WO2007147830A1

WO2007147830A1 - Method for producing a three-dimensional computer model of a town

Info

Publication number: WO2007147830A1
Application number: PCT/EP2007/056100
Authority: WO
Inventors: Jochen Hummel; Dietrich Charisius
Original assignee: Jochen Hummel; Dietrich Charisius
Priority date: 2006-06-19
Filing date: 2007-06-19
Publication date: 2007-12-27
Also published as: DE102006028086A1; EP2030173A1

Abstract

The invention relates to a method for producing a three-dimensional computer model of a real town from geoinformation for the real town which is to be modelled, where three-dimensional computer models of the town's buildings are produced on the basis of a predefined set of three-dimensional building types, and also on the basis of building attributes which can be used to vary the predefined building types.

Description

Method for generating a three-dimensional computer model of a computer

city

The invention relates to a method for generating a three-dimensional computer model of a city, in particular a computer model of an existing city in reality.

Recently, commercial and private users have high-resolution satellite and aerial images of the entire earth at their disposal. For example, NASA provides a freely available software under the name "World Wind"

Presentation of a virtual globe ready to digital satellite and aerial images in different resolution with geodesics superimposed and displayed on a digital elevation model of the earth as a three-dimensional graphics (3D graphics). The amount of data provided by World Wind is currently almost 5 terabytes. Since such a data volume can not be held locally on the user workstation, depending on the user-set image parameters, ie, in particular depending on the considered area of the virtual globe and the selected magnification factor, the required image data is automatically transmitted via a broadband Internet connection as possible Reloaded servers. A similar concept for the representation of a high-resolution virtual globe on the basis of geo-referenced satellites and aerial images follows the software "Google Earth" provided free of charge by the company Google Inc., at least in its basic version.The resolution lies with "Google Earth" displayable 3D graphics almost nationwide at about 15 m, with metropolitan areas and especially large cities often even with a resolution of less than 1 m, down to 15 cm can be displayed. Due to the combination of geographical coordinates, such as longitude and latitude, with height information can be generated, for example, in three-dimensional representations of a kind of bird's eye view. However, this elevation data only exists for the ground itself, so that elevations such as mountains or mountain ranges or depressions can be represented very well. In cities, however, height information is usually not available for individual buildings. so that the corresponding buildings remain flat even in bird's-eye view. For individual cities, "Google Earth" has provided three-dimensional building information, so that three-dimensional polygons can be superimposed on the floor plans visible in the aerial or satellite images, which also convey an impression of the shape and height of the building in the perspective view a three-dimensional, perspective view of the area around Times Square in Manhattan, created with "Google Earth". It can be seen that in this very schematic representation, no information about the design of the building facades or even about the building interior is obtained.

Some particularly important landmarks can now be modeled in more detail by the users with Google Software "Google Warehouse". This information is provided by means of so-called KML files. A disadvantage of this approach is that the objects created with "Google Warehouse" often have very different quality and therefore give an overall uneven impression of a scene. Furthermore, the superposition of three-dimensional objects with aerial images often leads to an unrealistic overall impression. Finally, such models lack any semantic information, i. H. each three-dimensional object is just a graphic that contains no information as to whether the object is, for example, a house, a street, a walkway or parts of a house, such as doors or windows.

In the currently very popular on-line roleplaying game "Second Life", it is also possible to implement smaller areas of real existing cities in "Second Life" using three-dimensional models, such as a virtual model of the German city of Frankfurt in "Second Life." However, the costs associated with manual modeling of individual buildings and the amount of time required are far too high to model an entire city, and the virtual city of Frankfurt in Second Life has only a few streets with some important buildings and landmarks in more detail modeled, so that they also have, for example, interiors and can be entered. Most of the buildings were modeled in less detail and can not be entered, for example. Although costly and time consuming, this approach creates an inconsistent overall impression at high cost.

In the German patent application DE-A-197 46 639 a method for the digital detection of spatial objects and scenes for a three-dimensional image card is described, which should allow a more accurate representation of the three-dimensional objects. For this purpose it is proposed to assign digitized aerial photographs of real objects or scenes to precise position coordinates and to make supplementary terrestrial images of the relevant real objects or scenes that provide views of the objects in question or scenes from different positions and / or directions that are not directly accessed from the original aerial images can. For example, information about the height of the objects in question and their surface design can be derived from the terrestrial images and linked by a computer to three-dimensional representations of the objects. This can create a photorealistic computer model of a city or complex building that can be manipulated using interactive programming techniques. For example, movements in the three-dimensional image map or changes in the viewing scale or the viewing angle are called. As typical applications DE-A-197 46 639 the use of such three-dimensional representations in realistic flight or driving simulators or the documentation and assessment of archaeological excavations or planned construction projects. However, the solution described in DE 197 46 639 A1 is disadvantageous for numerous other applications because a photorealistic representation of the three-dimensional objects entails the processing of large amounts of data. For example, while a three-dimensional computer model of an archaeological excavation site or a spatially limited construction project still manages with a manageable amount of data, requires a realistic three-dimensional representation of a real a volume of data that would normally not be available on a user's local workstation or downloaded in real-time from central servers.

Other proposals for three-dimensional computer modeling of cities also combine aerial photos with façade photos to create a realistic three-dimensional impression. These façade photos are to some extent projected onto computer-generated geometrically simple three-dimensional models of the buildings. However, this method also involves numerous disadvantages. Already the unique all-round detection of the facade of a building is time-consuming and cost-intensive. Therefore, the façade pictures are usually only in a version that was recorded under certain daily and seasonal conditions. It is therefore not possible in the computer model to represent daily and seasonal changes in the facades. One can not enter such three-dimensional building models either, since the facade pictures are not linked with any semantic information regarding individual elements such as doors or windows or other building details. The image information of the facades are usually present only in a certain resolution, which is optimized for a certain virtual viewing distance. When approaching a building, these images blur because an enlargement beyond the existing image information is required. For interactive three-dimensional computer games, such an approach is therefore not suitable because in the creation of such images often trees, people, vehicles and the like are included without the images get the appropriate semantic or three-dimensional information.

The present invention is therefore based on the technical problem of providing a method for generating a three-dimensional computer model of a real city, which enables a realistic generation of the computer model on the basis of as few initial data as possible, and immersing the user in the virtual reality ( "Immersion") and the amount of data required to model the city and the corresponding computational effort are kept as low as possible. For this purpose, it is only necessary to make use of the smallest possible amount of data to be kept locally and / or data to be transmitted by data connection.

This technical problem is solved by the method for generating a three-dimensional computer model of a real city according to the present claim 1. Advantageous developments of the method according to the invention are subject matter of the dependent claims.

The method according to the invention is based on the consideration that virtually every building can be represented realistically by combinations of a limited number of building types and building attributes. According to the invention, therefore, satellite images, aerial images and / or terrestrial images of buildings are not used directly for generating the three-dimensional building model, but rather serve as a basis for selecting a building type and providing it with suitable building attributes on the basis of this image information from a given set of building types and building attributes so that the resulting computer model matches the real building as well as possible. To solve this technical problem, the invention now proposes to specify a set of three-dimensional building types with corresponding building attributes and to define them by polygon surfaces or non-linear parametric functions and to represent the real buildings in the computer model determined by the geoinformation by one of the given building types with the respective building attributes

The invention therefore provides a method for generating a three-dimensional computer model of a real city, the method comprising the steps of: providing geo-information of the real city to be modeled,

Determining building data from the provided geoinformation, wherein the building data at least position information on the geographic coordinates, the height of the selectable buildings and image information associated with the selectable buildings and / or other building data,

Generating a predefined set of three-dimensional building types and building attributes that can vary the predefined building types and storing the predefined set in a computer-readable storage medium, wherein the predefined set of three-dimensional building types consists of objects defined by polygon surfaces or non-linear parameterized functions - assign a building type from the predefined set to at least one selectable building on the basis of the image information or other building data associated with the selectable building,

Setting building attributes from the predefined set for the at least one selectable building on the basis of the image information or other building data associated with the selectable building.

Generation of a three-dimensional computer model on the basis of the selected building type and the set building attributes and

Determine the position of the building model within the computer model of the city based on the position information associated with the selectable building.

A "selectable building" is understood to mean a building of the real city represented in the geoinformation that can be selected for modeling with the method according to the invention. To model an entire city, of course, one will endeavor to represent as many buildings as possible within the meaning of the inventive method by appropriate computer models.

Thus, according to the invention, a predefined set of three-dimensional building types is generated, so that the storage requirement can be limited, because not every single building has to be modeled independently, but instead a limited number of building types can be used. A significant advantage The method according to the invention can therefore be seen in the fact that, with just a few building types and building attributes, a large number of different buildings can be modeled realistically owing to the numerous possible combinations. As a rule, therefore, the number of buildings to be modeled in the computer model of a real city will be significantly greater than the number of building types and building attributes underlying the model. With the method according to the invention, for example, with 10 to 100 different types of buildings and 10 to 100 building attributes, where a building type, for example, several building attributes can be assigned to large cities with tens of thousands of different buildings are modeled so that a realistic impression of the city is created without Modeling larger amounts of data must be kept or transmitted.

Due to the definition of building types by polygon surfaces or non-linear parameterized functions, the modeled buildings can easily be scaled according to the desired level of detail.

The invention is based on the consideration that in many areas of application for generating a realistic, immersive character of a three-dimensional city model is less important to an exact reproduction of all details and textures of the real building, but rather on a coherent presentation, in particular discontinuities at the level of detail or even a change from one display type to another (eg from a polygon display at low LoD to a texture display at high LoD) This can be achieved in the inventive method by skillful choice of the given building types.

In the modeling according to the method of the invention, a certain heuristic in the form of assignment rules can also be included, which also provides the empirical knowledge of the town to be modeled and the probabilities of occurrence derived therefrom in the assignment of geographic information and image information to the individual predetermined building types certain building types. For example, if there is information that a certain part of the city to be modeled is dominated by prefabricated buildings, you can assign the prefabricated building type "prefabricated building" in this part of the city based on geographic information and automatically model the building accordingly - a two- or three-storey building Within a district known as the residential area, one can, for example, assign the building type "Art Nouveau villa" and in turn automatically generate the corresponding model.

Thus, with the method according to the invention, a 3-D model of a city can be largely automatically generated in any desired accuracy, even if only rudimentary geoinformation and image data for the real buildings of the city are available. A detailed heuristic can then take into account further information. If, for example, it is known from telephone directory data that a particular building has a retail store on the ground floor, then without providing detailed information about the real facade design, this information can be incorporated into the modeling, for example by assigning the attribute "fully glazed window front" to a particular building type.

Such an assignment then does not necessarily produce a perfectly exact representation of the real city to be modeled, but in any case a self-consistent and user-friendly representation arises, already on the basis of a few initial information and on the basis of a few predefined building types a detailed three-dimensional city model can be generated.

The extensive automation of building modeling also allows larger cities to be modeled quickly and reliably. All building models can be constructed as walk-in three-dimensional models. Each object of the virtual three-dimensional city can be geo-referenced and semantically contain data, that is, data about the type and properties of the object. For example, a building window can display the information ("can be opened", "mirrored Furthermore, within the entire virtual reality, a consistent and consistent impression is created, which supports the immersive character of interactive 3-D worlds, and ensures that the modeled buildings can be located at any distance, at different distances Day and season give a realistic impression.

Advantageously, the image information comprises at least one orthographic projection of the selectable building and an image of at least one building facade of the selectable building. From this information, the building dimensions in the sense of a floor plan, as well as the suitable building model for realistic modeling, as well as the appropriate building attributes can be determined.

The image information is advantageously obtained from aerial photos, satellite photos and / or terrestrial building pictures. From aerial photos or satellite photos one can directly determine the orthographic projection of the buildings to be modeled. However, these figures do not give any information about the façade designs of the buildings. In this case, additional images are required, which also represent the facade. Again, these may be aerial photographs obtained, for example, at an oblique viewing angle or purely terrestrially derived building pictures. Here, too, there are now commercial providers who, for example, for customers from the real estate industry, complete drive entire cities and thereby provide shots the busy streets. These images are, for example, with the help of the satellite positioning system for positioning GPS (Global Positioning System) of the US Department of Defense geographic again accurately referenced, so that a simple assignment of orthographic images and terrestrial images of the building facades is possible. For example, if only oblique images, so you can calculate the orthographic projection of the building and the building height at a known acceptance angle and also receives Façade information that can be used to later assign the predefined building types and building attributes.

Preferably, to compare a three-dimensional building model from the predefined set, one compares the image information of the selectable building to be modeled with the building types present in the predefined set and selects the type of building most closely closest to the building represented in the image information. Such an assignment can be carried out fully automatically by means of a suitable pattern recognition method. However, since ultimately the perceived by humans impression for the most realistic modeling of the building is crucial, preferably people are used as controllers to make the assignment of the image information and the most suitable building type or at least to check. In a particularly preferred variant of the method selects an automatic

Pattern recognition methods from the total set of available building types first a subset of building types, which already have some consistency with the building to be modeled, the human inspector ultimately makes the final selection of the most suitable building type of the subgroup for the building to be modeled.

In a next step building attributes such as facade texture, stucco, window types, façade colors and the like are assigned to the selected building type on the basis of the image information of the real building. Again, this mapping can be fully automated, or semiautomatic or manual with a human controller.

Alternatively or in addition to the image information, other building data can be used to assign a building type and building attributes to the building to be modeled. Such building data may include, for example, the age of the building, the use, the number of occupants, the District in which the building is located, etc. Such building data can be obtained, for example, from maps, cadastral data, phone books, such as the yellow pages, or from a variety of databases and information sources available on the Internet. From these alternative building data, plausible building types can therefore be assigned to individual buildings even in the absence of image information. Thus, even in such cases, a more realistic modeling of the real city is obtained than if the missing building is only replaced by a neutral placeholder.

The associated three-dimensional building type and the corresponding

Building attributes can be adapted dynamically to the orthographic projection of the selected building determined from the geographic information, so that, for example, the building floor plan corresponds to the floor plan determined from the orthographic projection.

The predefined set of three-dimensional building types may have a hierarchical structure. For example, the uppermost level of the hierarchical structure can be associated with facade types of the buildings, while the lower levels are assigned data relating, for example, to the facade texture, façade color, window details, extensions, roof types, roof texture or roof color.

Preferably, the predefined set of three-dimensional building types consists of objects that are defined by polygon area or so-called non-linear parameterized functions. Examples of such functions used in three-dimensional computer graphics are the so-called NURBS (Non Uniform Rational B-Splines). For example, the building types defined in this way can be easily scaled and displayed with different resolutions, generally referred to as Level of Detail (LOD), depending on the magnification scale selected by the user. The three-dimensional building types provided in the predefined set can, according to an advantageous embodiment of the method according to the invention, take into account country-specific or regionally typical characteristics of the real city to be modeled. Practically every city has architectural features that may arise, for example, from the geographical location, the historical urban development or the individual building regulations. Taking into account such special features, on the one hand the realism of the computer models can be improved and, on the other hand, the number of building types and building attributes to be traveled can be reduced.

The inventive method for generating a three-dimensional computer model of a real city can be used in a variety of applications. Preferred applications are, for example, mobile applications, such as in a navigation system, where on the one hand the storage space in the mobile terminal is limited and on the other hand, the necessary bandwidths for the transmission of large amounts of data are usually not available. Such applications benefit particularly from the method according to the invention, since only a little storage space-requiring basic data set of building types and attributes is required to produce the three-dimensional building models. Further applications can be found, for example, in aircraft or vehicle simulators.

However, the method according to the invention is particularly preferably used in interactive virtual realities which are accessible to one or more users within a computer network. A particularly popular form of such virtual realities is currently the so-called "Massively Multiplayer Online Roié Playing Games" (MMORPGs), which often several thousand users can simultaneously play together over the Internet. As an example, here are the online games called "World of Warcraft", "Second Life" or "Project Entropia", in which users can act via virtual, graphically three-dimensionally animated proxies called "avatars." Currently, the corresponding virtual realities are pure Computer-generated fantasy worlds The method according to the invention here offers the possibility of realizing real cities in a realistic manner simulate, so that such MMORPGs can be played in a realistic environment. The computer model of a real city can form a bridge between the virtual world and the real world. In addition to a low data volume, the method according to the invention has the particular advantage that even in the introductory phase of a new online game, the corresponding city can be modeled very realistically without large personnel resources. Then you can, for example, integrate the users of the game into the optimization of the computer model. For this purpose, according to a variant of the method according to the invention, the data on building types and building attributes of a larger number of users can be interactively made accessible for editing. Thus, for example, erroneous or less optimal modeling of individual buildings can be successively optimized over time in order to make the modeling of the virtual city more and more realistic.

According to a variant of the method according to the invention, the buildings are only generated when the virtual city is instantiated on the basis of the stored data of its geographical position, the building height, the assigned building type and the building attributes. According to a further variant, the buildings are only generated and visualized from a predetermined level of detail, for example at the close top view of the user.

A particular advantage of the method according to the invention is the fact that the modeled buildings can be produced as completely three-dimensional buildings, so that not only exterior views but also the internal structure of the building with individual floors and interiors in the desired

Detail accuracy can be generated. It is therefore possible, for example, for a user, for example with his avatar to enter the interior of a modeled building during an online role-playing game or by means of a computer-generated camera movement, to reach different floors or to look outside from the windows of the building. The invention also relates to a computer system which is adapted to the implementation of the method according to the invention.

The method according to the invention is explained in more detail below with reference to a schematic of the generation of a three-dimensional building model shown in the attached drawing.

In the drawing shows

FIG. 1 shows a three-dimensional computer model of a real city according to FIG

Prior art as currently represented by "Google Earth";

Figure 2 is a schematic representation of the inventive method for

Modeling a building using predefined building types and building attributes based on geographic information or

Image information;

FIG. 3 shows a schematic representation of the architecture of a computer system suitable for carrying out the method according to the invention.

The three-dimensional computer model of the area around Manhattan's Times Square shown in FIG. 1 has already been explained in the introduction.

FIG. 2 now shows a schematic representation of the method according to the invention:

The geo-information / image information provided includes, for example, an orthographic projection of a selectable building 10 to be modeled, which is obtained, for example, from an aerial image 11. In addition, a terrestrial image 12 of the building facade of the selected building 10 is available. On the basis of this information, the building dimensions 13 are first determined. Then the building type is defined, in the example shown a predefined set 14 of three Gebäudetpyen (band facade with flat roof 15, perforated facade with pitched roof 16, element facade with flat roof 17) are shown. In a hierarchical arrangement, only the facade types could be selected in a first step and the corresponding roof type in a later step. In the example shown, the assignment to the building type perforated facade with pitched roof 16 takes place, since this type of building has the greatest similarity to the real building. This building type is then assigned appropriate attributes from a predefined set 18 of building attributes, such as window type 19 (eg, sash windows, shutters yes / no, etc.). Further, extensions, such as balconies 20, may be associated, if any. In particular, in the inner city area, for example, also proven to define a separate building base 21, since often the design of the building base by store lines, among other things, may differ significantly from the rest of the building. The building 22 modeled in this way is precisely positioned in the three-dimensional computer model of the city on the basis of the position information obtained from the geographic information.

The computer system, generally designated by the reference numeral 100, is constructed in a client / server architecture. At the center is a server unit 110, which may, for example, be constructed as a load-balanced server cluster, which is connected to that of a database 111, on which the data necessary for generating the virtual reality are kept. Each virtual reality user has installed on his local machine a client 120 that includes a 3-D game engine 121 and a user interface 122. For example, each client may access the server unit 110 and the database 111 via an Internet connection 130. The data required to generate the virtual reality can be transmitted, for example, in the form of chunk files 123, from which the 3-D game engine constructs the virtual reality. If the files transferred from the database are not yet in one with the used game engine compatible format, the client 123 may include the respective libraries 124 for generating the 3-D models and a data converter 125. The creation of the city model is then largely independent of the currently used game engine.

Various programs, commonly referred to as authoring tools, such as the commercially available "LandXplorer" software package, may be used to generate the city database 111. The authoring tool may be implemented on a server and server cluster 140. Over a data interface 150, geoinformation is represented as two-dimensional and / or two-dimensional map data 151 and imported as additional geographic information 152. Such additional geographic information may include, for example, locations and types of trees present in the city, which are often electronically present in some kind of tree cadastre, or the locations and species Further urban infrastructure elements such as benches, traffic lights, etc. Furthermore, already existing three-dimensional graphics models can be imported 153. Often these are particularly well-known buildings and landmarks that are not compatible with the invention A method should be modeled automated, but is desired for the most realistic representation. Further, these three-dimensional graphics models 153 may contain prototypes of frequently recurring infrastructure elements within cities. As an example of another type of data to be collected, site specific facade heuristics 154 are mentioned.

Via a user interface 160, the user can operate the authoring tool and control the generation of the three-dimensional city model or edit the generated model. In the present example, the method according to the invention for the automated generation of three-dimensional building models is implemented as a so-called extension 170 for the commercially available authoring tool. This extension 170, which is a corresponding

User interface 171, which may be integrated into the user interface 160 of the authoring tool, contains the predefined basic models and in particular the rules by which the three-dimensional buildings / objects are to be generated on the basis of the existing geographic information.

The corresponding data are stored for a specific area of the virtual reality, for example in the form of binary transport files via an interface 141 in the database 111.

Claims

claims

A method of generating a three-dimensional computer model of a real city, the method comprising the steps of:

Providing geographic information of the real city to be modeled, determining building data from the provided geoinformation, the building data comprising at least position information about the geographic coordinates, the height of the selectable buildings and image information associated with the selectable buildings, and / or other building data;

Generate a predefined set of three-dimensional building types, as well as building attributes that can be used to vary the predefined building types, and save the predefined building types

A sentence in a computer-readable storage medium, the predefined set of three-dimensional building types consisting of objects defined by polygon surfaces or non-linear parameterized functions,

Assigning a predefined building type from the predefined set to at least one selectable building based on the image information or other building data associated with the selectable building,

Setting building attributes for the at least one selectable building based on the image information or other building data associated with the selectable building,

Generating a three-dimensional building model based on the assigned building type and the assigned building attributes and

Determine the position of the building model within the computer model of the city based on the location information associated with the selectable building.

2. The method of claim 1, wherein the image information comprises at least one orthographic projection of the selectable building and an image of at least one building facade of the selectable building.

3. The method according to claim 2, wherein the image information from

Aerial photos, satellite photos and / or terrestrial building images wins.

4. The method according to any one of claims 1 to 3, wherein for the assignment of a three-dimensional building model from the predefined set compares the image information of the selectable building with the building types of the predefined set and selects a building type, which is most similar to the building shown in the image information.

5. The method according to claim 4, wherein before the final assignment of a three-dimensional building model, first by means of an automated

Pattern recognition process forms a subset of building types from the predefined set and then assigns a building type of the subgroup to the selected building.

6. The method according to any one of claims 4 or 5, wherein building attributes, such as facade texture, stucco applications, window types, colors or attachments are set based on the image information.

7. The method according to any one of claims 4 to 6, wherein the assignment is completely automatic.

8. The method according to any one of claims 1 to 7, wherein other building data such as information about the age of the building, the use of buildings, the number of residents, the location of the building within a particular district and the like via publicly accessible

Data sources wins and out to the most likely building type and closes the associated building attributes.

9. The method according to claim 1, wherein the associated three-dimensional building type and the building attributes are dynamically adapted to the orthographic projection of the selected building determined from the geographic information.

10. The method of claim 1, wherein the predefined set of three-dimensional building types has a hierarchical structure.

11. The method according to claim 10, wherein facade types of buildings are assigned to the top level of the hierarchical structure, while data is assigned to the subordinate levels, such as the facade texture, facade color, window details, extensions, types of roofs,

Affect roof texture or roof color.

12. The method according to any one of claims 1 to 11, wherein in generating the set of predefined three-dimensional building types and building attributes regional characteristics of the real city to be modeled considered.

A method according to any one of claims 1 to 12, wherein the computer model of the city is used in a navigation system or in an interactive virtual reality generated by a computer network and accessible to a large number of users.

14. A method according to any one of claims 1 to 13, wherein the data on building type and building attributes is interactively made accessible to a larger number of users for editing.

15. The method according to claim 1, wherein the building model is only generated when the virtual city is instantiated on the basis of the stored data of the geographical building position, the building height, the associated building type and the building attributes.

16. The method according to any one of claims 1 to 15, wherein the building model is generated and visualized only at a predetermined resolution limit.

17. The method of claim 1, wherein the three-dimensional building model comprises the building interior.

18. The method according to claims 17, wherein the modeled building interior is formed virtually walkable.

19. Computer system with at least one server and at least one client, which is designed to carry out the method according to one of claims 1-18.