DE112017003426T5 - Method and system for reconstructing a three-dimensional representation - Google Patents

Abstract

The present invention relates to a method, a computer program and a system for reconstructing a three-dimensional image. Views of the image are captured and processed by means of a registration vector relative to the representative data of the pixels of a view corresponding to the pixels of the previous view and / or the representative data of the three-dimensional coordinates of given points in a view relative to the previous view.

Description

TECHNICAL FIELD OF THE INVENTION

The present application relates to the field of computer-implemented methods for reconstructing a three-dimensional representation. In addition, the application relates to the associated computer program products and systems.

TECHNOLOGICAL BACKGROUND OF THE INVENTION

It is known in the art to reconstruct three-dimensional images from a video or photo camera and perform computer processing of the captured images to reconstruct a three-dimensional "representation" of, for example, a scene or object, in particular in a virtual reality environment to study and / or use advanced reality.

Also known are various devices or three-dimensional data acquisition systems that generally use two sensors, at least one of which is an optical sensor (or optoelectronic, but no distinction is made here between the two terms). To improve the accuracy of the reconstructions, it is often suggested to use multiple cameras and to integrate the data into a computer. This type of solution leads to significant computation times and reduces the lifespan of the devices used for their implementation. Even if sometimes two optical sensors are used to perform stereoscopic detection, many other devices use a second sensor of a different type to provide three-dimensional data, often referred to as "depth data." As an example of a second type of sensor, the known infrared sensors may be mentioned, but other technologies may be used to obtain similar data, which generally represents the distance between the sensor and the detected target. Data of this kind, which originate from the acquisition of images and three-dimensional information representing the "three-dimensional detection", can advantageously be associated with data relating to the position of the sensors in space during the detection in relation to the three-dimensional data to have them placed in the room. These data make it possible, for example, to create representations (or reconstructions) of units such as rooms or environments as well as objects which are in particular reproducible, for example with three-dimensional printers (in the computer storage means).

In this optic, the accuracy of the reconstruction is a critical point and depends on the position of the optical sensor (generally a camera), but especially on the positions of that sensor in each successive shot and on the image quality of the image obtained by the camera. Moreover, this reconstruction generally requires ever more efficient computer systems for processing increasingly complex data to perform computations that allow characterizing various points within multiple images to reconstruct the three-dimensional representation of the acquired entity.

GENERAL DESCRIPTION OF THE INVENTION

The object of the invention is to propose a method which simplifies calculations and makes it possible, on the one hand, to reduce the computing times required for their execution and, on the other hand, to reduce the inaccuracy of the resulting three-dimensional representation.

• - repräsentative Daten von n Ansichten der Einheit, bestehend aus einer Vielzahl von Pixeln, die unter Verwendung mindestens einer ersten Erfassungsvorrichtung, optischer Sensor genannt, aufgenommen wurden, wobei n eine von Null verschiedene ganze Zahl ist,
• - repräsentative Daten von mindestens einer Entfernung, genannt Tiefe, gemessen mit mindestens einer zweiten Erfassungsvorrichtung, genannt Tiefensensor, zwischen mindestens einem Teil der Pixel in den n Ansichten und mindestens einer der Erfassungsvorrichtungen,
• - repräsentative Daten aufeinanderfolgender Positionen des Erfassungssystems bei jeder von n aufeinanderfolgenden Erfassungen, die unter Verwendung von mindestens einer dritten Erfassungsvorrichtung erfasst werden, genannt Positionssensor;

wobei das Verfahren durch ein System zum Rekonstruieren für eine dreidimensionale Darstellung implementiert wird, in dem die durch die Erfassungen erhaltenen n aufeinanderfolgenden Aufnahmen Daten liefern, deren Kombination Punktwolken darstellt, wobei das Verfahren dadurch gekennzeichnet ist, dass

• • jede der Punktwolken in Bezug auf eine jeweils gewöhnliche Referenz im Raum referenziert wird;
• • mindestens ein Vektor berechnet wird, um eine Neueinstellung zu definieren, die auf jede der aufeinanderfolgenden üblichen Referenz in Bezug auf die vorangehende Referenz anzuwenden ist;
• • zumindest ein Deskriptor mindestens einer Erfassung von Punktepaaren implementiert, die als Ähnlichkeitspunkte bezeichnet werden, die jeweils zu zwei aufeinanderfolgenden Ansichten gehören und eine bestimmte Ähnlichkeit zwischen den Werten von mindestens einem ihrer repräsentativen Daten haben, um als Referenz für die Neueinstellung zu dienen;
• • die Neueinstellung wird inkrementell auf die n aufeinanderfolgenden Erfassungen angewendet, so dass, einerseits jede Wiederholung es ermöglicht, die repräsentativen Daten einer gegebenen Ansicht und/oder die entsprechende Punktewolke in Bezug auf die vorangehenden zu registrieren und dass, andererseits die aufeinanderfolgenden Iterationen die Neueinstellung der Daten in Bezug auf das für die erste Punktwolke definierte übliche Referenz ermöglichen und als globale Referenz für die Rekonstruktion der dreidimensionalen Darstellung dienen,
• • die Paare, die als Referenz für die Neueinstellung dienen, werden durch mindestens einen der folgenden Schritte ausgewählt:
  • - Bildneueinstellung basierend auf den repräsentativen Daten der Pixel, die während der Erfassung der aufeinanderfolgenden Ansichten erfasst wurden, wobei diese Bildneueinstellung mindestens eine Auswahl der Paare unter Verwendung von mindestens einem Filter auf der Grundlage der repräsentativen Werte der Differenz zwischen den beiden Punkten eines Paares in Bezug auf die Translation und/oder Rotation;
  • • geometrische Neueinstellung basierend auf den Daten, die die Koordinaten der Punkte innerhalb der Punktwolken darstellen. und/oder
• • geometrische Neueinstellung auf der Grundlage der repräsentativen Daten der Koordinaten der Punkte in den aufeinanderfolgenden Punktwolken durch das Identifizieren für einen Punkt in einer Punktwolke, die einer gegebenen Erfassung entspricht, von mindestens drei Punkten, die in der Punktwolke der vorherigen Ansicht am nächsten sind, die im globalen Koordinatensystem festgelegt ist, so dass diese geometrische Neueinstellung diese drei Punkte, die als Referenzstandorte bezeichnet werden, als Bezugspunkt innerhalb der Punktwolke der vorherigen Ansicht verwendet, wobei diese drei Referenzstandorten ein Dreieck bilden, für das die Projektion des zurückzusetzenden Punktes der geringsten Entfernung entspricht.

In order to achieve this kind of goal, the invention proposes a method of reconstructing at least one three-dimensional representation of at least one examined physical unit based on detections corresponding to n successive recordings of a plurality of data which are preferably synchronized with each other in time and having:

• representative data of n views of the unit consisting of a plurality of pixels recorded using at least a first detector, called an optical sensor, where n is an integer other than zero,
• representative data of at least one distance, called depth, measured with at least one second detection device, called depth sensor, between at least a part of the pixels in the n views and at least one of the detection devices,
• representative data of successive positions of the detection system in each of n consecutive acquisitions detected using at least one third detection device, called position sensor;

the method being implemented by a system for reconstructing a three-dimensional representation in which the n consecutive exposures obtained by the captures provide data the combination of which represent point clouds, the method being characterized in that

• • each of the point clouds is referenced relative to a common reference in space;
• • calculating at least one vector to define a resetting to be applied to each of the consecutive common references with respect to the previous reference;
• • implements at least one descriptor of at least one detection of pairs of points, referred to as similarity points, each belonging to two consecutive views and having a certain similarity between the values of at least one of its representative data to serve as a reference for the readjustment;
• The readjustment is applied incrementally to the n consecutive acquisitions, such that, on the one hand, each repetition allows the representative data of a given view and / or the corresponding cloud of points to be registered with respect to the previous one and, on the other hand, the successive iterations Allow data in relation to the usual reference defined for the first point cloud and serve as a global reference for the reconstruction of the three-dimensional representation,
• • the reference reference pairs are selected by one or more of the following:
  • Image re-adjustment based on the representative data of the pixels acquired during the acquisition of the successive views, this re-adjustment of at least one selection of the pairs using at least one filter based on the representative values of the difference between the two points of a pair on translation and / or rotation;
  • • geometric resetting based on the data representing the coordinates of the points within the point clouds. and or
• • geometric resetting based on the representative data of the coordinates of the points in the successive point clouds by identifying, for a point in a point cloud corresponding to a given detection, at least three points closest in the point cloud of the previous view in the global coordinate system, so that this geometric re-adjustment uses these three points, referred to as reference locations, as a reference point within the point cloud of the previous view, these three reference locations forming a triangle for which the projection of the point to be reset corresponds to the shortest distance ,

According to a further feature, when the data obtained by the various devices of the detection system have been defined with reference to at least two different three-dimensional references, the method comprises, for each successive acquisition, at least one transposition of that data in a common three-dimensional reference in aligning the respective bases of these data relative to each other to obtain representative data of n point clouds referenced with respect to n consecutive common three-dimensional references defined in a common manner at each of the successive acquisitions to allow the registration of the point clouds To get reconstruction.

According to another feature, the method implements the selection of the pairs using at least one filter based on the representative values of the difference between the two points of a pair in translation, thanks to a search for a maximum number of pairs between which the translation is defined by a vector having a value which is lower than a certain translation threshold, for example an adjustable translation threshold.

• • mindestens eine Auswahl der Paare unter Verwendung von mindestens einem Filter basierend auf den repräsentativen Werten der Differenz zwischen den zwei Punkten eines Paares in Bezug auf die Drehung unter Verwendung von mindestens einer Auswahl von Referenzpunkten mit der größten Anzahl von Paaren, definiert als ihre zugehörigen Paare, mit denen sich ihre Registrierungsvektoren um einen Winkel unterscheiden, der kleiner ist als eine Winkelschwelle eines bestimmten Werts, beispielsweise ein einstellbarer Winkelschwellenwert,
• • die Auswahl basiert auf einer Abweichung, ausgedrückt durch ein Triplett von Drehwinkeln, zwischen den repräsentativen Vektoren von Paaren von Referenzpunkten in Bezug auf die repräsentativen Vektoren ihrer zugehörigen Paare.

According to another feature, the image registration comprises:

• At least one selection of the pairs using at least one filter based on the representative values of the difference between the two points of a pair with respect to the rotation using at least one selection of reference points having the largest number of pairs defined as their associated pairs with their registration vectors differing by an angle less than an angle threshold of a particular value, such as an adjustable angle threshold,
• The selection is based on a deviation, expressed by a triplet of rotation angles, between the representative vectors of pairs of reference points with respect to the representative vectors of their respective pairs.

According to another feature, the deviation corresponds to a function and the selection corresponds to a minimization of this function by successive iterations of individual rotations with a certain angle value, which is adjustable, for example.

According to another feature, the settable value is defined by a user or developer preparing the implementation of the method in the system.

According to another feature, the descriptor, which is responsible for the detection of pairs of points, corresponds to a Harris detector, which determines the similarity between the average values of the colors in their respective environments.

According to another feature, the selection step, which is based on the representative values of the difference between the two points of a pair with respect to the translation, uses a translation threshold of the order of 10%.

According to another feature, the selection step based on the representative values of the difference between the two points of a pair with respect to the rotation is based on an angle threshold on the order of one degree.

Another object of the invention relates to a computer program comprising sections of code representing instructions for carrying out the method according to one of claims 1 to 11, when the program is executed by means of a data processing means.

Another object of the invention is a system for reconstructing a three-dimensional image comprising data processing means executing the instructions stored in a storage device to implement the method according to one of the embodiments described in the present application.

A further subject of the invention relates to a system for reconstructing according to the preceding claim, characterized in that it comprises the detection means which provide the representative data of the consecutive recordings.

list of figures

• • 1 eine Datenerfassungsvorrichtung zeigt, die eine Einheit erfasst, um die Rekonstruktion einer dreidimensionalen Darstellung gemäß einigen Ausführungsformen zu ermöglichen;
• • 2A ein Beispiel einer Punktewolke darstellt, die von einer Datenerfassung stammt, die durch ein Erfassungssystem gemäß bestimmten Ausführungsformen durchgeführt wurde, und 2B zeigt ein Beispiel eines Satzes von ausgewählten Referenzstandorten, die durch ein Translationsfilter eines Systems zum Rekonstruieren einer dreidimensionalen Darstellung gemäß bestimmten Ausführungsformen ausgewählt wurden;
• • 3A ein Beispiel eines Satzes von Referenzstandorten zeigt, der durch ein Rotationsfilter eines Systems zum Rekonstruierens einer dreidimensionalen Darstellung gemäß bestimmten Ausführungsformen ausgewählt wurde, und 3B zeigt ein Beispiel eines Satzes von geometrisch neu eingestellten Punkten eines Systems zum Rekonstruieren einer dreidimensionalen Darstellung gemäß bestimmten Ausführungsformen;
• • 4 einige Ausführungsformen eines Verfahrens zum Rekonstruieren einer dreidimensionalen Darstellung darstellt, das innerhalb von zwei möglichen Ausführungsformen eines Systems zum Rekonstruieren einer dreidimensionalen Darstellung implementiert ist.

Further features and advantages of various embodiments of the present application will be described in the following description with reference to the accompanying drawings, in which:

• • 1 a data acquisition device that detects a unit to enable the reconstruction of a three-dimensional representation according to some embodiments;
• • 2A FIG. 4 illustrates an example of a point cloud resulting from data acquisition performed by a detection system according to certain embodiments; and FIG 2 B FIG. 12 shows an example of a set of selected reference locations selected by a translation filter of a system for reconstructing a three-dimensional representation according to certain embodiments; FIG.
• • 3A shows an example of a set of reference locations selected by a rotation filter of a system for reconstructing a three-dimensional representation according to certain embodiments, and 3B FIG. 12 shows an example of a set of geometrically reset points of a system for reconstructing a three-dimensional representation according to certain embodiments; FIG.
• • 4 some embodiments of a A method of reconstructing a three-dimensional representation implemented within two possible embodiments of a system for reconstructing a three-dimensional representation.

DESCRIPTION OF THE PREFERRED EMBODIMENTS OF THE INVENTION

The present application relates to a method and system for reconstructing a three-dimensional image. This type of method is generally implemented by computer means, and the system ( 1 . 1' ) for reconstructing a three-dimensional representation that allow its implementation, this computer means, in which data processing means ( 14 ) Execute statements written in storage means ( 13 ) are included (included in or retrievable by the system) so that the processed data is transformed and results in the result that the method can advantageously obtain. Thus, the present application does not describe the computer means in detail as they may vary as needed. In addition, processed, such as in 4 represented, the system ( 1 . 1' ) for reconstructing a three-dimensional representation of data from several detection devices ( 10 . 11 . 12 ) and can either be a system ( 1 ), which integrates these detection devices, or a system ( 1 ' ) containing only the data collected ( VE . PP . PS ) and the instructions to be used for the process to be reproduced. Thus presents 4 certain embodiments of a method for reconstructing a three-dimensional representation that in two possible embodiments of a system ( 1 . 1' ) are implemented to reconstruct a three-dimensional representation. It should be noted that the present application sometimes refers to the method by the term process, without particular distinction or limitation. In 4 possible steps are shown in the process, but sometimes some steps may be omitted, such as one of the choices ( 551 . 552 ) during image registration ( 55 ).

The method allows three-dimensional images of units ( e ) or to reconstruct elements, for example, to create expressions with three-dimensional printers or to examine the depth of the restored element.

This type of process is generally performed using a system ( 1 . 1' ) for reconstructing a three-dimensional representation that may or may not actually integrate the data acquisition device (s) for reconstruction, as in 4 that a system ( 1 ) with a plurality of detection devices ( 10 . 11 . 12 ) that benefit from their combination to provide the data that the execution of the method by data processing means ( 14 ) of the system ( 1 . 1' ), for example by means of instructions stored in storage means ( 13 ) and by these processing means ( 14 ) can be performed. By "detection device" is sometimes referred to an apparatus for capturing images, in particular the three-dimensional image to be reconstructed. For example, the first detection device ( 10 ) An optical data acquisition device, such as a video or photo camera. The second detection device ( 11 ) is able to measure depth data.

A problem in this area concerns the determination of a correspondence between the data acquired between a detection and the next acquired data within a three-dimensional detection system, in particular if the physical position of the detection devices varies between these two recordings. The present application therefore aims to reduce or even eliminate the inaccuracies of this estimated match, and / or to reduce the resources required to achieve such a result. Thus, we obtain data referenced in relation to a determinable global reference from a sequence of common references whose correspondence needs to be established.

This coincidence is particularly problematic in the frequent case of a shift of the usual successive references, that is, a displacement of the sensors relative to the detected unit. In addition, this shift may include at least one translation of the origin of the successive common references (as the optical sensor is moved) and / or rotation of the axes defining the basis of these successive references (if the orientation of the optical sensor is changed).

This displacement of the first and second sensors ( 10 . 11 ) during the acquisition of the successive recordings can advantageously be carried out with at least one position monitoring device ( 12 ), among which the gurometer is a non-limiting example, but is relevant in that it is inexpensive and widely used, and that it allows three-dimensional data to be collected in general are sufficient for the associated uses. In particular, it is generally desired to establish a relationship between data, such as optical or GPS data or others acquired during successive acquisitions. However, the inaccuracy of the gurometer (or even most devices for acquiring three-dimensional coordinates of various kinds) too often limits the quality of the three-dimensional representation that can be obtained by the methods known in the art or the cost to solve the problem is too great ,

The present application thus describes an invention which, moreover, offers a particular advantage that fits into this precise context by providing the data regarding the three-dimensional positions of the pixels or dots within the successive views or images (estimated by the three-dimensional coordinate detection devices) corrected while collecting these two types of data, which improves processing.

For this purpose, various embodiments of the present application propose to limit the error (detection of coordinates in space) possibly introduced by the positioning device during the estimation of the displacement of the optical and / or depth detection system.

Since the displacement of the sensors from one detection to another may involve translation and / or rotation of the detection system, in some embodiments, the suggested correction may be user adjustable and / or selectable within the suggested options (this determines whether at least one of these corrections must be used and / or how high the value should be to determine how this correction should be applied, for example via a user interaction interface ( 15 ), the variants of which are known to the person skilled in the art, and for which it is not necessary to go into detail here). Therefore, it may be advantageous to provide parameter flexibility since translation and rotation may result from various confounding factors but may have independent effects or consequences.

In addition, various embodiments of the present application advantageously relate to at least one problem relating to the correspondence between the bases and the origins corresponding to the various sensors that enabled the collection of data with respect to each other. This type of problem and solution generally implies that the bases and origins are readjusted relative to each other (preferably, if the relationships between these sensors are identified in terms of position and measured physical values). In certain embodiments, this method proposes, in a global reference, to transpose each of the successive references in relation to each other and then sequentially correct that translation by at least one translation of the origin of the successive conventional references relative to the others and / or at least one rotation Bases this successive conventional reference. For example, the method may be performed on the basis of data obtained with an image capture camera ( 10 ) and one on a system ( 1 ) for capturing or reconstructing attached depth camera ( 11 ) are simply an electronic tablet equipped with these devices. In addition, a position data acquisition device may be, for example, a gurometer and / or a GPS or other type of device that senses the movement of detection devices, although the method relates precisely to the need to register views when moving sensors one view to the other. Various embodiments of the present application have the peculiarity of being more advantageous when the successive movements measured during the recordings are inaccurate.

Preferably, the images are triggered by a program that the optical sensor ( 10 ), which is termed camera, controls without limitation, and the depth sensor ( 11 ), preferably synchronously, controls. In order to control the camera, various embodiments use computer modules (so-called "software development kit") that are capable of taking various pictures and / or acquisitions ( 51 ) at a given frequency and resolution. Thus, according to various embodiments, the program is preferably configured to control the camera in this manner.

The proposed method is based on data captured by these types of detection devices, but is not limited to this detection, since it preferably only relates to three-dimensional reconstruction, although it may include detection steps, such as in FIG 1 1, which shows an embodiment of a detection system ( 1 ) shows the representative data of a unit ( e ), while 4 two examples of systems ( 1 . 1' ), which differ in that the left one system ( 1 ) is to reconstruct and capture, while the right system is only one system ( 1' ) to reconstruct that is not a device ( 10 . 11 . 12 ) that captures the 51 ) allow the data used for the reconstruction.

• • repräsentative Daten (VE) von n Ansichten der Einheit (E), bestehend aus einer Vielzahl von Pixeln, die unter Verwendung mindestens einer ersten Erfassungsvorrichtung erfasst werden, optischer Sensor (10) genannt, wobei n eine von Null verschiedene ganze Zahl ist,
• • repräsentative Daten (PP) wenigstens einer Entfernung, gemessen mit mindestens einer zweiten Erfassungsvorrichtung, Tiefensensor (11) genannt, zwischen mindestens einem Teil der Pixel innerhalb der n Ansichten (VE) und mindestens einer der Erfassungsvorrichtungen;
• • repräsentative Daten (PS) von aufeinanderfolgenden Positionen des Erfassungssystems (1), bei jeder von n aufeinanderfolgenden Erfassungen, die unter Verwendung von mindestens einer dritten Erfassungsvorrichtung erfasst werden, Positionssensor (12) genannt;

Various embodiments relate to a method, an executable program or a system for reconstructing at least one three-dimensional representation ( RE ) of at least one examined physical unit ( e ) based on observations ( 51 ) corresponding to n successive recordings of a plurality of data, which are preferably synchronized with each other in time, and which comprises:

• • representative data ( VE ) of n views of the unit ( e ), consisting of a plurality of pixels detected using at least one first detection device, optical sensor ( 10 ), where n is an integer other than zero,
• • representative data ( PP ) at least one distance, measured with at least one second detection device, depth sensor ( 11 ) between at least a part of the pixels within the n views ( VE ) and at least one of the detection devices;
• • representative data ( PS ) of successive positions of the detection system ( 1 ), in each of n consecutive acquisitions detected using at least one third detection device, position sensor ( 12 ) called;

In various embodiments, a method is implemented by a system ( 1 . 1' ) for reconstructing a three-dimensional representation ( RE ), in which the data collected by the observations ( 51 ) successive recordings of data ( VE . PP . PP ) whose combination is representative of point clouds ( NP ).

In various embodiments, the method allows the use of acquired information to generate depth data ( PP ) of the acquired views ( VE ). In this case, the depths of certain points of the captured image are determined to create a corresponding depth map that is related to the detected values of the pixels of the image (color or black and white) of the device.

For example, computer modules can be used to create depth tables for each view. In parallel, these computer modules enable the synchronization of each view and the depth table with an image of a color sensor of the tablet.

The method may also benefit from position data related to the positioning of the detector during detection. By this step it can be determined at which position, in particular at which inclination the detection device was, when it performed the acquisition of the data of pixels and / or the depth.

For example, different orientations of the tablet are retrieved via fundamental change matrices relative to the rotation at a higher frequency than the acquisition frequency.

In various embodiments, the method further comprises generating representative data from point clouds ( NP ), which correspond to positions of the points of each view in space, which were combined from the acquired views, the depth data and the position data. In this step, the positions of at least some of the points acquired during the acquisitions are listed in order to perform the three-dimensional reconstruction.

• • jede der Punktewolken (NP) in Bezug auf eine jeweilige gewöhnliche Referenz (RN) im Raum referenziert ist;
• • mindestens ein Vektor berechnet wird, um eine Registrierung/Neueinstellung (53) zu definieren, die auf jede der aufeinanderfolgenden üblichen Referenz (RN) in Bezug auf die vorangehende Referenz anzuwenden ist;
• • mindestens ein Deskriptor mindestens eine Erfassung (54) von Paaren (CP) von Punkten, als Ähnlichkeitspunkte bezeichnet, implementiert, die jeweils zu zwei aufeinander folgenden Ansichten gehören und eine bestimmte Ähnlichkeit zwischen den Werten von mindestens einem ihrer repräsentativen Daten haben (VE, PP, PS), um als Referenz für die Registrierung (53) zu dienen;

In various embodiments, the method uses the fact that:

• • each of the point clouds ( NP ) with respect to a respective ordinary reference ( RN ) is referenced in the room;
• • at least one vector is calculated in order to register / readjust ( 53 ) to be defined on each successive common reference ( RN ) with reference to the previous reference;
• At least one descriptor at least one entry ( 54 ) of couples ( CP ) points, referred to as similarity points, each belonging to two successive views and having a certain similarity between the values of at least one of their representative data ( VE . PP . PS ) as reference for the registration ( 53 ) to serve;

Thus, in various embodiments, the registration ( 53 ) are applied incrementally to the n consecutive recordings so that each iteration records the registration of the representative data of a given view ( VE ) and / or the corresponding point cloud ( NP ) with respect to the previous ones. In addition, the successive iterations allow the re-adjustment ( 53 ) of the data relative to the usual reference, which for the first point cloud ( NP ) and uses it as a global reference ( RG ), in particular for the reconstruction ( 57 ) of the three-dimensional representation ( RE ).

• - Bildregistrierung/Bildneueinstellung (55) basierend auf den repräsentativen Daten der Pixel, die während der Erfassung der aufeinanderfolgenden Ansichten (VE) erfasst wurden, wobei diese Bildneueinstellung (55) mindestens eine Auswahl (551, 552) der Paare (CP) unter Verwendung von mindestens einem Filter auf der Grundlage der repräsentativen Werte (SRFT, SRFR) der Differenz zwischen den beiden Punkten eines Paares (CP) in Bezug auf die Translation und/oder Drehung; und/oder
• - geometrische Neuneinstellung auf der Grundlage der repräsentativen Daten der Koordinaten der Punkte in den Punktwolken durch das Identifizieren für einen Punkt in einer Punktwolke, die einer gegebenen Erfassung entspricht, von mindestens drei Punkten, die in der Punktwolke der vorherigen Ansicht am nächsten sind, die im globalen Koordinatensystem festgelegt ist, so dass diese geometrische Neueinstellung diese drei Punkte, die als Referenzstandorten bezeichnet werden, als Bezugspunkt innerhalb der Punktwolke der vorherigen Ansicht verwendet, wobei diese drei Referenzstandorte ein Dreieck bilden, für das die Projektion des zurückzusetzenden Punktes der geringsten Entfernung entspricht.

Additionally, in various embodiments, the pairs ( CP ), which is the reference for registration ( 53 ) are selected by at least one of the following steps:

• - Image Registration / Image Recalibration ( 55 ) based on the representative data of the pixels that are captured during the acquisition of the successive views ( VE ), this image resetting ( 55 ) at least one selection ( 551 . 552 ) of couples ( CP ) using at least one filter based on the representative values ( Srft . srfr ) the difference between the two points of a pair ( CP ) in terms of translation and / or rotation; and or
• Nine geometric adjustment based on the representative data of the coordinates of the points in the point clouds by identifying, for a point in a point cloud corresponding to a given detection, at least three points closest in the point cloud of the previous view that in the point cloud global coordinate system, so that this geometric resetting uses these three points, called reference locations, as a reference point within the point cloud of the previous view, these three reference locations forming a triangle for which the projection of the point to be reset corresponds to the shortest distance.

It is therefore understood, as mentioned above, that the invention can either use a registration based on the "image" data that is required for a captured view (FIG. VE ), or a registration based on the "spatial" data that represents the coordinates of the points of the cloud ( NP ) are representative, or the two types of registration. Preferably, the Invention both types of registrations successively. It is indeed preferable to combine the two registrations to get better results. In addition, it is preferable to use image registration ( 55 ) and then next with the geometric registration ( 56 ) continue. In fact, successive views are typically acquired at a frequency that provides the information necessary to determine the translations and rotations that have occurred between successive images. The after image registration ( 55 ) applied geometric registration ( 56 ) then allows to provide unequal accuracy. Nevertheless, depending on the application, the user may select a parameter that determines the use of the new setting (s) to be used.

On the other hand, in various embodiments, when the data ( VE . PP . PS ) generated by the various devices ( 10 . 11 . 12 ) of the detection system have been defined, at least with reference to two different three-dimensional references, the method comprises for each of the successive recordings at least one transposition ( 52 ) of this data ( VE . PP . PS ) in a common three-dimensional reference, the respective bases of these data ( VE . PP . PS ) are aligned relative to one another to obtain representative data of n point clouds ( NP ), which are referenced with respect to n consecutive ordinary three-dimensional references defined in a common manner for each of the successive recordings, in order to obtain the readjustment ( 53 ) of the point clouds ( NP ) to enable the reconstruction ( 57 ) to obtain.

For example, the data of an image with its depth table allows defining a raster of 3D points in a reference, called a "view", associated with the camera. This reference has the position of the camera as the origin, and its base is composed of a view axis (Z-axis) specific to each view. The axis perpendicular to the camera and the other axes (X, Y) are aligned with the tray. The term tray is used in this application to refer to the detection system (which may also constitute the system for reconstruction if it contains the executable instructions for implementing the method), but this term is not intended to limit the types of possible systems to interpret.

In this reference, the coordinates of the points of the grid are calculated according to the aperture angle of the camera covering the entire image.

In consideration of a first base change, these coordinates are expressed in the base of the tablet whose Z-axis is in the opposite direction to the Z-axis of the base view, and the axes (X, Y) are inverted in the opposite direction.

In particular, for each shot or image, the appropriate orientation of the tablet can be identified by looking at the next nearest distance under the last alignments stored by a gyro sensor.

By canceling the effects of this alignment over a second base change, the coordinates of the points are found in a so-called common or universal base, where the Z-axis is the vertical axis. The other axes (X, Y) of the usual base are defined by default and can be set by the user (for example, if the Y axis is seen as the north facing axis).

The second base change allows you to reach the same base for all acquisitions. Thus, with each capture, there is associated an ordinary (local) reference whose origin is the position of the camera and the base, the usual basis being defined by the two fundamental changes applied to the original base of the view.

In particular, due to the variation of the position of the detection device between detections, the points obtained are not located at the same location in space when the image is reconstructed at this stage by separate views. This results in a reconstructed image that is not very faithful to the original image, thereby distorting the depth of the image or rendering it unreliable in three-dimensional printing.

It is therefore necessary to perform at least one re-adjustment of the pixels to obtain a reconstructed image that is closer to reality. According to various embodiments, at least one type of correction is applied between two types of corrections of the representative data of the successive views that can be applied to them in order to make the reconstruction possible by virtue of at least one selection ( 551 . 552 ) of representative data for reliable registry values ( 53 ) to further improve one view in relation to the previous one.

According to various embodiments, the method implements the selection ( 551 ) of couples ( CP ) using at least one filter based on the values ( Srft ), which for the Difference between the two points of a pair ( CP ) are representative in terms of translation, thanks to a search for a maximum number of pairs ( CP ) between which the translation is defined by a vector having a value that is lower than a certain threshold, which is adjustable, for example.

In some embodiments, the image registration ( 55 ) at least one selection ( 552 ) of couples ( CP ) using at least one filter based on the values ( srfr ), which is the difference between the two points of a pair ( CP ) are representative of rotation, using at least one selection of a pair of reference points (P _r . Q _r ) having the largest number of pairs defined as their associated pairs (P _i , Q _i ) with which their orientation vectors differ at an angle which is lower than an angle threshold of a certain value, for example adjustable. In this type of embodiment, the selection is based on 552 ) on a deviation ( F expressed by a triplet of rotational angles (α, β, γ) between the representative vectors of pairs of reference points (P _r , Q _r ) relative to the representative vectors of their respective pairs (P _i , Q _i ).

In some of these embodiments, the deviation corresponds to ( F ) of a function, and the selection ( 552 ) corresponds to a minimization of this function by successive iterations of individual rotations of a certain angle value, which is adjustable, for example.

In general, the variables or adjustable values described in the present application may be defined by a user or developer who is responsible for implementing the method in the system ( 1 . 1' ) and the examples provided are not limiting. For example, in certain embodiments, the selection ( 551 ) based on the values ( Srft ), which is the difference between the two points of a pair ( CP ) are representative of translation, a threshold of 10%. On the other hand, in certain embodiments, the selection step ( 552 ), based on the values ( srfr ), which is the difference between the two points of a pair ( CP ) are representative of the rotation, on an angle threshold on the order of one degree.

According to various embodiments, the descriptor used for the collection ( 54 ) of pairs of points ( CP ), a Harris detector that determines the similarity between the averages of the colors in their respective environments. This detector, which is known per se, makes it possible to determine with sufficient accuracy which pixel of a previous view corresponds to a pixel of a next view. The determination of the vector then takes place on the basis of the pixels thus identified.

In some embodiments, as noted above, the geometric registration ( 56 ) are applied to points that are in a point cloud ( NP ) obtained during a given acquisition, which are based on spatial coordinates of a plurality of points within the point cloud ( NP ) and used as reference for the readjustment obtained during the previous acquisition. In some of these embodiments, the geometric registration ( 56 ) the plurality of points of previous detection as reference locations for this readjustment relative to the spatial coordinates of the points to be recalibrated with respect to at least three reference locations defining a triangle in which the distance corresponding to the projection of the point to be displaced is due to the identification for a point within a point cloud ( NP ), which corresponds to a certain detection, is the weakest of the three points that are in the next in the point cloud ( NP ) of the previous view described in the global reference ( RG ) was recalculated. It is thus understood that the geometric registration ( 56 ) a triangulation ( 562 ) of the data is used to improve the accuracy of readjustment with respect to a triangle of previous detection points, rather than being limited to a single point. This triangulation offers far greater accuracy than the methods known in the art, for example ICP-type ("Iterative Closest Point") methods. In other words, for a given view, the points of the previous view are defined as reference locations with respect to which the geometric registration is to be performed. This re-setting for a particular point in a view to be recalibrated is based on finding the three reference locations whose spatial position is closest to that of the given point. These three reference points thus form a triangle on which to search for the shortest projection of the given point. By using the two filters, the accuracy of the reconstruction can be further improved. In some of these embodiments, the geometric repositioning ( 56 ) a classification step ( 561 ) of the data using an accelerator grid ( GA ) to accelerate the triangulation step. This type of accelerator grid ( GA ) is known in the art and is additionally used in the present application to speed up the classification of points as reference points, in particular by filtering the reference points between them by using a threshold distance, for example an adjustable one Value. Searching for reference pages closest to a particular point will thus be speeded up.

The invention also relates to a program comprising code sections which represent instructions for the implementation of the method described in the present application, when the program is processed by data processing means ( 14 ) is performed.

On the other hand, the invention relates to a system ( 1 . 1' ) for reconstructing a three-dimensional image, the data processing means ( 14 ) executing instructions contained in storage means ( 13 ) are stored for carrying out the method described in the present application. In some embodiments, the system comprises ( 1 ) for reconstructing the detection devices ( 10 . 11 . 12 ), which provide the representative data of the successive shots.

The various embodiments described are intended to improve the consideration of depth data in a three-dimensional reconstruction and / or to correct the errors due to the movements of the sensors during the acquisitions or even to correct the inaccuracies of the sensors themselves.

example

Image based resetting

The first so-called "image-based registration" is based on the matching of pixels between two images. This vote is translated into corresponding 3D points that discretely cover the 3D environment.

In the first step of the so-called "image-based registration", for a particular view, only the previous view is considered and the re-adjustment is made with respect to it. In fact, this shot is usually the one that contains the most similarity points (pixels).

Image-based registration involves two consecutive phases. The first is a classical method for determining pairs of similarity points between these two images defined by a mapping of significant points in the two views.

These points are detected, for example, via a Harris detector, whereby the singularities of an image can be identified.

The coordination of these points between the two images is based on their average color value in their environment.

The pairs of similarity points may possibly contain erroneous similarities. To eliminate these false pairs of similarities, a second registration phase is proposed by using the 3D points associated with the pairs of similarity points.

At each pair of similarity points or pixels with (p, q) (where p belongs to the current view and q to the previous view), a 3D point pair (P, Q) is assigned, where P (or Q) of the 3D Point (where the coordinates are expressed in the usual reference), which corresponds to the pixel p (or q).

(da die 3D-Punkte alle in der üblichen Basis ausgedrückt werden).If the pair of similarity points (p, q) is valid, the translation sought is the same $\overrightarrow{PQ}$

(as the 3D points are all expressed in the usual base).

vom identischen Standard. Somit wird für jedes verbundene Paar (p, q) (von entsprechenden 3D-Punkten P' und Q') die Anzahl der Paare (p', q') so gezählt, dass $\Vert \overrightarrow{P\text{'}Q\text{'}}\Vert \approx \Vert \overrightarrow{PQ}\Vert .$

In der Praxis wird dieser Ausdruck so geschrieben: $$\Vert \overrightarrow{PQ}\Vert \left(1-\epsilon \right)\le \Vert \overrightarrow{P\text{'}Q\text{'}}\Vert \le \Vert \overrightarrow{PQ}\Vert \left[1+\epsilon \right]$$

wobei ε einen Prozentsatz darstellt (zum Beispiel 10%).First, a first filter is applied. The first filter is based on lengths and consists of identifying a maximum of pairs (p, q) with a corresponding vector $\overrightarrow{PQ}$

of the same standard. Thus, for each connected pair (p, q) (of corresponding 3D points P 'and Q'), the number of pairs (p ', q') is counted such that $\Vert \overrightarrow{P\text{'}Q\text{'}}\Vert \approx \Vert \overrightarrow{PQ}\Vert ,$

In practice, this expression is written as: $$\Vert \overrightarrow{PQ}\Vert \left(1-\epsilon \right)\le \Vert \overrightarrow{P\text{'}Q\text{'}}\Vert \le \Vert \overrightarrow{PQ}\Vert \left[1+\epsilon \right]$$

where ε represents a percentage (for example 10%).

The pair with a maximum number of connected pairs is maintained with its connected pairs.

von quasi identischer Richtung. In ähnlicher Weise werden für jedes Paar (p, q) (von entsprechenden 3D-Punkten P' und Q') die Anzahl der verbundenen Paare (p', q') so gezählt, dass $\overrightarrow{P\text{'}Q\text{'}}\approx \overrightarrow{PQ}.$

Then a second filter is applied. The second filter is angle based and consists of identifying (among the retained pairs) a maximum of pairs (p, q) with a corresponding vector $\overrightarrow{PQ}.$

of a nearly identical direction. Similarly, for each pair (p, q) (of corresponding 3D points P 'and Q'), the number of connected pairs (p ', q') are counted such that $\overrightarrow{P\text{'}Q\text{'}}\approx \overrightarrow{PQ},$

wobei 〈.,.〉 das Punktprodukt darstellt und θ eine Winkelschwelle ist (zum Beispiel 1 Grad).In practice this is expressed as: $$<\overrightarrow{PQ}.\overrightarrow{P\text{'}Q\text{'}}>\le \theta$$

where <.,.> represents the dot product and .theta. is an angle threshold (for example, 1 degree).

Vektoren. Es sei daran erinnert, dass diese letzten Vektoren gedacht sind, um die gesuchte Translation darzustellen.The pair with a maximum number of connected pairs, pair referred to as the reference pair, is maintained with its associated pairs. Thus, the retained pairs (p, q) have almost the same corresponding ones $\overrightarrow{PQ}$

Vectors. It should be remembered that these last vectors are intended to represent the translation sought.

Alternatively, the angle filter is a minimized function of three corresponding vector rotation angles. This aspect makes it possible to compensate for possible inaccuracies of the gyrosensor.

Therefore, the common base of the current view is slightly inclined relative to the base of the global reference. A correction method is proposed if the number of similarity point pairs retained above is sufficient. If the corresponding pair (p _r , q _r ) of the corresponding 3D points (P _r and Q _r ) is the reference pair, we denote (p _i , q _i ) the pairs associated with the points P _i and Q _i correspond.

wobei α, β und γ jeweils Drehwinkel um die Koordinatenachsen X, Y und Z in der üblichen Referenz der aktuellen Ansicht sind.We consider the function: $$F\left(\alpha .\beta .\gamma \right)={\displaystyle \underset{i}{\Sigma }{\Vert \overrightarrow{P_r{Q}_r}-\overrightarrow{P_l{Q}_l}\Vert }^2}$$

where α, β, and γ are rotation angles about the coordinate axes X, Y, and Z, respectively, in the usual reference of the current view.

relativ zum Referenzvektor $\overrightarrow{P_r{Q}_r}$

für ein bestimmtes Triplett von Drehwinkeln (α, β, γ) welcher auf die Koordinaten des Punktes P_r und der Punkte P_i wirkt.This function allows the quantification of the deviation of the vectors $\overrightarrow{P_l{Q}_l}$

relative to the reference vector $\overrightarrow{P_r{Q}_r}$

for a given triplet of angles of rotation (α, β, γ) which acts on the coordinates of point P _r and points P _i .

Ideally, if the usual base of the current view is aligned with the base of the usual reference, this function is essentially zero.

Therefore, to correct the inaccuracy of the gyroscope, it is proposed to minimize this function by iteratively applying small elementary rotations about the X, Y, and Z axes in the usual local reference of the current view to the usual local basis aligning the base of the global reference, resulting in the restoration of the local reference relative to the global reference.

Each iteration of the function F (α, β, γ) minimization scheme may comprise three consecutive rotations α, β, and γ about the X, Y, and Z axes of the local reference of the current view. Each rotation is also iteratively determined by small rotation steps as long as the function F (α, β, γ) becomes smaller.

The possible rotations (which minimize the function) that are examined in both the direct and indirect directions are cumulated. If no rotation is possible at the end of an iteration of the minimization scheme, the convergent scheme and the accumulation of rotations found represent the correction to be made on the usual basis to conform to the usual general basis.

entspricht, die den Paaren (p, q) der beibehaltenen Ähnlichkeitspunkte entsprechen.The first image base setting is made by translating the usual local reference of the current view with a vector that is the average of the vectors $\overrightarrow{PQ}$

corresponding to the pairs (p, q) of the retained similarity points.

Geometric registration

The first re-adjustment is completed by a second re-adjustment called "geometric re-adjustment" to better cover the environment. In fact, the geometric resetting allows a continuous exploration of the environment.

In this second step of geometric repositioning, for a given view, the corresponding 3D points (current points) are registered with respect to all the 3D points of the previous views.
These last points, called reference locations, are inserted into the global reference.

The method consists in calculating the distance of the actual points with the cloud of the reference locations in the overlapping zones and to define the desired translation from these distances.

For example, in a 3D point of a current view, the three nearest reference locations are determined. If these digits form a triangle, the projection of the 3D point on the support plane of that triangle is calculated.

If this projection falls into the triangle, it can be assumed that the point with the reference cloud, which is understood as the recovery zone, can be correlated via a local translation in order to bring the point back to its projection.

By looking at the mean of these local translation vectors, one obtains a global translation vector that represents the best correlation tradeoff of the current view with the reference sites.

In general, the algorithm of the program according to a variant of the invention for a given view then consists of iteratively applying this correlation method as long as the corresponding global translation vector is significant.

In fact, each time this correlation method is used, the 3D points of the current view are repositioned and the correlation is preferably rechecked. The final translation is applied to all 3D points in the current view, and this view is inserted into the reference locations.

For an efficiency question, an acceleration grid is set up to quickly identify the three locations closest to a particular point. Likewise, this or another raster is used to insert a view into the reference locations. In addition, the reference locations may be filtered through this or another grid with each other over an imposed threshold distance.

It is therefore understood that the various embodiments described in the present application achieve at least a portion of the goals by providing collected data (FIG. VE . PP . PS ) on a unit ( e ) and their transformation to make them more realistic through the use of specific data ( Srft . srfr etc.) by limiting the inaccuracies or errors of these data in relation to the actual physical values obtained by the (reconstructed) unit under investigation ( e ) being represented. So as soon as a registered point cloud ( NPR ), various embodiments of the method and / or system described in the present application may include a representation ( RE ) (in digital form) of the unit ( e ), for example in the form of data that can be used for many physical or virtual applications such as three-dimensional printing or virtual reality, augmented reality, or even the study of various features or parameters, for example, to study physical values such as the mechanical resistance of the examined one Unit whose less accurate representation does not guarantee satisfactory reliability.

It will be apparent to those skilled in the art that the present invention enables embodiments in many other specific forms without departing from the scope of the claimed invention. Therefore, the present embodiments are to be considered as illustrative, but may be modified within the scope defined by the scope of the appended claims, and the invention should not be limited to the details given above.

Claims (13)

A method for reconstructing at least one three-dimensional representation (RE) of at least one examined physical entity (E) based on acquisitions (51) corresponding to n consecutive acquisitions of a plurality of data, preferably synchronized with one another and comprising: • representative data (VE ) of n views of the unit (E) consisting of a plurality of pixels detected using at least one first detection device, called an optical sensor (10), where n is an integer other than zero, • representative data (PP) at least one distance, measured with at least one second detection device, called depth sensor (11), between at least part of the pixels within the n views (VE) and at least one of the detection devices; Representative data (PS) of successive positions of the detection system (1), in each of n consecutive acquisitions detected using at least one third detection device, called position sensor (12); the method being implemented by a system (1, 1 ') for reconstructing a three-dimensional representation (RE) in which the consecutive images obtained by the n acquisitions (51) provide data (VE, PP, PP) representative of their combination Point Clouds (NP) is. the method being characterized in that: • each of the point clouds (NP) is referenced relative to a respective ordinary reference (RN) in space; • calculating at least one vector to define a resetting (53) to be applied to each of the consecutive common reference (RN) with respect to the previous reference; • at least one descriptor implements at least one detection (54) of pairs (CP) of points, referred to as similarity points, each belonging to two successive views and having a certain similarity between the values of at least one of its representative data (VE, PP , PS) to serve as a reference for registration (53); • the new setting (53) is applied incrementally to the n consecutive shots, so on the one hand, each iteration allows the registration of the representative data of a given view (VE) and / or the corresponding point cloud (NP) with respect to the previous ones and, on the other hand, the successive iterations the nine setting (53) of the data with respect to the usual reference, that for the first point cloud (NP), which is used as a global reference (RG), in particular for reconstructing (57) the three-dimensional representation (RE), the pairs (CP) serving as a reference for the nine-setting (53), by a geometric registration (56) based on the representative data of the coordinates of the points in the successive point clouds (NP) by identifying, for a point in a point cloud (NP) corresponding to a given detection, at least three points located in the point cloud (NP) closest to the previous view set in the global reference (RG), such that this resequencing (56) uses these three points, referred to as reference locations, as a reference point within the point cloud (NP) of the previous view, these three reference locations forming a triangle for which the projection of the point to be reset corresponds to the smallest distance , Method according to the preceding claim, characterized in that the geometric resetting (56) is replaced by an image registration (55) or preceded by an image reset (55) based on the data representative of the pixels which occur during the acquisition of successive ones Views (VE) were detected, this screen reset (55) comprising at least one selection (551, 552) of the pairs (CP) that are used by means of at least one filter based on the values (SRFT, SRFR) that are used for the difference between the two points of a pair (CP) are representative in terms of translation and / or rotation. Method according to the preceding claim, characterized in that when the data (VE, PP, PS) obtained from the different devices (10, 11, 12) of the detection system have been defined at least with reference to two different three-dimensional references, the method comprises for each of the successive shots at least one transposition (52) of said data (VE, PP, PS) in a common three-dimensional reference, the respective bases of said data (VE, PP, PS) being aligned relative to each other to represent representative data of n point clouds (NP) referenced with respect to n consecutive ordinary three-dimensional references defined in a common manner for each of the consecutive exposures, to allow re-adjustment (53) of the point clouds (NP) to the Reconstruction (57). A method according to any one of the preceding claims, characterized in that the method implements the selection (551) of the pairs (CP) using at least one filter based on the values (SRFT) corresponding to the difference between the two points of a pair (CP ) are representative in terms of translation, thanks to a search for a maximum number of pairs (CP) between which the translation is defined by a vector having a value that is lower than a certain threshold, for example adjustable. Method according to one of the preceding claims, characterized in that the readjustment of the images (55) comprises: - at least one selection (552) of the pairs (CP) using at least one filter based on the values (SRFR) used for the Difference between the two points of a pair (CP) with respect to the rotation, using at least one selection of a pair of reference points (P _r , Q _r ) having the largest number of pairs defined as their associated pairs ( P _i , Q _i ), with which their orientation vectors differ by an angle lower than an angle threshold of a certain value, which is adjustable, for example - the selection (552) being based on a deviation (F), expressed by a triplet of angles of rotation (a, β, y) between the representative vectors of pairs of reference points (P _r , Q _r ) relative to the representative vectors of their associated P aare (P _i, Q _i). Method according to the preceding claim, characterized in that the deviation (F) corresponds to a function, and the selection (552) corresponds to a minimization of this function by successive iterations of individual rotations of a given angle value which is adjustable, for example. Method according to one of Claims 3 to 6 by defining the adjustable value by a user or developer preparing the implementation of the method in the system (1, 1 '). A method according to any preceding claim, characterized in that the descriptor responsible for the detection (54) of pairs of points (CP) corresponds to a Harris detector which determines the similarity between the averages of the colors in their respective environments. Method according to one of the preceding claims, in which the selection step (551) based on the values (SRFT), which are representative of the difference between the two points of a pair (CP) in relation to the translation, has a threshold value of approximately 10%. used. A method as claimed in any one of the preceding claims, wherein the selecting step (552) based on the values (SRFR) representative of the difference between the two points of a pair (CP) with respect to the rotation is on an angular threshold of the order of magnitude based on a degree. A computer program comprising code sections containing instructions for implementing the method according to any one of Claims 1 to 10 represent when the program is executed by data processing means (14). A system (1, 1 ') for reconstructing a three-dimensional image, comprising data processing means (14) executing instructions stored in memory means (13) for performing the method of any one of Claims 1 to 10 , System (1, 1 ') for reconstructing according to the previous claim, characterized in that it comprises detecting means (10, 11, 12) providing the representative data of the successive shots.

Images