DE2351511A1 - Position determination circuit - uses pulse radar stations, such as two marine D radar stations for strongly disturbed targets - Google Patents

Abstract

Automatic target tracking is performed. In order to save digital storage locations during clutter decorrelation after manual signal pre-processing, only the binary coded radar video signal evaluated over several aerail revolutions are added up in a nonrecursive integrator; after low-pass filters have reached steady state, the corresponding bearing signals are delivered; they are then processed in real time by a permanently allocated, or multiplex wandering window detector operated in time; target beginning and end signals are used by a following line estimator, together with single and double bearings; after a disjunctive linking over two consecutive range rings, they are stored in an intermediate bearing store with a suitable number of bits.

Description

Patent-Verwaltungs-GmbH 6000 Prankfurt (Main) 70, Theodor-Stern-Kai 1

Ulm (Danube), October 12, 1973

PT-UL / Schz / sa

UL 73/134

"Circuit arrangement for position estimation of

Radar targets "

The invention relates to one based on character recognition "based circuit arrangement for estimating the position of the pulse radar devices, in particular 2D ship radar systems, scanned, severely disturbed targets in order to enable automatic target tracking with as much smoothed as possible To enable trace display. This circuit arrangement is called the target coordinate stimator in the following.

In addition to target recognition, the task of radar devices is to determine the coordinates of a discovered target as precisely as possible, with polar coordinates Φ and E being used for the sake of simplicity (FIG. 2). In pulse radar devices, due to the relatively short transmission pulses T * and the inherently high distance resolution Δ R (T'ssA R) associated with it, the estimate of the amount of the radius vector R ₂ (»distance radar antenna to target) with a known target gravity line ( » Most likely hit line of a target representing.

5 09816/0618

- 2 - UL 73 / 134-

pattern) fewer difficulties than those of the azimuth angle Wy ₁ . By means of the invention, on the one hand, this target center of gravity (maximum correlated hit sequence of a range ring Δ E) is to be fixed and, on the other hand, the angle measurement errors "when determining the azimuth angle as the parameter of a stochastic distribution - despite the low azimuthal resolution of the radar antenna - are to be kept to a minimum the radar target information should only be available in the form of "binary quantized radar videos (quantization

. de

noise), the ClutteTEorrelation in a non-recursive Integrator "evaluates over several antenna revolutions summed up where "at the radar pulse repetition frequency is derived from the frequency of the azimuthal angle increments Δ φ, in order to obtain a spatially rigid angular grid. However, because of the statistical nature of the radar data, it is impossible to to localize the target position within the system time constant with any desired degree of accuracy, since the pure target recognition during this period because of the receiver noise and the target scintillations in many cases causes certain difficulties.

- 3 - 50981 6/06 18

From the literature and practice, many methods for measuring distance and azimuth Winkelbestiramung are known, especially the optimal estimation method according to the ordinance dea ^ Maximum Likelihood EstinatorsVnd the "binomial detector go - corresponding to the matched filter theory in the video area;." Transmit pulse length Γ equal Entfernungsinirenent λ / ϊ on the assumption that a target has an unknown azimuth ψ- is in a (U / range ring of width · δ £ at distance R _7. from the radar antenna via the hit sequence correlated in azimuth with the azimuth angles

(f. ... ψ , discovered as a sequence of transmitted signals emitted one after the other. If one denotes the sequence of discoveries at the distance R _-7 . with X = (X ₁ ... X) -, the basic problem consists in ^{ f estimate with minimal error from the sequence of discoveries X. If one now assumes a spatially fixed angular grid, with a symmetrical antenna function and neglecting the noise, the azimuth angle of the target is the center angle of the echo pulse group X. The target direction can consequently be determined from this that the pulse sequence X is summed up in the azimuthal direction in the corresponding distance ring and the target start ZA with the associated azimuth angle ^ ^{7 is} concluded if this hit sum exceeds a predetermined threshold or target end ZE is declared when the hit sum falls below this threshold again, with the associated aciniut angle is set with ψ ". The target center angle could now be γ ·> - ^ ^{7 4} ~ £ r calculated without errors, g if «β no noise, no target fluctuations, etc., that is, there would be no fluctuations in the echo pulses which ultimately lead to statistical errors in the azimuth determination. In these considerations, systematic angle errors, e.g. originating from the encoder or gearbox etc., are disregarded.

5 0 9 8 16/0618

State of the art criticism:

. 2 3 P I5 I I

However, none of these methods and approaches take into account several packages that have a strong influence on the results of the Vinkelbestinusung or make automatic target tracking impossible, namely: that the echo pulses extend over several (mostly 3-5) rings of distance, that due to interference none closed hit images exist that targets that are embedded in strong interference can only be discovered after an integration of the video signals over several antenna revolutions and their position can be estimated and that with repeated scanning of the same target its estimated value if ~ fluctuates by ψ ~ , ψ »Ε ί ψ) t where this fluctuation is due to the scattering

G '* (%) * £ $ ^t i ~%) ^Z J can be expressed

can and this standard deviation of the azimuth estimate but again a function of the receiver noise, the level of the quantization threshold, the scintillation spectrum, the antenna characteristics, the reflective cross-section of the target, the transmission pulse repetition frequency and the hit count is components that, even under ideal conditions, provide an accurate azimuth determination impede.

509816/0618 5 _

~ 5 -

The invention is based on the object of specifying a simple circuit that can be easily retrofitted to existing automatic extraction systems based on the principles of character recognition, which determines the probable target skew line at distance R_ and the estimated value ψ of the true target center angle Y ₇ with the smallest possible error. The performance of a radar observer as an optimal estimator is used as a comparison value.

A safe automatic target tracking even with difficult meeting criteria, combined with - according to the system time constant - maximum track smoothing in the case of the PPI display, should be guaranteed be.

If the target coordinate system is used in an automatic Extraction system retrofitted, so should for cost reasons the known function groups: -

Short-term control of the 1st threshold pulse center signaling
Determination of the distance to torment
Regulation of the 2nd threshold
Correlation detector

can be eliminated without replacement. Furthermore, the

a ■.

To enable the use of associative memories to increase the operating speed of the entire system, with the entire radar coverage area being able to be split into a limited number of target areas (frames) for further processing after signal processing to reduce the storage space requirement.

509816/0618 - 6 -

According to the invention, this object is achieved in that, in order to save digital storage space in the clutter decorrelation, after manual signal preprocessing, only the binary quantized radar video of each target occupied by a decorative expectation area is added up in a non-recursive integrator with a rigid angular grid over several antenna revolutions and evaluated according to dent settling this low pass filter and certain after fulfilling hit criteria per quantisierteia surface element (^ iR * ^ \ ip ~ n:> r / per Antenneaumlauf for each target expected area associated filtered fail he message s output, which then on the one hand associated in real-time over a fixed or time-division multiplexed traveling window detectors are processed, with the target start and finish messages (ZA, ZE) obtained therefrom ± in the following line emitter, which is made up of an analyzer with extraction logic and hit image memory and a subsequent line emitter This classifier with decision logic and target coordinate memory can be used together with the single and double hits of a target after performing some arithmetic and logical operations to determine the most likely target center of gravity at the distance R_, on the other hand, after a disjunctive link over two successive range rings, are stored in a hit buffer whose number of bits corresponds to a target expectation area the number of quantieierten surface elements and that target expectation zone is fed only the most likely Ziölschwerlinio from dom results latch in d n same VTanderfensteisdetektor on the row selector of Zeilenestiraators with the arrival of Rahmenendeirapulsxeines, now these results consequence also on

509816/0618

ZA and ZE connections are examined and ία uninterrupted operation supplies the associated angle approach values or »final angle values * f , or ψ ", based on the azimuthal start of a target expectation area, to the target computer for calculating the target center angle Ψ " . whereas in the disturbed operation these two angle values are modified by the line structure and made available to the target computer and that, for reasons of time, the target coordinate measurements are carried out in parallel processors or, however, associative memory is used.

509816/0618

- O -

■ · ΐ

Further configuration: 2351511

To a target coordinate stimator according to the invention to make it particularly advantageous in terms of price, according to a further embodiment of the invention in in many cases forego the maximum speeds that can be achieved with parallel processors and all of them Use circuit units with the exception of the memory in multiplex mode. In addition to a noticeable With this solution, cost reduction is also an increase in circuit reliability due to the Reduction of the number of components achieved «, however, the processing speed drops noticeably. However, a certain amount of time can be saved through the use of associative memories in the target extractor as well as in the target coordinate stimulator.

509816/0618 - 9 -

The advantages that can be achieved with the invention are, in particular, that the target coordinate _{stimulator considerably refines the azimuthal (V z} ) and distance (R_) position determination of a target and at the same time the target detection probability by approx. 5 db is increased. In addition, considerable savings can be achieved in terms of circuitry, since the assemblies fine distance, pulse center determination, correlation detector, short-term control of the 1st threshold (binary quantization) and control of the 2nd threshold are omitted. * -

Furthermore, the separability of two meeting targets and consequently the automatic target tracking is increased in an advantageous and considerable way, since the target computer no longer delivers the first hit line, which delivers the ZA message via the moving window detector and, as experience has shown, is mostly roughened and therefore leads to larger target center jumps The cause is, processed, but the most probable one of a target consisting of several hit lines one behind the other in terms of distance, namely the so-called target center of gravity. Likewise, the disjunctive connection of the quantized videos of two distance rings Λ. R the quantization effect and the pulse width modulation effects of the 1st threshold are reduced and a certain smoothing occurs at the target edges in the azimuth direction. Since the distance estimate., A target position is also not Mohr as previously derived from the in azimuth more or less correlated results succession of differentiated unit impulses of a target, but rather on the Trefferhäufigkeitavertoilungen (single hit f _T and Doppeltreffor FVJ) ^{of the} * ⁿ unchanged its pulse length

509816/0618

Echo hits, the target fine structure is lost, which leads to a further gain in information known circuit arrangements and results in. Smaller scattering of the angle and deformation errors precipitates; i.e. all information losses by the loss of knowledge of the hit length a target echoes are eliminated and thereby also a gain achieved when integrating hits over several antenna revolutions, in contrast to when the roughened hit images, which are differentiated from the Unit hits of the circuit for determining the lapuls would result in the middle. Furthermore we should also point out the loss of information caused by the so-called shadowing effects of the short text control occur for the 1st threshold and cause great difficulties with the automatic target tracking to lead.

Another advantage of the invention, which should not be underestimated, is ultimately that it is circuit-wise easy to already existing target extractors can be upgraded.

509816/061 8

OO Γ -JC -Ι Λ

A. Embodiment of the invention is in the block diagram and will be described in more detail below described .. '.

Show it: /

Fig. 1 Digital target coordinate stimulator retrofitted in an automatic target extractor «

Fig. 2 Target expectation area in the coverage area a panoramic radar system.

Fig. 3 Non-recursive before integrator. . "· Fig. K Hit patterns in target expectation areas.

Fig. 5 Disjunction of the hit lines with hit buffer.

Fig. 6 Line stimulator of a target expectation area.

Fig. 7 Allocation of hit image and target coordinate memory.

Fig. 8 Determination of the target gravity.

9 Schematic diagram of the target coordinate stimulator.

For the automatic acquisition and processing of radar video data are increasingly automatic Target extractors used that as likelihood ratio detectors based on the theory of Neyman and Pearson. Despite it being used for radar applications many times over optimal properties, they also have disadvantages on, since they have a present .Trefferbild for Ziclcxtraktion (see Fig. 2) only evaluate line by line in azimuth, a rigid, predetermined number of samples

509816/0618

- Λ2 -

need, and ultimately, the statistical distributions of disturbances on which their conception is based
and signals (echoes) must be known, since deviations from the distribution functions assumed when they were distributed also lead to serious deviations in the expected mode of action.

Advantageously, an automatic detector of this type is therefore subsequently supplemented by a target coordinate based on character recognition, together with auxiliary devices, for target recognition, the estimation of the target position and the automatic
Goal tracking to get the best results.
1, the principle of the target coordinate stimulator is shown together with the essential circuit parts of a digital target extractor (based on the principle of the other window detector); the heavily framed blocks represent the most important assemblies of the target coordinate controller
weakly lined boxes, on the other hand, parts of what is known as; required extraction systems. For the sake of simplicity, the following text only refers to the processing of one (1) target expectation area with 8 distance rings (4th R) and 12ß azimuth steps ( & γ )
received; but it almost goes without saying "
that the method according to the invention also applies to such expectation areas and thus in the borderline case “to the entire coverage area of a panoramic radar system
(Fig. 2) is applicable. The individual building blocks according to FIG. 1 will be described later in detail; for the time being, only the common thread for understanding the
entire, extended extraction system.

The analog radar video is integrated in the circuit ¹¹ I. Threshold ¹¹ binary and in the peak detector (peak detector ^ temporally quantisicrt and for the C Iu 11 er decor relation over several antenna runs on, whereby Verabr-edungsgeiriäß only a single target expectation area are examined more closely in each case βο11 -

509816/0610 ^ ^Γ '

After meeting certain hit criteria Integrator now distributed hits over a large area according to the composition of the target echoes, whereby for automatic acquisition and evaluation these hit patterns representing one or more targets, in addition to target extraction by binomial detectors, the methods of character recognition are used. The quantized radar video therefore runs coming from the integrator, in real time via the switch Sl into the moving window detector, which hits the '· ■■ follow their azimuthal correlation per range ring and after successful target recognition as the first step of processing target start and end of goal messages (ZA, ZE) to the line stimulator supplies. 6 and 9 show its structure, which essentially consists of an analyzer and classifier ■ fceil together with the associated memories.

The analyzer examines the binary quantized radar video signals for certain features, so-called form elements, while the classifier assigns one of the possible meanings to the respective feature leveler of the analyzer according to certain decision rules; From the type (geometric nature) and the chronological order of the recognized formula components (length, latitude, number and difference of the single and double hits, distance, etc.) a classification is made in this way, expanded by the comparison with the old hit pattern description in the target coordinate memory, carried out and the number of that distance in the relevant target expectation area passed on to the Trofferzwiechenspeicher, which should represent the most likely target center of gravity, or makes use of the possibility of returning the hit pattern. then exactly defined and the 2nd step of the task done by transferring the frame-related distance value R _{z to} the target computer.

509816/0618

Parallel to the wall window detector, the hit reports from the integrator are also disjunctively linked in the line veroderuns over two successive distance rings (= lines) and also read in real time into the hit buffer, which has the same number of bits as a target expected area. Kit for the end-of-frame pulse - the moving window detector then processed the real-time video of an expected area - switch S to position 2 and the moving window detector now uses the second pass to read and delay the radar video of the probable hit lines to determine the azimuthal, frame-related target start and end angles ( f , ^ r «) processed. In the undisturbed case, the angle values found are communicated directly to the target computer via the line stimator, which then calculates the estimated value of the center angle ^ f _{_ via the frame angle 7 D.}

If, on the other hand, there are disturbances (e.g. encounters with several targets, clutter, etc.), the target computer is transmitted from the arithmetic unit of the line determiner - appropriately modified angle values for each target, so that the target computer can then still separate targets in azimuth which, with the same transmission and receiving antenna have an angular difference smaller than 0.75 Ω when θ represents the half-width of the antenna. The third and final step in the estimation of the target position is thus made.

The target coordinates are used at the end for smoothing the track output via an interference-adaptive smoothing low-pass filter.

In this context it must also be mentioned that in the hit image memory all hit lines whose angles at initial values ψ are not more than * /? "" Angle increments (^ ς?) Differentiate »can be stored with the same target only, whereby the first incoming ZA message is arbitrarily per antenna cycle

509816/0818 - ¹ > -

Destination number 1 is assigned. All ZA messages arriving later are compared with respect to their rahiaen-relatedeaea angle start value ψ. Rait all ψ values already stored in order to be able to assign the same target user to all related hit lines of a hit pattern (use of an associative memory is very advantageous!) - Simultaneously With the ψ -values, all distance values R * of the hit lines of a target expectation area (= frame) recognized via the ZA messages are stored in the hit image memory, as well as the sums of the single hits f., the double hits f__ or the difference between the two values Arrival of the ZE report of each hit line created and saved. The type of end-of-destination message is also stored in the memory. Furthermore, the line overwrite bits record how often each range ring of a target area was filled with a hit line and a comparison of the characteristic sizes of a hit line (^. Ι ' _E i R ¹ ) from antenna circuit to antenna circuit is carried out (hit line validation) End of frame "the content of the hit memory is then processed further in the analyzer using the extraction logic, the content of the coordinate memory is further processed by the classifier using the decision logic according to certain algorithms.

1. Determine the target number ZN that has most ZA messages having. .

2. Determine the middle distance ring: (hit line) of a hit cmster in the target expectation area.

3. Search for the hit line with the largest single

hit sum ^. f _T.

max L

-. ' ■ ■ - 16 -

509816/06 18

4. Compare these two hit lines and assign ZV as target processing:

ZV = 00 £ The two distance rings according to points and 3 are not identical and their distance is greater than 2-xäR.

ZV = 0 TL, = the middle hit line has a smaller distance value R ¹ .

ZV 3 = L 0 "= The middle hit line has a larger distance value R ¹ .

ZV ss L L ^ The distances R * between the two distance rings are identical.

5. Search at least two hit lines per target number with single hit sums (ETS), their difference smaller d istj d is <^ / £ »but otherwise freely selectable.

ETS a 0 's No 2 hit lines can be found according to the above condition.

ETS = L = Fulfill two or more hit lines this condition.

(This program step also automatically examines the hit lines for approximately identical ZE messages solved.)

6. Search per target user at least 2 hit lines, where the hit difference TDS is two lines (f _L - f _LL ) is less than ^ / V.

TDS = 0 = The equation is not true. TDS «L ^ The equation is fulfilled.

509816/0618

7 · Determine the difference between single and double hits in a hit line in the hit comparator TK.

= ° - ■ * L- ^f LL > ■

- ■ ^Z L ¹ LL ^ v

8. Determine the end-of-goal declaration ZEE.

ZEE ȣ 000 Not used.

ZEE = 0OL = 2 consecutive zeros in AZ

ZEE s OLO ύ 3 "« »n

ZEE = OLL- = ^^ "* ⁿ "

ZEE = LOO = end of frame command is available ZEE a LOL = The hit line is too long ZEE = LLO = No ZE-M? / A'ung is from the moving window

detector reported
ZEE = LLL = ZE message is from the Vander window

detector reported.

9. Start the program of the classifier for processing of the target coordinate memory.

This program consists of the following steps:

9Λ. Compare the position values ψ _% VL, R ^{1 of} a selected, most likely hit line of the hit image memory from the antenna circuit U. ORed values stored with the ± xa target coordinate memory ψ ", 9 ^"»R"

A 'L

from the antenna circuit U for assistance

of the system of equations.

/ R "- R '/ <LAR

■ - 18 -509816/0618

The values ψ ", 9V ¹ » ^R ", which are crossed out twice, represent the frame-related /}, ored? Angle start, angle end and distance values. If these equations are fulfilled and there are no frame or target collisions, the values Ψ. ', C ^ ₀ ¹ , R' to target

A ' Ej

position determination are passed on to the target computer.

If rejected, the next Ziolnuramer is and restart the analyzer program.

5.2 · If, on the other hand, there is a frame or target collision (RK or ZK) - here, "frame collision" means the extensive overlap of two target expectation areas and "target collision" ±% t means two hit patterns characterizing two targets to calculate the following system of equations for any rack shift (RV) initiated by the target computer:

- ^ A (N) / ^ R ⁺ SV (N ₊ I) ^ N) _; ^ R ⁺ 5 ⁶ E.

R _R + R »= R _{Z (N)} . R ₁₁₊ R ^ _{n + 1} J - ^r _Z ( _{N +} D

The following applies:

RV = 00 = there is no frame shift RV = O L j = there is an azimuthal distance

advance
RV ss LO = There is a range shift in

the distance in front of RV s = L L = There is an azimuthal and distant

normal frame shift

5098 16/06 18

In these cases of conflict, must in fact take place with the relative, that is, frame-related target coordinates, now be calculated with the North-related that absolute destination coordinates to be able to perform a Zielseparierung or the assignment of associated goals and Zielerwartuagsgebieten (Rahaon) including the destination symbols. The goal-

Icoordinates ψ. , ψ ', R_ of the newly recognized A .15 L

Target are indexed rait (N + 1) _s the coordinates from the previous antenna cycle U ^ have the index N. -

3. Compare the target coordinates ^ κ / w ju γ p / vT ₁ \ i Ry / vj -ι-λ of the newly discovered target with the. corresponding coordinates of all targets "discovered in the antenna run N" which are stored in the target coordinate memory. If, during this comparison, a coincidence within the tolerances of azimuth angles uia fy ^ 'A'-p and of distance around ΖΆ R is determined for the frame being processed , these newly determined target coordinates are output by the target coordinate estimator to the target computer, taking into account the target quality. If, on the other hand, there is no coincidence of coordinates in the sense required above, or if a coincidence is established in another target expectation area, this knowledge is made by setting a target quality bit in this At the same time, the next target number, which contains the next largest number of ZA messages, is requested for processing, in order to still lead to a target discovery or target clarification in this or a far Another pass - until all targets have been processed.

509 8 167 06 1 8

4 »The target quality is in a 2 bit word with the following assignment

ZQ β GO = No target validation is possible in a frame

ZQ = O L = target validation is in-house Frame possible

ZQ = L O β target validation is in others Frame possible

ZQ = "LL = target validation is possible in your own and in other contexts

5 · In the event of a rejection of all targets of a target expectation area set a target bridging bit and give the command "target failure with target bridging" to the target computer.

6. If the target override counter reaches a predetermined one Threshold so report "target failure with alarm" to the target computer.

The program sequence of the target coordinate stimulator is now again fixed by four subroutines:

1. Assignment of target and expected target area manually or automatically.

2. Free target movement.

3 · Collision of target expectation areas.

4. The hit oysters characterizing the targets run into one another (= bit rougher).

■ whereby the target computer does not recognize these operating states. ^{It only calls the target position values 4} ^ ', Lf- * and R' from the estimator at the azimuthal end of a target expectation area

^{A E} ^

in order, among other things, to assign the polar target coordinates <f ~ and R ₂ as estimated values of the target position

509816/0618 " ^2Λ ~

_ 21 _

calculate and at the same time a target validation perform to eliminate "outliers". For the better. Understanding, some more Explanations are given in the following text, alternatively using the words target expectation area and frame *. . -. »W can be used. ·

To 1. -.

With the manual setting, any frame is hand over; the target to be pursued is set, ie target and target expectation area coupled and then the automatic button is pressed. No contact (collision) (= frame) manually or automatically between two expected areas' take place to all the difficulties caused by the response of several moving window detectors due to empty Trefferbild- and Zielkoordinatenspeichcr to avoid .. The to be tracked target - in principle may during d & a setting operation should be clearly recognized in the frame assigned to it in the setting stage and assigned the appropriate symbol.

A free target movement exists when no frames touch or overlap, i.e. a conjunction Any two Rahrcenflip-flops must not be envied be. Such a frame flip-flop will last as long as set, as the radar antenna, the target expectation area in question strokes. If there is no framework meeting, then no two goals can be pursued. occur in an expectation area and there will only be one - namely the most likely hit line (Fig. 8) - as a target line of gravity at the hit intermediate. memory output, although- but due to the target criteria several goals can be recognized within the framework, but all of them rated a lower target quality C ^ Sum of all target criteria) ■

5098 16/06 18 -'22 -

If one or more conjunctions are fulfilled via the frame flip-flops, there is a collision of the target expectation areas and the analyzer's extraction logic no longer reports the most likely hit line to the hit buffer, since this hit line is yes -n & h to the hit pattern of a colliding one Frame embedded target could belong. Rather, the frame collision bit is now set in the target coordinate memory with the effect that before the output of the. Classifier determined the most likely target coordinates, these with those stored in the target coordinate memory. . can be compared resulting from the previous antenna circuit and then forwarded to the target computer if they match. If, on the other hand, there is no coordinate coincidence, then these previously determined coordinates are compared with all the coordinates of all target coordinate memories in which a frame collision bit is set and, when a coincidence is encountered, output as target coordinates of the relevant target expected area.

Then the next probable target number becomes retrieved from the target coordinate memory and the program sequence is repeated as described above. Can after processing all target cursors in the hit image memory no probable hit line and thus

will
no target recognized "J KG 'a target failure bit is set in the associated target coordinate memory and the polar coordinate ψ. ' = * / * -, · = 0 is transferred to the target computer,

A J-

eo that it decides on bridging the target and outputs the target coordinates from the previous computing cycle to the smoothing low pass.

If the following system of equations is not satisfied, then can refer to a merging of the hit image, i.e. a Target collision (F sharp 4) are closed:

-23 509 816/0618

ZA, k = f-

^R ' _A 1 - ^RI A2. * ^A * ^R

^R1 E1 - ^RI E2 ^ ^{2Δ R} ~>

with f = radar pulse repetition frequency θ β. Antennae half width
U «= antenna revolutions per minute

As a result, only target coordinates that were not already stored in the target coordinate memory of another frame are stored in the target coordinate memory permanently assigned to the target maintenance areas processed as belonging, but that j is the most likely line with the best features belonging to the frame target.

To k,. ·.

Is the previously mentioned system of equations no longer exactly fulfilled, the classifier decides Target collision and in every expectation area involved in the collision is stored in the ziol coordinate memory the target collision bit is set. It should be noted that the simultaneous response of several moving window detectors - every target expected area is a wall window detector assigned - not to be evaluated as a target collision criterion.

The set ziol collision bit lifts the odorization Radar video in the direction of the range, but still as before the target position values ^. "l '/' β" ι R "are in each target coordinate memory - but accordingly modified to the following equations ^ - filed.

- 24 -

5098 16/06 18

If there is a collision between four frames RA1 and RA2, so applies to the

Frame Ij The target angle starting value V _A '»reported by the moving window detector is multiplied by a factor of 2 and the target angle value ψ _Ε ^% liull is set.

Frame 2: £ /> «-. 2 ^ ⁰

E.
= 0

This means that the apparent target angle ^ ₇ '- in the first target discovered in the azimuthal direction of a target expectation area ^ ₁ = 7 _zl and in the second

Target φ ^ = £ ^ _Z2 set.

The corresponding assignment of the angles beginning or -endverte as the target ridge angle is determined by comparing the angle values supplied by the respective V / other window detectors 7. 'received with the values stored in the corresponding target coordinate memories, · still missing: the end of target message in a hit line or if it is recognized as incorrect, then ψ \ Ξ ψ' is set for this target, but the ZA message is missing or is falsified (is determined by comparing the target coordinate memory contents), then ^ g ¹ = V _{z is} set. The last criterion in the hit image memory is the largest initial angle value ψ. or the smallest Winlcelend-

value ^ _ - determined for the one goal and for this one Efflxn

Case ψ * = ψ _z ^% set and for the other. The aim is the smallest starting angle value ψ, and. the largest final angle value ψ ~ and then set VL ¹ = Ψτ * for this target.

509816/0618

As the range value R 'of each of the two targets, the mean range ring of all hit lines beginning and ending with an unadulterated ψ :' or 7_, respectively, is determined (see FIG. 4). In even-numbered rows of the hit of a target, the number of matching rows is divided by 2 and d / eses result fixed distance in the direction of de »output the distance ring.

If, on the other hand, there is a collision between three frames, then the following applies to the target with the largest starting angle value

^ Amax ' Ψζ ^Ξ ? L'
for the target with the largest final angle value

and for the target embedded in the center, for which none of the angular values V » ⁻¹ « V ₁ , ¹ transmitted by its moving window detector match those in its

A Γ »

Target coordinate values stored in the memory ^{is identical.} tp! + <* '·

(0 * ■ ~ t

The determination of the distance is carried out as described above, with half the sum of all evaluated hit lines used for the middle target - in the distance direction - at $ R ¹ - Οίαίλ in the case of an even number, the last hit line in terms of distance of the first half of the hit pattern is output.

The detailed description of the individual function groups that follows does not bring any fundamentally new ones Knowledge, but only serves for a better understanding of the entire circuit (Fig.'I). The analogue, which for reasons of dynamism is usually hard limited, rectified wheel video is binary quantized by the regulated 1st threshold, although strictly It is important to ensure that this regulation on shading effects does not result in any loss of information,

509816/0818.

This amplitude quantization is followed by the temporal quantization of the radar echoes 2.5 db higher probability of detection by the peak detector in front of the edge detector preference is given, the echo sequence is thus synchronized with the machine cycle.

In order to ensure that the signal-to-noise ratio between useful and interfering target echoes is large enough for automatic signal evaluation to get, the digital echo sequence is used to suppress the statisiioi distributed interference and further processed for clutter decoration via a digital, non-recursive integrator (FIG. 3) and the strongly correlated useful signals are separated from the less correlated interference signals. at of this temporal decorrelation, the decorrelation time of the sea echoes must be taken into account up to 100 ms. -

A pulse to pulse integration of the radar memory lying within an antenna half-width; enlarged the signal-to-noise ratio not because the pulse repetition frequency is too high, i.e. the time interval between successive ones Memory * is too small. The prerequisite for integration over several antennae runs (Image to image integration) is that the goal is in the same resolution cell during the integration is located. It should therefore with a spatially fixed angular grid and with the repetition frequency the radar pulse repetition frequency derived from the azimuthal angle increments be worked (Fig. 2). The tracking of the relatively small integrator memory r

8χ128 ^ bit} for moving targets or also Wandering target expectation areas (Fig. 4) is carried out by the target computer, as is the corresponding necessary azimuthal or distance-related shift the information stored in these memories. For reasons of complexity, this non-coherent integration in the Vxdeo area are accepted, with «11erdings the ^ loss fields are coherent in the moving window detector

509816/0618 _2?

Integration (collapsing losses) partially "compensated: wearden» ■ · ..-'- · ■ - ■ "" -. ·, · - -... -.- ".. ■ ..-. <:

Despite the greater memory requirements compared to a recursive integrator, a low-pass filter with a symmetrical impulse response ripi Ze / ib he was given preference because of the smaller target center angle error, whereby this situation in Fig. 3 cs with the target expectation area El *, the η-times ( n = 5) is called up , is illustrated. Each target expectation area now has surface elements on Ci = 8, .j = 128), 'which are covered with the echo pulses of the echo sequence when the radar antenna sweeps over the target expectation area. An echo pulse then becomes a hit pulse declares if it fulfills the hit criterion set in each case, ie within a certain number of antenna runs U must have been counted at least F echoijapul; se. per. investigated surface element. ¹ ^ .... -.-. I. ·. · .- · · (From D possible echoimt / ulsen must... Ita

cv-at least

Considered period of time TF be present.) In Integrator takes place, i.e. the conversion of the echo pulse sequence (disturbance and useful target echoes) into a hit pulse sequence (mainly useful target echoes), whereby during the conversion process due to different temporal evaluations of the echo pulses of a surface element a different reacting, the respective Uta world conditions well adaptable system can be realized. 'This hit pulse sequence of a target expected area (see Fig. K) is then disjunctively linked in real time in the circuit part "line decoding" (see Fig. 5) via two successive distance increments Δ R (£ distance ringc ^ line), with switch S2 optionally unfiltered or filtered quantized video can be selected for processing. Another switch S3 cancels this ORing again before the gate 01 before the real-time video in the hit buffer

509816/0618 - 28 -

is read in. This memory has the same number of bits as a target expectation area (e.g. 8x128) and has the task of delaying the eroded, mostly filtered operating video during the real-time processing of a target expectation area and after the end of the frame of this video it is selected according to the most likely hit lines in the moving window detector to determine the frame-related, i.e. relative target start and target end angle ^ f. ¹ or ψ '' . This disjunctive linking of the radar data over 2 range increments & R eliminates quantization effects and integration losses and the hit patterns of a target that are torn at the azimuthal edges are smoothed. This ORing may only be carried out in the distance, but not in the azimuthal direction, since otherwise it would simulate transmission pulses that are not present at all. The angle values Q * then determined by the moving window detector. '

and ψ * of a target are called up by the estimator at a given time, temporarily stored in the target coordinate memory (FIG. 7) and transferred to the target computer in undisturbed operation to calculate the target center angle in the target area. In addition, the target computer uses the position of the expected target area known to it to determine the estimated coordinate values R n and-of the target, which are then output via an interference-adaptive smoothing low pass for smoothing the track.

At the same time, the undelayed, not eroded Radar video but still in a known way in a moving window detector and processed according to the invention in the line destinator. Are in the Wanderfonster detector in Corrcled the individual range rings over the azimuth Hit pulse sequences recognized, then for each hit sequence after certain target criteria have been met Target start and target moon messages (ZA, ZE) are delivered to the estimator and the associated angle

509816/0618

-29 -

values V. ¹ and 4 ^ * one after the other in the hit

A L

Image memory stored word by word (see Fig. 7) »whereby up to 16 ZA-ZE messages including the associated distance increments a ^ l R, which correspond to a maximum of 3 ^ destinations, can be saved per target area. Furthermore, the sizes of an extraction logic determined by the analyzer, such as destination number, difference f the number of single hit in this memory. W ~ OCE the double hit f _TT (for de-tuning the degree of correlation of a matching row), type of target end of generation (ZE, over long matching row , closed sequence of zeros, end of frame), line validation values and the number of line overwrites are stored; That is to say, all the physical features of a hit pattern that corresponds to one or more targets, some of which have run into one another, are provided as binary components of a feature vector by the Vander window detector and the analyzer for the occupancy of this hit image memory. Each hit pattern (FIG. k) is thus represented by a sentence of features described, with the help of which a classification can be carried out in the following, together with the previous events. These events are stored in the so-called target coordinate memory (Fig. 7) as target quality, ored angle start and angle end values (^ V '»^ r") " ^Distance S ^R " ^{in the} target waiting area, number of target bridges, target collision, frame collision, frame shift and the estimated values of the position of the target waiting areas V _R , Rp from the previous antenna cycle.

In the classifier (s. Fig. 6) is subsequently calculated by a decision logic according to certain rules from the feature vector with seineu binary components, the result vector, whereby it is, however, observed to that the probability density functions that describe the class frequency of Merkraalsvektoron statistically, either in their analytical form

509816/0618 -. 30-

are still known in their statistical parameters, since the mathematical formulation of the sea clutter alone has not yet been successful.

In summary, the following applies: The analyzer uses the extraction logic to examine the radar signal with regard to certain features and fixes the most likely hit lines (see Fig. 9) of each target found in the target area, i.e. the features are determined in such a way that the difference between the radar signals of different useful targets is particularly evident ι during the classifier with the decision logic with these feature vectors, the special meanings as known destination, collision opponent disorders, targets with coordinate transformation, non-identifiable objects, non-separable objectives, etc. allocates so that achieved by the Zielkoordinatenestimator for an optimum decision rule a maximum of reliability of recognition and the three types of errors: failure to recognize, mix-up and superimposition of disruptive or useful targets is printed to a minimum .

In order to keep the hardware and software expenditure within reasonable limits. hold, only Merkraal vectors with a few, binary-coded components are considered, whereby in principle each digit of these binary numbers is to be regarded as an independent feature. In contrast to character recognition, however, there is no assignment here, instead a standard character (^ hit window of a target S "rectangle") is to be recognized and the position of the recognized standard character (which can change in its geometry to a limited extent) in the character recognition generally no blind plays, determine in the picture element grid of the target expectancy area as "ones-median line".

509816/0618 - ₃₁

Claims (10)

- 31 - in »73 / 134- Claims (1.! Scarf device arrangement for estimating the position of the 'pulse radar devices, in particular 2D ship radar systems, scanned, severely disturbed targets to enable automatic target tracking with as far as possible smoothed track display, characterized in that to save digital storage space in the Clutter decorrelation after manual signal preprocessing, only the binary quantized radar video of each target assigned a target "expectation area" is added up evaluated in a non-recursive integrator with a rigid angular grid over several antenna revolutions and that after this low-pass filter has settled and after certain hit criteria have been met per quantized surface element (ΔE . Δ Ψ = resolution cell) and the associated filtered hit reports are output per antenna cycle for each target expectation area, which then on the one hand in real time via a permanently assigned or time-division multiplexed Wanderfe nsterdetektor are processed, whereby the target start and target end messages (ZA, ZE) obtained therefrom in the following line stimulator, which is composed of an analyzer with extraction logic and hit image memory and a subsequent classifier with decision logic and target coordinate memory, together with the single and double hits of a target after performing some arithmetic and logical operations to determine the most likely target center of gravity at distance IL. can be used, on the other hand after a disjunctive connection over - 32 509816/0618 - 32 - UL 73 / 134- two successive range rings are stored in a hit buffer, the number of bits of which corresponds to the number of quantized surface elements of a target expectation area and that with the arrival of the frame end pulse of a target expectation area via the line selector of the line destimator, only the most likely target gravity line from the hit buffer is now fed into the same wandering window this hit sequence is also examined for ZA and ZE messages and, in normal operation, the associated angle start and angle end values Φ ¹ <. or Ψ'-ρ, based on the azimuthal start of a target expectation area, to the target computer for calculating the target center angle Ψ _ζ , whereas in the disturbed operation these two angle values are made available to the target computer modified by the line specifier and that the target coordinate estimations 9 for time reasons ^, 3 * 2 ^ ⁿ parallel processors or associative memories are used. 2. Circuit arrangement according to claim 1, characterized in that that in addition to the azimuth, the elevation of a target is also processed according to the same principles according to the invention. 3. Circuit arrangement according to claim 1 or 2, characterized in that that instead of using parallel working estimators there is also a serial working method. 4. Circuit arrangement according to one of claims Λ to 3 »characterized in that for better target separation in the case of target collisions, the line stimulator for the target first recognized in azimuthal sighting only has the initial angle value multiplied by the factor 2 (2. Φ» ¹ ) and for the target last - 33 -509816/0618 - 33 - DEATH 73 / 134- recognized target only outputs the final angle value multiplied by the same factor (2. Ψ _Ε ') - both related to the start of the frame - to the target computer for determining the target center angle Φ «. 5. Circuit arrangement according to one of claims 1 to 4-, characterized in that an evaluating, non-recursive integrator in a spatial one for clutter decoration fixed angular raster is used, the memory of which is preferably designed as a holographic memory. 6. Circuit arrangement according to one of claims 1 to 5i characterized in that, to reduce storage space, the entire detection area is divided into individual, appropriately designed, Target expectation areas is divided. 7. Circuit arrangement according to one of claims 1 to 6, characterized in that a peak detector is used to increase the detection probability for the time quantization is used, which is queried with a scanning grid that corresponds to a quarter of the transmission pulse length will. 8. The circuit arrangement according to one of claims 1 to 7, characterized is that the contents again, but zeilenverodert each target expectation area over the same Vander window detector once in real-time and after the "end of frame", processed, wherein in each case only the wa h rscheinlichste matching row to Processing arrives. 9. Circuit arrangement according to one of claims 1 to 8, characterized in that to increase the Arbeitsgeschwin- 509816/0618 - 34- - UL 73 / 134- the target extractor and target coordinate stimulator whose memories are constructed as associative memories. 10. · Circuit arrangement according to one of claims 1 to 9 »characterized in that in each case for the quantized single: sional VBroaerbe 70 -a. / video of two resolution cells a two-dimensional matrix window with kxk (e.g. 3x3) elements and that this Penster created a new hit pattern for saving in forms the hit buffer in such a way that a logical one ("L") goes into that of the respective matrix window position corresponding resolution cell of the target expectation area is only enrolled if the hitsurce of the window exceeds a predetermined threshold, this matrix corresponding to the speed of propagation the transmission pulse and the angular velocity of the antenna is guided across the expected target area. 50981 6/061 8