1 (en) Artificial-intelligence-assisted method for providing urban design form and layout with improved wind environment
Döllner et al. Continuous level-of-detail modeling of buildings in 3D city models
CN104077447B (zh) 基于纸质平面数据的城市三维空间矢量建模方法
CN108171780A (zh) 一种基于激光雷达构建室内真实三维地图的方法
CN109117745B (zh) 一种基于建筑信息模型的云端人脸识别与定位方法
CN102799714A (zh) 基于AutoCAD图纸的地铁综合管线数据提取方法
Kong et al. Enhanced facade parsing for street-level images using convolutional neural networks
Gao et al. Large-scale synthetic urban dataset for aerial scene understanding
CN117934705A (zh) 建筑模型批量生成方法、装置、可读存储介质及电子设备
Utica et al. Integrating laser scanner and bim for conservation and reuse:“the lyric theatre of Milan”
Dogan et al. CitySeek: towards urban daylight models based on GIS data and semi-automated image processing
Stavric et al. From 3D building information modeling towards 5D city information modeling
Tiede et al. A two-way workflow for integrating CAD, 3D visualization and spatial analysis in a GIS environment
Sharma et al. Developing a 3D geometry for urban energy modelling of Indian cities
Semeraro et al. Technologies and techniques offering new interpretations of the landscape evolution
Xiao et al. Efficient Building Category Classification with Façade Information From Oblique Aerial Images
Yudono et al. The Three Dimension (3D) Spatial Urban Heritage Informatics of Malang City, Indonesia
Anca et al. A workflow for UAV’s integration into a geodesign platform
CN113268872B (zh) 一种基于全息三维的国土空间规划仿真方法及系统
CN116958476B (zh) 一种基于bim数据的建筑物可视化建模方法及系统
Zhang Implementation of Landscape Design Instructional System Based on CAD

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20952202

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20952202

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20952202

Country of ref document: EP

Kind code of ref document: A1