Method for detecting radar lobes from a plurality of electromagnetic pulses collected by an electronic warfare antenna system

Abstract

Method for detecting radar lobes from a plurality of electromagnetic pulses collected by an electronic warfare antenna system. This method consists of: selecting a pulse ( ) whose amplitude ( ) is a local maximum in a time sequence of pulses, as the reference pulse ( ); defining a search domain (D) in the parameter space of a parametric family of parabolas modeling pulses emitted by the passage of a radar lobe through an observed electromagnetic scene, the search domain taking into account the arrival time and amplitude of the reference pulse; identifying parameters of an optimal parabola from the parametric family of parabolas by a simulated annealing process traversing the search domain; and grouping pulses close to the optimal parabola as pulses belonging to the same radar lobe. Figure for the abstract: Figure 1

Description

Title of the invention: Method for detecting radar lobes from a plurality of electromagnetic pulses collected by an electronic warfare antenna system

[0001] The invention has as its general domain that of electronic warfare - GE. Its particular domain is that of deinterlacing processing methods.

[0002] In GE, an antenna system listens to an electromagnetic scene - EM. The antenna system collects, over an acquisition time window, a plurality of pulses. Each pulse is characterized by a time of arrival - TOA ("Time of Arrival") and an amplitude - PA ("Pulse Amplitude").

[0003] The collected pulses are mixed in the sense that they come from an unknown number of sources present in the observed EM scene, each source emitting pulses according to one or more unknown waveforms (the number of waveforms also being unknown).

[0004] The objective of deinterlacing is then to partition the collected pulses into several groups, each group being associated with a particular emitting source.

[0005] Among the emission sources of interest are scanning radars.

[0006] The problem to be solved then consists of detecting the passage of a lobe of a scanning radar in the observed EM scene, in order to group the pulses which correspond to this lobe, so as to inform a deinterlacing processing implemented downstream of such a detection.

[0007] The object of the present invention is to solve this problem.

[0008] For this purpose, the invention relates to a method for detecting radar lobes from a plurality of electromagnetic pulses collected, over an acquisition time window, by an electronic warfare antenna system, each pulse being characterized by an arrival time and an amplitude, the method being characterized in that it comprises the steps of: selecting a pulse whose amplitude is a local maximum in a time sequence of pulses, as a reference pulse; defining a search domain in the parameter space of a parametric family of parabolas modeling pulses emitted by the passage of the lobe of a radar in an electromagnetic scene observed by the electronic warfare antenna system, the search domain taking into account the arrival time and the amplitude of the reference pulse;identify parameters of an optimal parabola from the parametric family of parabolas by a simulated annealing process traversing the search domain and using the; collected pulses; and group pulses close to the optimal parabola, as pulses belonging to the same radar lobe.

[0009] According to particular embodiments, this process comprises one or more of the following characteristics, taken individually or in all technically possible combinations:

[0010] - the parametric family of parabolas f is described by a parametric model:

[0011] =

[0012] where the parameter set H' comprises a lobe width of -3dB, L, and the coordinates of a lobe peak, respectively a along a time dimension and fi along an amplitude dimension;

[0013] - the search domain is defined as the product of three intervals elementary, respectively a first interval of values ​​for the lobe width, a second interval of values ​​for a peak arrival time, Da, and a third interval of values ​​for the peak amplitude, ;

[0014] - the reference pulse xRef being defined by an arrival time tj and a amplitude yj: the first interval DL is chosen of constant width; the second interval D„ is chosen of the form: [G- ôt, tj + ôt], with a half-width equal to ôt; and the third interval is chosen of the form: [Ty - ôy, yj + ôj+], with a lower “half-width” equal to 5y and an upper “half-width” equal to <5v+;

[0015] - The simulated annealing process is implemented to optimize a score function which takes a parabola f defined by a set of parameters w and returns a number of impulses xi close to said parabola:

[0016] s

[0017] with f 1 if Xf is close to f sw(xi) = ] n . 1 0 otherwise and N the number of pulses collected;

[0018] - an impulse xi is said to be close to the parabola f when it is at a distance less than a predefined constant e: -yj f, with the momentum of reference xRef being defined by an arrival time tj and an amplitude yj;

[0019] - at the current iteration k of the simulated annealing process, a new set parameters are sampled from the search domain and then compared to a set of parameters retained from the previous iteration, using the The following criteria: if SVk > then the parameter set is retained as the parameter set retained wk at the current iteration; otherwise the set of parameters wk is retained as the set of parameters retained at the iteration current with a probability of the form ; and otherwise, when the parameter set wk is not retained, the parameter set is retained as the parameter set wk retained at the current iteration;

[0020] - the set of parameters retained at iteration K is the optimal solution w* — K being a predefined integer;

[0021] - the collected pulses L \ which are close to the optimal parabola / xj\lr yj] are grouped according to the relation: | fw( [ < £, where " a predefined constant.

[0022] The invention also relates to a computer program comprising software instructions which, when executed by a computer, implement a process in accordance with the previous process.

[0023] The invention and its advantages will be better understood upon reading the following detailed description of a particular embodiment, given solely by way of non-limiting example, this description being made with reference to the accompanying drawings in which:

[0024] [Fig-1] Fig. 1 is a time-of-arrival (TOA) - amplitude-of-axis (PA) graph on which the pulses collected during an acquisition window by the GE antenna system are plotted; and,

[0025] [Fig.2] The [Fig.2] is a block representation of a preferred embodiment of the process according to the invention.

[0026] The radar lobe detection method according to the invention is based on the assumption that the pulses emitted by a scanning pulse radar, one lobe of which, primary or secondary, passes through the electromagnetic scene observed by the GE antenna system, are distributed according to a parabola, in the arrival time-amplitude plane.

[0027] A family of parabolas is defined by the following parametric model:

[0028] f^t) = _n(t_a)2 + p

[0029] where the parameter set w includes the width of the lobe at -3 dB, L, and the position coordinates of the lobe vertex S, respectively a according to the time dimension and P according to the amplitude dimension.

[0030] The next step is to identify the parameters of an optimal parabola of this family which fits a subset of the acquired pulses and which can therefore be grouped together as emitted by the same radar source.

[0031] More specifically, with reference to [Fig.2], the method 100 according to the invention begins with a step 110 of collecting, by the GE antenna system, a plurality of pulses, during an acquisition time window, for example of 120 ms.

[0032] N pulses are collected. N is not constant and varies from one acquisition window to another.

[0033] Each pulse xi, i an integer between 1 and N, is characterized by an arrival time and an amplitude

[0034] Step 120 consists of defining a research area.

[0035] Empirically, it is observed that the pulse with the highest local amplitude is close to the vertex of a parabola. For example, the pulse xRef corresponds to a local maximum amplitude. For example, the pulse x_j is selected if its amplitude is greater than m previous pulses and m subsequent pulses (m being an integer greater than one). Alternatively, it is possible to iterate the process. In this case, the global maximum of the pulse set is used as the reference pulse for the current iteration. The pulses clustered around the parabola optimized for the current iteration of the pulse set are then removed before the process is repeated for the next iteration on the remaining pulses.

[0036] However, it is not sufficient to consider this reference pulse as the vertex of a parabola. Not only are pulses generally subject to noise, but some radars also use a waveform in which successively emitted pulses have different frequencies. Now, the attenuation by air of the amplitude of a pulse between its emission by the radar and its detection by the antenna system depends on the frequency. It may therefore happen that the pulse xj corresponds to a strongly absorbed frequency and that its amplitude therefore does not correspond to the vertex of the parabola to be identified.

[0037] A search domain D is therefore defined in the three-dimensional space of the parameters w of the parametric model. This domain results from the product of three elementary intervals, respectively a first interval of values ​​for the width of the lobe, a second interval of values ​​for the time of the lobe apex, Da, and a third interval of values ​​for the amplitude of the lobe apex, Dp,

[0038] The first interval DL is, for example, chosen to have a constant width, between predefined lower and upper bounds: [L-, L+], respectively.

[0039] The second interval Da is for example chosen of the form: [tj - ôt, tj + ôr]. It is symmetric around the arrival time of the reference pulse, tj, with a half-width equal to ôt, a predefined parameter.

[0040] The third interval Dp is, for example, chosen to be of the form: [Jy - Ôy, yj + 5j+]. H is preferably asymmetrical around the amplitude of the incoming pulse, Jy, with a lower "half-width" equal to Ôy and an upper "half-width" equal to 5y+. The lower and upper half-widths, which are predefined, may be different from each other. This is particularly to take into account the effects of a frequency variation between pulses originating from the same source.

[0041]

[0042]

[0043]

[0044]

[0045]

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057] Step 130 involves optimizing the parameters of the parabolic model to identify the parameters of the optimal parabola. This optimization step is carried out by implementing a simulated annealing process. Simulated annealing, which is presented in detail in the article Scott Kirkpatricket et al., "Optimization by simulated annealing", Science, 220(4598):671-680, 1983, is an optimization metaheuristic, used here to estimate the parameters of the parabola that best fits a portion of the pulses collected in the EM scene. Simulated annealing is implemented here to optimize a score function Sw, which, given the parabola defined by the parameters w, returns the number of pulses close to that parabola. A pulse i is said to be close to the parabola f when it is at a distance less than a predefined constant e. Therefore: With : f 1 Sï 1 / (t)-j.|<£ I 0 otherwise Thus, in the search domain D, we seek the optimal parabola f* whose parameters w” are given by: / = argmax5„, (wel) Simulated annealing is an algorithm for finding an optimum by traversing the search domain D. Simulated annealing is an iterative process. For initialization, a set of parameters is sampled, for example by random sampling, in the search domain D. Then, iteration k begins with a substep 132 during which a new set of parameters wk is sampled in the search domain D. Then, in substep 134, it is compared to the set of parameters retained wk in the previous iteration: If Swk > Swk^ then the set wk is retained as the parameter set retained wk at the current iteration; and, Otherwise, the set wk is retained as the parameter set retained wk at the current iteration with a certain probability: / >(&,, T) =exp(-^A h ) If the set wk is rejected, the set of parameters retained wk[ from the previous iteration is kept:

[0058] Then, at step 136, the integer k is incremented by one unit for the next iteration.

[0059] According to the simulated annealing process, the so-called "temperature" variable, T, with positive values, follows a decreasing evolution law with each iteration. In step 138, a new temperature is defined for the next iteration.

[0060] The stopping condition of the iterative process is for example a maximum number of iterations K (test carried out during substep 135).

[0061] The returned solution is then considered the optimal solution. Thus: & AW ~WK'

[0062] This allows us to obtain an estimate of w* despite a large number of impulses outside the model to be optimized.

[0063] Step 140 then consists of grouping the pulses.

[0064] The pulses collected at step 110 that are close to the optimal parabola are associated in the group Gw^, i.e., the pulses j such that:

[0065] [f^-y^E

[0066] It is possible to extract the pulses belonging to the group Gw* from the set of pulses considered as input to step 120 and to iterate steps 120 to 140 around another local maximum among the remaining pulses, in order to seek to constitute other groups, in particular groups associated with other radar lobes.

[0067] Alternatively, the proximity condition between optimal parabola and impulse could take another form, such as a quadratic distance, or with a parameter £ which is not constant, but evolves for example as a function of the amplitude.

[0068] The invention therefore makes it possible to identify, in an observation scene composed of radar pulses from different emitters, the lobe passage, without prior knowledge about this source.

[0069] Existing methods have the disadvantage of either not being robust to impulses belonging to other sources, or being too sensitive to their proportion.

[0070] On the contrary, the invention has the advantage of being robust to impulses outside the parabolic model thanks to the scoring function. Furthermore, the algorithm has a computation time independent of the proportion of such impulses.

[0071] An iteration of the process, with extraction of grouped pulses at each step, makes it possible to detect several lobes, and therefore to distribute the pulses into several groups.

[0072] It should be noted that several groups can be constructed for the same scanning radar, each group corresponding to the passage of the main lobe or one of the secondary lobes of this radar.

[0073] Furthermore, at the end of the implementation of the lobe detection method according to the invention, a certain number of pulses may remain outside the radar lobe pattern.

[0074] Downstream of the detection process according to the invention, it is the deinterlacing process itself which performs, from the pulses and the groups to which these pulses could be attached, a processing of identification and characterization of the different sources present in the EM scene listened to.

[0075] The method according to the invention is implemented by computer, for example by a computer associated with the GE antenna system and suitably programmed for this purpose.

Claims

Demands

1. A method (100) for detecting radar lobes from a plurality of electromagnetic pulses collected, over an acquisition time window, by an electronic warfare antenna system, each pulse (x0) being characterized by an arrival time (¾) and an amplitude (^,), the method being characterized in that it comprises the steps of: - selecting a pulse whose amplitude is a local maximum in a time sequence of pulses, as a reference pulse (xRef); - defining a search domain (D) in the parameter space of a parametric family of parabolas modeling pulses emitted by the passage of the lobe of a radar in an electromagnetic scene observed by the electronic warfare antenna system, the search domain taking into account the arrival time and the amplitude of the reference pulse;- to identify parameters of an optimal parabola from the parametric family of parabolas by a simulated annealing process traversing the search domain and using the collected pulses; and, - to group pulses close to the optimal parabola, as pulses belonging to the same radar lobe.

2. A method according to claim 1, wherein the parametric family of parabolas ft is described by a parametric model: = -^At-aŸ + H where the parameter set w comprises a lobe width at -3dB, L, and the coordinates of a lobe peak, respectively a in a time dimension and fi in an amplitude dimension.

3. A method according to claim 2, wherein the search domain is defined as the product of three elementary intervals, respectively a first interval of values ​​for the lobe width, DL, a second interval of values ​​for a peak arrival time, Da, and a third interval of values ​​for the peak amplitude,

4. A method according to claim 3, wherein, the reference pulse xRef being defined by an arrival time tj and an amplitude J / : - the first interval DL is chosen to have a constant width; - the second interval Da is chosen to be of the form: [tj - ôt, tj + ôt], with a half-width equal to ôt; and, - the third interval Dp is chosen to be of the form: [Jy - 5y, yj + 5j+], with a lower "half-width" equal to 8y and an upper "half-width" equal to 5y+-

5. A method according to any one of claims 1 to 4, wherein the simulated annealing process is implemented to optimize a score function which, to a parabola fw defined by a set of parameters w, returns a number of pulses xi close to said parabola: with [ 1 if X; is close to f and N the number Sw(,Xi) — | „ . 1 0 otherwise of pulses collected.

6. Method according to claim 5, wherein a pulse xi is said to be close to the parabola f when it is at a distance less than a predefined constant e: ) - y | < E, with the reference pulse xRef being defined by an arrival time tj and an amplitude J j.

7. A method according to any one of the preceding claims, wherein, at the current iteration k of the simulated annealing process, a new set of parameters is sampled in the search domain and then compared to a parameter set retained at the previous iteration, using the following criteria: - if Swk > S^j, then the parameter set wk is retained as the retained parameter set wk at the current iteration; - otherwise, the parameter set wk is retained as the retained parameter set at the current iteration with a probability of the form: / \ ; and, exP\ ' T ) - otherwise, when the parameter set wk is not retained, the parameter set is retained as the parameter set wk retained at the current iteration.

8. Method according to claim 7, wherein the parameter set retained at iteration K is the optimal solution w* ~ K being a predefined integer.

9. A method according to any one of the preceding claims, wherein the collected impulses \ which are close Xj\J" to the optimal parabola / , are grouped according to the relation: ] / „,( / / ) - yj £, where £ is a predefined constant.

10. A computer program comprising software instructions which, when executed by a computer, implement a method for detecting radar lobes from a plurality of pulses collected by an electronic warfare antenna system according to any one of the preceding claims.

Images