FR2763147A1 - Method for refining images in polar co-ordinates acquired by a mobile radar system. - Google Patents

Abstract

method for processing remote polar co-ordinates (R) and a bearing angle ( eta ) of images acquired by a radar or sonar system on board a moving reference. The method permits passing from the moving reference to a remote local reference, those co-ordinates acquired during successive sweeps each consisting of a series of quanta bearing angle, themselves made-up of a recurrence formed from a series of distance quanta. The polar co-ordinates in the local reference (R', eta ) are determined closer and closer and by successive approximations for each quantum of distance. This is done by applying or not, for each quantum of distance, to each distance or bearing correction a pre-determined distance or bearing increment so as to reject the preceding correction value around the theoretical exact value of this correction. The initial value of the correction is made equal to, for this first quantum of distance, the mobile carrier displacement projected on the horizontal plane and , for the first bearing quantum, to the angle ( eta '- eta ) of orientation of the mobile carrier with respect to the sweep.

Description

1 'Method and device for registering pictures in polar coordinates

  The present invention relates to a method of processing polar coordinates acquired by a radar or sonar boarded on board.

a mobile carrier.

  Such processing is particularly necessary when it is desired to perform temporal filtering processing from a plurality of images acquired successively. This temporal filtering results, for example, from image averaging or image integration techniques which make it possible to improve the image quality by increasing the contrast of the fixed or weakly mobile echoes with respect to the background noise. An example of such an image integration processing is the suppression of the sea clutter (radar noise produced by the waves) that

  it can be decorrelated after a few seconds.

  When the carrier is stationary, all the points of the same coordinates correspond from one image to the next, and the treatment on the

  : D successive images can be executed without difficulty.

  On the other hand, when the wearer is in motion, the images of the same scene obtained during successive scans will be at different angles and at different distances, and thus the problem of the registration of the points of the new image on the points will arise. correspondents of

the old image.

  This problem is particularly acute for planes flying at high speed, when it is desired, for example, to process images delivered by a panoramic radar or an advanced radar before: for such radars, if the scanning is for example carried out in one second and if the aircraft flies at a speed of Mach 1, it will have moved, during this time, about 300 meters, a value much higher than the resolution in distance of the radar, which is typically of the order of About 15 meters

  (for sampling every 100 ns).

  - As will be explained in more detail later in the description, the

  calculation of the geometric corrections to be made is done by modifying the addresses in the image memory, these addresses respectively corresponding to the values of the two polar coordinates, that is to say the distance R and the angle 0 (reference angle referred to relative to the geographic North). Moreover, the filtered image obtained from the plurality of correlated successive images must in turn be referenced in a mobile coordinate system related to the current instantaneous position of the carrier, which means that the addressing corrections will also have to be made. to be applied during the

  reading phase of the image memory.

  These different operations require, in addition to a point-to-point random access memory, computation times lower than the periodicity of the video quanta (the distance quanta are the resolution cells

  temporal - therefore metric - corresponding to the duration of sampling

  radar calibration; an image can thus be obtained for example from 1000 angular quanta or recurrences of 1 ms each-or 0.4 grade in angular resolution for a panoramic radar, each recurrence being itself formed of a series of 10000 quanta of

  distance of 100 ns each - ie 15 m in metric resolution).

  In the general case, the calculation of the geometric corrections requires, as will be seen below, a division and a square root extraction for each quantum of distance, that is to say, in the example indicated above, a division and a square root extraction all

the 100 ns.

  Given the current technology, it is not possible to perform any of these operations in such a short time in the case of a serial architecture, which excludes geometric corrections in real time (The state of the art in calculation circuits does not allow

  perform a root division or extraction only in ls at best).

  To be able to perform these operations in real time, it would be necessary to provide a highly developed parallel architecture, which would lead to very large circuit volumes, incompatible with a hardware

  embedded, including aeronautical equipment.

  One of the aims of the present invention is to propose a method and a device for processing polar coordinates allowing the registration of images acquired during successive scans, both in real time and with very small circuits, in free from everything

  division calculation or root extraction.

  As will be seen below, the method of the present invention uses only the three basic operations of digital electronics (addition, subtraction and multiplication), which can be performed

in a very short time.

  The use of the method of the present invention therefore makes it possible to considerably increase the processing speed compared to conventional methods, perfectly allowing a real-time calculation with circuits of acceptable volume, since it is not possible to no need to

  use a parallel architecture.

  For this purpose, according to the invention, the method is characterized in that the polar coordinates in the local coordinate system are determined step by step and by successive approximations for each quantum of distance by applying or not, for each quantum of distance, to each distance or bearing correction a predetermined increment of distance or bearing so as to readjust the value of the previous correction around the exact theoretical value of this correction, the initial value of the correction being made equal, for the first quantum of distance, to the movement of the wearer projected on the horizontal plane and, for the first quantum of deposit, to the angle

  orientation of the wearer with respect to scanning.

  We will give further details of the complete calculation algorithm

  to implement this method.

  Moreover, according to a certain number of advantageous embodiments of this method: when the current direction of the recurrence of the current scan is close to the direction of movement of the wearer, the angle correction is systematically set to zero deposit; the bearing angle step A (n) is a variable quantity as a function of the orientation angle of the recurrence with respect to the direction of movement of the carrier and / or the rank of the distance quantum considered in the recurrence; a correction of altitude and / or of minimum zone not visualized is carried out, using as projected distance R on the horizontal plane R, for each of the calculations of correction of distance and angle of correction, the value previously calculated according to the relation R = [(Rr + Rm) 2 - H2] 1/2, Rm being the minimum distance not displayed, corresponding to the first quantum of the radar recurrence, Rr being the visualized radar distance, between the first quantum of the radar recurrence and the distance quantum being processed, and H being the altitude of the carrier relative to the projection plane; in this case, the distance value corrected as a function of the altitude and / or the minimum zone not displayed is performed only at each new radar recurrence, preferably by a first-order limited development; periodically updating said local coordinate system so as to correspond substantially to the movement of the carrier, so that the latter is always substantially in the center of the radar map obtained and updated; a filtering is carried out by processing multiple images on p consecutive images referenced with respect to the same reference frame and in which, after acquisition of the Nth image of N successive images, said processing is carried out on the images N to (N-p + l), the common reference being the image (N-p + l1), then, after acquisition of the (N + 1) th image, we recalibrate the (pl) last images already acquired and the (N + 1) th image acquired on the image (N-p + 2) constituting the new common reference for the

next treatment.

  The invention also relates to a device for the implementation

of this process.

  According to the invention, this device essentially comprises at least one circuit for calculating addressing corrections, at least one memory comprising at least one memory plane capable of storing the image corresponding to a complete scan, and a circuit for managing this correction. memory, receiving as input the corrected coordinates calculated by the circuit for calculating the addressing corrections, and outputting to the memory read and write addresses modified according to the calculated corrections of distance and bearing angle, the coordinates correspond to memory addresses according to the rank of the quantum

  of angle of bearing and the rank of the distance quantum.

  According to a number of advantageous features, considered alone or in combination, of this device: 2D - the memory comprises a plurality of memory planes capable of storing a corresponding plurality of images acquired during a series of consecutive scans, and there is provided a filter circuit for performing multiple image processing on said plurality of consecutive images; the memory is a memory that can be used in such a way that it can, alternately, read the data and then retranscribe them to new addresses; this retranscription is performed with a predetermined offset of the memory addresses, so as to reactualize periodically said local coordinate system so as to correspond substantially to the movement of the carrier, so that the latter is always substantially in the center of the radar map obtained and put up to date; after acquiring the Nth image of N successive images, performing said image processing for filtering on the images N to (N-p + 1), the common reference being the image (N-p + 1), then, after acquisition of the (N + 1) th image, said predetermined offset of the memory addresses is made so as to readjust the (p-1) last images already acquired and the (N + 1) th image acquired on the image (N-p + 2) constituting the new common reference for the next treatment; the memory comprises two symmetrical zones forming together a ping-pong memory, said zones being used alternately, so as to be able simultaneously to read the data of one zone and to transcribe them in the other; - In the latter case, it is very advantageously provided two circuits, symmetrical, calculating addressing corrections, and the data read in the ping zone or the pong zone at the rate of calculation and delivery addresses corrected by the first on the one hand, calculation circuits are sent to the filter circuit and on the other hand sent to the pong zone or the ping zone, respectively, to be entered, at the rate of calculation and delivery of the addresses corrected by the second calculation circuit, alternately with respect to the registration of the current image, this inscription of the current image taking place at the rate of calculation and delivery of addresses corrected by the first calculation circuit, and conversely on the next scan; the circuits for calculating the addressing corrections comprise means for permanently storing tabulated values; counting means are furthermore provided which make it possible to define, in terms of distance quantum rank, the radar distance displayed as a function of the clocking of the video radar clock, as well as means for combining the visualized radar distance thus defined with the minimum distance not visualized and for

  apply the result to the calculation circuits.

   Other features and advantages of the present invention

  will appear on reading the detailed description below, made in

  reference to the accompanying drawings, in which: FIG. 1 is a view, projected on the horizontal plane, of the various reference marks used showing the coordinate values and the various parameters acquired or to be calculated, FIG. 2 is an elevation view explaining the The manner in which the altitude correction and the area correction are displayed, FIG. 3 is a block diagram of the circuit making it possible to implement the method of the present invention; FIG. 4 explains the way in which the memory is managed; Image

  ping-pong, with introduction of a slippery local landmark.

  ID Position of the problem We will first explain how the problem is posed

  Attaches the invention, with reference to Figure 1.

  At a given moment, the carrier (the mobile) is at the point O and his radar scans the horizon or a part of the horizon, each point A of the scanned zone being represented by its polar coordinates, namely the distance OA and the angle of bearing (Ox, OA), that is to say the angle that forms

  the radius OA with respect to the geographic North.

  The wearer having made a displacement (AX, AY), he is then found at the point Om; AX and AY are the distances projected on the horizontal plane, respectively along the zx axis and the y axis and, as a first approximation, we will assume that the mobile and the scanned zone are in the same plane (we will indicate later how to perform

  the altitude correction of the wearer).

  Point A is then seen by the wearer at a distance R and with an angle

deposit e.

  The carrier having moved from O to Om between the two successive sweeps, it is therefore a question of transforming the coordinates (R, O) obtained in the movable coordinate system (Oxm, Oym) into corresponding coordinates (R ', o'). determined in the reference (Ox, Oy) initial, independent of the movement of the mobile. This marker, which will be called "local coordinate system" will in the first case be considered as an absolute reference, fixed relative to the ground, but it will later be seen that this local coordinate system is not necessarily an absolute reference, and that it can be constituted by a

  sliding marker, movable so as to follow the movements of the wearer.

  By thus performing a registration, on the same local coordinate system, of a plurality of images obtained during successive scans, it will be possible to carry out an operation, itself known, of averaging or integration of these different images to improve their performance. contrast and

2o decrease the noise level.

  Such an operation assumes, however, that all the images are related to the same reference, namely the local coordinate system (Ox, Oy), whatever may have been the movement of the carrier during the acquisition of the successive images. To determine the recalibrated coordinates (R ', o'), we consider the

triangle OAOm.

  In this triangle, we have the following relation: R'2 = R2 + D2 + 2.R.D. cos (o "-e), with: D = (Ax2 + Ay2) 1/2 (projected distance traveled on the horizontal plane), and:

e "= arctg (& x / Ay) (heading angle).

  We deduce: R '= [R2 + D2 + 2.RDcos (e "-e)] 1/2, and: e' = e + arctg [D. sin (O" -O) / R + D. cos (o "-o)] If, given the available technology, one could perform the calculation of these last two expressions at a rate higher than the rhythm of the quanta of distance, we could get directly in time

  real the recalibrated coordinates (R ', O').

  However, as mentioned above, the state of the art in computing circuits does not allow for square root extraction or

division in such a short time.

  Indeed, the basic operations of digital electronics are

  addition, subtraction and multiplication. Any other operation -

  especially the division or the extraction of a square root - must be programmed by implementing an appropriate algorithm, and the execution of such algorithms requires significant computation times, much higher than the temporal resolution corresponding to a quantum of

  distance (of the order of 100 to 300 ns).

  On the other hand, the trigonometric lines can be tabulated, and therefore

obtained immediately.

   The present invention proposes a calculation method making it possible to determine the quantities R 'and e' mentioned above by avoiding any division or extraction of square root, and thus allowing the registration of

  real-time images acquired from one scan to the next.

  Basic Principle of the Invention The method of the invention is based on a number of remarks

  preliminary, that we will first expose.

  The corrected coordinates (R ', o') being given by the relations: R '= R + AR, and = e-Ae, the following properties are to be noticed: - for each first quantum of distance, the value of the correction in distance AR is AR (1) = D, that is to say that it is independent of the angle of bearing; it can therefore be deduced

directly.

  similarly, always for each first quantum of distance, the value of the deposit correction Ao is Ae (1) = e-o ", that is to say that it is independent of the distance;

  therefore, like distance correction, be deduced directly.

  -for the other quanta of distance, the corrected coordinates can be obtained from the coordinates of the previous quantum, by decrementing AR (for the distance correction) and by incrementing or decrementing Ao, according to the orientation of the wearer (for the angular correction angle of bearing). In this way, if we operate step by step instead of operating globally, the corrected coordinates can be directly deduced without complex calculation involving division or

Square root extraction.

  Details of the process of the invention These increments and decrementations and therefore the corrected coordinates corresponding to each quantum of distance - are given

  by applying the calculation algorithms that we will give now.

  2D0 Distance correction The distance coordinate R '(n) corresponding to the nth quantum is given by the relation R' (n) = R (n) + AR (n), R (n) being the uncorrected coordinate for the nth quantum and AR (n) being the value of the correction in

  distance to this uncorrected coordinate.

  We know, as we have just noticed, that for the first quantum we have AR (1) = D (displacement of the carrier, projected on the horizontal plane); knowing the initial value, we can therefore determine, from

  near, the values corresponding to all video quanta.

  The algorithm for calculating AR (n) is as follows: first quantum: AR (1) = D second quantum: calculation of: A = R (2) + AR (1) and: B = R (2) 2 + D2 + 2.R (2) .D.cos (o "-eo), then: if B <A AR (2) = AR (1) -K

if B> A AR (2) = AR (1).

  (K being a given value of the decrement in distance, which value is generally constant, and chosen once and for all 495 as a function, in particular, of the sampling interval, for example

  a decrement value of 1/4 quantum).

  in the same way, for the nth quantum: calculating: A = R (n) + AR (n-1) and: B = R (n) 2 + D2 + 2.R (n) .D.cos (o "-e), then: if B <A AR (n) = AR (n-1) -K

if B> A = AR (n) = AR (n-1).

  We note, provided we have chosen for K an appropriate value,

  that the algorithm always follows the corresponding theoretical curve.

  Bearing Correction The bearing coordinate (angular coordinate) 0 '(n) corresponding to the nth quantum is given by the relation 0' (n) = 0 (n) - A0 (n), where 0 (n) is the uncorrected coordinate for the nth quantum and AO (n) being the value

  the deposit correction to be made to this uncorrected coordinate.

  2D It is known, as noted above, that for the first quantum we have AO (1) = OPB (OPB = ee "in Figure 1 being the angle of the carrier with respect to scanning) knowing the value initially, we can determine, step by step, the values corresponding to all

the video quanta.

  The algorithm for calculating Ao (n) is as follows: first quantum: Ae (1) = OPB 309. second quantum: computation of: A = log (I tgAo (1) I) and: B = log (I D .sin (o "-o) I) - log (IR (2) + D.cos (o" -0)) (It should be noted that, to be able to be obtained immediately, the trigonometric lines and the logarithms will have to be tabulated - this which poses no particular difficulty technically.) Three cases are then possible: -if OPB = 0 or 180 * A (2) = 0 -if OPB <180: if B <A AO (2) = Ao (1) -L if B> A = no (2) = Ao (1) -if OPB> 180: if B <A Ae (2) = A (1) + L if B> A = Ae (2) = Ae (1) L being a given value of the value-bearing decrement which can be fixed but which is preferably variable according to the orientation of the scanning with respect to the direction of movement of the carrier and depending on the rank of the quantum considered in the recurrence, as will be explained more in

  detail in the following description.

  in the same way, for the nth quantum: a:) 0 computation: A = log (I tgAe (n-1) I) and: B = log (I D.sin (e "-e) I) - log ( IR (n) + D. cos (o "-e) I), then: --siOPB = 0 or 180 * AO (n) = 0 - if OPB <180: if B <A m Ao (n) = AO ( n-1) -L if B> A As (n) = Ae (n-1) -siOPB> 180: if B <A = AO (n) = A (n-1) + L if B> A o A (n) = Ae (n-1) Process Improvements and Improvements Boundary Offset Correction When the orientation of the radar antenna (ie, the current direction of recurrence of the current scan) is close the direction of displacement of the carrier, the correction in the bearing becomes negligible so that, to simplify the process, it can force it to zero for a certain angular range, for example for a scanning angle between +5 and -5 grades on both sides of the direction of

moving the carrier.

  On the other hand, when the direction of scanning is in opposition to the direction of movement of the wearer (that is to say, when "looking back"), the calculated function has a discontinuity on either side of this direction, since we go from the OPB case <180 above

  in case OPB> 180 on both sides of this direction.

  To preserve a satisfactory calculation despite the important dynamics,

  a pitch angle step is preferably provided (parameter As (n)

  above) which varies when one gets closer to this "difficult" area.

  For example, we can choose a pitch angle angle As (n) of 1.25

grade instead of 0.25 grade.

  It is also possible to apply this augmented value only for distance quanta having a certain rank in the recurrence, for example only for the first 80 quanta following the first quantum of the recurrence. 2D correction of altitude and / or visualized zone Another correction that may be necessary is that due to the fact that, especially in the case of an airborne radar, it is not the projected distance that is known, but the distance altitude radar echoing the ground, which deviates even more from the projected distance than the altitude of the

  carrier above the horizontal plane containing the target is high.

  Furthermore, the area displayed does not start at the zero distance, because there is always a minimum distance not displayed at least equal to the corresponding distance, in the time domain, to the duration

  the radar pulse (for example 300 m for a pulse of 2.ls).

  The various parameters that come into play are illustrated in FIG. 2, where the point A corresponds to the target and the point O 'to the position of the carrier: R is the projected distance on the horizontal plane (that is to say the value that one wishes to determine), Rm is the minimum distance not visualized (distance corresponding to the first quantum of the radar recurrence), Rr is the visualized radar distance (distance between the first quantum of the radar recurrence and the quantum of distance in

  course of treatment) and H is the altitude of the wearer.

  In this case, the value of R in the calculations of AR and Ae previously exposed is given by the relation: R = [(Rr + Rm) 2 H2] ln In a radar application, the altitude H of the carrier being low in front of the Radar distance / target (Rr + Rm), a first-order limited expansion is generally sufficient for the calculation of R. It then comes: R = (Rr + Rm) - [H2 / 2 (Rr + Rm)] One s' thus frees from the extraction of the square root (the value of

  [H2 / 2 (Rr + Rm)] may further be tabulated).

  Using a sliding marker So far, it has been assumed that the local coordinate system (Ox, Oy) was a reference

fixed relative to the ground.

  Nevertheless, since the processing and use of the imaged images is carried out on the carrier and not on the ground, it is desirable to periodically update this local coordinate system so as to match it.

  substantially to the movement of the carrier, so that the latter is always D substantially in the center of the radar map obtained and updated.

  Suppose for example that we have already acquired ten successive images and

  that we operate the averaging on ten consecutive images.

  In this case, after the acquisition of the image n 10 will be carried out the averaging on the images n 1 to 10, the common reference being the image n 1. In the same way, after the acquisition of the image n 11, the averaging will be carried out on the images n 2 to 11, the common reference being now the image n 2, and no longer the image n 1 as previously,

  and so on after the acquisition of each subsequent image.

  In this way, during the storage of the image n 11 - which will have to be referenced on the image n 2 and not on the image n 1 -, it will be necessary to simultaneously transcribe the images n 2 to 10 - which were until now referenced with respect to the image n 1 - to reset them on the

  new reference, that is to say that of the image n 2.

  In practice, this operation results in a dual address correction calculation (distance and field correction on the one hand, local reference registration on the other hand) and the use of a double memory

  images, the structure and use of which will be described in detail below.

  Description of the circuits allowing the implementation

of the process of the invention

  We will now give the description of an exemplary embodiment

  of a circuit for implementing the method according to the present

  invention that has just been described.

General configuration of the circuit

  The general synoptic of this circuit is illustrated in FIG.

  It essentially comprises two circuits for calculating the addressing corrections 10 and 10 'which are identical, with a counting circuit 20 making it possible to define, in terms of rank of distance quantum, the displayed radar distance Rr as a function of the timing of the clock.

  video radar, this information being combined with the minimum unviewed distance Rm (which is a constant value) in an adder 21 to apply it to the computing circuits 10 and 10 '.

  The circuits 10 and 10 'are circuits of conventional type in themselves, and therefore the internal structure will not be described in detail.

  It is possible to use, for example, 16-bit fixed-point calculation circuits, TTL technology circuits comprising multipliers

  fast, e.g. 16 bit / 50 ns multipliers. The tabulated values will be stored for example in PROMs or EPROMs included in these circuits 10 and 10 '.

  These 16-bit calculations will make it possible to easily obtain a final resolution of 11 useful bits, corresponding to an angular resolution of 0.25 grade, a value which is in practice largely sufficient. The corrected coordinates calculated by the circuits 10 and 10 'are applied to a circuit 30 for managing two identical memories 40 and'. In practice, these coordinates correspond to memory addresses depending on the rank of the quantum of bearing angle and the rank of the distance quantum. The memories 40 and 40 'together form a so-called "ping-pong" type memory, that is to say a memory comprising similar zones ("ping" zone and "pong" zone) used alternately, in such a way to be able simultaneously to read the data of one zone and to transcribe them in the other, this transcription possibly being carried out with a predetermined offset of the memory addresses. Such memory configuration is also known as "even / odd memory" or "alternate memory". Each of the memories 40 or 40 'consists of a plurality of memory planes, each memory plane corresponding to a radar map acquired during a complete scan. It will be seen that, if it is desired to carry out image averaging processing on N images, it is necessary for each of the memories 40 and 'to comprise at least N + 1 memory planes, in order to be able to contain both the radar map in acquisition price and the N radar maps previously acquired on which the averaging process is carried out simultaneously; for example, for this purpose, for each of the memories 40, 40 ', a series of ten memories each corresponding to a memory plane can be used, the averaging then being performed on nine images; each of these memories can be for example constituted by

  fast static or dynamic RAMs.

  Calculation of addressing corrections As indicated above, the calculation circuits 10 and 10 'form two

  paths a and b identical, only the input variables are not the same.

  In the case of the channel a (circuit 10), R'a and O'a are calculated by applying the algorithms indicated above as a function of: on the one hand, the projected distance corresponding to the displacement D of the carrier between the instant current and the start of acquisition of the Nth image, and - on the other hand, the difference of course calculated according to the projected distance corresponding to the displacement of the aircraft between

  the current moment and the beginning of acquisition of the Nth image.

  With regard to the first point, D is given by the following relation: D = [(Ax-AxN) 2 + (Ay-AyN) 2] 1/2 Ax being the distance projected on the horizontal plane along the axis of x, corresponding to the current instant, AxN being the distance projected on the horizontal plane along the x-axis corresponding to the beginning of acquisition of the Nth image (N being the number of images to be averaged), Ay being the distance Projected on the horizontal plane along the y-axis, corresponding to the current time, and AyN being the distance projected on the horizontal plane along the y-axis

  corresponding to the beginning of acquisition of the Nth image.

  With regard to the second aforementioned point, the parameter e "is given by the following relation: 0" = Arctg [(Ay-AyN) / (Ax-AxN)] In the case of the circuit 10 '(channel b), the calculations are identical, only changes the reference image, which is no longer the Nth image, but the

(N-) th picture.

  The parameters are then as follows d = [(Ax-AxN-1) 2 + (Ay-AyN1) 211/2, and O "= Arctg [(Ay-AyN-1) / (Ax-AxN-1)] Ax being the distance projected on the horizontal plane along the x-axis, corresponding to the current instant, AXN-1 being the distance projected on the horizontal plane along the x-axis corresponding to the beginning of acquisition of the (N- 1) i Me image, Ay being the distance projected on the horizontal plane along the y-axis, corresponding to the current moment, and AyN-1 being the projected distance on the horizontal plane along the axis of the

  y corresponding to the beginning of acquisition of the (N-) th image.

  It should be noted that, although the calculations of D and d involve a root extraction, this is not a problem since this calculation can be performed at the rhythm of the radar recurrences - that is to say every 1 ms or so,

  2 enough time - not the rhythm of video quanta.

  This calculation, like those of the input parameters involving trigonometric lines (namely D.cos (e "-o), Dsin (e" -e), d.cos (o "O) and d. sin (e "-O)) can be carried out conventionally, for example by a microcomputer, the calculated values being input to the circuits 10 and 10 'of the invention, which in turn will calculate the corrected addresses R'a, o 'a and R'b, eb to the rhythm of quanta video according to the

previously discussed process.

Management of the image memory

  We will now describe how the ping memory is managed.

  pong to allow the introduction of a sliding landmark in the way

previously indicated.

  Such an architecture is necessary to transcribe the (N-i) preceding images, referenced on the Nth preceding image, in images referenced on the (N-1) th previous image. This transcription is made at the same time as the inscription of the current image

  referenced acquisition, meanwhile, on the (N-1) th previous image.

  The data read in the ping zone or the pong zone at the rhythm of the calculation and delivery of the addresses R'a and O'a are on the one hand sent to the video filtering circuit 50 (which performs, in a manner in itself, even known, the operation of averaging the N images thus read), and secondly sent to the pong zone or ping zone, respectively, to be registered, at the rhythm of the addresses R'b and O'b, so alternating with the registration of the current image, which is carried out at the rhythm of the calculation and delivery addresses R'a, O'a - and vice versa during the next scan. A simplified example of such ping-pong management is given in FIG. 4, which shows diagrammatically the ten memory planes, numbered from 1 to 10, of the ping zone (on the left of the figure) and the ten memory maps homologous to the pong zone (on the right of the figure), with

  the evolution over time of the content of these memory plans.

  The content of the memory plans has been designated by a number 1, 2, 3, ...

  corresponding to video cards successively acquired, and these figures have been assigned an index (from 1 to 5 on the example) to follow them in successive steps, each incrementation of an index corresponding to the completion of the acquisition of a new video card (the content of the corresponding information is not changed at

  course of time, but only its addressing).

  As indicated above, each of the ping and pong zones comprising

  ten memory planes, we can perform the averaging on N = 9 images.

  The succession of the different phases is as follows: Step 1: Let us suppose first of all that we are in the presence of nine images 11 to 91 previously acquired and stored at the addresses of the memory planes 2 to 10, corresponding to a referencing by

compared to image n 1.

  step 2: the eight images 21 to 91 are transcribed at the same positions from the ping zone to the pong zone, these images of the pong zone being

now designated 22 to 92.

  Simultaneously, a new image is acquired, the image 102, which will be inscribed in the pong zone at the rank following that of the image 92, that is to say in the memory plane n1 since the image 92 is

  found at the last address (memory map # 10) of the pong zone.

  Finally, the image corresponding to the memory plane n 2, which has not been registered, is forced to zero so as to cancel automatically

  the previous image that could have been there.

  Thus, the writing cycle in the pong zone has been completed.

  We can then perform the averaging calculation on all the images in the pong zone (reading cycle), that is to say on the nine images 22 to 102, since the content of the plan

  Memory No. 2 has been erased and plays no role in averaging.

  step 3: repeat the two write / read cycles of the step

  previous, but in the ping zone.

  Thus, the eight images 32 to 102 of the pong zone are transcribed at the same positions to the ping zone, these images of the ping zone being

now designated 33 to 103.

  At the same time, a new image is acquired, the image 113, which will be inscribed in the ping zone at the rank following that of the image 103,

  that is to say in the memory plane n0 2.

  Finally, the image corresponding to the memory plane n 3 is forced to zero.

  The averaging calculation is performed on all the images in the ping zone, that is, in fact on the nine images 33

at 113.

  The following steps are repeated identically so as to reset the

  calculation reference for each new acquisition of a complete image.

2D

Claims (16)

  1. A method for processing the polar coordinates of distance (R) and angle of bearing (e) of images acquired by a radar or sonar on board a carrier defining a movable marker, a method for moving from this mobile coordinate system at a reference local reference mark distinct from the mobile marker, these coordinates being acquired during successive scans each consisting of a series of deposit angle quanta themselves each consisting of a recurrence formed of a series of quanta of distance, characterized in that the polar coordinates in the local coordinate system (R ', e') are determined step by step and by successive approximations for each quantum of distance by applying or not, for each quantum of distance, to each correction of distance or deposit a predetermined increment of distance or of deposit so as to readjust the value of the preceding correction around the exact theoretical value of this correction. one, the initial value of the correction being made equal, for the first quantum of distance, to the displacement of the projected carrier on the horizontal plane and, for the first 2 quantum of bearing, to the angle (o "- o) d ' carrier orientation by compared to the scan.   2. The method of claim 1, wherein determining the correction AR (n) of the distance coordinate R (n) corresponding to the distance quantum of rank n by iteratively executing, for each quantum of distance, the following steps: a) calculating the quantities A = R (n) + AR (n-1) and B = R (n) 2 + D2 + 2.R (n) .D.cos (e "-eo) D being the displacement of the carrier projected on the horizontal plane, O being the bearing angle of the point to be corrected and e "being the bearing angle of the carrier, b) comparison of the quantities A and B thus calculated, and c) if B <A, assignment to AR (n) of the value AR (n-1) -K and, if not, assigning to AR (n) the value AR (n) = R (n-1), K being said predetermined increment of distance , and AR (1) being rendered   equal to the movement of the intended wearer on the horizontal plane.   3. The method of claim 1, wherein the correction AO (n) of the bearing angle 0 (n) corresponding to the distance quantum of rank n is determined by iteratively executing, for each quantum of distance, the following steps. : a) calculation of the quantities A = log (ItgAo (n1) l) and B = log (I D.sin (o "-eo) I) - log (IR (n) + D.cos (o" -e) (1) where D is the motion of the intended carrier on the horizontal plane, where o is the bearing angle of the point to be corrected, and a is the heading angle of the carrier, and b) comparison of the quantities A and B thus calculated, and ) in the case where the orientation angle of the wearer relative to the sweep equals 0 or 180, assignment to A @ (n) of the value zero; 2 [) d) in the case where the orientation angle of the carrier with respect to the scan is less than 180, if B <A assignment to Ao0 (n) of the value AO (n-1) - L and, if not, assignment to Ao (n) of the value AO (n -1); e) in the case where the angle of orientation of the wearer with respect to the sweeping is greater than 180, if B <A assignment to Ao (n) of the value AO (n-1) + L and, if not, assignment to AO (n) of the value AO (n-1). L being said predetermined increment of bearing, and AO (1) being rendered   equal to the orientation angle of the carrier with respect to the scan.   4. The process of one of claims 1 to 3, wherein when the   current direction of recurrence of the current scan is close to the direction of movement of the wearer, systematically deposit angle correction.   5. The process of one of claims 1 to 4, wherein the step   angle angle Ao (n) is a variable quantity depending on the orientation angle of the recurrence with respect to the direction of displacement of the carrier.   The method of one of claims 1 to 5, wherein the step   angle angle A0 (n) is a variable quantity depending on the rank of   quantum of distance considered in the recurrence.   7. The method of one of claims 1 to 6, wherein   a correction of altitude and / or minimum area not visualized, using as projected distance R on the horizontal plane R, for each calculation of distance and correction angle, the value previously calculated according to the relation: R = [(Rr + Rm) 2 _ H211] 2 Rm being the minimum distance not displayed, corresponding to the first quantum of the radar recurrence, Rr being the visualized radar distance, between the first quantum of the radar recurrence and the distance quantum under treatment, and   H being the altitude of the carrier relative to the projection plane.   The method of claim 7, wherein calculating the corrected value of R as a function of the altitude and / or the minimum non-standard area.   visualized is performed only at each new radar recurrence.   2. The method of one of claims 7 and 8, wherein the calculation of   the value of R corrected as a function of the altitude and / or the minimum zone not displayed is made by a first-order limited development,   with R = (Rr + Rm) - [H2 / 2 (Rr + Rm)].   25.10. The process of one of claims 1 to 9, wherein   periodically updates said local coordinate system so as to correspond substantially to the movement of the carrier, so that the latter is always substantially in the center of the radar map obtained and updated.   11. The method of claim 10, wherein filtering is carried out by processing multiple images on p consecutive images referenced with respect to the same reference frame and in which, after acquisition of the Nth image of N successive images, said image is processed. processing on the images N to (N-p + 1), the common reference being the image (N-p + 1), then, after acquisition of the (N + 1) th image, the (p-1) ) last images already acquired and the (N + 1) th image acquired on the image (N-p + 2)   constituting the new common reference for the next processing.   12. A device for carrying out the process of one of the Claims 1 to 11, comprising:   at least one circuit (10, 10 ') for calculating addressing corrections, at least one memory (40, 40') comprising at least one memory plane capable of storing the image corresponding to a complete scan, and ----------   a circuit (30) for managing this memory, receiving as input the corrected coordinates calculated by the circuit for calculating the address corrections, and outputting modified read and write addresses to the memory according to the calculated corrections distance and bearing angle, the coordinates correspond to memory addresses depending on the rank of the quantum of bearing angle and the rank of the quantum of distance. The device of claim 12, wherein: the memory comprises a plurality of memory arrays capable of storing a corresponding plurality of images acquired during a series of consecutive scans, and a circuit is provided ) for performing multiple image processing on said plurality of consecutive images. 14. The device of claim 13, wherein the memory is a memory that can be used so as to be able, alternately, to read the   data and then retranscribe them to new addresses.   The device of claim 14, wherein said retranscription is performed with a predetermined offset of the memory addresses, so as to reactualize periodically said local coordinate system so as to correspond substantially to the movement of the carrier, so that the latter is still substantially in the center of the radar map obtained and updated.   D 16. The device of claim 15, wherein, after acquiring the Nth image of N successive images, performing said image processing for filtering on the images N to (N-p + 1), the common reference being the image (N-p + 1), then, after acquisition of the (N + Ii) image, said predetermined offset of the memory addresses is made so as to readjust the (p-1l) last images already acquired and the (N + 1) th image acquired on the image (N-p + 2) constituting the new reference common for the next treatment.   17. The device of one of claims 14 to 16, wherein the   memory comprises two symmetrical zones (40, 40 ') forming together a ping-pong memory, said zones being used alternately, so as to be able to simultaneously read the data of a zone and transcribe them in the other.   18. The device of claim 17, wherein: - there are two circuits (10, 10 '), symmetrical, calculating addressing corrections, and -the data read in the ping zone or the pong zone to the rhythm the calculation and delivery of the addresses corrected by the first of the calculation circuits are first sent to the conditioning circuit (50) and secondly sent to the pong zone or ping zone, respectively, to be registered. at the rate of calculation and delivery of the addresses corrected by the second calculation circuit, alternately with respect to the inscription of the current image, this inscription of the current image taking place at the rate of calculation and delivery of addresses corrected by the first circuit of   calculation, and vice versa during the next scan.   19. The circuit of one of claims 12 to 18, wherein the circuits   (10 and 10 ') for calculating addressing corrections comprise means   permanent storage of tabulated values.   The circuit of one of claims 12 to 19, further comprising:   2) - counting means (20) making it possible to define, in terms of distance quantum rank, the visualized radar distance (Rr) as a function of the clocking of the video radar clock, and - means (21) for combining the radar distance thus visualized A25 defined at the minimum distance not displayed (Rm) and for   apply the result to the calculation circuits (10 and 10 ').