Doctoral report of first year

A Doctoral report is due at the end of the first and of the second academic year, before the final year examination and presentation of achievements to the PhD board. This form provides an index and brief description of the items that it should contain. Please leave blank if not relevant and do not change ordering and numbering of sections.

LAST NAME Filippini

NAME Francesca

CURRICULUM Radar and Remote Sensing

DOCTORAL CYCLE XXXII

Current year First

Supervisor Fabiola Colone

Co-supervisor

PhD Advisory board P. Lombardo, R.Seu, M.Scarpiniti

Double-degree

1. Compliance of activities with doctoral program form of current year

During this year, I attended the advanced courses "Project Management ‘Prince 2’ Foundation" and "Elaborazione delle Immagini Radar", the seminars on "Scrittura tecnico-scentifica", "Biologically Inspired Radars: Lessons from Nature", " Retrieval of soil moisture using SAR data and surface water fraction with radiometer data from the L-band SMAP satellite" and a seminar on the activities of the company Space Technology S.r.l. . Moreover, I attended the 9th International Summer School on Radar/SAR and the 1st IEEE AES Radar Summer School.

The research activities carried out during this year, as planned in the doctoral program form, concerned the development of advanced processing techniques for Passive Coherent Location (PCL) systems, aiming at improving their performance. Specifically, the exploitation of polarimetric and frequency diversities has been considered. The following activities have been carried out:

- A low cost processing scheme has been proposed and compared to existing techniques, thanks to an analysis against experimental data

- The proposed approach has been extended to a multi-frequency scenario and its performance has been studied

- The benefits of the joint exploitation of polarimetric diversity have been studied also when using DVB-T signals as illuminator of opportunity, in a collaboration with Fraunhofer FHR.

Sapienza PhD in ICT Doctoral program in Information and Communications Technologies at Sapienza Università di Roma, Rome, Italy

2. Courses and seminars During this year I attended two advanced courses:

Course Duration/Period CFU Exam

Project Management ‘Prince 2’ Foundation

19-21 June 2017 6 Yes

Elaborazione delle Immagini Radar 26 September 2016 - 21 December 2016

6 Yes

A final exam is planned for both these courses. While the first exam has been done, the exam for "Elaborazione delle Immagini Radar" will be taken during the next month. Also I attended the following seminars:

Seminar Duration/Period CFU Exam

Scrittura tecnico- scientifica (Prof. Emilio Matricciani)

January - February 2017 4 Yes

Biologically Inspired Radars: Lessons from Nature (Prof. Alessio Balleri)

February 17 2017 1/3 No

Retrieval of soil moisture using SAR data and surface water fraction with radiometer data from the L-band SMAP satellite (Dr. Seung-bum Kim)

April 4 2017 1/3 No

Space Technology S.r.l. (Ing. Paolo Tabacco)

July 11 2017 2/3 No

3. Other activities During the year I attended two summer schools. They both allowed me to improve my knowledge on radar and SAR topics thanks to distinguished lecturers and workshops held by experts in the field. Last May I attended the 1st IEEE AES Radar Summer School, which was held in Seattle (WA), USA and hosted by the organizers of IEEE Radar Conference 2017. Last July, I attended 9th International Summer School on Radar/SAR that was held in Bonn, Germany. The host of the International Summer School was the Fraunhofer Institute for High Frequency Physics and Radar Techniques FHR.

Other activities Duration/Period CFU

9th International Summer School on Radar/SAR July 14 - 21 2017 - Bonn, Germany 5+

1st IEEE AES Radar Summer School May 6 - 7 2017 - Seattle (WA), USA 3+

Moreover, I attended two international conferences (one of them will be held at the end of the current month) and two international workshops. They are listed below.

Conference / Workshop Duration/Period

IEEE Radar Conference 2017 May 8-12 2017 - Seattle (WA), USA

International Conference on Radar Systems 2017 October 23-26 2017 - Belfast, UK

6th PCL Focus Days May 22-24 2017 - Wachtberg, Germany

1st GTTI Radar and Remote Sensing Workshop 2017 May 24-26 2017 - Naples, Italy

4. Research activities The core of my research activity is the development of advanced techniques for the improvement of Passive Coherent Location (PCL) performance, both in terms of detection capability and reliability.

PCL technology has been receiving conspicuous interest in the last decades and has rapidly reached a point of maturity in many surveillance applications. However, several experimental programs conducted with such sensors demonstrated that, despite the effectiveness of the signal processing techniques, target detection and localization capability of a PCL system may be strongly limited by many effects that are not under the control of the PCL designer (e.g. interfering signals, time-varying characteristics of the employed waveforms of opportunity). Hence, the relevant technical literature continuously reports new attempts to identify advanced solutions, which aim at increasing the reliability of such systems against the aforementioned undesired effects.

Recently, the possibility of combining the returns from differently polarized antennas has been considered as a way to improve the target detection capability in PCL systems and different signal processing schemes have been proposed.

By resorting to a Generalized Likelihood Ratio Test (GLRT), a locally adaptive polarimetric detection scheme was derived and it was shown to be able to exploit the polarimetric differences between the target and the competing disturbance (e.g. cancellation residuals or interfering signals). The experimental results obtained against FM radio based PR experimental data proved that, thanks to its capability to locally adapt the filter weights in the bistatic range-Doppler domain, the conceived P-GLRT scheme allows to mitigate the effects of the interferences thus significantly improving the target detection performance of the system.

As an example, in Fig. 1 (a-b) we compare the raw detection results (in green) provided by the employed PCL receiver across 50 consecutive acquisitions, after the application of a simple non coherent integration of the polarimetric contributions (P-NCI) (a) and the P-GLRT (b) detection schemes, respectively. The considered experimental receiver features two dual-polarized antennas, each one equipped with two independent, one vertical (V) one horizontal (H) polarized, outputs. The available air-truth data are sketched in black colour on the same figure. As is apparent, the P-GLRT is able to increase the target detection capability of the system while controlling the amount of false alarms. This analysis had been included in my master thesis.

During the last year, we have further extended the results obtained according to different strategies:

Time -domain adaptive polarimetric scheme

We have investigated possible methods to limit the computational load required by a locally adaptive approach as the P-GLRT, by looking for alternative approaches able to provide a reasonable trade-off between achievable performance and computational complexity. To this purpose, we considered the possibility of globally adapting the weights of the polarimetric whitening filter by estimating the disturbance covariance matrix in the time domain. This requires the estimation and inversion of a single covariance matrix at each coherent processing interval with a dramatic reduction of the computational burden. By observing Fig. 1, we can easily verify that:

(i) the P-NCI detection scheme (Fig. 1a) offers the worst capability of controlling the false alarm rate, confirming the weakness of the assumption of statistically independent disturbance affecting the different polarimetric channels;

(ii) the P-GLRT (Fig. 1 b) and the newly proposed approach (Fig. 1c) approaches yield quite comparable results

Joint exploitation of frequency and polarimetric diversity

Based on the promising results obtained with the globally adaptive polarimetric scheme, we investigate the possibility of extending the proposed approach to a multi frequency scenario. Assuming the availability of multiple signals simultaneously collected at different carrier frequencies by at least two surveillance antennas with different (orthogonal) polarizations, a multi-frequency polarimetric detection scheme was derived by extending the time-domain polarimetric adaptive approach.

The available experimental data allowed us the analysis of the performance offered by the proposed scheme, referred as MF&TD-APF when N=4 frequency channels and L=2 surveillance antennas with orthogonal polarizations (H and V) are available. Furthermore, we compared its performance with the MF&P-GLRT, which had been proposed previously by resorting to a GLRT approach.

In Fig. 2(a-b) the false alarm rate control curves and the ROC curves of the MF&TD-APF approach are sketched (in red) and compared with the MF&Pol-GLRT (in dark blue) and the MF&Pol-NCI (in black). In addition, the Pol-GLRT at the best and worst performing single frequency (SF) channels (in dotted light blue) and the best and the worst SF single-pol solutions (in dotted green) are included in the comparison so as to understand the improvement.

By observing Fig. 2, few considerations are in order:

(i) The MF&Pol-NCI detection scheme yields the worst false alarm control capability with respect to all considered approaches and it results in the scarcest target detection capability with respect to the MF solutions that include a polarimetric whitening stage. The rate of detection is even decreased with respect to the best performing SF Pol-GLRT, confirming the effectiveness of the adaptive cancellation stage to counteract the disturbance contributions.

(ii) The MF&TD-APF and the MF&MP GLRT detection schemes yield all-around almost identical

performance, both in terms of false alarm control capability and target detection capability.

However, we notice that both approaches result in a slightly worsen false alarm control

capability of the system.

(iii) The improvement in terms of target detection performance that both MF approaches yield is

substantial with respect to the best performing SF Pol-GLRT and tremendous with respect to

the worst performing SF Pol-GLRT.

(iv) The joint exploitation of polarimetric and frequency diversity is shown to be the preferable

solution with respect to a SF and single-pol approach. In fact, the improvement in terms of

detection capability is terrific with respect to the vertically polarized surveillance signal at

FMCh1.

a b c

Fig. 1 Detection results for 50 consecutive acquisitions using

(a) P-NCI; (b) P-GLRT; (c) TD-APF

In conclusion, the analysis conducted demonstrated that the multi-frequency polarimetric operation provides substantial mitigation of the interference effect on the detection stage and increased robustness with respect to the time-varying characteristics of the exploited waveforms. Moreover, it allows to better recover the frequency diverse and randomly polarized target echo by integrating the returns at different polarimetric and frequency channels.

Experimental Results of Polarimetric Detection Schemes for DVB-T based Passive Radar

Based on the successful results obtained in the FM radio band, there was a strong motivation in understanding the potential benefits of the use of polarization diversity also when exploiting different sources of opportunity. A collaboration between our research group and Fraunhofer FHR, started during my master thesis yielded some preliminary results of a PCL system exploiting DVB-T signals as Illuminator of Opportunity.

In the frame of the same collaboration, we could carry out a more extensive analysis on experimental data collected by the experimental PCL system PARASOL, developed at Fraunhofer FHR and equipped with two linearly orthogonally polarized surveillance antennas. Both aerial and maritime cooperative targets were employed in the experimental tests and enabled a reliable comparative analysis of the achievable detection performance.

Specifically, the acquisition campaign has been carried out in Eckernförde, Germany (see Fig. 3a). The exploited IO is located in Kiel (approximately 20 km away from the receiver site) and broadcasts DVB-T signals in a single frequency network. The system PARASOL was mounted on a moving platform, namely a military boat approx. 15 meters long (see the black trajectory in Fig. 3a). A total amount of 168 consecutive data files (i.e. a total acquisition duration of approx. 95 seconds) have been extracted from the entire acquisition. Three cooperative targets, equipped with GPS receivers, were available during the trials.

a b

Fig. 2 Performance analysis of different polarimetric detection schemes: Pol-GLRT at worst and best performing

frequency channels, MF&Pol-NCI, MF&Pol-GLRT and MF TD-APF

a Actual false alarm rate vs nominal false alarm rate curves

b Empirical ROC curves

a b c

Fig. 3 Acquisition geometry and cooperative targets

(a) sketch of acquisition geometry and targets trajectories (b) Delphin aircraft (c) One of the two identical

speedboats

Specifically, one ultra-light aircraft from Fraunhofer FHR (Delphin, see Fig. 3b), and two identical 7 m long rubber boats (see Fig. 3c). Their true trajectories are reported in Fig. 3a in blue, red and green, respectively.

In Fig. 4 we report the sequence of detected plots (black dots) provided by different detection schemes. Specifically, we compared the corresponding velocity measurements to the bistatic velocity tracks of the cooperative targets, as a function of time. In Fig. 4, the truth-data of the aircraft Delphin and of the two speedboats are sketched as blue, red, and green continuous lines, respectively. First, by observing Fig. 4(a-b), we can observe that the use of a single-pol operation leads to quite different results depending on the employed polarization. For this particular data set, we can say that, on average, the single-pol H yields better results in terms of number of correct detections against all the considered targets. Nevertheless, this trend is not stable and it may change with time, namely since targets echoes typically show random polarizations, the single-pol V occasionally provides correct detections at scans where the single-pol H experiences a missed detection. Thus, it is not possible to establish a priori the best performing polarimetric channel on average.

Then, the two available polarimetric channels are jointly processed according to different polarimetric detection schemes. The results obtained are reported in Fig. 4 (c-d), where the P-NCI and the P-GLRT detection schemes are compared. First, we can observe that the P-NCI still experiences some missed detections at scans where the best performing single-pol channel provides correct detections. Moreover, thanks to its capability to reject adaptively the disturbance based on the polarimetric information, the P-GLRT detection scheme (see Fig. 4d) offers the best performance with respect to any other approach.

a b

c d

Fig. 4 Extracted sequence of correct detections reported in the velocity-time domain for 168 consecutive

scans and nominal 𝑷𝑭𝑨 = 𝟏𝟎−𝟔 with:

(a) single-pol H

(b) single-pol V

(c) P-NCI

(d) P-GLRT

Ongoing work and future activities

The activity that will be carried out during the second year will pursue the attempt of increasing the performance of a PCL systems and to enable some advanced features. Specifically, it will be focused on two different topics:

1. Threshold region performance of multi carrier ML DOA estimator The capability of estimating the direction of arrival of a target is a very important feature in radar applications. Assuming the availability of a sensor array, collecting returns on multiple carriers simultaneously, a theoretical study of the performance of a maximum likelihood estimator will be carried out. In particular, we are interested in describing its performance in a low SNR scenario, namely in the threshold region. This theoretical study will guide us in the optimization of the system parameters in order to achieve the best performance possible.

2. Long CPI : Range and Doppler migration compensation With the aim of increasing the detection of low RCS targets, or in order to broaden the coverage area, long coherent processing interval (CPI) could be considered. Typically, this value is limited in order to avoid both range and Doppler walk. In order to properly correct the range and Doppler migration effects in PCL, some techniques will be studied and some new approaches will be proposed. This activity will be carried out along with Leonardo Finmeccanica S.p.A and the effectiveness of the considered approached will be demonstrated against experimental data.

5. Software

The research activities have been the developed in Matlab witch is a proprietary product of MathWorks.

6. Periods abroad 7. List and description of applications/patents 8. Prizes and awards

During the 1st GTTI Radar and Remote Sensing Workshop 2017, I won the prize for the Best Oral Presentation from Under 35.

9. List of publications International Journals

F.Filippini, F.Colone, D.Cristallini, G.Bournaka: “Experimental results of polarimetric detection schemes for DVB-T based passive radar”, IET Radar, Sonar & Navigation, vol. 11, n. 6, (2017), pp. 883 – 891

Conference Proceedings

F.Filippini, F.Colone, D.Cristallini, G.Bournaka: “Preliminary experimental results of polarimetric detection schemes for DVB-T based passive radar” Proceedings IEEE Radar Conference 2017, Seattle, WA, USA, May 2017, pp. 383-388

F.Filippini, F.Colone: “A practical approach to polarimetric adaptive target detection in Passive Radar”, International Conference on Radar Systems 2017, Belfast (UK), October 2017

10. Specific list of publications that include coauthors of foreign institutions International Journals

F.Filippini, F.Colone, D.Cristallini, G.Bournaka: “Experimental results of polarimetric detection schemes for DVB-T based passive radar” IET Radar, Sonar & Navigation, vol. 11, n. 6, (2017), pp. 883 – 891

Conference Proceedings

F.Filippini, F.Colone, D.Cristallini, G.Bournaka: “Preliminary experimental results of polarimetric detection schemes for DVB-T based passive radar” Proceedings IEEE Radar Conference 2017, Seattle, WA, USA, May 2017, pp. 383-388

11. Double-degree section

Supervisor signature for approval

Signature of Doctoral student

Date 17-10-2017