WO2006003268A1

WO2006003268A1 - General method for determining a list of elements potentially visible by region for three-dimensional large scenes representing virtual variable altitude cities

Info

Publication number: WO2006003268A1
Application number: PCT/FR2004/001371
Authority: WO
Inventors: Loïc BOUGET; Christian Bouville; Alexandre Cotarmanac'h
Original assignee: France Telecom
Priority date: 2004-06-03
Filing date: 2004-06-03
Publication date: 2006-01-12

Abstract

The invention relates to a method for determining a list of potentially visible objects in a three-dimensional scene which comprises objects (100) placed on a ground (101) of variable altitude and in which a user can virtually move through a displacement area corresponding to a totality on grounds areas devoid of objects (100). The inventive method consists in forming ground objects (104) each of which pertains to a ground area, in dividing the user's displacement area into a network of sight cells and in determining for each sight cell a list of objects potentially visible by the user therefrom, wherein said list is determined by testing the visibility of the scene objects (100) and the formed ground objects (104).

Description

General method for determining a list of potentially visible elements by region for 3D scenes of very large sizes representing virtual cities of variable altitudes

The present invention relates to the general technical field of synthetic imaging, and more particularly the field of 3D scene representation.

GENERAL PRESENTATION OF THE PRIOR ART

The general object of the invention is to propose a visibility calculation method making it possible to determine lists of potentially visible objects by region for 3D scenes of very large sizes representing virtual cities of variable altitudes. In synthetic imaging, the role of visibility calculus is to determine which objects of a 3D scene are visible from a point and an observation direction.

Historically, visibility computation has always played a crucial role for reasons of display performance of 3D scenes. Indeed, when a 3D scene is displayed, some parts are not visible: parts outside the field of view, for example, or objects, which are hidden by opaque objects in front.

In order to speed up the display of a 3D scene, we must avoid unnecessary drawing of these "hidden" parts. Thus, it is necessary to eliminate as soon as possible these parts of the processing processes that will generate (draw) the 3D scene.

This is why we often apply pretreatment to the scene to determine what will ultimately be visible in a given view of it. This pretreatment is a method of calculating visibility. It provides a list of objects in the scene that will be visible in a given view. Only objects in this list will be drawn by the means processing and displayed on the display means, which is faster than drawing all the objects of a scene.

Many methods of calculating visibility allowing the elimination of non-visible objects were thus proposed in the 70s, in particular the most famous and most used is the z-buffer method [3]. This particularly powerful process is implemented on all modern 3D accelerator cards.

In recent years, these visibility methods have regained interest because of the increasing complexity of 3D databases that makes it impossible to use conventional z-buffer methods.

Therefore, in a real-time client-server perspective, it becomes necessary to use processes (off or on-line) allowing according to a point or an observation area to reject part of the scene before to use the conventional methods of the z-buffer type allowing the calculation of the synthesis image.

In visibility calculation processes, work is differentiated on the exact visibility and work on the conservative visibility.

An exact visibility calculation method aims to determine at an observation position and direction the exact list of visible objects. This technique generally uses a structure called an aspect graph.

A conservative visibility calculation method searches at a position or an observation area for the smallest list of potentially visible objects that will be named in the LOPV suite. The list of visible objects is included in the list of potentially visible objects.

In the following, we will focus on processes that treat conservative visibility better adapted to navigation in large urban environments. Indeed, the calculation of exact visibility of a cell of view is very complex, and therefore heavy to implement. We also differentiate the work on visibility that takes place in the 3D object space of the scene, those that take place in the discretized 2D image space obtained by projections of 3D objects from the scene on the image plane. For reasons of performance, we will focus only on the first category.

It has already been proposed a number of methods for establishing lists of potentially visible objects to eliminate the non-visible objects. A first visibility calculation method for establishing lists of potentially visible objects is based on the technique of eliminating hidden faces. This technique consists in rejecting the faces of objects whose normal has an angle between [-Pi / 2; Pi / 2] with a given observation direction [1, 4, 8, 9, 13, 16]. A second method of calculating visibility for establishing lists of potentially visible objects is based on the technique using the pyramid of view. This technique starts from the observation that all the objects located outside this pyramid at a given instant are rejected during the rendering phase. Several methods using hierarchical structuring of the scene have been coupled to this method.

A third method of calculating visibility starts from the following property: to test the occultation of each polygon being too slow, it is more efficient to use techniques based on the use of hierarchical structures of the scene and bounding volumes [9 ]. The three processes that have just been presented make it possible to reject part of the scene of the 3D rendering phase.

However, for performance reasons, these methods are not applicable to real-time use for overly complex databases. The method of the present invention is usable in various 3D applications, but is particularly suitable for applications proposing a ground-level 3D navigation in a 3D scene representing a virtual city.

In this type of application, the most important property is that at a given observation position, only a very small part of the buildings constituting the 3D scene is visible because of inter-building occultations.

In the following, we will focus on the various processes implementing the principle of inter-object occultation.

A first method of calculating visibility using the inter-object occultation principle is the Coord and Teller process [5, 6]. This process is based on the use of large convex blackout objects.

This method determines occultation planes valid relative to a certain displacement of the observer. Several plans are defined with respect to a set of observation points. There are several evolutions of this process based on different structuring of the 3D scene. Other visibility calculation methods implementing the principle of inter-object occultation have also been proposed. These methods are based on the property of the occultation volumes (shadow volume) [2, 11] generated by a point of observation and the outline of some objects that obscure the rest of the scene. Indeed, at a given point of observation, we can define a shadow volume centered on the point of observation and based on the outline of occulting objects guaranteeing that any object included in this shadow volume will be not visible from the observer. Several processes mixing scene structuring and this property have been proposed. Concerning the choice of occulting objects, several dynamic or static strategies have also been proposed.

The methods presented above calculate the visibility with respect to a point of observation. This requires, when moving the user in the 3D world, to recalculate the visibility at each new position. These methods are therefore heavy to implement and are not applicable to real-time use for databases that are too complex.

Another approach in the visibility calculation is to make a partition of the 3D navigation space and for each element of the partition, called the view cell, of. calculate the list of potentially visible objects.

Thus, rather than performing a visibility calculation at each position of the user, this visibility calculation is performed for areas, or cells of view, of the 3D space. In the following we will focus particularly on this type of visibility calculation method for determining lists of potentially visible objects by region (that is to say by view cell).

As in the method of Coord and Teller, there are methods that select large convex occult objects [15] with respect to the view cell.

However, the strong constraint on the selection of occulting objects takes into account that these processes are not most interesting in real cases.

Another method is proposed by Durand and Al. This method [7, 10, 14, 19] tests the non-visibility of objects by projecting the occulting objects and the objects on a 2D plane and checking the inclusion of the objects projected on the projections of the occulting objects.

However, this process is difficult to implement because of the importance of the choice of projection planes and occulting objects taken into account.

One can also mention a method [12] dynamically calculating virtual occulting objects. The interest of this type of process is to take into account the merger of shadow volumes of the occulting objects.

Recently Peter Wonka [17, 18] proposed a very promising visibility calculation method that uses the characteristics of the wired z-buffer of modern maps. This method can be used for example on urban databases, that is to say 3D scenes representing virtual cities. These scenes include objects that are buildings on a lot.

This process assumes that the urban database is flat, that is, the land on which the buildings are built is flat. This method also assumes that each building is represented by a 2D model ¹ A

The 2D ^1/4 model of a building includes an altitude (null by default), a footprint defined by a list of points, and a height. From this information (footprint E, altitude A, height H), it is possible, as illustrated in FIG. 1, to draw a 3D building by extruding a prism 24 at altitude A, the base of the prism 24 corresponding to the impression E and the height of the prism being equal to the height H of the building.

Peter Wonka's process consists of different stages: - a first step is to make a partition of the navigable space into a view cell,

a second step consists in calculating visible minimum heights in order to determine a map of the minimum heights visible with respect to each view cell, and a third step is to test the visibility of the objects of the scene with respect to each view cell. according to the map of the minimum visible heights of the view cell.

The calculation of the map of visible minimum heights ensures that an object will not be visible from a view cell if its height is less than the minimum visible heights of the region where it is located. This map is obtained by a grid of the surface occupied by the urban environment and containing for each rectangle r the minimum visible height Hmin (r) in all the rectangle with respect to the cell of view.

An object entirely included in a rectangle r and of height less than Hmin (r) will not be seen from the view cell. This map is represented using a 2D matrix and is called the visibility matrix. In the following, the rectangles will be called pixels because of the use of 3D graphic imaging techniques to speed up processing.

The calculation of the map of the visible minimum heights uses the resources of the 3D graphics cards and links the dimension of the 2D zone for calculating the visible minimum heights with the number of pixels of the 3D graphics card and the surface covered by a pixel.

For example, for a graphics card of [1024, 768] pixels and a recovery value of 1 meter, a 2D area for calculating the minimum visible heights of 1024 * 768 square meters is obtained. So the map of the minimum visible heights will cover 1024 * 768 square meters of the 3D urban scene.

When calculating the map of the visible minimum heights, the method imposes to reduce the occulting objects of an erosion value denoted "e" in order to limit the errors related to the method used to calculate the visible minimum heights of said map. Due to modeling

2D ^1/2 buildings, the erosion operation is performed on the footprint of the buildings. The erosion value e (corresponding to the erosion amplitude) is related to the recovery value (corresponding to the size of a pixel) denoted k by the following relation: e> k / + Jl. In practice to take into account the occultation phenomena, the erosion value must be low.

Indeed, if the value of erosion is strong, the occulting objects will be too eroded or reduced, and thus the phenomena of occultation will not be taken into account. For each cell of view, the visibility test on the objects of the 3D scene is as follows.

For the objects contained in the area covered by the map of the minimum visible heights, the visibility test is done by comparing the height of an enclosing box of the object and the minimum visible heights of the pixels having an intersection with the footprint. on the ground of the object. For objects that are outside the area covered by the map of the minimum visible heights, the visibility test is shown in Figure 2. Wonka proposes the following modification of his process: the boundaries of the map of the minimum visible heights 1 make it possible to define a horizon of visibility 4. This skyline 4 assures us that every object

3 outside the region covered by the map of the minimum visible heights 1 and whose projection 5 is below the horizon line 4 is not seen from the view cell.

Peter Wonka's method therefore does not deal with the inter-building concealment that is outside the overlap area, ie outside the area covered by the map of the minimum visible heights. .

The Wonka process has the following disadvantages: i. The process does not take into account field data

Indeed, the method of Peter Wonka assumes that the ground is flat. This implies that the Peter wonka process does not take into account the field data neither for the occultation phenomena, nor for the visibility tests.

In practice, however, the terrain is very often in relief. Thus, some parts of the field may obscure areas of the scene. For example, a user in front of a hill does not see what is behind the hill.

In addition, some areas of the scene may be visible from some parts of the terrain. For example, the number of buildings visible to a user at the top of a hill will be greater than the number of visible buildings for a user who is at the bottom of a hole. ii. The process does not take into account buildings of different altitudes.

Indeed, as previously described, the method of Peter Wonka assumes that the ground is flat, and therefore that the buildings are all at the same altitude. iii. The process does not take into account urban furniture type data, vegetation, road signs and specific buildings.

Indeed, the process of Peter Wonka takes into account the buildings represented with a 2D model ^1A Thus, it does not take into account the urban furniture type objects such as bus shelters, benches, billboards or road signs. It does not take into account vegetation such as trees, or groves. For more realism an urban database should take into account all this data.

The process of Peter Wonka does not take more into account the specific buildings, that is to say of modeled buildings using models other than the 2D model ^1A Indeed, for types of buildings such as churches or monuments, the 2 D / 4 model is not suitable for viewing. To represent all these specific buildings, for example, a 3D facet model is used.

For more realism, these specific buildings should be taken into account for the visibility tests. iv. The area of recouyrement of the map of the minimum heights visible only a part of the city.

Indeed, not to reduce too much the faces of the blackout buildings, e is generally equal to one meter, which requires that the recovery value of a pixel (k) is less than or equal to ^ β meter. Indeed, the values k and e are linked by the relation e ≥k / y / 2. So if we assume an erosion value e equal to one meter and a graphic card of 1280 * 1024 pixels, then the map of the minimum visible heights will cover a maximum of 1817 * 1454 square meters, and in real cases, the cities are of the order 8km * 8km.

To cover a larger area it is necessary to increase the value of k, which implies an increase in the erosion amplitude. For example for k equal to

e must be greater than or equal to 2 meters. Therefore, the larger the value of e, the smaller the blackout faces will be, and the less occultation will be taken into account. For example, an erosion amplitude of 2 meters removes 4 meters per building face. v. The Wonka process does not take into account the inter-building occultation on the entire urban database.

Indeed, as previously described, the inter-building concealment is not taken into account outside the area covered by the map of the minimum visible heights. vi. The erosion operation generates visibility.

Indeed, one of the disadvantages of the erosion operation is that this mathematical operation generates visibility between the buildings having at least one stop in common, even if this visibility does not exist in the initial database. Thus, as represented in FIG. 3, an object 6 which is situated, with respect to the current view cell, behind two buildings 7 and 8 having a stop 9 in common, and which is not visible before the erosion operation. (step

10) will become visible after the erosion operation (step 11). vii. The size of the list of potentially visible objects increases exponentially by taking urban databases of increasing size.

In fact, the inter-building concealment is not taken into account outside the area covered by the map of the minimum visible heights.

Thus, any object 3 located outside the area covered by the map 1, and whose projection 5 on the horizon of visibility 4 has a height greater than the minimum visible heights will be considered potentially visible. However, it can be hidden by another building located outside the area covered by map 1.

For example, assuming, as illustrated in FIG. 4, four buildings 12, 13, 14, 15 of the same width and an observation point 18 aligned in an observation direction, the heights of the buildings 12, 13, 14, 15 being respectively 5, 25, 20 and 10 meters from the nearest to farthest from the observation point 18.

If it is also assumed that these four buildings 12, 13, 14, 15 are located outside the area 16 covered by the map 1, and the height of the horizon of visibility 17 is 6 meters.

Then building 12 (5 meters) closest to the viewpoint will be considered hidden, and the following three buildings 13, 14, 15 (25, 20 and 10 meters) will be considered as potentially visible, their projections on the visibility horizon 17 having heights greater than the height of the horizon of visibility 17 (6 meters).

These three buildings 13, 14, 15 will therefore be part of the list of potentially visible objects although two of them are not visible (the buildings 14, 15 of 20 and 10 meter being hidden by the building 13 of 25 meters). . An object of the present invention is to provide a general method for determining a list of potentially visible objects by region for 3D scenes of very large sizes representing virtual cities of varying altitudes to overcome most of the aforementioned drawbacks.

PRESENTATION OF THE INVENTION ¹

The invention relates to a method for determining lists of potentially visible objects in a 3D scene comprising objects on a terrain of variable altitude and in which a user is able to move virtually on a displacement zone corresponding to the whole. areas of the ground not covered by the objects, characterized in that it comprises the steps of:

- create terrain objects each corresponding to a part of the terrain, - partition the user's movement zone into a network of view cells, determining, for each view cell, a list of objects potentially visible to the user from the view cell, said list being determined by testing the visibility of the objects of the scene and the created terrain objects. The present invention thus makes it possible to treat the cases of real urban environments for which the effect of the natural relief is far from being negligible. The creation of paving stones makes it possible to take into account the occultation generated by the relief of the terrain. The present invention makes it possible to take into account complementary urban data (street furniture, vegetation, signage and field data) in visibility. The method of the present invention is completely independent of the memory resources with respect to the size of the bases since the size of the rectangles of the grid is adapted according to the available memory resources for the calculation of the map of the visible minimum heights. Some methods of calculating visibility can be very effective but they require storage in local memory of a data structure describing the entire scene. This is not the case of the method of the present invention since the only data structure that must be in RAM is the visibility matrix (or map of the minimum visible heights) whose size is fixed in advance, independently of the size of the urban environment to be treated. In addition, this data structure is stored in the memory of the graphics card and not in the central memory of the computer.

Preferred, but not limiting, aspects of the method according to the invention are the following: the method further comprises a step of:

calculating a minimum altitude zmin of the terrain in order to define a plane Pmin at the minimum altitude zmin; the step of creating terrain objects comprises the substeps of: - Make a grid in M rectangles of the plane Pmin on all the surface of the ground,

associating a prism called a terrain object with each rectangle of the grid, the prism being defined by a base, an altitude and a height, the base of the prism being the rectangle, the altitude of the prism being equal to the minimum altitude of the surface terrain having a non-zero intersection with the prism, and the height of the prism being equal to the maximum altitude of the terrain surface having a non-zero intersection with the prism; the step of partitioning the user's moving area into an array of view cells comprises the substeps of:

translating the user's movement zone in the plane Pmin; sharing the translated of the movement zone into a triangle network;

extruding a prism representing the view cell on each triangle, the height of the view cell being equal to a reference height to which is added the maximum value of the heights of the terrain objects having a non-zero intersection with the prism; the scene comprises generic objects defined by a footprint, an altitude and a height, and in that the method further comprises a step of transforming the generic objects, and field objects into modeled 2D ^1/2 occulting objects by: - extending vertically the faces of the generic objects up to the altitude zmin, each occulting object thus obtained being defined by a fingerprint corresponding to the projection of the generic object on the plane Pmin, an altitude equal to zmin and a height equal to the sum the altitude and height of the generic object, - vertically extending the bases of the terrain objects to the altitude zmin, each occulting object thus obtained being defined by an imprint corresponding to the projection of the terrain object on the Pmin plane, an altitude zmin and a height equal to the altitude of the terrain object of which it is the extension; the scene comprises specific objects modeled by an assembly of polygons, and in that the method further comprises a step of transforming the specific objects, the generic objects and the field objects into 2D ^1/2 extended objects, said step comprising :

- the definition of an enclosing volume for each specific object of the scene,

the transformation of generic objects, field objects and bounding volumes into 2D ^1/4 extended objects, said transformation consisting in extending the faces of generic objects, field objects and bounding volumes to the Pmin plane, each extended object thus obtained being defined by a footprint equal to the projection on the Pmin plane of the generic object, of the terrain object or of the enclosing volume associated with it, by an altitude equal to zmin and a height equal to the sum of the altitude and height of the generic object, the terrain object, or the enclosing volume associated with it; the step of determining, for each view cell, a list of potentially visible objects comprises the substeps of:

calculating a first map of visible minimum heights, said map being obtained by: o carrying out a grid in N rectangles of the plane Pmin, the area occupied by each rectangle being a function of a recovery value k1 chosen so that the grid covers all the surface of the ground, o selecting occulting faces, in particular among the faces of the occulting objects, and by eroding the occulting faces selected by an erosion value e1 proportional to the recovery value k1, o calculating for each rectangle the minimum height visible from the view cell, - determining a temporary list of potentially visible objects by comparing, for each object of the scene and each field object, the height of the associated extended object and the minimum heights of the first map of the rectangles intersecting with the imprint of the extended object, - calculating a second map of the visible minimum heights, said map being obtained by: o carrying out a grid in N rectangles of the plane Pmin on part of the terrain, said part including the base of the view cell and the area occupied by each rectangle being a function of a recovery value k2 less than k1, o selecting the occulting faces, in particular among the faces of the occulting objects, and by eroding the occulting faces selected by an erosion value e2 proportional to the recovery value k2, o calculating for each rectangle the minimum height visible from the view cell,

determining the list of potentially visible objects by comparing, for each object in the temporary list of potentially visible objects, the height of the associated extended object and the minimum heights of the second map of the rectangles intersecting with the the prolonged object; the method further comprises a step of merging related blackout objects to obtain related sets, said merging step being performed prior to the step of selecting the blackout faces, and consisting of: - to determine groups of occulting objects satisfying conditions of connectivity,

Then for each group of related blackout objects, determine a related set by: - deleting fingerprints of objects the edges they have in common in order to obtain a unique fingerprint that is assigned to the related set,

- determine, among the objects in the group, the one whose height is the smallest and assign its height to the related set; the conditions of connectivity to be verified by a set of objects are that their fingerprints have at least two consecutive vertices in common; the recovery value k1 is calculated using the relation: k1 ≥max (Δx / Nx, Δy / Ny). Or :

- Δx and Δy are the width and the length of the ground,

Nx, Ny are the number of rectangles usable in row and in column for the map of the visible minimum heights, so that k1 is greater than or equal to the maximum between the ratio of the width of the land by the number of rectangles usable in line, and the ratio of the length of the land by the number of rectangles usable in column, and in that the erosion value e1 is deduced from the preceding calculation by using the relation relating the erosion value el to the recovery value k1: el ≥k1 / _\ 2; the erosion value e2 is chosen to be substantially equal to 1 meter, and in that the recovery value k2 is deduced from the value of e2 chosen by using the relation linking the erosion value e2 to the recovery value k2: a >k2Λβ; the step consisting in selecting occulting faces, in particular among the faces of the occulting objects (300, 301), consists of:

selecting all the faces of the occulting objects (300, 301); selecting all the faces of the sets of related occulting objects; the step of eroding the blackout faces with an erosion value e consists in eroding the footprints of the blackout objects and sets of obscuring objects related to the erosion value e;

The invention also relates to a system capable of determining lists of potentially visible objects in a 3D scene comprising objects on a terrain and in which a user is likely to move virtually on a displacement zone corresponding to all the zones. land not covered by the objects, characterized in that it comprises:

means capable of creating terrain objects, each corresponding to a part of the terrain,

means capable of partitioning the user's movement zone into a network of view cells,

means capable of determining, for each cell of view, a list of objects potentially visible by the user from the view cell, said list being determined by testing the visibility of the objects of the scene and the created terrain objects.

PRESENTATION OF FIGURES

Other features, objects and advantages of the present invention will emerge from the description which follows, which is purely illustrative and nonlimiting and should be read with reference to the accompanying drawings, in which:

- Figure 1 is a perspective view of a model object 2OV ₂ ; FIG. 2 is a perspective view illustrating a visibility test of the method of Peter Wonka applied to an object of a scene, the object being located outside the area covered by the map of the minimum visible heights; FIG. 3 is a front view of three objects before and after an erosion operation;

FIG. 4 is a front view illustrating the visibility test of the method of Peter Wonka applied to four objects of a 3D scene, the four objects being located outside the area covered by the map of the minimum visible heights;

FIG. 5 is a perspective view of a scene on which a user is located, said scene comprising an occulting face of an occulting object, a visible object and a non-visible object;

FIG. 6 is a perspective view of a scene comprising objects on a terrain of variable altitude;

FIG. 7 is a front view of a scene comprising objects on a terrain of variable altitude;

FIG. 8 is an illustration of a triangulation step and comprises top views of a scene before and after triangulation, and a perspective view of a view cell obtained by extruding a prism on a triangular partition obtained by triangulation;

FIG. 9 is a perspective view of a view cell that overlaps two terrain objects; FIG. 10a is a front view of modeled blackout objects

2DY ₂ obtained from generic objects of the stage and terrain objects;

FIG. 10b is a front view of 2DVz modeled extended objects obtained from the objects of the scene and the field objects; FIG. 11 is a first illustration of the step of fusing related blackout objects and comprises a view from above of three footprints of related blackout objects, and a top view of a fingerprint resulting from the fusion of the fingerprints of the three related blackout objects;

FIG. 12 is a second illustration of the step of fusing related blackout objects and comprises a front view of three related blackout objects, and a front view of the object resulting from the fusion of the three related blackout objects;

FIG. 13 is a top view of a scene on which there is a user and two blackout faces of occulting objects; FIG. 14 is a flowchart illustrating the various steps of the method.

DESCRIPTION OF THE INVENTION

An object of the present invention is to provide a general method for determining a list of potentially visible objects by region for very large 3D scenes representing virtual cities of varying altitudes.

Thus, the visibility calculation method presented below makes it possible to determine lists of objects that are potentially visible in urban scenes of real size of the order of 8km * 8km including objects, particularly representing buildings, on a parcel of land. variable altitude. In addition, this method takes into account the effects of the terrain for occultation and visibility calculations. Finally, the process incorporates for all visibility tests all types of objects (street furniture, vegetation ...).

For this, the inventors started from the observation that at ground level very few objects of a scene like street furniture are visible because of the inter-object occultation.

Indeed, as shown in FIG. 5, at an observation point 20 located near a face 21, a large volume 22 of the 3D scene, called shadow volume, is not visible from the point of view. observation 20. Thus, an object 23 included strictly in this shadow volume 22 will not be visible from the observation point 20. Therefore, during a ground navigation in an urban scene representing a virtual city, very few objects 19 will be visible from an observation point 20. The determination of a minimal subset of potentially visible objects per region in a scene is a complex problem that can not be solved without hypothesis.

With reference to FIG. 6, the hypotheses of the method are as follows: the objects of the scene (buildings, signage, tree, ...) are

represented in a direct orthonormal coordinate system (O, X, y, Z);

- the set of objects of a scene, is composed of: o generic objects 100, that is to say objects represented by a 2D / 4 model (namely a footprint, an altitude and a height), these generic objects 100 are buildings; Moreover, the imprints of objects

-> - »generics are parallel to the plane (O ₁ -X y) and their altitude is variable; o specific objects, that is objects modeled by assembling polygons (3D facet model), these specific objects are specific buildings (churches, monuments ...), street furniture (bus shelters) , benches, fires, stops ...), and vegetation (trees); o a terrain 101 (digitized relief) modeled by assembling polygons;

each generic object 100 may be associated with a roof 102 modeled by assembling polygons;

- the objects of the scene are included in a box of dimension (Δx, Δy); the number of pixels that can be used for the map of the visible minimum heights is equal to (Nx, Ny);

As those skilled in the art will have understood, the visibility calculation method makes it possible to process digital data. These digital data are characteristic of virtual objects, for example buildings, bus shelters, or terrain.

To virtually represent a city in 3D, a mobile or fixed terminal is used as a work station, said terminal comprising memory means for storing a database comprising these digital data.

For example, as far as terrain is concerned, it is modeled by assembling polygons. To define a polygon, it is necessary to know certain information. This information can for example be

the coordinates in the coordinate system (O, X, y, Z) of three characteristic points of the polygon.

For each polygon, the numeric data associated with it includes this information. These digital data are stored in the workstation database, along with other information about the polygon such as its texture for example. From these digital data, the processing means of the central unit are able to calculate the position of all the points of the polygon to be displayed. By displaying all the points of the set of polygons on display means of the workstation, a representation of the terrain is obtained. In the following, we will present the visibility calculation process by referring to the objects (land, buildings, street furniture) rather than referring to the associated numerical data on which the treatments are carried out. With reference to FIG. 7, the first step of the method consists in calculating the minimum altitude of the terrain of the stage. In the following, this minimum altitude will be called altitude zmin.

The second step of the method consists in defining a plane Pmin 103 at the altitude zmin. This plan will be used in some of the following steps of the process.

The third step of the method consists in calculating the field 104 modeled 2D A ¹ These field objects objects are calculated in particular to take into account the concealment of objects generated by the field. To create these terrain objects 104, a 2D regular grid of the Pmin plane 103 is defined. To each rectangle of the grid, a 3D block is associated, of which: the lower base is the rectangle itself, the minimum altitude 105 is equal to the minimum altitude of the terrain surface having a nonzero intersection with this tile, the height 106 is equal to the maximum altitude of the terrain surface having a non-zero intersection with this tile.

These 3D blocks are called field objects and will then be used in the visibility calculation process. The fourth step of the method consists of partitioning the near-ground observation space into regions called view cells. Indeed, the exact calculation of visible objects from a 3D volume is very difficult and very expensive in terms of calculation time and material resources.

As illustrated in FIG. 8, in a visualization application of urban environments, the space 25 in which the virtual observer can move is equal to the complement of the union of all the objects 26, 27, 28, 29 of the scene.

The partition of the user's moving area is obtained very simply as follows. We translate the terrain and the objects of the scene in the plane Pmin 103. Then, we carry out a 2D triangulation of the complementary

(translated at the altitude z = zmin) footprints on the ground of the objects of the scene. This triangulation consists of sharing the user's zone of displacement in a network of triangles 30.

The 3D partition is deduced from the triangulation by extruding a prism on each triangle previously obtained. The prisms thus created are called view cells 31.

As illustrated in FIG. 9, the altitude 108 of a cell 107 is equal to at least 109 of the altitudes of the lower bases 109, 110 of the terrain objects 111, 112 having an intersection with the view cell 107, and the height of the a view cell 107 is equal to a reference height 113 (between 0 and 2 meters at the city scale) to which is added the maximum value 114 of the heights 114, 115 of the field objects having an intersection with the cell .

With reference to FIG. 10a, the fifth step of the method consists of transforming the generic objects 100 and the field objects 104 of the scene into occulting objects 300, 301 of the same altitude zmin. These occulting objects are modeled 2D ¹ A Indeed, the model 2D ^1/4 simplifies the calculations of visibility. These occulting objects 300, 301 will be used when calculating maps of the minimum visible heights.

To transform the generic objects 100 into occulting objects 301, the faces of the generic objects 100 are prolonged in the plane Pmin 103 of altitude zmin. We then obtain prisms modeling extended generic objects. These prisms are called occulting objects 301.

The imprint of an occulting object 301 corresponds to the projection on the plane Pmin of the generic object of which it is the extension. All these occulting objects 301 have the same altitude zmin. The height of an occulting object

301 is equal to the sum of the altitude and the height of the object 100 associated with it.

To transform the terrain objects 104 into occulting objects 300, the prints of the terrain objects 104 are vertically extended in the plane Pmin 103. Prisms are then obtained modeling the extended field objects. These prisms are called occulting objects 300. The footprint of an occulting object 301 is equal to the footprint of the terrain object 104 translated at the altitude zmin, the altitude is equal to zmin, and the height is equal to the altitude of the terrain object 104.

We then obtain the set of occulting objects 300, 301 as shown in Figure 10a.

Then, for each cell of view, one computes the list of the potentially visible objects (LOPV). This list results from a minimalist evaluation of all visible objects (objects of the scene and field objects) since at least one of the points of the current view cell. The calculation of the list of potentially visible objects is divided into two phases.

First, a first phase effectively resolves the far-end inter-object occultation and calculates, for each view cell, a temporary list of potentially visible objects using a first map of minimum heights. visible. This map is obtained by carrying out a grid in the plane Pmin of the surface of the ground, the dimensions (length, width) of this first map being a function of a recovery value k1.

Then a second phase makes it possible to effectively take into account the close inter-object occultation, and calculates, for each view cell, the list of potentially visible objects using a second map of the visible minimum heights. This map is obtained in the same way as before. The dimensions (length, width) of this second card are a function of a recovery value k2. In this second phase, the visibility tests are done only on the objects of the temporary list calculated in the first phase.

The first step in the first phase of calculating the list of potentially visible objects is to choose the recovery value k1. This recovery value is chosen according to the area occupied by the terrain. The area occupied by a pixel depends on this recovery value k1. In order to be able to take into account all the phenomena of inter-object occultations, one chooses the recovery value k1 so that the first map of the visible minimum heights covers the whole city. For this, the recovery value k1 must be such that: k1> max (Δx / Nx, Δy / Ny). Or :

- Δx and Δy are the width and length of the ground,

- Nx, Ny are the number of pixels usable in line and in column for the calculation of the maps of the visible minimum heights.

Knowing the value of recovery k1, we deduce the erosion value el which must be greater than or equal to YΛIsJl. Indeed, the recovery value k1 and the erosion value el are linked by the relation: el ≥k1Λ £. It should be noted that the value of the erosion to be applied will be great. For example, if we assume a city of dimension 8km * 8km, and if we assume that we have a means of 3D graphics processing allowing us to use 1280 * 1024 pixels for the calculation of the first map of minimum heights visible. Then the values of k1 and el become: k1 = max (8000/1280, 8000/1024) = 6.25 meters el = 6.25 / ^ = 4.42 meters.

The second step in the first phase of calculating the list of potentially visible objects is to choose blackout faces.

In his process, Peter Wonka takes as faces all the faces of generic objects.

In the present method, all the faces of the occulting objects 300, 301, that is to say the faces of the extended generic objects and extended field objects, will be selected as the blackout faces. To this set, we also add the faces of groups of merged occulting objects. To obtain groups of merged obscuration objects, the fusion method described below is used. "Let a set B of objects modeled 2DA ¹ , then one can realize a partition (in the sense ensemblist), of all the objects, by the criterion of connectivity of the footprints on the ground following:

Let Bi and Bj be two objects of set B, then Bi and Bj are said to be related if their footprints on the ground have at least two consecutive vertices in common. "

From this definition we deduce a partition of the set B by related subassemblies:

B = {Uj Bci for i ranging from 1 to N} Bci is therefore one of the subsets of related occulting objects, that is to say in which any two objects taken in the same subset Bci necessarily belong to at least one connected component of Bci. BICs will be called "related sets".

Let be a connected set of occulting objects as defined above. Then the result of the merging of all the related occulting objects is a related set defined as follows:

- its footprint is the result of merging the footprints of the occulting objects obtained by deleting the edges in common,

- its height is equal to the minimum height of all the heights of all the occulting objects.

Indeed, as shown in Figure 11, three blackout objects whose footprints 32, 33, 34 have stops 35, 36 in common will be merged by deleting their edges 35, 36 together to obtain the footprint 37 of the related assembly. Moreover, as represented in FIG. 12, three blackout objects

38, 39, 40 having faces 41, 42 in common will be merged by deleting their faces in common 41, 42. The height of the connected assembly 141 obtained by merging the occulting objects 38, 39, 40 will be equal to the height of the occulting object 38 which is the lowest of the three occulting objects 38, 39, 40. Thus, the step of selecting the occulting faces consists in selecting all the faces of the occulting objects 300, 301, and the faces of the related sets.

In the third step of the first phase of the calculation of the list of potentially visible objects for the current view cell, an operation of erosion of the blackout faces is performed. This makes it possible to correct errors related to the method used to calculate the maps of the visible minimum heights.

A mathematical erosion operation is defined by the following property:

"A 'is said to be eroded from A by e if A' is included in A and if for any pair of points (a, a ') such that' a '' belongs to A 'and' a 'belongs to A, the distance from "a" to "a" is greater than or equal to e. "

Due to the 2D ^1/4 modeling of the occulting objects and related sets, erosion by e1 is done on the footprint of the blackout objects and related sets.

The fourth step in the first phase of calculating the list of potentially visible objects is to calculate the first map of the minimum visible heights for the current view cell. Since the calculation of a map of the visible minimum heights is a difficult and time-consuming problem, a minimal approximation of this calculation is proposed. This approximation consists in sampling with points the edges of the upper face of the current view cell and calculating a map of the minimum heights visible at each of these points.

The calculation of the map associated with the current view cell is then simply deduced from the maps associated with the sampling points of the current view cell.

To calculate a map of the minimum visible heights Hp (r) from a given observation point p, we define the notion of shadow volume as follows: Let either an observation point 20 (see FIG. 5) and a face 21 of an occulting object modeled by a rectangle perpendicular to the ground, then called shadow volume 22 the truncated pyramid defined by the four half-lines passing through observation point 20 and the four peaks of face 21. "

This shadow volume 22 ensures that any object 23 entirely contained in this 3D volume is not visible from the observation point 43.

Consequently, the map Hp (r) of the minimum heights visible with respect to the observation point p 43 (see FIG. 13) is deduced from the shadow volumes 45, 48 by orthogonal projection on a matrix 46 of pixels of the volumes of shadows 45, 48 generated by the occulting faces 44, 47.

To calculate the map Hp (r) taking into account all the occulting faces with respect to the observation point p, the procedure is as follows. The set of maps Hp is determined, where o (r) takes into account only one occulting face o of the set FO of the occulting faces with respect to the observation point p. Then, the map Hp (r) is calculated.

Indeed, the map Hp (r) associated with the set of maps Hp, o (r) of all the occulting faces o is given by the formula:

Hp (r) = max {Hp, o (r), where ε FO}, where:

Hp (r) is the function representing the minimum visible height of a pixel r with respect to the observation point p,

- Hp, o (r) is the function representing the visible minimum height of a pixel r with respect to the observation point p and taking into account only the occulting face o,

- FO represents the set of blackout faces.

The maps Hp, o (r) are calculated by orthogonal projections of the shadow volumes on the pixel matrix 46 located in the plane Pmin.

But the discretization on a grid of pixels gives rise to errors. This is the reason why the mathematical erosion operation is carried out before the calculation of the first map of the minimum visible heights.

Once all the maps Hp (r) of all the sampling points have been calculated, the map Hmin (r) associated with the current view cell is determined.

The function Hmin (r) giving the visible minimum heights associated with a view cell knowing the functions Hp (r) of all the sampling points p is given by the following formula:

Hmin (r) = min {Hp (r), PE}, where: PE represents the set of sampling points taken on the edges of the upper face of the view cell.

The condition for this calculation to be valid is that the sampling interval is less than or equal to the erosion amplitude e of the blackout faces. The fifth step of the first phase of calculating the list of objects. potentially visible is to determine the temporary list of potentially visible objects in relation to each view cell.

First, we first convert generic objects (buildings 100), specific objects, and extended modeled objects field 2D objects ^1A

For this purpose, for each specific object modeled by polygon assembly, a bounding volume is determined.

Next, and as illustrated in FIG. 10b, the faces of the generic objects 100, the enclosing volumes, and the terrain objects 104 are prolonged in the minimum altitude plane Pmin 103 zmin. Prisms are then obtained modeling the generic objects 100, the enclosing volumes, and the field objects 104.

These prisms are called prolonged objects 305. The imprint of an extended object 305 corresponds to the projection on the plane Pmin of the generic object 100, the field object 104 or the enclosing volume of which it is the extension. The altitude of an extended object 305 is equal to zmin. The height of an extended object 305 is equal to the sum of the altitude and the height of the object (generic object, field object or enclosing volume) associated with it.

Thus, unlike the step of determining the first map of the minimum visible heights where it was the altitude of the terrain object that was taken into account in the occultation phenomena, here it is the height of the terrain object that is taken into account for visibility tests.

The set of extended objects 305 is obtained as illustrated in FIG. 10b. 305. For each extended object 305, this visibility test is done by comparing the height of the extended object 305 and the minimum visible heights of the first map of the minimum visible heights of the pixels. intersecting with the footprint of the object. For each view cell, a list of the potentially visible extended objects 305 is then obtained. Since each extended object 305 in the list is associated with an object of the scene or a field object, the temporary list of potentially visible objects is deduced.

At the end of the first phase, we thus obtain for each view cell a temporary list of potentially visible objects which is a sur¬ set of the minimum list since computed with a large e1. Thus, during the second phase, only the visibility of the objects belonging to the temporary list of potentially visible objects obtained in the first phase will be tested. The second phase of calculating the list of potentially visible objects includes a first step of choosing an erosion value e2. This erosion value e2 is chosen low (in practice less than 1 meter) in order to reduce as much as possible the phenomenon of reduction of the occulting faces. The recovery value k2 corresponding to the size of the pixels is deduced from the erosion value e2 by the formula connecting e and k: e ≥VJ-JZ. Since the erosion value e2 is small, then the recovery value k2 is small (less than k1). As a result, the second visible minimum height map only covers part of the terrain. This part covered by the second card includes the current cell The second step of the second phase consists in choosing the blackout faces.

The occulting objects determined during the fifth step of the method are used here.

The faces of the occulting objects and the faces of the pairs of related occulting objects are selected as occulting faces.

The merger of pairs of occulting objects is a special case of the fusion of groups of related occulting objects. Let there be a pair of related occulting objects, then we define their fusion as being the connected set defined by: - its footprint on the ground which is equal to the merging of the footprints, with deletion of the edges in common of the two related occulting objects - its height which equals at least the heights of the two related blackout objects. The third step of the method consists in eroding the blackout faces by the erosion value e2. The method used to erode the faces is identical to that used in the first phase.

The fourth step of the second phase is to compute the map of the minimum heights visible to the current view cell. This calculation is done in the same way as in the fourth step of the first phase. In this way, for each view cell, the second map of the visible minimum heights is obtained.

The fifth step of the second phase is to test the visibility of the objects in the scene to determine the list of potentially visible objects. This test is performed with respect to each view cell as a function of the second map of the visible minimum heights (associated with said cell). This test is performed on the objects of the temporary list of potentially visible objects and is a function of the distance of the objects relative to the area covered by the second map of the minimum visible heights.

Indeed, if the extended object 305 associated with the object (generic, specific or field object) tested has an intersection with the second map of the minimum visible heights, then the object is said to be potentially visible if the maximum height of the object The associated prolonged 305 is greater than one of the pixel heights of the second map intersecting with the object's footprint. If the extended object 305 associated with the object under test does not intersect with the second map of the visible minimum heights, an encompassing volume of the extended object 305 is projected (see FIG. 2) over a visibility horizon. If the projection of the object is above the horizon of visibility 4, the tested object (generic object, field or specific) is said potentially visible.

In summary, the visibility calculation method for determining lists of potentially visible objects by region for very large 3D scenes representing virtual cities includes the following steps: a) calculating the minimum height of the urban scene : zmin, and the definition of a plane Pmin at the altitude zmin (step 70 of figure 14), b) the creation of terrain objects (step 71 of figure 14): The field objects are obtained by grid of the plane z = zmin. For each grid rectangle, we associate a 3D block whose: • the lower base is the rectangle itself, the minimum altitude is equal to the minimum altitude of the terrain surface having a non-zero intersection with this block, the height is equal to the maximum elevation of the terrain surface having a non-zero intersection with this tile. This altitude is therefore also that of the upper base of the pavement, c) The partition of the user's moving area (step 72 of FIG. 14):

The partition in view cells is obtained by extrusion of prisms on the triangles resulting from the triangulation of the user's displacement zone translated at the altitude z = zmin. The altitude of a cell is equal to the minimum of the minimum altitudes of terrain objects intersecting with the view cell, and the height of a view cell is equal to the reference height to which the maximum value of the objects is added. pit heights intersecting the cell, d) selecting a view cell (step 73 of Fig. 14):

The upper edge of the current view cell is sampled by e, e) The calculation of the size of a pixel (k1) and the associated erosion value (e1) (step 74 of FIG. 14): k1 is calculated so that the first map of visible minimum heights covers the entire city. The erosion value e1 is deduced from the recovery value k1, f) The selection and erosion of the occulting faces (step 75 of FIG. 14):

For this, transforms generic objects and field objects modeled blackout 2D objects ^1A

Then, the blackout faces are selected. The faces of the occulting objects are chosen as the occulting faces and the faces of the related sets resulting from the merging of the groups of related occulting objects are added thereto. g) Calculation of the first map of visible minimum heights

Hmin1 (r) associated with the current view cell (step 76 of FIG. 14): The computation is done by the operation of union and intersection of the shadow volumes generated by the occulting faces and the sampled points of the cell of view. The calculation is done using the hard-wired functions of the 3D graphics cards accessible by the OpenGL graphic library. To calculate the first map of the minimum heights, a grid is made in N rectangles (pixels) of the plane Pmin, the area occupied by each rectangle being a function of the recovery value k1, then for each rectangle the minimum height visible from the view cell (31, 107), h) determining the temporary list of potentially visible objects (step 77 of Fig. 14):

To determine this list, we perform a visibility test of the extended objects (transformed 2D Vz field objects, bounding volumes and specific and generic objects) using the first map of visible minimum height.

The visibility test of the extended objects is as follows: an object is said to be potentially visible if its maximum height is greater than one of the heights of the pixels of the card having an intersection with the footprint of the object.

If a terrain object is potentially visible then, the entire soil surface contained in that terrain object will be considered potentially visible.

For generic objects raised with a roof, the test is done on the surrounding volume of the union of the object and the roof. i) The choice of a low value of erosion e2 and the deduction of the size of a pixel k2 (step 78 of FIG. 14): In practice, the erosion value e2 is less than 1 meter. j) The selection and erosion of the blackout faces (step 79 of FIG. 14):

The selected faces are the faces of the occulting objects and the faces of the fusions of all the pairs of related occulting objects. k) The calculation of the second map of the visible minimum heights Hmin2 (r) associated with the current view cell (step 80 of FIG. 14): The calculation is done in the same way as in step g). The length and the width of the second card are a function of the recovery value k2. I) Determining the list of potentially visible objects (step 81 of Figure 14):

For this, a visibility test is carried out using the second map of the minimum visible heights. The visibility test is only performed on the extended objects associated with objects in the temporary list of potentially visible objects according to the distance of the objects relative to the area covered by the map of the minimum visible heights. Indeed, if an object has an intersection with the area covered by the second map, then the object is said to be potentially visible if the maximum height of the extended object associated with it is greater than one of the heights of the second map of the minimum visible heights of the pixels having an intersection with the footprint of the object. In the opposite case, the enclosing volume of the object is projected on the horizon of visibility. If the projection of the object is above the horizon of visibility, the object is said to be potentially visible.

It should be noted that operations d) to I) are iterated for all view cells.

The method presented above can be implemented in a system for performing visibility calculations.

This system is for example a personal computer comprising memory means (RAM, ROM) connected to processing means such as a processor, display means such as a display screen and input means such as a keyboard and a mouse. This system can also be an Internet server. In this case, the server makes it possible to determine the lists of potentially visible objects, and forwards them online for example to a remote personal computer that will display the 3D scene.

REFERENCES [I] UIf Assarsson and Thomas Mller. Optimized view of frustum culling algorithms for bounding boxes

[2] J. Bittner, V. Havran, and P. Slavik. Hierarchical visibility with occlusion trees. [3] Edwin E. Catmull. Algorithm for Computer Display of Curved Subdivision

Surfaces.

[4] J. H. Clark. Hierarchical geometry models for visible surface algorithms

[5] Satyan Coorg and Seth Teller. Temporally consistent conservative visibility

[6] Satyan Coorg and Seth Teller. Real-time occlusion culling for models with large occluders

[7] Frédo Durand, George Drettakis, Joelle Thollt and Claude Puech.

Preserving visibility preprocessing using extended projections

[8] James D. Foley, Andries van Dam, Stephen K. Feiner, and John F.

Hughes. Computer Graphics, Principles and Pratice Second Edition [9] B. Garlick, D. Baum and J. Winget. Parallel Algorithms and Architectures for 3D Image Gegeration

[10] Ned Greene, Kass and Gary Miller. Hierarchical z-buffer visibility

[I] T. Hudson, D. Manocha, J. Cohen, M. Lin, K. Hoff, and H. Zhang. Accelerated occlusion culling using shadow frustra. [12] Vladelen Koltun, Yiorgos Chrysanthou, and Daniel Cohen-Or. Virtual occluders: An efficient pvs representation

[13] Subodh Kumar, Dinesh Manocha, William Garrett, and Ming Lin.

Hierarchical back-face computation

[14] Hong Lip Lim. Toward has fuzzy hidden surface algorithm. [15] Boaz Nadler, Gadi Fibich, Shuly Lev-Yehudi, and Daniel Cohen-Or. A qualitative and quantitative visibility analysis in urban scenes

[16] Mr. Slater and Y. Chrysanthou. View volume culling using a probabilistic cashing scheme

[17] Peter Wonka and Dieter Schmalstieg. Occluder shadows for fast walkthroughs of urban environments [18] Peter Wonka, Michael Wimmer and Dieter Schmalstieg. [19] Hanson Zhang, Dinesh Manocha, Hudson Thomas, and Kenneth E. Hoff III. Visibility culling using hierachical occlusion maps

Claims

A method of determining lists of potentially visible objects in a 3D scene including objects (100) on a terrain (101) of variable altitude and in which a user is likely to move virtually on a moving area (25). ) corresponding to all the areas of the terrain not covered by the objects (100), characterized in that it comprises the steps of: - creating terrain objects (104) each corresponding to a part of the terrain,

partitioning the user's movement area (25) into a network of view cells (31, 107),

determining, for each view cell (31, 107), a list of potentially visible objects by the user from the view cell, said list being determined by testing the visibility of the objects of the scene and the created terrain objects.

2. Method according to claim 1, characterized in that it further comprises a step consisting in:

calculating a minimum altitude zmin of the terrain (101) in order to define a plane Pmin at the minimum altitude zmin.

The method of claim 2, characterized in that the step of creating terrain objects (104) comprises the substeps of:

- Make a grid in M rectangles of the plane Pmin over the entire surface of the ground (101),

associating a prism called field object (104) with each rectangle of the grid, the prism being defined by a base, an altitude and a height, the base (105) of the prism being the rectangle, the altitude prism being equal to the minimum altitude of the terrain surface having a non-zero intersection with the prism, and the height of the prism being equal to the maximum altitude of the terrain surface having a non-zero intersection with the prism.

The method of claim 2 or 3, characterized in that the step of partitioning the user's moving area (25) into a view cell array (31, 107) comprises the substeps of: translating in the plane Pmin the movement zone (25) of the user,

- to share the translated of the zone of displacement in a network of triangles,

extruding a prism representing the view cell (31, 107) on each triangle, the height of the view cell (31, 107) being equal to a reference height to which is added the maximum value of the heights of the terrain objects having a non-zero intersection with the prism.

5. Method according to any one of claims 2 to 4, characterized in that the scene comprises generic objects (100) defined by a footprint, an altitude and a height, and in that the method further comprises a step of transformation of generic objects (100), and field objects (104) blackout objects (300.301) 2D modeled ^_half in:

vertically extending the faces of the generic objects (100) to the altitude zmin, each occulting object (301) thus obtained being defined by an imprint corresponding to the projection of the generic object (100) on the plane Pmin, a altitude equal to zmin and a height equal to the sum of the altitude and the height of the generic object; - vertically extending the bases (105) of the terrain objects (104) up to the altitude zmin, each occulting object (300) thus obtained being defined by an imprint corresponding to the projection of the terrain object (104) on the plane Pmin, an altitude zmin and a height equal to the altitude of the terrain object of which it is the continuation.

The method according to any one of claims 2 to 5, characterized in that the scene comprises specific objects modeled by an assembly of polygons, and in that the method further comprises a step of transforming the specific objects, the generic objects and field objects in extended objects (305) modeled 2D ^1/2 , said step comprising:

- the definition of an enclosing volume for each specific object of the scene,

the transformation of the generic objects (100), the field objects (104) and the bounding volumes into objects, extended (305) modeled 2D14, said transformation 'consisting in extending the faces of the generic objects, the field objects (104) and the bounding volumes up to the Pmin plane, each extended object

(305) thus obtained being defined by a footprint equal to the projection on the Pmin plane of the generic object, of the terrain object or of the associated bounding volume, by an altitude equal to zmin and a height equal to sum of the elevation and height of the generic object, field object, or enclosing volume associated with it.

The method according to claims 5 and 6, characterized in that the step of determining, for each view cell (31, 107), a list of potentially visible objects comprises the substeps of: calculating a first map of visible minimum heights, said map being obtained by: o carrying out a grid in N rectangles of the plane Pmin, the area occupied by each rectangle being a function of a recovery value k1 chosen so that the grid covers all the surface of the ground (101), o selecting the occulting faces (21), in particular from the faces of the occulting objects (300, 301), and eroding the selected blackout faces by an erosion value e1 proportional to the recovery value k1, o calculating for each rectangle the minimum height visible from the view cell (31, 107),

determining a temporary list of potentially visible objects by comparing, for each object of the scene and each terrain object, the height of the associated extended object (305) and the minimum heights of the first map of the rectangles intersecting with the object; imprint of the extended object,

calculating a second map of the visible minimum heights, said map being obtained by: o carrying out a grid in N rectangles of the plane Pmin on a part of the terrain (101), said part including the base of the view cell and the area occupied by each rectangle being a function of an overlapping value k2 less than k1, o selecting the blackout faces (21), in particular from the faces of the blackout objects (300, 301), and by eroding the blackout faces (21) selected by a value of erosion e2 proportional to the recovery value k2, o calculating for each rectangle the minimum height visible from the view cell (31, 107), - determining the list of potentially visible objects by comparing, for each object of the temporary list objects potentially visible, the height of the associated extended object and the minimum heights of the second map of the rectangles intersecting with the imprint of the extended object.

8. Method according to the preceding claim, characterized in that it further comprises a step of merging related occulting objects (38, 39, 40) in order to obtain related sets, said merging step being carried out beforehand. step of selecting the blackout faces, and consisting in: - determining groups of blackout objects (38, 39, 40) satisfying conditions of connectivity,

Then for each group of related occulting objects, determine a related set in:

- removing from the fingerprints (32, 33, 34) of the objects (38, 39, 40) the edges (35, 36) that they share in order to obtain a unique fingerprint (37) which is assigned to the set related (141),

determining, among the objects (38, 39, 40) of the group, the one whose height is the smallest and assigning its height to the related assembly (141).

9. Method according to claim 8, characterized in that the connectivity conditions to be verified by a set of objects (38, 39, 40) are that their indentations (32, 33, 34) have at least two consecutive vertices in common. .

10. Method according to one of claims 7 to 9, characterized in that the recovery value k1 is calculated using the relation: k1 ≥max (Δx / Nx, Δy / Ny). Or :

- Δx and Δy are the width and the length of the ground, - Nx, Ny are the number of rectangles usable in row and in column for the map of the minimum visible heights, so that k1 is greater than or equal to the maximum between the ratio of the width of the land (101) by the number of usable rectangles 5 in line, and the ratio of the length of the land (101) by the number of rectangles usable in column, and in that the erosion value e1 is deduced from the previous calculation by using the relation binding the erosion value el to the recovery value k1: el ≥k1 / -v2. 0

11. Method according to one of claims 7 to 10, characterized in that the erosion value e2 is chosen substantially equal to 1 meter, and in that the recovery value k2 is deduced from the erosion value e2 chosen using the relation binding the erosion value e2 to the recovery value k2: e2 ≥Yliφ.

12. Method according to one of claims 7 to 11, characterized in that the step of selecting blackout faces, in particular 0 among the faces of the occulting objects (300, 301), consists of:

selecting all the faces of the occulting objects (300, 301),

select the set of faces of the sets of related blackout objects. 5

13. Method according to one of claims 7 to 12, characterized in that the step of eroding the blackout faces by an erosion value e consists of eroding the footprints of the blackout objects and sets of blackout objects related by the e 0 erosion value e.

A system capable of determining lists of potentially visible objects in a 3D scene comprising objects (100) on a terrain (101) and in which a user is likely to move virtually on a movement area (25) corresponding to all areas of the field (101) not covered by the objects

(100), characterized in that it comprises:

means capable of creating terrain objects (104) each corresponding to a part of the terrain, means capable of partitioning the user's displacement zone (25) into a network of view cells (31, 107),

means capable of determining, for each view cell (31, 107), a list of objects potentially visible to the user from the view cell, said list being determined by testing the visibility of the objects of the scene and the ground objects created.