Digital determination of spatial objects and scenes for 3D photographic map

The invention relates to a process for the digital acquisition dimensional objects and scenes as well as a 3D image map and applies in particular to the presentation, Katalogi tion and subsequent evaluation of real three-dimensional Structure, wherein, starting from photographs, which the real representing formations, a rendering a dreidi is dimensional mapping model buildable.

The three-dimensional mapping model is then used to use for documentation and interpretation. Possible applications are For example, the multimedia documentation of archaeological Excavations or of monuments, generating models for robot operations, flight and driving simulators, interactive 3D games as well as the control of unmanned inspection satellites space, where for all the aforesaid objects as inputs photos, movies or current are video camera recordings used.

From DE 43 38 280 C1 discloses a method for image-based Position sensing and mapping of underground, close to the surface Objects with an optical instrument and geophysical Sensors for detecting, evaluating and interpreting geo physical data known. With a highly accurate Peilvor direction in conjunction with an orientation platform is a continuous, accurate plan and three-dimensional aboveground Position determination is made, the relative position of the Coordinate systems with high accuracy have to be recorded. Overground and underground object data is then in a entered the central computer and subsequently to a digital transfer card schemes appropriate format and to view brought.

The necessary sensors and the location, respectively. coordinates However, assignment leads to a considerable cost, so that the field of application of the solution described above, is limited.

For navigation, a mobile body such as a motor vehicle along a pre surrounded itinerary, according to DE 195 44 921 A1 proposed a solution in which a streets card in an external storage unit of large capacity z. B. a CD-ROM is stored. The quasi there in a zweidimen dimensional plane stored information is on one screen screen graphically presented, a certain portion of Details of the road map data in accordance with the attached showed position of the screen relative to the display format can be changed such that the graphically Drawn Neten road maps in bird perspective view three-dimensional with a switched for the viewer Depth feeling can be detected. A three-dimensional reference However, in two dimensions only by means of conversion created data, and leaving it to actual reality-based forms of representation arrives.

According to DE 195 38 133 A1 is a method for 3D visualization tion of data on the screen of a computer known the visualizing complex relationships within large amounts of data is used. A plurality of data of a multi- dimensional data space is interactively graphically in a three-dimensional data space implemented, in which the new data ronal, unsupervised learning networks are fed, the genetic algorithms, the data for abnormalities in analyze the topological similarity transformation and transform groups of data networks.

Structural similarities of the data of the multidimensional Data space are in geometrically representable similarities Groups of records transformed, this being with the aim of occurs, a topographical feature map as spatially together menhängende or of related entities on the screen To make a monitor graphically visible.

In such a method, however, there is a Difficult ness that is not present in three-dimensional reality Relationships are created artificially, whereby from this the Disadvantage is that the records thus obtained for a Subsequent documentation and evaluation, for. example, in archaeo logical excavations are unsuitable. In addition to the Input data sets as quasi continuous record with internal reference among themselves available as otherwise a three-dimensional mapping of individual points without external connection elements, which are in turn pretending is impossible.

It is therefore an object of the invention, a method for digital Detecting spatial scenes for a 3D image map and a specify such 3D card, and that allows, in easily any natural scenes by individual position recordings capture without a priori passport predefined points or a single coordinate system must be, wherein on the basis of the thus detected 2D scenes be created subsequently a geometric spatial image can, which in a corresponding database in turn trans is formierbar.

The object of the invention is carried out with a View drive as defined in claim 1 and with a 3D image card of claim 9, wherein the sub-claims at least expedient embodiments and developments include.

According to the invention the method side for obtaining data volume for 3D image cards already present digitized Retired aerial photographs of real objects or scenes attacked or there are made such recordings. The digitized aerial image data are precision position assigned coordinates, so that the desired location accuracy for subsequent combination with other image data sets given is. Ready-made terrestrial recordings of Subject Author fenden real objects or scenes provide views from below different positions and / or directions so that particu particular, the surface that is not accessible from the aerial images may be present as a supplementary image data set.

The digitized image data and also in digital form lying image data sets are to scale in a classified and provide 3D image map coordinate system, in a corresponding storage system is stored, one of a computer managed represent database anywhere on the visualization can be used. be using the above written procedure succeeds, 2D data with scale Object images, the information about the third axis ent keep to combine, so that when viewing data from the image map coordinate system the impression of actually existing 3D image is created.

So according to the invention is an integrated processing made aerial images and terrestrial images, the exact location assigning individual scenes in particular with a The method succeeds, as described in claim 2nd

represents the present invention 3D image map Therefore a data file, the objects and scenes of real Environment in their spatial and surface texture or shape coordinate closely in the aforementioned three-dimensional represents Bildkarten- or image data coordinate system. By Storing the 3D image map in an associated memory or a database and, by appropriate access options the scene or the real object in the scene from different virtual viewpoints with different scales Help of a computer monitor to be considered. Similarly, a virtual moving in the scene possible and it can 3D Koor dinatenmessungen be performed.

In one embodiment of the invention, synthetic 3D graphical objects are shown in the scene, what For example, for the evaluation of projects in a natural environment of substantial advantage is.

In a further embodiment of the invention is not only the spatial structure of the objects, but the entire upper area detected and evaluated with image processing techniques and then on position with the given texture and complete scale on the corresponding areas of the computer transmitted stored 3 D object structures.

On the basis of the aforementioned method steps can at for example a realistic computer model of a city or a complex building arise under which helps Interactive program techniques different Operations are carried out. Thus, the data on the can Bank accessing users in any given scene moving 3D image map, drive up to individual objects, the Larger viewing scale or from one location to Change view in any direction. There is also the possi sensitivity to ride around existing objects, so that a Back can be considered. Likewise, there is a Option when mounting indoors in the database Also the interior of buildings to take virtually inspected.

With the given possibilities of handling using the 3D image map, it is therefore ensured, various practical tasks to a realistic data model to edit. These include the visual assessment and Interpretation of scenes and spatial arrangements of Objects in these scenes, the measurements of distances and Premises or part of objects of the scene, examining the aesthetic and environmental impact in planning Changes of scenes, eg. As at ambient or new buildings, as well as The use of realistic model scenes in flight or Driving simulators.

Inventive applications for example also in the 3D presentation of buildings, building complexes or . like carried out a modeling of the presentation objects given on the basis of plans and realistic Description of the existing scene. As a natural Scene is resorted, is of a higher detail case of com presentation at low cost go.

The picture card of the invention may also be a 3D cityscape map be performed to a virtual experience of Worth seeing to liabilities with recourse to originating from the reality Images to accomplish. In addition, there is a field of application in which Picture card as 3D building plan for building complexes with sensitive provide usage. So z. B. the 3D building plan is a wesent pending aids in the operational control of police, fire military and other rescue workers in hazardous situations, particularly for hospitals, schools, hotels, theaters or . like

Another application is in traffic management systems for motor vehicles, in addition to the already known Vector Design aided maps and information about the current position the pictorial 3D representation of the plan from the perspective of Driver a substantial improvement in Orientierungsmög opportunities are.

According to one aspect of the invention is the ausgestaltenden scene of interest from different positions z. B. with received a digital camera, the image data obtained are stored in appropriate files. Thereafter are corresponding reference or link elements in the individual recordings determined, and this also interactive can take place. The reference or link elements be trademarks, which are arranged in each scene, or striking points that the image from the scene or Scene images themselves originate.

The recording or recording sequence can both kontinuier Lich and discontinuously, ie in the form of individual Screenshots are made.

The resulting digital data sets are now a automatically determining the internal spatial geometry of the System linking elements subjected, for which purpose partially known methods of digital video and / or are pattern recognition applicable. In addition, there is a possi probability that external coordinates or control points provided are, but this is a not a stringent requirement.

Also, a mutual association of points in ver different images or scenes and automatic tracking of structures that points from these within these Images shown are made. Also, start values ​​in be given the pictures, so there is no exact match in the following images is required, but only a Exact Matches must be given. alternative can an optimum search be carried out with the objective of Determining the maximum correlation of the pixels or Link elements. For this purpose, a point in one of the Images defined and recorded. In the other images will a correlation to this selected pixel performed in which case, starting from the respective positions of a local Maximum search in a given search area, ie before given search steps, for. example, 10 × 10 pixels is performed. The optimum values ​​thus obtained then deliver a correspondence to the selected position in the first image, so that Convention provisions are reachable in Subpixelgenauigkeitsbereich.

Thereafter, there is the possibility of movement of the so determined pixel in the selected first image to a small change size z. B. .DELTA.x or Dy <10 pixels and a re-determining the corresponding positions in the other, other images. In the case of possible incorrect assignments leads the thus made image tracking to contradictions, so that eliminating erroneous initial conditions is given.

A correlation mentioned comparison is characterized, that rectangular image content of a fixed size pixel by pixel ver be adjusted. A further implementation form is the internal correlation in the individual images and the Cross to determine correlation between the images mathematically.

Eventually, after the assignment of the link elements of different images and derive Bezie relationships for determining the spatial geometry in the system the link elements, the spatial geometry of the concrete System linking elements determined, based on the created such spatially three-dimensional (3D) model dimensional characteristics of the individual images in the 3D model can be transferred to a three-dimen complete dimensional digital description of the scene or the scene obtain located objects.

Ultimately, is described as a result of the above Steps a complete three-dimensional digital Description of the output side only two dimensions existing Data or derived from them digital files before.

When determining the link elements to the mentioned methods of edge detection, on Korrelationsver same, ie finding the best match between two images with the same image sections or a combination resorted nation the methods described above. As linking elements picture elements are used, the be characterized in that the element in question min is present in least two images and wherein the relevant features clearly distinguishable from the surrounding area. Ultimately must for the picture elements a characterization the corresponding features through 2D and 3D coordinate to be possible.

As linking elements, as in photogrammetry known to be made of points. In various Applications could inventively lines and polylines with good results are used. Sites with pronounced Textures are also suitable. The by Bildverar beitungsverfahren input link elements can list through interactive user works on image process expanded or supplemented or even modified processing system will. In such an embodiment, used automatic assignment of potential linking elements different images, the aim of reducing the number of assignments, for which purpose use a number of criteria are the wrong assignments and allocation of quantities close and highlight the correct assignments.

The following inclusion criteria are preferred.

An association has with the known or already in advance continuous operations certain constraints such. B. approximate relative or absolute camera position pos SHORT- known spatial relationships in the scene - his model compatible;the link elements must be mutually transferable;it is a test of the connection elements on similarities / Correlations in the image environment necessary with only such Elements are assigned, their direct environments aufein are other mappable; It also includes a test of several independent assignments for consistency in solving resulting system of equations with the remaining Ver knüpfungselementen;alternatively, the possibility of interactive mentioned Engagement by assigning shortcuts items by Operator using the present wealth of experience, and this also with a fuzzy logic, in turn, automatically return under handle on database experience is possible.

The necessary in the process described independent Computing and decision steps can be particularly preferably execute by Paralellrechnersystemen so that overall time accelerating the transformation given is.

In a further embodiment of the invention can ver drive moderately integration of object or scene Koor dinatendaten from external data sources for determining the 3D model done. There may be, for example, known linear and angular coordinates of the real scene, to Measurements of external sensor systems, eg., Laser removal knife or laser radar or the like, but also to infor mation on the absolute or relative positioning of the camera act. The position of the camera when taking the object scenes can using accelerometers from a calibration position, movements in x-, y- and z-direction ver Following are detected.

The inclusion of coordinates from external data sources easier method described above embodiment, Find exact assignments and increases the speed at the determination of the 3D model.

In one embodiment, the method is for detecting archaeological excavation situations, ie for documentation use of such excavations.

Specifically, in this usage, the respective Gra detects ambient situation in the complex, and by single shots a obtain documentation of the temporal excavation progress. The recording of the exposed finds from various Per prospects is supplemented by part shots particularly inter essierender finds in situ, so that a correspondingly comprehensive 3D image map exists.

During the preparation of the image data, a highlight is from Details, eg. As soil profiles and finds and edges with means digital image processing, as in the description already mentioned. Detail view of the excavation situation can be divided into the image data obtained models integrate, where also a mark identical image sections on the individual is shooting during image processing possible.

The so existing, edited image data is used for Erstel ment of the 3D model and the 3D map, and this then ideal means of documentation of the exposed finds, and advantageous only in the wake of excavation drawings from any manufactured views using the existing 3D data can be.

Compared to conventional diagnostic data record by transmis tion of coordinate and make sketches before Place z. B., in Public Works, the interruption of the Construction progress can be reduced to a considerable extent without valuable archaeological information will be lost.

The invention shall hereinafter of embodiments are described in detail.

In the embodiment described below is of an investigation of archaeological situations out addressed, in which case the digital capture of spatial scenes involving certain texture information with geological background is the desired goal, so that Finally, a 3D model displayed on a computer system, is present.

First, a recording of the scene being played from done different, essentially freely selectable positions and Directions, the individual coordinates of the mounting points need not be known. After a determination of korre sponding reference or control points (link elements) in the individual recordings is an automatic determination the internal spatial geometry of the system of control points or the linking elements by a recursive solution of nonlinear system, z. B. comparable to a so-called bundle adjustment.

In the presence of external, definable control points or by external link elements, these known Koor have coordinates, a policy may be primary or supplementary take place on or in this coordinate system.

Furthermore, according to the method, an automatic determination relevant features in the individual images, which is this may involve points, lines and segments. With a interactive engagement can be based on experience of the archaeologist or make the image processing system staff back be gripped. Of course, according to the method a Feature analysis automatically from a database or be made by means of a fuzzy logic.

The present 2D features from the individual images then through the now existing spatial geometry system a 3D model converted, with the associated description of the 3D features, eg. As colors and textures, automatically from the Data existing 2D images is obtained. tasks specific data, eg. as signatures or indices may inter are active input, so that the later investigation simplified view of the available 3D documentation.

As a result, the method provides a digital spatial Description of each scene for as a basis a CAD or virtual reality archiving system be used may, with the data obtained in a known manner in common Data formats can be transferred.

In a further embodiment of the inte grated processing of aerial images and terrestrial assumed images to or on the basis of this integration Combination of image data a three-dimensional image map to obtained, which can be stored in a database. With this 3D image map can then use an appropriate Computer system subsequently visualizations of buildings in be made of their natural environment. As a result of that the embodiment of the dataset of the 3D image map from taken photographs, especially aerial photographs of real scenes is, is the reality situation of high detail. The Surface structure is the custom built further terrestrial shots displayed with high precision, whereby the actual precise location coordinates of certain Object points by appropriate positioning systems can be detected and reproduced later.

For the 3D image map are natural objects, such as individual Buildings, streets or neighborhoods based on high resolution precision photos in a combination of the mentioned aerial images and terrestrial recordings faithfully measured in its three-dimensional spatial structure and coordinate closely through appropriate image data structures in Computer system saved. In addition to the spatial Structure of the objects is detected surface thereof with image evaluated processing techniques and with their complete Texture on position and scale in the corresponding areas transfer of data stored in computer 3D object structures. The result is an accurate 3D computer model of the mapped Scene, or mapped object before, the scale neutral abge managed and the data structure of the computer system is stores. An interactive program of a called Viewer the user can arbitrarily in moving scene of the 3D image map to z. B. to certain objects : approaching, the viewing scale or the line of sight to change. Ultimately, a visual assessment and Interpretation of scenes and spatial arrangements of Objects are made, the measurement of distances and spaces or part objects of the scene based on the from the Database conjunction with the information that visually depicting are bar, is possible.