CA2354569C - Aiming aid method and system for a light weapon - Google Patents

Abstract

The invention relates to a method and a system (1) for assisting the aiming of a light weapon (2) held in the hand by a shooter. The system (1) comprises a first imaging sensor (C1), said small field, fixed on the barrel (2) of the weapon (20) and whose optical axis (.DELTA.C1) is harmonized by the construction with the axis of view (.DELTA.20) and a second imaging sensor (C1), said large field, integral with the head (Te) of the shooter. Each of the sensors (C1, C2) is associated with devices (5, 6) for measuring the attitude and heading, so as to make a rough first estimate of the difference between the optical axes (.DELTA.C1, .DELTA.C2) by calculating a rotation matrix and the estimated deviation between the sensors (C1, C2). Knowing this matrix and the fields, an image processing device (3) performs the correlation between the images, so as to display a reticle in a visual helmet type visualization device (4) to materialize the image axis. referred.

Description

METHOD AND SYSTEM FOR ASSISTING LIGHT ARMS
The present invention relates to a method of assisting aiming for a light weapon, more precisely a process aiming aid for a shooter carrying such a weapon at the hand.

The invention also relates to a system for assisting the aim for the implementation of such a method.

It is well known, to get a very accurate shot with a light weapon, that is to say in the context of invention a portable weapon, to equip it with a telescope referred. This glasses can also be coupled to a light intensifier, which allows night shooting.
However, this provision requires a shoulder weapon, or even requires a stable support. The shot may then be accurate, but the aim requires in this case a not insignificant time.

However, in many situations, the shooter must to be able to react very quickly and therefore must be able to use his weapon without assisting him, while ensuring accurate shooting.

To do this, it has been proposed, for example, to provide the weapon of a laser pointer, that is to say a generator of collimated beam propagating parallel to the axis of shot of the weapon or line of sight. The impact on the target is translated by a spot of light of small diameter. The wavelength used may also be in the field of not visible, for example in the infrared. The shooter is not therefore no longer obliged to shoulder his weapon. It is enough that he the position of the spot, either directly (domain of visible), or using special glasses (infrared), to get a good aim.

2 However, in all cases, the major flaw of this process is the lack of discretion. Indeed, the target can detect radiation, either directly (in plain view, for wavelengths in the visible), or with the aid of a detector appropriate to the wavelength used. Else On the other hand, accuracy is limited by discrimination visualization of the spot, the dispersion of the beam and the limit visual perception of the shooter.

The aim of the invention is to overcome the defects of the processes and devices of the known art, some of which have just been recalled. It allows to get a quick shot, do not not needing to support the weapon, and yet accurate, this which appears antinomic a priori. It also provides a sighting process not detectable by the opponent. In in other words, the system remains entirely passive, that is, ie does not generate radiated energy.

Given the distances of use of the system according to the invention (typically in a range of distances greater than 25 m and less than 100 m), and the typical dimension of the targets (1.5 m depending on the dimension vertically and 0.5 m depending on the horizontal dimension), it is necessary to offer the shooter better accuracy than 2.5 mrad on the direction of the barrel of his weapon, and this by a measuring device ensuring a discretion complete, as it has just been recalled.

To do this, the method according to the invention comprises determining the position of the line of sight of the weapon by image correlation and the restitution of a sighting symbol (for example a reticle embodying the line of sight) on a display unit, advantageously of the type helmet visual type of helmet. The symbol can be displayed in collimated form to infinity.
It can be superimposed on the scene observed by the shooter, in direct vision or through a camera. For the determination of the position of the line of sight,

3 correlates an image obtained from a sensor carried by the weapon and an image obtained from a sensor carried by the shooter's head. Orientation data in space of the two sensors are located relative to each other by auxiliary means and serve to estimate the differences orientation between the two images to facilitate the correlation of images. The correlation of images allows determine in the second image the position of the first picture. We can then determine the position in the second image of a reticle representing the line of sight of the weapon which carries the first sensor. The reticle thus determined by calculation is displayed on a display solidarity with the head the shooter and the sensor of the second image. It is displayed at a position that corresponds to its place in the second image, and it is displayed in overlay with the scene observed by the shooter. The shooter can then point his weapon on a target by aligning the reticle on this target without having to shoulder the weapon.

The first sensor is preferably field plus reduced than the second sensor.

The system for implementing the method essentially comprises a first image sensor fixed on the weapon, a second image sensor attached to the head of the shooter, electronic image processing circuits to calculate the position of the field of the first sensor in the image of the second sensor, and a device infinitely collimated visualization for embedding in the shooter's field of vision a reticle or symbol similar materializing the line of sight of the weapon.
The whole is completed by orientation sensors, independent of the image sensors, respectively fixed on the gun and on the shooter's head, allowing an estimate (preferably continuous over time) Orientation of First and Second Fields of Sight image sensors to help position determination of the field of the first sensor in the image of the second. The

4 signals from these orientation sensors are also treated by the aforementioned electronic circuits.

The subject of the invention is therefore a method of assisting the intended for light weapons carried by a shooter, characterized what it includes the following steps:

- the acquisition of a first image using a first image sensor, associated with a first field and fixed weapon, whose optical axis is mechanically linked to the axis of the barrel of the weapon; said first image representing a front scene, seen from said gun;

- the acquisition of a second image using a second image sensor, associated with a second field and worn by the head of the shooter, so as to encompass all or part of of a scene before observed by it and likely to contain a target;

- acquisition by worn orientation sensors respectively by the weapon and the shooter's head, estimation of the angular differences between the field observed by the shooter and the field seen by the sensor carried by the weapon;

- the correlation of the images of the two sensors for determine the position of the field seen by the first sensor image in the field seen by the second sensor, and for determine in particular the position in the image of the second sensor of a point of the first image representing the line of sight of the weapon, the correlation using the angular difference determined previously;

- display on a helmet visual, at a position corresponding, a symbol representing this line of referred.

The symbol is preferably collimated to infinity. he is preferably superimposed on the scene observed in vision direct, but we can also consider that it is superimposed on an image provided by the second sensor or by another camera carried by the shooter's head.

WO 00/3787

5 PCT / FR99 / 03185 Orientation sensors can be of the type magnetometer, for example with a heading sensor and a two-axis inclinometer, these sensors being provided on the one hand 5 on the shooter's head and on the weapon.

Correlation is facilitated particularly in a acquisition phase where you have to find a portion of the image of the second sensor coinciding with the image of the first sensor. Orientation sensors make it possible to develop a rotation matrix to make turn the two images relative to one another quantity corresponding to the (approximate) indication given by the orientation sensors. The images are found then oriented substantially identically and the correlation can continue more easily. In the phases subsequent pursuit, the image correlation is the main element used to move the reticle in the display field.

In particular, it is determined from the indications of the two orientation sensors a rotation matrix representing the relative positions in the space of two reference trihedral linked to the weapon, the other to the head.
We can also, to help the correlation, use a translation vector, limited a priori to about 50 cm, representing the positional gap between the head and the weapon.

The subject of the invention is also a system for assisting the aim for the implementation of this method.

The invention will be better understood and others features and benefits will appear on reading the description which follows with reference to the appended figures, among :

FIGS. 1A and 1B schematically illustrate the architecture of a sighting aid system according to the invention;

6 FIG. 2 is a geometric diagram allowing to explain the general functioning of the system of figures lA and 1B;

FIG. 3 illustrates an exemplary device modular signal processing and image processing used in the aid system for the purpose of the invention;

- Figure 4 illustrates the nesting of images provided by imaging sensors used in the help system to the aim of the invention;

FIG. 5 is an embodiment example one of the modules of the device of FIG. 3;

FIGS. 6A and 6B schematically illustrate a steps of the method according to the invention;

FIG. 7 is an embodiment example more detailed description of a correlation module used in the device of Figure 3;

- and FIG. 8 is a diagram illustrating the last step of the method according to the invention, consisting of a so-called fine correlation.

We will now, to explain the process of referred to according to the invention, describe an embodiment of a sighting system architecture 1 implementing it, with reference to Figures 1A and 1B. Figure 1A illustrates schematically.its overall architecture and Figure 1B is a detail figure showing one of the components used, in the occurrence a display device 4, advantageously of the so-called visual type of helmet.

The system 1 comprises first of all a sensor with C1 imaging in the visible or infrared domain, attached to the weapon 2. In the first case (visible light), can use a standard matrix sensor, or component of the type "IL-CCD" (light intensifier with load coupling). In the second case, we can use a component of the type "MWIR" (Medium Wave Infrared.

7 infrared in the range of wavelengths included between 3 and 5 m) or LWIR (Long Wave Infrared: infrared in the ranges of wavelengths between 8 and 12 m). The direction of the optical axis, AC1, of the sensor C1 is harmonized, by construction, with the axis, A20, of the barrel 20.
If this is not the case, it is taken into account in the calculations.
A device for measuring the position of the center of the field of the C1 imaging sensor of weapon 2. This device is composed of a second C2 imaging sensor fixed on the shooter's head and an electronic device processing of signals and images 3 provided by the C2 sensor, this device 3 having in particular the function of calculate the position of the field of the sensor C1 in the image of the C2 sensor. Indeed, the field of the sensor C1 is chosen advantageously smaller than the field of the sensor C2.

The system 1 also comprises a device 4 of infinitely collimated rendering allowing to embed in the shooter's field of view a reticle 41, or any other symbol adapted, materializing the aim. This device is advantageously constituted by a display member of the so-called visual type of headphones.

Finally, a device consisting of two devices, 5 and 6, which are orientation sensors allowing, by combining the information they deliver, to determine the difference between the orientation of the weapon and the orientation of the head. The simplest is "- to provide that each of the orientation sensors comprises two sensors:
a heading sensor and a two-axis inclinometer (no represented separately). These devices, 5 and 6, which generate heading and trim data, are set, respectively, on weapon 2 and on the shooter's head Te in order to give a continuous preference estimate of the orientation deviation of the two axes of aim, A1 and A2r C1 and C2 sensors. These sensors are independent of image sensors.

8 This double device, 5 and 6, provides, in cooperation with specific circuits of the device electronic processing of signals and images 3, a estimation of the MR rotation matrix of passage between the C1 sensor and the C2 sensor. The translation vector T is not known. Knowing that the weapon is held by the shooter, one can estimate the norm of T at an average value of 0.7 meter and an upper limit of 1 meter. The weapon can for example to be held at the hip.

The electronic signal processing device and images 3 receives the signals generated at the output of the Cl imaging sensor and sensors 6, fixed on the head The shooter's Te, via the link 60, and receives the signals generated at the output of the C2 imaging sensor and the sensors 5, carried by the weapon 2, via the link 50. The device 3 outputs the processed signals for display on the helmet visual display unit 4 (in the example described), as well as the power supply required for functioning of this body.

FIG. 1B illustrates the viewing element 4 in front view, that is to say on the side observed by the shooter. he allows to embed in the scene before 40, actually observed by the shooter, a reticle 41 embodying the targeted target, or other indications, for example an arrow indicating which side he should turn the weapon 2, during a initial period of the sighting process, as it will be shown below. The image provided by the large sensor C2 field encompasses all or part of the scene observed by the shooter.

FIG. 2 is a geometric diagram illustrating the relations existing between a TRC2 reference trihedron associated with the devices carried by the shooter's head Te, and in particular to the sensor C2, and a reference trihedron TRC1 associated with the devices carried by the weapon 2, and in particular Ci sensor.

9 This diagram shows the vector translation T and superimposed on the reference trihedron TRC1 a reference trihedron TR'C2 (in dotted lines), representing the reference trihedron TRC2, after translation following the vector T. This superposition makes it possible to calculate the rotation matrix MR.

The sighting process consists of two stages initials, carried out using the two positioning sensors 5 and 6.

The first step, which we will call "rallying automatic "is to use the signals generated by the orientation sensors 5 and 6, on weapon 2 and on the head Shooter's T, to help the shooter point his weapon into the same direction as his head. This is done by the signal and image processing device 3. The processed signals include heading information and TRC1 and TRC2, related to sensors, C1 and C2 with respect to a terrestrial reference. In output, the processing device 3 generates signals to estimate the orientation difference of the fields of referred to sensors, C1 and C2, at least in direction and direction.
These signals are transmitted to the display member 4, to display a specific symbol on the screen, for example example an arrow 42, in place of the reticle 41 materializing the aim. The shooter knows then that he must move the barrel of his weapon 2, following the direction general symbolized by the arrow 42, so that the reticle 41 symbolizing the aim enters its field of vision and be displayed on the screen of the organ of viewing. It is therefore an automatic help, very rough, to get an approximate weapon score 2, so that the steps of the process of helping to actual purpose can begin. However a score initial approximation can be done manually without the assistance of any symbology.

The second initial step is to calculate a estimation of the aforementioned rotation matrix MR, still at using the data provided by the two sensors guidance, 5 and 6. This estimate of the matrix of 5 rotation MR is obtained by comparing the data (plate and cap) provided by the two devices, 5 and 6, and possibly using the estimate of the translation vector T.

Once this initial step or steps have been completed, the correlation of the images is performed using the

10 rotation matrix to facilitate the acquisition of zones identical images in the images provided by both image sensors; by rotating the image of the first angle sensor defined by the rotation matrix, one gives the images of the two sensors substantially the same orientation, which facilitates correlation.

This correlation will then make it possible to determine in the field of the head sensor what is the position of the line of sight, which is represented by a point well determined (eg a central point) of the field of weapon sensor Ci and the weapon sensor field being localized, by the correlation, in the field of the sensor of head.

So we can display on the screen of the organ of visualization 4 a reticle 41 materializing the aim and stomping on the scene 40 observed by the shooter, and therefore overlaying the target he wants to reach, since the reticle 41 will move in synchronism with the movements of the weapon 2 (phase of pursuit after the phase acquisition of the correlation operation).

The electronic circuits of the device 3 can be cut into modules for the execution of different necessary steps.

Figure 3 illustrates the modular configuration of a image processing device 3 for implementation

11 of the aiming method, according to a preferred embodiment of the invention. We did not show the circuits allowing the execution of the initial step or steps mentioned above.

In the following, a number of conventions concerning signals and other parameters used by the method of the invention will be respected. These conventions are entered in the "TABLE OF RATINGS" annexed at the end of this description.

The device 3 comprises two acquisition modules analogue / digital, 30 and 34 respectively for acquisition channels of the output signals of the sensors C1 and C2.

The output signals of the acquisition modules analog / digital, 30 and 34, are transmitted to extraction of contours from homogeneous zones, or objects, contained in digital images, modules 31 and 35, respectively.

A module 32 then calculates the contour pyramid of the image provided by the small field sensor C1, after contour extraction.

Similarly, the signals at the output of the module 35 are transmitted to a first imaging module 36 distances to the contours of the large field image, followed of a module 37 for establishing the pyramid of so-called images "chamfer", that is to say, distances to the most close.

Finally, a double correlation module called "Gros-End "33, receives the output signals of the modules 32 and 37, as well as gradient signals provided by the module contour extraction 35, and, provided by the modules respective 30 and 34 image signal acquisition, iP
and IG, whose definition is given in the table of aforementioned notation.

12 The calculations made by the so-called "Gros" part of the correlation module 33 are based on a type algorithm valued correlation.

The calculations made by the so-called "End" part are based on an algorithm for estimating the difference in positioning between the center of the small sensor image C1 field and the center of the FG positioning area in the C2 wide-field image, calculated by the module "large".

The different modules making up the device 3 will now be detailed.

Once the model and type of sensors, C1 and C2, are selected, the respective fields of these sensors are known. These are data provided by the constructor, as well as other characteristics associated: resolution, etc.

Analog / digital conversion of signals of images generated by sensors C1 and C2, conversion carried out in modules 30 and 34, is ensured in a synchronous by an image scanning device classic, on a typical dynamics of 28 levels of gray.
The estimation of the rotation matrix MR, and its application to images from sensor C1, as well as the knowledge of the sensor fields, C1 and C2, allows build, by a spatial resampling processing, from the scanned image from the small sensor C1 field. two derived images IF and I'P of parallel axes to the axes of the IG image, assuming that it is rectangular, or more generally a repository orthonormed linked to it. These images, IP and I'F, are such that the following relationship is satisfied CIG = QxCIP

with Q integer, and

13 CIG = CI'P (1a), with CIG, CIP and CI'P, the instantaneous fields of the IG images, IP and I'P, respectively.

The spatial re-sampling is performed by a classical interpolation function, bilinear or bicubic.

The value of Q is chosen in the manner indicated below.
below.

The desired field of the "raw" image CIGBrute in C2 sensor output is less than the instantaneous field of the "raw" image CIPBrute of the small field sensor Cl.
relations linking these two "raw" images are the following:

CIGBrute =; ~ xCIPBrute (2), with k a real number larger than unity and Q part whole of ~.

Figure 4 is a diagram illustrating schematically the IG, IP and I'P images. The cutting of the useful area of the re-sampled images is such that the I'P image is included in the IG image of the large sensor C2 field. Diagrammatically, in FIG.
in dashed lines, the pictures of the IP and I'P images before applying the rotation matrix MR. The XYZ axes symbolize the reference TRC1 (Figure 2).

Digital signals representing IP images and IG are passed to module 33.

At the end of this operation, we define by the maximum apparent apparent roll of I'P in the IG image.

14 The modules 31 and 35 perform the extraction of contours objects present in the two scanned images. The process is similar in both modules.

Figure 5 illustrates in more detail, under form of block diagrams, one of the modules, for example the module 31.

The extraction parameters are chosen so to introduce an identical low-pass filtering level on the two scanned images.

An input module 310 calculates images of gradients GX and Gy along two orthonormal axes X and Y
attached to the image. This operation can be performed by any conventional method of the field of image processing. We may have recourse, for example, to recursive like those described in the article by Rachid Deriche, entitled: "Using Canny's detector to derive a recursively implemented Optimal Edge Detector "," Computer Vision 1987 ", pages 167-187, article to which we will refer profitably.

Gradient images calculated according to the X axes and Y, are transmitted to a first module 311 for calculate, at each point, the norm of the gradient NG, according to the following relation NG (i, j) = IGx- (i, .1) + IG ,. (i, .i) i (3), with i and j pixel coordinates in the image, according to two orthonormal axes.

Calculated gradient images are also transmitted to a second module 312, intended to calculate in each point the orientation of the gradient OG, according to the following relation:

OG (i, .1) = Ai'Ctg (- Gr (T, .J) ~ (4).
Gx (J, J) The output signals produced by the 311 modules and 312, represent, respectively, the images of the standard and the orientation of the gradients, are transmitted to a module 314 which produces a list of contour points for 5 objects detected in the scanned image, in the form of numerical coordinates. This step includes the sub-steps following binarization, deletion of non-maxima and application of a threshold by hysteresis. The sub-step of binarization requires knowing the high and low thresholds.
10 The estimation of these thresholds is carried out by a module additional 313, according to an iterative process. The module 313 receives the data associated with the list contour points in feedback. He also receives on a second entry the image of the gradient standard

Calculated by the module 311.

From the list of extracted contour points of the previous image and standard gradients of the current image, the 313 module aims to enslave dynamically, image area by area of the image, the high and low thresholds.

The current image is divided into size areas equal. On each zone, of arbitrary index z, the module 313 calculates a high threshold, which will be referenced SHZ, and a low threshold, which will be referred to as SBZ, for the image in course of treatment at an instant t arbitrary, this in function of the high and low thresholds obtained on the area at a moment t-1 corresponding to the previous image. The relationship following is used:

SHZ (t) = ax (3xSHZ (t-1) + (1-a) xSmall (5), with a E [0, 1], the exact value being fixed by the application, depending on the type of sensor selected, the value a = 0.8 being a typical value. j3 depends on the number contour points extracted in the previous image, the moment t-1. If the number of points is too high for keep a real-time pace, ~ is chosen lower than

16 unit. On the other hand, if the number of points is insufficient to get a contour information, we choose ~
greater than unity. Typically the value of P is included between 0.8 and 1.25.

Snorme is the threshold value to apply on the gradient standard to perform a threshold operation on n ô points of the zone. The number n is chosen in function of sensor characteristics (wealth space of the image) and is typically between 5 and 15%.

The following relationship is satisfied SBz = E (, vx (SHz (t)) (6), with E whole value and fixed there once and for all by the application. y is usually less than 0.5.

To avoid the effects of interzonal discontinuities, a process of smoothing the thresholds obtained is set artwork. Each zone z is divided into four sub-zones contiguous and the high and low thresholds are recalculated for the subfields by a conventional interpolation algorithm, using the zones adjacent to the zone z.

The module 314 binarizes the outlines and the suppression of local non-maxima, according to a method hysteresis. This module 314 determines the outline chains (that is, sets of related contours according to the principle of connectivity to eight) such that at least one point each chain exceeds the local high threshold of its zone and all other points in the chain exceed their threshold local low.

In addition, each point of the chain is kept if it constitutes a local maximum of the norm. For the check, we compare the norm of the gradient at this point, that we'll call P, with the gradient standard of two more points, Pi and P2, which are closest to the point P,

17 following a direction that achieves the best approximation on a 3x3 neighborhood of the normal to the outline at point P. The direction of this normal is calculated from the orientation of the gradient at point P which gives the direction from tangent to point P.

The process just explained is illustrated very schematically in FIGS. 6A and 6B. On the FIG. 6A shows an image representing the scene seen by the large field sensor C2, comprising, in illustrated example, two remarkable objects made up of an Ar tree and a Bt building. At the end of the process contour extraction, we have a list of points contours CAr and CBt, represented, for example, by zeros. The two objects, symbolized by these outlines, can represent potential targets that are distinguishable from the background.

Outside of the outlines, the points or pixels in the image are represented by numbers different from zero (1, 2, 3, ...) and whose value is all the more so since these points are far from contours (connected distance to 4 of a pixel at the edge of the closer, the distance being expressed in pixels).

Naturally the zeros in the list are associated with digital coordinates, according to a repository orthonormed XY, whose axes are advantageously parallel at the edges of the image, if it is rectangular. These coordinates are stored, at least temporarily, in conventional memory circuits (not shown), so as to be used by the next module, 32 or 36, according to the track considered (Figure 3).

We will refer again to Figure 3.

On the upper track (image processing issue of the small field sensor C1), the module 32 enables

18 the establishment of the contour pyramid, starting from list computed by module 31.

For reasons of speed of execution, the constitution of the pyramid is to use the image of C'p contours determined by the module 311 at the maximum level of resolution and to constitute the sequence of images Ck of the pyramid according to the following relation (7) Ck (i, j) = Ck, (2i, 2j) vCp 1 (2i, 2j + 1) vCk 1 (2i + 1,2j) vCp 1 (2iT1,2j + 1) -relation in which k E (O, K), with K integer, such that 2 K> _ tg (I 0 '~) T'â ~, and the maximum value of the roll residual previously defined.

Pixels are considered as booleans (the proposition is true if there is presence of outlines, false on the other hand).

The TâX parameter is the maximum of the IP sizes along the orthonormal axes X and Y.

The number of non-zero points (that is, the number of outlines) in the image Ck "is noted _ (2k ' On the lower channel (image processing issue of the large field sensor C2), the gradients calculated by the module 310 (FIG. 5) are transmitted on an input of module 33.

The module 36 makes it possible to establish the image of the distances to the nearest contour. This picture is formed from the image of the outlines of the large image field given by the C2 sensor. To do this, we can use a known algorithmic method, for example that disclosed by the article by Gunilla BORGEFORS, entitled: "Hierarchical Chamfer Matching: A parametric edge matching algorithm ", published in" IEEE Transactions on Pattern Analysis and Machine Intelligence ", November 1988, Vol 10, No. 6, pages 849-865, to which reference will be made profitably.

19 Module 37 makes it possible to establish the pyramid of images "chamfer" calculated from the image provided by the previous module 36.

For reasons of speed of execution, the constitution of the pyramid is to use the image of "chamfer" distances constituted from the image of contour of the large field image D ~ determined by the module 36 at the maximum level of resolution and to constitute the following images D of the pyramid according to the relation (8) next DG
(i, j) = Max (DG
1 (2i, 2d)> DG, (2i, 21 + 1), D ~ 1 (2i + 1.2d), D ~, (2i + 1.2d + 1)) diti ~ 2, relation in which r E [0, R], with R = K + q, and div represents the entire division function.

The correlation module 33 is actually subdivided in two modules, as indicated.

The so-called "big" module, 33G, is represented in form block diagram in Figure 7, and includes three sub-cascading modules: 330 to 332.

The 33G module performs a valued correlation, at same level of resolution, between the images of the pyramid outline of the small field image and the images of the pyramid of distance images with outlines obtained at from the large field image.

The input sub-module 330 allows the constitution of the correlation web Nk (u, v), according to the following relation:

NK (DR (i, J) - C; (i; - 11, j + 1 ')) (9), relation in which, u and v are coordinates orthogonal in the correlation web, with Ukf <U <Tlkup and i ~ inf <v ~ VfsUP ~ inequalities in which VKnf and UKnf are typically equal to -2, and VK "p and UK" p are typically equal to +2.

Module 331 is intended for the selection of minima local. This selection is made in two stages.

First, couples (u, v) are selected such that N (u, v) has a local minimum and N (tr ~ P i ~) _ <S. S is a K
threshold that depends on the spectral density of the image input. The threshold S is determined by a procedure conventional calibration of the image processing device.

10 The set of local minima at level K of the pyramid is noted EEK

Finally, in module 332, one carries out, by a iterative process, the rise of pyramids levels and a filtering of the values.

15 For any level n of the pyramid of CA images, one constitutes a set ---: r, as shown below.

For any pair (u, v) of n + 1, a table is calculated correlation Nn reduced to 4 points, at level n, such that Nn (W'V) - ~~ (Dny (', .I) - Cn (lTW, .l + -'') 1 (10), 1 day

In which relation (u ', vI) E [2u, 2u + 1] x (2 v, 2 v + 1J.

A point P is kept for each N-sheet.
point P is the one for which the value of the tablecloth is minimal if it verifies the relationship Nn (P) <_ S
nP (11), not with S the previously defined threshold.

2t Thus we constitute a series of sets ~ j for I [O, K-1].

The point selected by the "big" module 33G is the point II such as:

IIE-- ~ o and 'dpE ~ ,,, p # II => Np (p)> No (n) (12).

The correlation process can stop at this step. Indeed, point II is the best point selected. It can be displayed in the device visualization 4 a reticle 41 corresponding to the impact of aiming axis A20 of the weapon 2 on the targeted target, and superimposing on the scene observed 40 by the shooter through of the device screen 4.

However, it is possible to further improve the precision of aiming. Also, according to another aspect of present invention, applicable to other contexts or a correlation is necessary, it is proposed to proceed to a so-called "fine" correlation operation, using a estimate of a subpixel gap between a window selected in the image of the first sensor and a window corresponding selected in the image of the second sensor.

we will consider, as illustrated by the diagram of FIG. 8, a window F'P of a few pixels, centered in the I'G image, and an IT center FG window, also a few pixels, in the IG image. The pixels are delimited by lines, vertical and horizontal, in dashed lines, the pixels being assumed to be square. Of more precisely, the point TI is the center of the pixel center of the FG window, and a point O'p, the center of the central pixel of the F'P window. The vector E represents the offset between the two "grids" of pixels, on the axis II-O'P, that is to say in amplitude and direction.

The size of the window F'1 'is chosen from such to ensure the presence of minimal information in this window FP. In other words, it does not have to be a uniform area. An indication of the information can be obtained by checking the value of the sum of gradient norms on this window.

The FG window guarantees a margin of 1 pixel around from the FP window.

A priori, there is no reason for the pixels of the two windows coincide. We suppose, indeed, that the small iP field image has an intrinsic resolution superior to the large-field IG image.

We will therefore try to estimate the sub-pixel gap (E1 E = I sJde positioning between the center of the window FG and the,.
the center of the FP window.

Given the compatibility of the bands of the two imaging sensors, C1 and C2, uses an equivalence between the image of the FG window and the image of the FP window.

In general, if we consider an image any numerical I, a few pixels in both dimensions, and a point M also ~ included within image I, we can define a unicolon vector IV (M) containing the values of image I on neighborhood V of the If the point M corresponds to an entire pixel, the values are directly the pixel values of I, while only where the point M has sub-coordinates pixels, the IV (M) vector values are obtained by a classic mechanism of interpolation. The precise choice of interpolation mechanism should be done on a case-by-case basis, according to constraints specific to the specific application envisaged: speed of execution, precision of the result, etc.

In the case of the process of the invention, the neighborhood V work of the same sizes according to X and Y that the F'P window. The total number of neighborhood points V is equal to LV.

The following identity is obtained: FG (II) - = F'P (-E). The IWIJ offset is less than one pixel.

P
We define H = v V, matrix of Lv lines and 5 ~ o ' of 2 columns. We then obtain the following relation F '(-E) = F'P (OP) + H - (- E) (13), and E = - (HT H) = HT = (Fc '(OG) - F' ~ (OP)) (14).

The point OG is equivalent to point II, but is an origin of a landmark for F =

To take advantage of the intrinsic resolution superior of iP, a magnification operation is carried out of the FG window with a Q factor. This operation is realized by a classical interpolation process, by bilinear or bicubic example. The resulting window is called GG.

Fp is a centered window of IP, of size in X
(respectively in Y) equal to the size of F'P multiplied by Q.

From the difference E calculated previously, find, by translation, a point II1. II1 corresponds to center of a pixel of the GG window whose coordinates integers are the best possible approximation of E. The coordinates of this point E1 are the following:

XIT1 = rounded (Q.Ex) (15), and YII1 = round i (Q. Ey) (16).

We will try to estimate the sub-pixellic distance E, position between rIl and the center of the FP window, in using the same approach as before.

To do this, we define a working neighborhood W
of the same sizes according to X and Y as the window Fp.

The total number of points of neighborhood W is equal at I, W = We obtain then the identity: G (rl ~ II, p, E,) As previously we define H = I a Gold matrix of LW rows and 2 columns.

We then obtain the following relation Fw (-El) = FW (OP) + H 1 = (-E1 (17), and E, = - (HT H). HT = (F, ç (OG) - F, p (OP)) (18).

E1 is a small deviation from E. It can be display, on the screen of the viewing device 4, a reticle 41 positioned at the point (II-E-E1) with respect to a point that corresponds to the center of the large sensor field C2 field.

The point! I is obtained with a typical precision of the order of 2 to 3 pixels on the large field image. The map corresponds to 1 pixel of the small image resolution field. We can thus obtain a precision in the reticle 41 positioning typically in the range of 2 at 3 mrad, which corresponds to the goal set by the process according to the invention.

Upon reading the above, it is easy to see that the invention achieves the goals it has set for itself.
The shooting aid method allows all at once to achieve great aiming accuracy and a great speed, since it does not require a shoulder of the weapon. In 25 ' Moreover, according to another advantageous characteristic, the process, since it is not accompanied by of radiant energy, remains quite discreet. In others terms, even if the target has a sensor sensitive to wavelengths used, it will not be able to detect the shooter, at least because of the implementation of the method specific to the invention.

It must be clear, however, that the invention is not not limited to only examples of achievements explicitly described in particular in relation to FIGS. 1A to 8.

In particular, numerical values have not been specified only to fix ideas. They depend essentially of the specific application aimed at.

Similarly, the components usable. sensors imaging, electronic circuits for the treatment of digital signals, etc., are part of a simple choice technology within the reach of the skilled person. Especially, as has been indicated, several types and technologies of imaging sensors can be used, particularly in depending on the choice of wavelengths implemented (visible or infrared).

TABLE OF RATINGS

VS; Image at k level of the contour pyramid the small field image IG Image Grand Champ IP Image Small Field (initial resolution and axes parallel to the axes of IG) I'P Small Field Image (IG resolution and axes parallel to the axes of IG) CIG Instant field of the IG image CIp Instant field of the IP image CIGBrute Raw image of the large field sensor CIPBrute raw image of the small field sensor GX, Gy Images of gradients in X and Y

R Maximum level of the image outline pyramid small field NG Image of Gradient Standard OG Image of gradient orientation Di Image at k level of pyramid images of distances to the contours of the large-field image NV Correlation map valued at level v of the outline pyramid of small field image çlA Number of non-zero points in the CA image Set of minima at v level of the pyramid of outline of the small field image

Claims (17)

1. ~ Method of assisting aiming for light weapons carried by a shooter, characterized in that it comprises the following steps :

- ~ acquisition of a first image with the help of a first image sensor (C1), associated with a first field and attached to said weapon, and whose optical axis is linked mechanically to the axis of the barrel of the weapon; said first image representing a scene before, view of said gun;

- ~ acquisition of a second image using a second image sensor (C2), associated with a second field and carried by the shooter's head, so as to encompass all or part of of a scene before observed by it and likely to contain a target;

- ~ acquisition by orientation sensors (5,6) carried respectively by the weapon and the shooter's head, an estimate of the angular differences between the field observed by the shooter and the field seen by the sensor worn by the weapon;

- the correlation of the images of the two sensors for determine the position of the field seen by the first sensor image in the field seen by the second sensor, and for determine in particular the position in the image of the second sensor of a point of the first image representing the line of sight of the weapon, the correlation using the angular difference previously estimated;

- ~ display on a helmet visual, at a position corresponding, a symbol representing this line of referred. 2. Process according to claim 1, characterized in what the symbol is displayed in collimated form to infinity. 3. ~ Method according to one of claims 1 and 2, characterized in that the symbol is displayed in superposition with the scene directly observed by the shooter. 4. ~ Process according to one of claims 1 to 3, characterized in that the orentation sensors each comprise a heading sensor and an inclinometer two axes. 5. ~ Method according to one of claims 1 to 4, characterized in that it comprises a preliminary step based on the estimated differences angle, in the estimation of the orientation difference of the lines of sight (.DELTA.C1, .DELTA.C2) of said first (C1) and second (C2) image sensors, at least in direction and direction, of display way on said display member (4) a second determined symbol (42) indicating to said shooter the direction in which he must move said gun (20) of the weapon (2). 6. ~ Process according to claims 1 to 5, characterized in that said weapon (2) being held at the hand by the shooter, the correlation uses an estimated of a translation vector (T) between the positions of the two image sensors, this vector having a terminal superior estimated at about 1 meter. 7. ~ Process according to any one of Claims 4 to 6, characterized in that said first field of said first image sensor (C1) is smaller than second field of said second image sensor (C2), in what, said first and second images being analogue, the said acquisition includes the conversion Analog / digital (30, 34) output signals said first (C1) and second (C2) image sensors, way to build two digital images associated with first (TR C1) and second (TR C2) repositories orthonormed, respectively representing the images from said first (C1) and second (C2) imaging sensors, in that a step consisting of construction, from the digital image resulting from said first imaging sensor (C1), and by application of said rotation matrix, of at least a first image Derived Digital System (IP), with the same ortho-referenced the digital image (GI) issuing from said second sensor (C2), whose instantaneous field is proportional, by a Q factor, at the instantaneous field of said digital image from the second imaging sensor (IP), Q being a whole, and in that steps are taken subsequent image processing methods including correlation (33) of said derived digital image and said digital image from said second sensor in imaging, to determine the position of said first field said first image sensor (C1) in said second image acquired by said second image sensor (C2) and the position of said line of sight in this second picture. 8. Process according to claim 7, characterized in what, said digital images from said first (C1) and second (C2) imaging sensors comprising non-uniform areas (Ar, Bt), the treatment steps at least the following steps:

a step (31, 35) for extracting contours (CAr, CBt) said non-uniform areas (Ar, Bt), so as to generate a list of coordinate points from said digital image corresponding to said contours and to associate with a given numerical value;

for the digital image originating from said first sensor (C1), a step (32) of establishing, from the image said contours (CAr, CBt), a series of images forming a pyramid of contours, on a number of levels determined;

for the digital image issuing from said second sensor image (C2), a step (36, 37) of establishing, from the image of said contours (CAr, CBt) and for each point of the image, an image of the distances to the closest contour and, from this image of distances to the nearest contour, a sequence of images forming a pyramid of contours distances, on a number of levels determined; and a valued correlation step (33) between said sequence of images forming a pyramid of contours and said suite of images forming a pyramid of distances from contours, so as to select a point from the digital image from said second image sensor (C2) constituting the position closest to the center of the field of this image, and to display, on said organ of visualization (4), said first symbol (41) determined, at this position of said second image, superposition of said observed forward scene (40). 9. Process according to claim 8, characterized in what said edge extraction step (CAr, CBt) said non-uniform areas (Ar, Bt) of said images numbers from said first (C1) and second (C2) Image sensors includes the following substeps:

the determination (310) of the following gradient images two axes of orthonormal coordinates related to said digital images;

the determination (311) of the norm of said gradients in each point of said images;

determining (312) the orientation of said gradients at each point of said images; and - establishment (313, 314), from those standards and said gradient orientations, from a list of outline points. 10. Process according to claim 8, characterized in that what said valued correlation step (33G) comprises the following substeps:

the constitution (330), from said sequence of images forming a pyramid of the outlines and said sequence of images forming a pyramid of contours distances, a correlation web;

the selection (331, 332) in said set of values representing local minima below a threshold determined, for one of said pyramid levels, and the constitution, at each level of said pyramids, of a set of local minima; and - selection of a point (.pi.) corresponding to the smallest of minima to estimate the position of the center of a zone of positioning said digital image resulting from said first image sensor (C1) in said digital image said second image sensor (C2), this position representing said line of sight. 11. ~ Process according to claims 8 or 10, characterized in that said correlating step valued (33G) is followed by a correlation step additional, so-called "fine, implementing a algorithm for estimating the positioning error (E), amplitude and direction between the center (O'P) of said digital image from said first image sensor (C1) and the center (.pi.) Of said positioning zone of this digital image in said digital image said second image sensor (C1), the positioning of said center (.pi.) being obtained during said step of valued correlation (33G), and in that at the end of additional step, a new position of the said center is determined by a translation of amplitude and direction consistent with said gap (E). 12. ~ Targeting system for a weapon light range by a shooter, characterized in that includes:

a first image sensor (C1) making it possible to acquire said first image, said first image sensor (C1) being associated with a first field and being attached to said weapon (2), the optical axis (.DELTA.C1) of this sensor (C1) being mechanically linked to the axis (.DELTA.20) of the barrel (20) of the weapon (2) and representing the line of sight of that weapon (2);

a second image sensor (C2) making it possible to acquire said second image, said second image sensor (C2) being associated with a second field and fixed on the head (Te) said drawer, so as to encompass all or part of a scene before (40) observed by it, said scene (40) being capable of containing a target for said shoot ;

- orientation sensors (5,6) allowing determine the angular difference between the direction of the weapon and direction of the shooter's head;

- a device for processing signals and images (3) received from said first (C1) and second (C2) sensors imaging and orientation sensors (5,6) for correlate images received from two image sensors, using the angular gap determined, in order to determine the position of the first image sensor in the field of the second image sensor, - and a display member (4) for displaying a determined symbol (41), embodying said position of the line of sight in said second image, or a second determined symbol (42), indicating to said shooter a direction in which he must move the barrel of said weapon (2), said determined symbols (41, 42) being superimposed on said scene before (40) observed by said shooter. 13. System according to claim 12, characterized in that said viewing member (4) is a visual helmet including a screen on which are displayed said determined symbols (41, 42) and through which said shooter observes said front scene (40). 14. System according to claims 12 or 13, characterized in that said first (C1) and second (C2) Image sensors are load-type sensors coupled to radiation in the spectrum of the visible. 15. System according to claims 12 or 13, characterized in that said first (C1) and second (C2) image sensors are sensors sensitive to a radiation in the infrared spectrum. 16. System according to any one of Claims 12 to 15, characterized in that the sensors each include a heading sensor and a two-axis inclinometer, so as to generate data cap and trim of the weapon and the head. 17. System according to any one of Claims 12 to 16, characterized in that, said first and second images being analog, said signal and image processing device (3) comprises acquisition modules (30, 34) comprising a analog / digital converter for each of said images.