NL1024654C2 - Doppler polarimetric detection and classification methods. - Google Patents

Description

Doppler polarimetric detection and classification methods

The present invention relates to radar technology and, more particularly, to a detection method and a classification method adapted for polarimetric radar. The detection and classification methods of the invention are extremely suitable for detecting and classifying small objects in an environment with much more power.

Identification and discrimination of objects can be achieved by polarizing a transmitted RF signal and checking the co-linear and cross-polarized scattering characteristics of an object. For simple radar objects, the ratio between co-linear polarized and cross-polarized scattered energy reflected by the target is largely constant, while for more complex radar objects, the co-linear and cross-polarized ratios vary according to variations in the distance between the radar system and the object. In this case, radars apply a polarimetric, non-coherent measurement, for example for the weather.

20

Double (separate) antenna-based non-coherent polarimetric systems are used for this purpose. Compared to radar that works with identical transmit and receive polarizations, polarimetric radar provides a more complete description of the reflection behavior of the environment of the radar. Instead of the one-quantity RCS (Radar Cross-Section, radar return plane), a 2x2 complex scattering matrix is estimated, which is usually referred to as S. Reflection properties for every possible send and receive polarization can be derived from this matrix. To obtain this additional information, however, a relatively advanced antenna system is required that supports multiple polarizations. Unfortunately, such polarimetric systems do not work correctly when Doppler shifts occur.

In other cases, radars may use coherent Doppler methods in FMCW (Frequency Modulated Continuous Wave, 1024654 2 frequency modulated continuous wave) or pulse waveforms. Doppler radars are used to detect objects in locations where echo signals have a high nap or noise content, which, when using conventional radar signal processing techniques, hinders the detection of the presence of an object.

Doppler radars may be able to detect moving objects under conditions where conventional processing is ineffective. A conventional MTI (Moving Target Indication) radar subtracts the echo signal received from one pulse from the echo signal for the next pulse, as a means to remove the echo signals from stationary objects (dutter), and only the echoes of moving objects. Doppler radar is dependent on the frequency shift in the echo signal, which is caused by the speed of an object that reflects the transmitted radar signal. In general, the objects that reflect the clutter components of the echo signals are stationary or moving very slowly. Consequently, clutter echo signals have essentially the same frequency as the emitted radar signal. If the object sought or followed is moving at a relatively high speed, a sufficiently clear Doppler frequency shift occurs in the signals reflected back by that object, so that these can be separated from the clutter signals by narrow-band Doppler filtering.

25 Due to the development of very fast aircraft that are able to fly very low, Doppler radar is now of increasing importance thanks to these characteristics. Such an aircraft could escape detection by conventional radar and MTI radar, because clutter signals reflected from the ground, vegetation or the surface of the sea can be strong enough to disrupt the detection of the signals reflected by the aircraft. These problems are significantly reduced by the use of Doppler radar.

1024654 3

The SQUIRE ground radar is a high-resolution radar that is used for this other purpose. The SQUIRE ground radar uses Doppler but is unable to apply polarimetry.

5 The same applies to all systems in use: none of them is suitable for simultaneous application of Doppler and high resolution polarimetry. That is why these systems do not detect small objects in an environment (for example, the sea) with much nap.

The present invention eliminates the aforementioned drawbacks by simultaneously using Doppler and high-resolution polarimetric data.

An object of this invention is a method for the detection of an object in an environment, from a range Doppler spectra obtained by processing received polarimetric data, comprising: - The selection (11) of a primary polarization channel Ci.

- Filtering (13) the range Doppler spectra of this primary channel so that only those values that correspond to the range Doppler bins in the primary polarization channel that have exceeded a threshold remain.

- Filtering (13) the range Doppler spectra of a secondary channel so that only the values that correspond to the range Doppler bins of the primary polarization channel remain that have exceeded a threshold.

The detection method may include a feature extraction function, in case an object to be detected is not clearly separated from a clutter phenomenon after filtering (13).

30

Another object of this invention is a classification method that utilizes the characteristics extracted by the detection method, including the probability processing of the extracted characteristics that include polarimetric and non-polarimetric characteristics belonging to an object category.

102 46 '·.

4

Further features and advantageous features of the invention will become apparent from the following description of examples of embodiments of the invention, with reference to the illustrations showing specific features of the invention, and from the claims. The individual properties can be realized separately, all or in any desired combination in one of the possible embodiments of the invention.

Figure 1, a function diagram of a polarimetric system 10 according to the invention.

The Doppler polarimetry is considered in a search radar for the maritime surface image. The task of this radar is to detect objects with a low RCS in coastal areas. The system produces 15 estimates of the scattering matrix per Doppler bin range. A number of physical quantities are discussed that can be exploited to. distinguish object echoes from sea clutter responses and give an impression of their effectiveness.

A radar beam is made in a number of directions and the reflected (polarimetric) energy from small objects and the environment is received and extracted. By utilizing Doppler polarimetric information, the object can be isolated from its environment, using a number of different features and techniques, so that it can be detected and classified.

The radar is designed to promote the simultaneous reception of two polarizations. Upon transmission, circular polarization can be effected by properly feeding the radiation probes 30 with signals shifted 90 ° in phase. There are also two reception modes: linear (H and V) and circular (L and R). The send polarization and the receiving mode can be programmed in advance and can be changed from sweep to sweep.

1024854 5

Range processing is performed as pre-processing. The starting sequences form range spectra. The range spectra undergo Doppler processing.

The Doppler spectra are subjected to a calibration correction. The purpose of the calibration correction is multiple. Cross-talk between receiver channels is compensated, inadequate tuning of the radiators and / or of the antennas is remedied, the transmission functions (phase and amplitude response) of the transmission paths and also those of the two receiver channels are rectified.

The output of the process includes calibration-corrected scattering matrix estimates per range Doppler cell for the range Doppler spectra in the so-called linear basis, i.e. for the polarization combinations HH, W, HV and VH. The first letter indicates the receive polarization, the second letter characterizes the send polarization.

Figure 1 shows a detection system 10 according to the invention, which receives this calibration-corrected scattering matrix estimates per range Doppler cell. The explanation of the detection below is given on the basis of the example of a polarimetric ship radar. The clutter-rich environment Is therefore a sea clutter environment.

In a first step, depending on the wave direction Dw in relation to the viewing direction of the radar Dr, the primary polarization selection means 11 select a primary polarization channel Ci. The other channel is defined as the secondary channel.

The scattering matrices can be viewed in different polarization bases (for example the following bases - rectangular: range Doppler spectra for HH, W, HV and VH; Pauli: range Doppler spectra for HH + W, HH-W, VH + HV and VH-HV ; circular: range Doppler spectra for LL, RR, RL and LR). In the following, the linear basis serves as an example.

10246 54 "6

The information pc, vc, aCl in the dutter map shows the current position of the radar relative to the sea waves. Using this information, the optimum polarization for peak detection can be determined and thus the optimum polarization is selected by the primary polarization selection means 11.

Depending on the wave direction, sea clutter will be more present in one of the two channels, resulting in more hits. The purpose of the further detection steps is to separate object and sea clutter hits by means of a characteristic or a combination of characteristics.

The object response is traced at a given range of Doppler spectrum. A two-step procedure is followed to identify object responses.

15

First, a CFAR (Controlled False Alarm Rate) set to the noise level is applied. The primary channel Ci is the channel that is passed through the CFAR filter 12. Cells with a signal power that exceeds a certain threshold 20 may be clutter responses. Therefore, the CFAR filter 12 receives the primary characteristics, namely a range Doppler matrix of the primary channel R1p and filters this with the CFAR threshold. Cfar · The primary characteristics used by the CFAR filter 12 are the power p, the distance r and the Doppler speed v.

25

To decide which spectrum is to be used as the threshold by the CFAR filter 12 and which of the feature group filters FS1, 2 and 3 is to be used for detection, a dutter map 14 is required. The following information is stored in the dutter folder 14: power pCl sea clutter radial speed spectrum vc, dutter covariance matrix oc.

An existing power distribution of the object is not assumed, since the RCS is not known nor where the object 35 is located and what its speed is. By using interpolation 1024654 7, a better estimation of distance and radial velocity of objects can be obtained. The interpolated radial tendency improves the detection of objects that are in the vicinity of a sea clutter with radial tendency.

5

The CFAR filter 12 is connected to hit selection means 13, such that only the values in the range Doppler bins that have exceeded the threshold are stored for both channels (the primary channel and the secondary polarization channel). These exceedances are called 10 hits h. In this way, dominant scattering is detected by comparing each of these cells with, for example, the eight adjacent range of Doppler cells. If the response in the cell is higher than that of the eight neighboring cells, it is considered to be a dominant scatterer and is considered an object response.

15

Depending on the object position with regard to distance pt and radial speed vt relative to the sea clutter position, the hits h are applied to one of the following three feature group filters FS1, FS2 and FS3, in order to extract appropriate features.

20

For each hit (either sea clutter or object response), a series of features is extracted from the Doppler polarimetric images (D-HH, DHV, D-VH, D-VV), which can help to distinguish between the two categories. The most significant features include the following eight: • Power p, with a specific polarization combination, for example W.

• Distance r, for this specific polarization combination.

• Width in the range domain σΓ, for this specific polarization combination.

30 · Distance difference Ar. The distance between the distances determined for the specific polarization combination and its opposite, for example W and HH.

• Radial speed difference Av. The difference between the radial rates for the specific polarization combination and the opposite thereof is determined, for example W and HH.

1024834 8 • Polarization ratio HVR. This is the power with this opposite polarization combination, divided by the power with this specific polarization combination, for example HH divided by VV.

5

The feature groups include the following features:

Primary characteristics are the power p, the distance r and the Doppler speed v. These primary characteristics are utilized by the CFAR filter 12. From the peak-fitting procedure, the interpolated Doppler speed is used to distinguish between the nap and object. Information from the dutter folder is used to set the Doppler speed limits between which this is useful.

A series of features can be extracted for each peak detected in the range Doppler domain. The extension (extension) of small objects in terms of distance and radial speed is limited. Furthermore, the maximum difference in distance and radial speed between the maximum different polarimetric channels is small. However, in relation to sea clutter, it is known that the response is caused by different physical phenomena, which are characterized by different radial speeds that do not necessarily occur at identical distances. This results in an extension of the individual retrenching in terms of distance and radial speed, and the peaks in the polarimetric channels do not coincide in terms of distance and speed. Two categories of features make use of this: features that represent the extension in terms of distance and radial speed, and features that indicate a difference in the location of maxima in different polarimetric channels. The secondary characteristic group contains a non-polarimetric group and a polarimetric group: 30 - Non-polarimetric secondary characteristics: σι and öz, spectrum widths.

• σι is given by_ | jb + 2 - + 2 -. m + 2 σι = i Σ - Σ * Ptoiai = Σα. rn is the Doppler bin 1 / »« - 2 Ptotal J i = «- 2 Ptotal i« m-2 1 02 4-8 B4 9 containing the maximum, pi is the power and vi is the radial speed corresponding to Doppler bin i.

• σ2 is given by σ2 = 1 Pmax =% p (, Pt0BÜ = jpt, m is the Ρ, οωΐ ‘° m-2

Doppier bin that contains the maximum, pi is the power and vi is the radial speed corresponding to Doppler bin i.

• Polarimetric secondary characteristics: Δν and Ar, spectrum widths.

• Difference in radial decay between the primary channel and the secondary channel, Δν (m / s): Δν = vp- vs..

10 · Difference in distance between the primary channel and the secondary channel, Ar (m): Ar = rp-rs.

Secondary property Av is used. When headwind and crosswind are measured, the polarimetric properties of sea clutter overlap the polarimetric properties of most artificial objects with regard to the properties used by P-CFAR (Polarimetric CFAR). Some other polarimetric and non-polarimetric properties can be used to increase the detection performance. Av can be combined with one of the attributes σι and σ2 or Ar to further increase the detection performance, depending on the object of interest

Scattering properties of each send and receive polarization can be derived from the scattering matrix S, which is represented as follows: Ssr® Sflv. in the indexes gives the first

L ^ vw J

letter the receive polarization and the second letter characterizes the send polarization. These scattering properties are displayed in 3D range Doppler signal strength images.

The algorithm based on the polarimetric properties within the scattering matrix that is used to distinguish between object and dutter is known as P-CFAR FS3.

This algorithm estimates the polarimetric properties of the dutter from the 1024654 measurements, without assuming specific a priori polarimetric properties for objects. For a correct calculation of the correlation matrix, it is necessary that enough sea clutter echoes are present.

The correlation of sea clutter can be described by a 3x3 correlation matrix Σα. Matrix Σ0 can be estimated from the scattering matrix of the clutter measurements, denoted by Sc, according to I, c = —-—--- where N is the number of measurements and μ «is the average of the available feature vectors of clutter measurements after 1 o application of the CFAR filter 12.

After determining Σο, a scalar is calculated for each peak according to Q = (S-mc) hΣ ^ (£ - // J. The scalar Q is used to calculate the probability ratio HVR for use in the P-CFAR filter FS3.

15

When measuring with the wind, the polarimetric properties of sea clutter can be clearly distinguished from artificial objects. These polarimetric properties are used by the P-CFAR filter FS3 to filter sea clutter from the objects of interest.

20

When the object is clearly separated from the dutter, these hits are immediately applied to plot extraction function 20. After the feature group filters, all other hits are also applied to plot extraction function 20.

* The plot extraction function 20 combines all necessary information (including the scattering matrix) from the detection process into a beta, the so-called plot, which is sent to the tracking function 30.

Plots including the complex scattering matrices are supplied to the tracking function 30. The tracking algorithm can be based on the Kalman filtering process. In the follower 30, the polarimetric information is used to optimize the tracking results. This procedure corresponds to that of the classification process.

In the classification function 40, the scattering matrices included in the voig paths are compared with an object database 41, which contains the polarimetric information of defined objects. The best match gives the most likely identification of the track followed.

Classification function 40 re-uses the polarimetric and non-polarimetric characteristics, similar to the detection procedure 10, which is actually a one-category classification between sea clutter and no sea clutter. Classification is based on probabilities. The probability, for example, of the radial velocity of an object if it is given that the object belongs to class C, is assumed as follows: p (y \ c) = ï ·· 1 exp- where e {v} c is the expected value from

4 {nacJ 2ac J

v is for class C and σ * is the variance of the speed for this particular class. C, Σς and PC must be known in advance for each class.

This information is stored in a classification database 41.

20

More generally, such detection and classification methods can be used to detect and / or classify any small object in an environment with a lot of naps (on the water: zodiacs, canoes, swimmers, yachts ...; in the air: birds, 25 planes, hang gliders, balloons ....).

These Doppler polarimetric detection and / or classification methods can be applied in many ways such as, for example, in terrain and airspace surveillance, detection of and protection against floating objects or persons, border or zone patrolling, coastal protection, search and rescue operations (via the air) or at sea ...), tracing .drug couriers and assistance with interception, etc.

1024654

Claims (11)

Method for the detection of an object in an environment, from a range of Doppler spectra obtained by processing received polarimetric data, characterized in that it comprises: Detection method according to the preceding claim, characterized in that it comprises a feature extraction function, in case an object to be detected is not clearly separated from a clutter phenomenon after filtering (13). Detection method according to the preceding claim, characterized in that the feature extraction function comprises processing (FS1) the power p, the distance r and the Doppler speed v from the filtered Doppler spectra of the two (primary and secondary) channels. Detection method according to claim 2 or 3, characterized in that the feature extraction function comprises processing (FS2) non-polarimetric and polarimetric features from the filtered range of Doppler spectra of both (primary and secondary) channels. Detection method according to one of claims 2 to 4, characterized in that the feature extraction function comprises a polarimetrically controlled false alarm filter function (FS3), which receives the filtered range of Doppler spectra from both (primary and secondary) channels and a 3x3 matrix that is a Doppler polarimetric image. 1024654 5. The selection (11) of a primary polarization channel Ci. - Filtering (13) the range Doppler spectra of this primary channel so that only the values that correspond to the range Doppler bins in the primary polarization channel that have exceeded a threshold remain. io - Filtering (13) the range Doppler spectra of a secondary channel so that only the values that correspond to the range Doppler bins of the primary polarization channel that have exceeded a threshold remain. Detection method according to one of claims 2 to 5, characterized in that the feature extraction processing function is selected from a number of processing functions (FS1, FS2, FS3), depending on the object to be detected and the environment. 5 Detection method according to one of claims 2 to 6, characterized in that it comprises: a hit detection function (12) which detects the range Doppler bins in the primary polarization channel that have exceeded a threshold between the channel selection (11) and the filtering . 10 Detection method according to one of claims 2 to 7, characterized in that the channel selection (11) and the hit detection (12) use dutter map information. A classification method that uses the features extracted by the detection method of any of claims 2 to 8, characterized in that it comprises: the probability processing of the extracted features that belong to an object category, polarimetric and non-polarimetric 20 features. The classification method according to the preceding claim, characterized in that the extracted probability feature uses a 3x3 matrix, which is a Doppler polarimetric image. 25 A classification method according to claim 8 or 9, characterized in that the object category information is stored within a classification database (41). 1 ü 2 4 6 '6 4