Car simulator scene based on real world geographical data[1]
Ing. Pavel Hrubeš
Joint Laboratory of System Reliability, Department of Control Engineering and Telematics,
Faculty of Transportation Sciences, Czech Technical University, Prague
Konviktská 20, Prague 1, 11000
Czech Republic
http://www.lss.fd.cvut.cz
Abstract: - In this paper is presented the idea of usage the two-dimensional geographical data as a base for the building more realistic three-dimensional virtual scene. It is discussed here the background of the geographical information system (GIS) and its stay of art. Then is here defined a framework architecture with description of system blocks and their functions to realize two to three dimensional transformation.
Key-Words: - GIS, virtual scene, car simulator
1 Introduction
The main reasons, why I am interested to realize two to three dimensional transformation, come from my participation on two projects running in the Joint Laboratory of System Reliability (JLSR) of the Department of Control Engineering and Telematics of the Faculty of Transportation Sciences, Czech technical University in Prague. The first one concerns the research of the human-machine interaction and the second is oriented into the field of in-car services and applications.
As concerns the first mentioned laboratory project, this is supported by two scientific projects No. ME 478 “Neuroinformatics” and ME 701 “Development of Neuroinformatic Databases and the relevant Datamining”, of the Ministry of Education of the Czech Republic. Similar research problems are also in the focus of the research cooperation contract between JSLR and VW Co. and Skoda-Auto Co. For simplicity I shall denote this area as the “Micro-sleep” project.
Here we use at present three levels of the car simulators. These form the technical, methodical and software basis for simulation of the real driver situation.
Fig.1: Skoda Superb car simulator at CTU in Prague
The main components of these simulators are:
Ø Skoda Superb or front panel of Skoda Superb
Ø CAN interface,
Ø master application to manage all the simulator components,
Ø sound application for a preparation of engine sound simulation,
Ø virtual scene application for preparation the scenario, which is projected on the screen, observed by the tested human subject.
The simulator data and video acquisition application are running on the recording state.
One of our important tasks, when developing these car simulators is to prepare the 3D virtual scenes. These scenes will be based on the measurements of tested subject significant attention parameters in the course of the experiments. We would like to create these scenarios on the base of the most realistic approximation of the real world conditions.
As concerns the second mentioned project, running in the JLSR, we would like to prepare a new kind of the in-car navigation system, which there will be shown to the driver the predicted route in the three dimensions scene.
For both these tasks, we need some knowledge, methods and tools of the conversion of two-dimensional GIS data to three-dimensional scene.
2 GIS overview
The GIS is in principle the system for input, storage, manipulation, and output of geographic information. It integrates spatial and other types of information and provides a consistent analysis framework for the geographically referenced data. It provides a new and insightful ways of manipulating and displaying of data and allows viewing and analysis of these data based on the geographical proximity and relationships.
The representation the real world in GIS brings the serious problems with the infinite complexity of the external world. Each spatial database (which content in naturally limited) allows only a partial view on simulated reality - the spatial database is a model of reality. The user sees the real world through the medium of such database, which includes besides some digital imaginations of real objects (e.g., houses, roads, forests etc.) also some digital versions of fictitious (i.e., invented) objects (e.g., political boundaries).
2.1. Spatial and attribute data
The spatial data can represent the objects or entities that are referenced by their location:
Ø Latitude / longitude coordinates and altitude
Ø x / y / z coordinates
Ø Street address
Ø Administrative unit
The attribute data are linked to the spatial objects and contain the value corresponding to the object. For example: census data by administrative unit, land parcel ownership records, soil or vegetation characteristics, health records by medical center and road quality information
The current data sources for GIS are of the following two types. The primary source represents the data, measured directly by surveys, field data collection and remote sensing. The secondary source represents the data reached from existing maps, tables or other data sources.
2.2. Layers and themes
Paper maps use symbolism to distinguish between layers, to compensate for the limitations of the technology. GIS data is represented in logical layer, in which are place related objects to each theme. This architecture allows assemble maps for different purposes by combining themes. New themes and results are based on analysis and examine interaction between input themes.
Fig. 2: GIS Layers
2.3. GIS data models
The main part of common data models is based on relationships between basic geographical objects stored in the database (Georelational). The linked tables include a list of information items related to a single theme or to a subject. The specific items of information associated with a single record form the fields or items. Each such field may contain data of different types:
Ø numeric (integer or real),
Ø string or text,
Ø date / time or
Ø logical data.
Any stored geographical information is described by its
Ø geometry,
Ø topology and
Ø attributes.
The geometry could be of the two major types:
Ø vector and
Ø raster form.
2.3.1. Raster data model
Some geographical space is represented as a grid of cells. Each cell is represented by numerical values. The advantages of raster approach consist mainly in simpler data structure. More over - certain analytic operations can be easily implemented, which is good for representing continuous variation and which also allow the import/export from/to other raster GIS systems.
Improved are also the analytical capabilities, overlaying, explicit knowing of neighborhoods and good representation of continuous quantities and pixellated data.
2.3.2. Vector data model
The vector objects are defined by their coordinates in the planar or another coordinate system. The attributes of objects are attached as database tables. Real world objects are represented as points, lines, areas and volumes. Points identify locations, lines connecting points and areas (polygons) are consisted with connected line segments. Volumes are assembled from corresponding areas. The precision of coordinates is virtually infinite (it is dependent only on the used computer power), however the accuracy of each calculation is, of course limited.
Advantages can be seen in compact data storage, in the scalable presentation, in the object based database linkage and in the possibility of network analysis.
The typical two-dimensional GIS application map-objects are the points, lines or poly-lines, areas, regions or polygons and texts.
In contrast to the 2D GIS, an outstanding feature of 3D GIS is the realistic visualization. The 3D visualization techniques allow to the users to get better interactive interface, the operation with which requires much shorter training time and which also involve much more spatial information. Another important feature of 3D GIS is the huge amount of involved data and their complex nature. The 3D GIS is not only a simple extension from 2D to the 3D space, but it represents a new system in which many new data types and their spatial relations emerge. The huge amount of data requires of course the perfect database management.
The research in 3D GIS is intensive and covers all aspects of the collecting, storing and analyzing real world phenomena. Among all, 3D analysis and the issues related are mostly in the focus of investigations.
Fig. 3: The raster image draped over a digital elevation model with a 3D objects
To main topics of interest are:
Ø 3D topological model
Ø Formalism for detecting spatial relationships
Ø 3D visualization
3. Problem formulation
The current technology for collecting of the GIS data is based on analyzing of some orthographic photo or on the direct measurement realized by GPS or other similar methods. Many companies collect the 3D data, but do not maintain or supply it. All those approaches are very expensive if they have to be applied on large areas in required accuracy and precision.
It is evident, that the collection of three-dimensional data is more complex and therefore is still used only for solution of small and closed areas (the town guides e.g.).
Our vision is to prepare the low-coast solution, which could allow to get the 3D virtual scene from the 2D GIS data, and which will be not the substitute of perfect real word model, but will be a suffice for given tasks.
Therefore I try to formulate the following basic summary of the required functions for 2D to 3D GIS conversion:
Ø The input for 2D GIS sources and DTM formats
Ø The output scene will be variable upon the theme or objects time life
Ø The output scene will be variable upon level of object details (LOD)
Ø The output scene will be variable upon relationships between themes
Ø The output scene will operate in open data format
4. Outline of the problem solution
4.1. The process of transformation
It is obvious that it is not possible to achieve universal transformation of 2D data into 3D. The problem is complexity of the object, which we want to reach and the input data don’t meet the requirements. Other problem is the mistakes in input data and imperfection of existing algorithms.
Based on me work I proposed a process of transformation consisting of the following steps:
Ø Classification
o semantic
o vector
o attribute
Ø Modelling
o replacement
o transformation
o creation of a new model
4.1.1. Semantic classification
In this process we supplement the information for creation of 3D model based on knowledge of basic attributes of shown objects. For example the height of a forest equals about 30m, the water level of a lake 0m to TIN model. Based on the knowledge of the cross section of a road and a material we detect requirements for representation in 3D, which results in the selection, or possibly creation of a model and selection of a texture.
4.1.2. Vector classification
In this process we fill in the information for creation of 3D model based on the knowledge of graphic representation of an object and its surroundings. The classification of a prototype is differentiated. For example an individual tree and forest, road, crossroad and grade crossing. Based on this information we also detect the requirements for representation in 3D.
4.1.3. Attribute classification
In this process we supply information for 3D model creation based on the knowledge of the attribute part of the object, where the selection of a model is possible, based on a value of control attributes of the geometric models and also the selection of a model. For example the height, number of floor in a house, area and number of offices defines the demand in the 3D.
4.1.4. Modelling
This part of the process is crucial for the final image of virtual objects, where based on classification, we select a representative model, transform it or possibly we create new model.
4.2. Framework architecture
The area of possible problems in proposed framework is really very wide. For solution of this task, I expect therefore to compose a project team with participation of several students from our faculty.
My main personal interest and contribution, which I would like to realize in my Doctoral thesis, I see in the field of classification models based on the object graphics expression and linked descriptive attributes. These classification functions will form a substantial part of the Theme analyzer, which plays an important role in the framework architecture.
Fig. 4: The framework architecture
The designed framework architecture consists of three basic functional blocks and of the input and output interfaces. Each component will have a graphic user interface for managing their functionality.
The theme analyzer is responsible for analyzing input two-dimensional data and for finding the corresponding class of objects in the framework catalog. The object catalog involves a general knowledge about real word necessary for preparation of realistic 3D virtual scene. To this information are linked theme and graphic/attribute meta-description, which enable to classify the corresponding input data and select sufficient scene modelling method. The virtual scene modelling processes use the linked 2D data and the knowledge involved in the Object catalog, which all is used to set up the required output scene.
In the course of the development of the framework I expect that each its component will undergo all design phases, from manual, semiautomatic to automatic functionality.
4.3. Framework data flow
The following data flow diagram describes the framework process in the time and shows the variables that are transferred between individual components.
Fig. 5: Framework data flow
5. Conclusion
Reliability of man-machine interaction is in a front of interest not only our group. The need to minimize losses from transportation accidents is the dominant motivation for activity in this area.
The progress in this respect could be reached by combination of the following 4 main approaches:
Ø Improvement of the training the drivers with respect to their higher resistance to disturbing factors causing decrease of their attention,
Ø Improvement of the interior of the car cockpit with respect to minimizing the influence of disturbing factors causing the decrease of drivers attention,
Ø Development of micro-sleep warning systems and their installation in car cockpit
Ø Improvement of the traffic control systems with respect to wide scale detection of risky and aggressive driving and of its punishment.
I believe that the prospective way to the understanding and forecasting of driver behavior is in the good simulation of traffic condition and measurement of driver feedback.