Target characterization method for a detection device such as radar or multi-panel electronically scanned sonar

Abstract

The invention relates to a target characterization method for a multi-panel electronically scanned radar or sonar detection device, comprising the steps of: - generating a plurality of pulses on a plurality of antenna panels (PE1, PE2, PE3) of the detection device according to a temporal and angular interlacing pattern, so as to perform a scan over the entire bearing domain of the detection device; - generating a plurality of detection maps by acquiring a plurality of observations combined by coherent or incoherent integration of the echoes corresponding to the plurality of pulses, each detection map being obtained in a given direction (EL1, EL2, EL3) corresponding to the width of the main lobe of the antenna panel; - combining the detection maps so as to detect the presence of a target within the bearing domain of the detection device. Figure for the abstract: Fig. 4

Description

Title of the invention: Target characterization method for a detection device of the electronic scanning multi-panel radar or sonar type Technical field

[0001] The invention relates to a target characterization method for a detection device of the electronic scanning multi-panel radar or sonar type. It is applicable to any set of antenna arrays (each set forming a "panel") equipped with the capacity to phase shift the signal applied to each of the antennas or each of the subgroups of antennas constituting the array, with the aim of modifying the antenna pattern ("electronic depointing"). The systems concerned by the invention are in particular radars and sonars comprising phase-controlled panels.

[0002] The design of detection modes (a mode is composed of a waveform to be emitted, a listening duration during which the network digitizes the signals received, and a set of processing operations applied to the signals received in order to decide - or not - the presence of a target) is an active field of research, particularly in complex environments.

[0003] Complex environments are characterized in particular by the presence of ground echoes in the received signal (ground / ground, air / ground or air / air modes when the system is looking downwards), by the presence of sea echoes in the received signal, or more generally by the presence of any real echo (i.e. not being thermal noise), powerful, and / or statistically distributed according to a complex law.

[0004] One method that helps with target detection is to estimate the Doppler shift experienced by the emitted waveform: ground and sea echoes have a low-amplitude, predictable Doppler shift. For example, in order to overcome sea clutter in target detection, it is known to combine two observations of the same area, spaced approximately one second apart. This delay corresponds to the sea clutter correlation time. Any powerful echo with a Doppler shift that is significantly higher or lower than those predicted for the ground and sea can be considered as coming from a target.

[0005] When developing the waveform, the designer faces a set of constraints posed by the hardware capabilities of the detection system (radar / sonar) as well as by the scanning pattern used, limiting the possible choices on the waveforms. One of the constraints, well known to those skilled in the art, is illustrated by [Fig.l].

[0006] The left part of Figure 1 represents the time sequence of Nnop emitted pulses. Each pellet symbolizes a pulse, that is to say a waveform of duration L, TR. After reflection on a distant object, a sequence of Nnop pulses received at the period TR is recovered. The Fourier transform of these Npop pulses makes it possible to obtain a frequency spectrum, a realization of which is presented in the right part of Figure 1. This spectrum comprises points (since calculated on Nnop pulses), and has the unambiguous spectral width Fa = FR, in application of the sampling theorem.

[0007] On the right part of [Fig. 1], we can see a peak around 200 Hz which corresponds to the ground clutter. During target detection, if the Doppler echo of the target is located in the peak, it will be difficult to detect the echo, because in this case the target echo must be more powerful than the ground echoes. On the contrary, a target echo outside the peak can be easily detectable. Outside the peak, the frequency spectrum corresponds to the thermal noise. Thus, at the border between the two zones, the false alarm regulation is complex. Noises of very different natures (sea or ground clutter, or thermal noise) can mix in the middle of the peak (at 200 Hz in [Fig. 1]). The fact that the peak shifts in frequency and spreads out or thins from one draw to the next, complicates the regulation of the false alarm at the border.Indeed, from one draw to another on a "border" frequency box, there will sometimes be pure thermal noise, sometimes very powerful sea clutter. The canonical algorithm which chooses the threshold to apply on each frequency box is disturbed by this "all or nothing", and chooses a moderately high threshold, which therefore generates a false alarm when the peak shifts to the box considered.

[0008] The coherent modes of the prior art (see in particular the article “Radars Advanced Radar Signal Processing” J. Darricau, Engineering Techniques) based on known scanning methods, canonically exhibit the following properties:

[0009] Fa = Fr(l)

[0010] Da«^(2)

[0011] The ambiguous frequency Fa of the integrated pulse sequence is equal to the repetition frequency Fr of these pulses, and the ambiguous distance Da is inversely proportional to Fr, with co the propagation speed of the pulse in its medium. In the prior art, Fa and Da are therefore linked by an inverse proportionality relationship, limiting the possibilities of coherent mode design.

[0012] Figure 2 illustrates the constraint of relation (2). The left part represents the two operating phases of a pulse radar / sonar: transmission of duration Li and reception of duration TR - L,.

[0013] The right part illustrates the consequence of this mode of operation: without distance ambiguity removal processing, the instrumented domain is given by the disk contained between the circles Cl and C2. The circle Cl corresponds to the minimum instrumented distance Dmin around the detection device 1 and the circle C2 corresponds to the maximum instrumented distance Dmax ( = D^. The radius of the circle C2 is given to within a factor by the constraint (2).

[0014] In the prior art, reducing the spectral width Fa necessarily leads to increasing the instrumented domain Da. Given the roundness of the Earth, it is impossible to observe a target of given altitude beyond a certain distance represented in Figure 2 by the circle C3 corresponding to the horizon line. The enlargement of Da beyond this horizon therefore results in an unnecessarily large listening duration TR - Lj, which represents wasted time during which no new useful information is acquired.

[0015] The scanning method described in patent application WO 2013 / 149828 A1 uses a mechanical scanning antenna to scan the environment with a rotation speed of 60° / s. The antenna also has electronic depointing capabilities for two predefined angles relative to the boresight axis of the mechanical scanning (-30° and +30°). The radar thus revisits each area of ​​interest after a short period of time (approximately one second). This “look-back” capability (look-back mode) allows the natural decorrelation of sea clutter to take place between the initial and retrospective samples of the monitoring area.

[0016] Thus, the scanning method described in patent application WO 2013 / 149828 A1 allows the creation of modes subject to the constraint Fa < FR in the case of a radar / sonar system equipped with a single electronically scanned panel, mounted on a mechanical positioner. This scanning method is however not compatible with a quasi-instantaneous 360° coverage objective, due to its single-panel design (a typical electronically scanned panel generally has an angular coverage of 120° maximum).

[0017] The use of a mechanical positioner can make it possible to gradually achieve 360° coverage, but this solution is not satisfactory due to the constraints of the mechanical positioner (mechanical inertia limiting the pointing agility). In addition, the production of Doppler modes covering the entire bearing domain in a time of less than a second (sea clutter correlation time) proves very difficult to achieve due to the forces imposed on the rotating mechanism at these scanning speeds.

[0018] There is therefore a need to provide a target characterization method for a radar or sonar type detection device, allowing greater freedom of mode design, and making it possible to develop Doppler modes with very low ambiguous frequency Fa, covering the entire bearing domain in a potential time. slightly less than a second, and observing all directions almost simultaneously. Summary of the invention

[0019] An object of the invention is therefore a method of characterizing a target for a detection device of the electronic scanning multi-panel radar or sonar type, comprising the steps of: - generating a plurality of pulses on a plurality of antenna panels of the detection device according to a temporal and angular interleaving pattern, so as to perform a scan over the entire bearing domain of the detection device; - generating a plurality of detection maps, by acquiring a plurality of observations combined with each other by coherent or non-coherent integration of the echoes corresponding to the plurality of pulses, each detection map being obtained in a given direction corresponding to the width of the main lobe of the antenna panel; - combine the detection maps so as to detect the presence of a target in the field of the detection device.

[0020] Advantageously, the interlacing pattern comprises a sequential emission of pulses of different emission frequencies in a plurality of directions, each direction being associated with a set of emission frequencies, the emission frequencies being used cyclically in the same direction, the pulses being spaced by a predefined repetition period from one direction to the other, the ambiguous period between two pulses of the same frequency in the same direction at the end of each scan on the bearing domain being greater than the repetition period.

[0021] Alternatively, the interleaving pattern comprises a sequential emission of pulses of the same emission frequency in a plurality of directions, the pulses of the same frequency being spaced apart by a predefined repetition period, the emission frequency being modified at the end of each scan over the bearing domain, the ambiguous period between two emissions of the same frequency in the same direction being greater than the repetition period.

[0022] Alternatively, the interlacing pattern comprises a sequential emission of pulses having different emission frequencies in the same direction, the pulses being spaced apart by a predefined repetition period, the ambiguous period between two emissions of the same frequency in the same direction, at the end of each scan over the bearing domain, being greater than the repetition period.

[0023] Advantageously, the repetition period and the ambiguous period are determined so that:

[0024] Ag > 0^ [00251 With AG = «-S,. and Nb = rnd(% )

[0026] In which Gmax - Gmin corresponds to the bearing domain to be covered, and 03dB corresponds to the width of the main lobe.

[0027] Advantageously, a coherent integration is carried out on the pulses of the same direction and the same transmission frequency.

[0028] Advantageously, the interlacing pattern is repeated, applying, at each repetition, an angular offset equal to the width of the main lobe.

[0029] Advantageously, the interlacing is spread over three phase-controlled panels, the field area to be covered being equal to 360°.

[0030] Advantageously, the pulses are generated on a number of transmission frequencies between two and six inclusive.

[0031] The invention also relates to a detection device of the electronic scanning multi-panel radar or sonar type, the detection device being configured to: - generating a plurality of pulses on a plurality of antenna panels of the detection device according to a temporal and angular interleaving pattern, so as to perform a scan over the entire bearing domain of the detection device - generating a plurality of detection maps, by acquiring a plurality of observations combined with each other by coherent or non-coherent integration of the echoes corresponding to the plurality of pulses, each detection map being obtained in a given direction corresponding to the width of the main lobe of the antenna panel; - combine the detection maps so as to detect the presence of a target in the field of the detection device. Description of figures

[0032] Other characteristics, details and advantages of the invention will emerge from reading the description given with reference to the appended drawings given by way of example.

[0033] [Fig.l], already described, illustrates the link between waveform and frequency spectrum.

[0034] [Fig.2], already described, illustrates the link between waveform and instrumented domain.

[0035] [Fig.3] illustrates an interlacing pattern according to a first embodiment.

[0036] [Fig.4] illustrates an interlacing pattern according to a second embodiment, on several panels, so as to perform a complete scan.

[0037] [Fig.5] illustrates an interlacing pattern according to a third embodiment.

[0038] The method according to the invention comprises a first step in which a plurality of pulses is generated on a plurality of antenna panels of the detection device, in a temporal and angular interleaving pattern, so as to cover the entirety of the detection device's deposit area. "The entirety of the deposit area" means at least one complete scan over the deposit area to be covered.

[0039] The detection device advantageously comprises three panels, so that each panel has a coverage substantially equal to 60°. The invention is however not limited to three panels, and other configurations can be envisaged, for example with four panels.

[0040] In a second step, several detection maps are generated. For this, the observations of the echoes corresponding to the pulses emitted in the first step are integrated. The integration can be coherent (coherent phase between transmission and reception) or non-coherent. Each detection map is obtained in a given direction corresponding to the width of the main lobe of the antenna panel.

[0041] In the case where the integration is non-coherent, the detection map is a distance / recurrence map, and not a Doppler map (no Fourier transform). In the case where the integration is coherent, the detection map is a distance / frequency map (also called a Doppler map in the prior art). In the present application, mention is made of detection maps, which includes distance / recurrence maps and distance / frequency maps.

[0042] The detection maps are then combined so as to detect a presence of a target in the bearing domain of the detection device. The combination of the detection maps, or post-integration processing, is well known to those skilled in the art. The detection maps established for the same direction can be combined, for example, by summing the squared modulus of each detection map, or by any other function leading to the formation of a test statistic maximizing the probability of detection for a given false alarm rate.

[0043] Thus, obtaining a detection map almost simultaneously, in several directions, thanks to temporal and angular interleaving, makes it possible to design new Doppler modes at a very low ambiguous frequency Fa, lower than the pulse repetition frequency Fr.

[0044] The three steps are repeated as many times as necessary depending on the mission entrusted to the detection device.

[0045] A first embodiment of the invention is shown in [Fig.3].

[0046] According to this first embodiment, the interlacing pattern comprises a sequential emission of pulses of different emission frequencies in a plurality of directions. The EL ellipses schematize the width of the main lobe of the radiation pattern of the emitting panel, and therefore the pointing direction.

[0047] Each direction is associated with a set of transmission frequencies, the transmission frequencies being used cyclically in the same direction. In Figure 3, each set is composed of two emission frequencies and forms a unique pair of emission frequencies (Fel / Fe4 for the ELI direction, Fe2 / Fe5 for the EL2 direction, and Fe3 / Fe6 for the EL3 direction). The embodiment is not limited to a set of two different frequencies, and can be extended to a greater number of different emission frequencies. Each pellet corresponds to an emitted pulse, with an emission frequency among Fel, Fe2, Fe3, Fe4, Fe5 and Fe6. The ellipses ELI, EL2 and EL3 schematize the width of the main lobe of the radiation pattern of the emitter panel. The pulses are spaced by a predefined repetition period TR from one direction to the other. The ambiguous period Ta between two emissions of the same frequency in the same direction at the end of each scan on the bearing domain is greater than the repetition period TR.

[0048] The sequence shown in Figure 3 comprises twenty-four successive pulses emitted at the rate FR. These pulses are interlaced angularly (regular change of pointing direction, here with a rate FR so as to observe all the directions almost simultaneously during the integration time) and temporally (change of pointing and / or emission frequency at a rate much higher than the ambiguous frequency of the detection cards).

[0049] At the initial instant, a pulse at frequency Fel is emitted. After a period FR, a pulse at frequency Fe2 is emitted with the same panel, the main lobe being defocused. After a new period FR, a pulse at frequency Fe3 is emitted with the same panel, the main lobe being defocused, and so on. The ambiguous period corresponds to the delay between two pulses of the same frequency, in the same main antenna lobe (same direction).

[0050] In the context of the invention, it is not necessary for different frequencies to be emitted. Emitting the pulses on different frequencies is preferable because this makes it possible to obtain independent target echoes, which makes post-processing more efficient. With a single transmission frequency, the target echoes, which would be weakly separated in time, would be highly correlated with each other, and little information could thus be extracted from the echoes by post-processing.

[0051] The method according to the invention operates optimally with a number of transmission frequencies between 2 and 6.

[0052] Thus, the invention implements angular, temporal, and optionally frequency interlacing.

[0053] Thus, coherent integration (usually a Fourier transform) can be performed only on pulses of the same direction and the same emission frequency.

[0054] It is recalled that the coherent integration is carried out on a single transmission frequency, and the post-processing (post-integration) is carried out on a plurality of transmission frequencies.

[0055] The spectrum thus obtained results from samples integrated at the rate Fa = FR / NbFe (where NbFe corresponds to the number of emission frequencies), while emitting the pulses at the rate FR.

[0056] Thus, Ta > TR, that is to say Fa < FR

[0057] In Figure 3, = 6 ms and TR = 1 ms

[0058] [Fig.4] illustrates another embodiment, in which the entire domain deposit is covered, using several panels. The different emitting panels (PE1, PE2, PE3) are separated, in [Fig.4], by a vertical dotted line.

[0059] In [Fig.4], the pulses are generated with a first frequency Fel, in different directions, and using several panels, so as to scan the entire deposit area. When the complete scan has been carried out, the pulses are generated with at least a second frequency Fe2, in different directions, and using several panels, so as to scan the entire deposit area.

[0060] In [Fig.4], the angular interlacing covers 360°, and is spread for the example over three electronically scanned panels, each panel covering a 120° sector of bearing. During the integration period, all directions are therefore observed almost simultaneously, without having to go back.

[0061] The successive pointings are spaced in bearing of AG = 360 ° / Nbpojutages-

[0062] With AG = and Nh« = rndl ) points ' «

[0063] In which Gmax “ Gjjjjjj corresponds to the bearing domain to be covered (360 or less), $3dB corresponds to the width of the main lobe (width commonly measured at -3 dB), and md( ) corresponds to the function which returns a rounding to the unit.

[0064] For example, to hold an ambiguous set frequency Fa = 55 Hz, with nine pointings over a complete antenna scan, the spacing between successive pointings can be fixed at Ag = 40 °.

[0065] When AG > 03ciB, it is no longer necessary to separate the different successive pointings in frequency, which makes it possible to significantly reduce the number of transmission frequencies used. This embodiment advantageously makes it possible to use only two transmission frequencies Fel and Fe2, whereas in the example of [Fig.3], six transmission frequencies are used.

[0066] The determination of AG, to be much greater than 63^, depends in particular on the antenna diagram, the objective being to be insensitive to a potential second recurrence echo which would be received through a secondary lobe of the antenna diagram. This parameterization can be carried out by a person skilled in the art using his general knowledge.

[0067] Advantageously, a coherent integration is carried out on the pulses of the same direction and the same emission frequency. The integration is carried out over an integration period Tintegration defined by the following relation:

[0068] Tjutegration — NbpOjntages^Nbfe.Nj)Op.T^

[0069] NDop corresponds to the number of pulses on which the Fourier transform is calculated.

[0070] In the example of Figure 4, Integration — 72 lUS.

[0071] At the end of an integration period Tjintegration, we thus obtain a set of detection maps of spectral width Fa << FR, distributed over 360°, constructed almost simultaneously, and angularly separated by AG — 360 ° / NbpOjntages.

[0072] With independent NbFe detection maps in each pointing direction, these NbFe detection maps can be combined in order to characterize the targets in the coverage area.

[0073] In order to fill the blind zones between two successive detection cards, the interleaving pattern, of duration Tjintegration, is repeated, by applying, at each repetition, an angular offset Go equal or substantially equal to the width of the main lobe 03dB. We thus obtain, at the end of the complete interleaving pattern of figure 4, NbFe detection cards in each direction, without any observation gap, with a monotonic scanning in bearing, that is to say without any backtracking.

[0074] When designing the Doppler mode, it is possible to play on the duration of the observation holes of width ATrou ~ AG - 03dB to scan the entirety of the deposit domain over the single duration Tjntermation, while respecting the instructions linked to the repetition frequency FR and Fa < FR.

[0075] According to another embodiment, illustrated by [Fig. 5], the interlacing pattern comprises a sequential emission of pulses having different emission frequencies in the same direction (Fel and Fe2 in [Fig. 5], generalizable to more than two emission frequencies), then a scan so as to emit the pulses having different emission frequencies in another direction.

[0076] Thus, the embodiment illustrated by [Fig.5] provides for changing the pointing direction once all the pulses on transmission frequencies have been transmitted, whereas according to the first embodiment, the transmission frequency changes once all the pointings have been carried out with the same frequency.

[0077] In the embodiment illustrated by Figure 5, two successive pulses of different frequencies are spaced apart by a repetition period TA. (difficult to see in Figure 5 because it is very small), and the pulses of the same frequency, in the same direction defined by the main lobe of the panel, are spaced apart by an ambiguous period Ta. The pulses are emitted in the same direction with the same emission frequency after complete scanning of the deposit area.

[0078] The embodiments illustrated by Figures 3 and 4 allow the observations to be well decorrelated, and therefore to be better averaged in post-integration. Indeed, considering a pointing direction, the delay between two pulses of different frequencies is higher in the embodiments illustrated by Figures 3 and 4 than in the embodiment illustrated by [Fig.5]. Indeed, in the embodiment illustrated by [Fig.5], the detection maps which are shifted in frequency (emission on Fel and Fe2), but are very little shifted in time. However, when the observations are very close together in time, the observations are highly correlated and therefore carry very little information.

[0079] The method described according to either of the embodiments makes it possible to design a new class of Doppler modes with a very low ambiguous frequency Fa, with a high recurrence frequency FR > Fa, and exploiting the agility of electronic scanning to very quickly produce detection maps over the entire bearing domain (360°). Thus, the detection maps can be produced after a delay Tintegration (approximately 100 ms) if an angular spacing of the detection maps equal to ATrou is tolerated, or after a delay TintéaratiOn-ATrou / 03dB (approximately 1 s) if the angular holes are eliminated.

[0080] An example of the design of the Doppler mode illustrated by [Fig.4] is described below.

[0081] The designer of the mode first defines:

[0082] - A maximum instrumented distance Da =< c0 / ( 2FR ), with co the speed of pro pagation of the emitted pulse

[0083] - A set spectral width Fa < FR

[0084] - The number of integrated pulses Nnop to form each DaVa map (map of detection or Ambiguous Distance Speed ​​Ambiguous map)

[0085] - The refresh time y _ nd»pj, with Rat DaVa- y ct.0G ' FR y

[0086] — NbDaVa the number of DaVa cards desired in each direction, for the purpose of post-integration or any other processing

[0087] — Gmax - Gmin The deposit domain to be covered

[0088] — a > 0 a scalar fixing the angular spacing between the different directions observation

[0089] — a < 1 ; No blind directions

[0090] — a > 1 ; Blind directions tolerated so as to reduce TRaf

[0091] These parameters being fixed, the secondary parameters can be obtained:

[0092] NbFe - NbDaVa

[0093] Nbpointa^e = nid ( ' with rn^ ( • ) °rounding operator

[0094] inlays

[0095] Tintegration = NdoP / Fa + ((Nbpointings -1).NbFe + NbFe -1) / FR

[0096] The elementary pattern of duration Tm^gration is repeated _ rnj / ..AG. \ times, in mO 1 y / applying to each repetition i an angular offset Q ( jj = AG...., so as to ” 7 Nbmoüf fill in the blind spots.

Claims

Claims

1. A method of target characterization for a detection device of the electronically scanned multi-panel radar or sonar type, comprising the steps of: - generating a plurality of pulses on a plurality of antenna panels (PE1, PE2, PE3) of the detection device according to a temporal and angular interleaving pattern, so as to perform a scan over the entire bearing domain of the detection device; - generating a plurality of detection maps, by acquiring a plurality of observations combined together by coherent or non-coherent integration of the echoes corresponding to the plurality of pulses, each detection map being obtained in a given direction (ELI, EL2, EL3) corresponding to the width of the main lobe of the antenna panel; - combining the detection maps so as to detect a presence of a target in the bearing domain of the detection device.

2. Method according to claim 1, in which the interleaving pattern comprises a sequential emission of the pulses of different emission frequencies in a plurality of directions (ELI, EL2, EL3), each direction being associated with a set of emission frequencies (Fel / Fe4, Fe2 / Fe5, Fe3 / Fe6), the emission frequencies being used cyclically in the same direction, the pulses being spaced by a predefined repetition period (TR) from one direction to the other, the ambiguous period (Ta) between two pulses of the same frequency in the same direction at the end of each scan on the deposit domain being greater than the repetition period (TR).

3. Method according to claim 1, in which the interleaving pattern comprises a sequential emission of the pulses of the same emission frequency in a plurality of directions, the pulses of the same frequency being spaced by a predefined repetition period (T^), the emission frequency being modified at the end of each scan on the deposit domain, the ambiguous period (Ta) between two emissions of the same frequency (Fel) in the same direction (ELI) being greater than the repetition period (TR).

4. The method of claim 1, wherein the interleaving pattern comprises sequential emission of pulses having different emission frequency (Fel, Fe2) in the same direction (ELI), the pulses being spaced by a predefined repetition period (Tr), the ambiguous period (Ta) between two emissions of the same frequency in the same direction, at the end of each scan on the deposit domain, being greater than the repetition period (T

5. Method according to one of claims 2 to 4, in which the repetition period (TR) and the ambiguous period (Ta) are determined so that ^G &3dB With AG = and = rnd(% ) pointing \ K / In which Gmax - Gmin corresponds to the bearing domain to be covered, and @3dB corresponds to the width of the main lobe.

6. Method according to one of the preceding claims, in which a coherent integration is carried out on the pulses of the same direction and the same transmission frequency.

7. Method according to one of the preceding claims, in which the interlacing pattern is repeated, by applying, at each repetition, an angular offset (Go) equal to the width of the main lobe j.

8. Method according to one of the preceding claims, in which the interlacing is spread over three phase-controlled panels (PE1, PE2, PE3), the field area to be covered being equal to 360°.

9. Method according to one of the preceding claims, in which the pulses are generated on a number of transmission frequencies between two and six inclusive.

10. Detection device of the electronically scanned multi-panel radar or sonar type, the detection device being configured to: - generate a plurality of pulses on a plurality of antenna panels of the detection device according to a temporal and angular interleaving pattern, so as to perform a scan over the entire bearing domain of the detection device - generate a plurality of detection maps, by acquiring a plurality of observations combined with each other by coherent or non-coherent integration of the echoes corresponding to the plurality of pulses, each detection map being obtained in a given direction corresponding to the width of the main lobe of the antenna panel; - combine the detection maps so as to detect a presence of a target in the field of the detection device.

Images