Short-Range UWB Radar Application: Problem of Mutual Shadowing between Targets

Transcription

1 VOL., NO., DECEMBER Short-Range UWB Radar Application: Problem of Mutual Shadowing between Targets Dušan Kocur, Jana Rovňáková, Daniel Urdzík Department of Electronics and Multimedia Communications, Technical University of Košice, Slovak Republic {dusan.kocur, jana.rovnakova, Abstract Ultra wideband (UWB) radar system allows detect and track multiple moving targets in critical situations and environments. However, the results of application of single radar have shown the reduced ability to detect the other targets if some target is located between the radar antennas and the rest of targets. In this paper, we will outline the origin of this effect as the consequence of the mutual shadowing between people. This explanation will be verified by the experimental analyses based on processing of measured radar signals. Introduction The short-range localization capability is becoming one of the most attractive features of UWB radar systems. UWB radars, as promising technologies for short-range detection and tracking of multiple moving persons, have a variety of potential applications including through wall and through fire detection and tracking of moving targets during security operations, protection of reservoirs, power plants and other critical infrastructures against to a terrorist attack, detection of trapped people after an avalanche or earthquake, etc []. In the case of detection and short range tracking of moving persons for rescue and military purposes, the UWB radars have to be deployed immediately. For these purposes the UWB radar devices should be built in small sizes with an antenna array usually consisting of one transmitting and two receiving antennas. The UWB radar system of that kind is usually referred to as handheld UWB radar system. The problem of short-range detection and tracking of moving persons have been studied e.g. in []-[]. However, the problem of multiple human tracking in real complex environment has been less well addressed. Our experiences with the short-range UWB radars received at a number of measurement campaigns dedicated to detection and tracking of multiple moving persons (e.g. [7]) have shown that the target located nearby the radar antennas is very often visible only. The other targets can be also detected but with less reliability than that of the target located the most closely to the radar antennas. This frequently arising effect results in essential reducing of radar performance efficiency. In this paper, we will show that this effect is due to mutual shadowing of moving targets within multiple target tracking. We will show also that the origin of this effect consists in the frequency band employed by the UWB radar, dielectric and magnetic properties of a human body and clothes and mutual positions of the targets and radar antennas. In order to fulfill this intention, the presented paper will have the following structure. In the next section, the reduced ability of the UWB radar equipped with transmitting and receiving antennas to track two moving persons will be illustrated. Then, the explanation of this UWB performance based on shadowing effect is presented in Section 3. The correctness of the analyses presented in Section 3 is verified in Section by means of experimental measurements. Obtained results are discussed and evaluated in Section 5. Finally, some proposals for the solution of the problem of multiple moving target tracking are presented in the last section. Problem Statement In order to illustrate the fundamental problem of multiple moving target tracking, let us consider the following scenario of two moving persons tracked by an M-sequence UWB radar (Scenario ). The measurements took place in the school dining room depicted in Fig.. As the moving targets to be detected and tracked, two persons have been selected. The first person was walking from the position P through the all remaining positions up to the position P5 (Fig. ). At Figure : The setup of the M-sequence UWB radar system in the school dining room: experimental radar device, measured area. 37

2 VOL., NO., DECEMBER m P3 P P5 P P P P7 P P9 P P5 P P P P3 m.. Rx Tx Rx m Figure : Scenario : The simplified scheme of the measurement with the radar antennas, all reference positions and outlined target trajectories. the same time, the second person was walking the opposite trajectory from the position P5 up to the position P (Fig. ). The raw radar signals are represented by the impulse responses of the environment through which the signal emitted by the radar is transmitted from transmitting to receiving antennas. The raw radar signals were acquired by the UWB pseudo-noise radar using maximum-length binary sequence (M-sequence) as the transmitting signal [] (Fig. ). The system clock frequency of this M-sequence UWB radar is.5 GHz, which results in the operational bandwidth from DC to.5 GHz. The M-sequence order emitted by radar is 9, i.e. the impulse response covers 5 samples regularly spread over ns. This corresponds to an observation window of ns leading to an unambiguous range approximately m. The instantaneous dynamic range (IDR) of radar is about db [9]. The used radar device was equipped with one transmitting (Tx) and two receiving (Rx, Rx) horn antennas. The positions of the particular antennas are outlined in the scheme of the measurement (Fig. ). All antennas have been set on the table in the height of 75 cm from the floor. The number of antennas, their mutual positions and the distance among them correspond approximately to an application of a handheld UWB radar system. In order to estimate the moving person trajectories, the complex radar signal processing procedure for through wall multiple target tracking proposed in [] was applied. It consists of seven phases, namely background subtraction, weak signal enhancement, detection, time of arrival (TOA) estimation, wall effect compensation, localization and target tracking. Because the measurement was arranged without any obstacle located between radar antennas and targets, the wall effect compensation phase was omitted. In Fig. 3, two processing results obtained for Scenario are depicted. The former one given in Fig. 3 represents the detector output for the first receiving channel (receiving antenna Rx). In this figure, the vertical and horizontal axes are related to the propagation and observation time of the radar signals. The detector output consists of binary data, where the value (white points in Fig. 3) represents the reflection from moving target. These samples of the detector outputs create so called primary target traces. In Fig. 3, two traces corresponding to reflections from both targets walking inside the room can be clearly identified only in the bottom part of the detector output. After the propagation time instant of approximately ns they are both very weak, what results in incomplete estimation of the target tracks (Fig. 3). It can be clearly identified from this figure that the target moving close to the radar antennas is still visible very well. On the other hand, the latter target moving far from the radar antennas is almost absolutely obscured. Figure 3: The processing results of Scenario : detector output from Rx, estimated tracks. 3

3 VOL., NO., DECEMBER Rx Dead zone T-A Tx T-B Dead zone T-B Rx T-A Figure : Simplified illustration of the mutual shadowing effect and the dead zone creation: target in front of Tx, target in front of Rx. This effect, i.e. the reduced ability of the radar system to detect the other targets if some target is located between the radar antennas and the other targets, has been identified not only for Scenario, but also for a number of a wide spectrum of different scenarios intent on short-range tracking of multiple moving targets. Because this effect arises very often, it results in considerably limited efficiency and reliability of a handheld UWB radar system application for short-range tracking of multiple moving targets. In the next section, we will outline the origin of this effect firstly. Then, we will sketch a possible solution of this problem. 3 Shadowing Effect According our observations, the phenomenon described in the previous section is caused by mutual shadowing of targets within multiple person tracking scenario. With regard to the frequency band employed by a standard UWB radar applied for moving persons tracking, and permittivity, permeability and conductivity of a human body [] and clothes, a person reflects and absorbs the energy of electromagnetic waves emitted by the radar transmitting antenna or reflected by another object in such a way as that only a negligible part of electromagnetic wave energy is transmitted through/around a human body to a region located behind his/her. This effect referred to as shadowing results in creating of limited region behind the person where the electromagnetic waves emitted by the radar transmitting antenna or reflected by another object are not able to penetrate. This region is referred to as a shadow region or a dead zone. Let us consider very simplified model of the dead zone creation, where the electromagnetic waves are transmitted by an isotropic antenna. Let us assume also that principles of geometrical optics can be applied for the description of the electromagnetic wave propagation. Then, if the person is substitute by a simple ellipse in D space, the shape of the dead zone in D is given in Fig.. There are two basic forms of shadowing effect manifestation at multiple moving target tracking scenario. Firstly, Tx let us consider scenario, when a target A (T-A) is located in front of transmitting antenna. It results in the creation of the dead zone behind T-A (Fig. ). Then, if a target B (T-B) is located in this dead zone, it cannot reflect the electromagnetic wave and hence it is not able to detect it. The second form of the shadowing effect manifestation is presented in Fig.. At this scenario, T-A reflects electromagnetic wave transmitted by the transmitting antenna Rx. However, between T-A and the receiving antennas, T- B is located. It creates also the dead zone. Because the receiving antenna is located in this dead zone, T-A cannot be detected. Experimental Results In order to confirm the idea presented in the previous section, we have accomplished the set of the four measurements hereinafter referred to as Scenario, Scenario -Tx, Scenario -Rx, and Scenario -Rx. The scheme of the measurements is shown in Fig. 5. The person creating the dead zone was staying in a motionless way at the position m before the Tx, Rx, and Rx for Scenario -Tx, Scenario -Rx, and Scenario - Rx, respectively. In Scenario, the shadowing person was not considered, i.e. it is a single target scenario without dead zones. In the case of all scenarios, the moving person was walking along the trajectory of the S shape from the position P, through the all positions up to the position P9 (Fig. 5). The same radar device, antennas setting and radar signal processing procedure as for Scenario have been used for all scenarios considered in this section. Theoretical expectations of dead zones for Scenario - Tx, Scenario -Rx and Scenario -Rx are depicted in Fig.. For creation of these figures, the same simplified assumptions like in Section 3 (an isotropic antenna, principles of geometrical optics, ellipse shape of target) were applied. The first form of mutual shadowing (illustrated in Fig. ) is represented by grey-colored dead zone in Fig.. The second form of shadowing (Fig. ) is given by blue-colored dead zone for receiving antenna Rx and by green-colored dead zone for receiving antenna Rx. From Fig. is obvious that the dead zones created in such a manner depend a lot at the width of the shadowing person and her/his position in regard to location of radar antennas. Theoretically, not considering other sources of errors, the result of moving target tracking for analyzed scenarios should be those parts of trajectories from Fig. which are out of the dead zones. In the case of Scenario -Tx (Fig. ), it coresponds with the beginning of S trajectory and in Scenario -Rx (Fig. ) and Scenario -Rx (Fig. (c)) with the left and right side of S trajectory, respectively. Selected experimental results obtained by the processing of measured radar signals are given in Fig. 7. First two figures for every scenario have form of the detector outputs, the last picture represents the final track estimated 39

4 VOL., NO., DECEMBER P7 P P9 P P5 P m P P P3 m.. Rx Tx Rx m Figure 5: Scenario : The simplified scheme of the measurement with the radar antennas, all reference positions and outlined target trajectory m Rx m Tx Rx Shadowing person m for moving target. From the detector outputs in Fig. 7 - Fig. 7 it can be observed that the radargrams from Rx and Rx are quite similar. The only difference consists in a small time shift along the propagation time axis due to a small difference of the distance between the receiving antennas. As we can see from these figures, the trace corresponding to reflections from the person walking along the S trajectory can be clearly identified. It starts at the propagation time instant of ns and ends approximately at 5ns. The target locations, presented in Fig. 7(c) by thin red plus signs, were obtained by the detector output processing using the signal processing phase such as TOA estimation and localization mentioned in Section. The final target track, depicted in Fig. 7(c) by thick red circles, was estimated based on tracking methods. They are responsible for the partial trajectory completion and smoothing. It can be seen from this figure that the target track estimated for Scenario corresponds well with the true target trajectory. figure The results obtained for the Scenario -Tx are depicted in Fig. 7(d)-7(f). As can be observed from the first two pictures, both antennas received strong reflections from the target only when it was moving in the vicinity of the radar antenna system. The reflections creating the rest of the primary trace of moving target have been weak or none. It has resulted in incomplete estimation of target track (Fig. 7(f)). The standing target was detected by both antennas, too. Its primary trace, in the form of a horizontal stripe, is visible in the bottom part of radargrams around propagation time instant of ns (Fig. 7(d)-7(e)). The target location was estimated correctly (m in front of Tx), but was mixed with the track of moving person (Fig. 7(f)). Theoretical expectation from Fig. coincides with the first part of track estimated in Fig. 7(f). Additional target m Rx m Rx m Tx Rx m Tx Rx (c) Shadowing person m Shadowing person m Figure : Theoretical expectation of dead zones for: Scenario -Tx, Scenario -Rx, (c) Scenario -Rx.

5 VOL., NO., DECEMBER (c) (d) (e) (f) (g) (h) (i) (j) (k) (l) Figure 7: Experimental results for Scenario : detector output from Rx, detector output from Rx, (c) estimated target track; Scenario -Tx: (d) detector output from Rx, (e) detector output from Rx, (f) estimated target track; Scenario -Rx: (g) detector output from Rx, (h) detector output from Rx, (i) estimated target track; Scenario -Rx: (j) detector output from Rx, (k) detector output from Rx, (l) estimated target track.

6 VOL., NO., DECEMBER locations calculated between positions P-P7 indicate that some target reflections was able to penetrate also to dead zone of Rx (blue shadowed region in Fig. ). The Scenario-Rx is analyzed in Fig. 7(g)-7(i). In this case, the reflections received by the antenna Rx, are remarkably shadowed by the standing person (Fig. 7(g)). On the other hand, the antenna Rx has seen almost whole target trace (Fig. 7(h)). As the target locations are calculated based on combination of data from both receiving antennas, the resulted track is again very bad estimated (Fig. 7(i)). This result, except for few locations near position P, corresponds with the theoretical analysis from Fig.. The results obtained for the last scenario (Scenario-Rx) are demonstrated in Fig. 7(j)-7(l). Here the output from the antenna Rx was surprising because Fig. 7(j) should be similar to Fig. 7(h). Instead of it, the primary target trace was even stronger as in the basic scenario without shadowing person (Fig. 7). Fig. 7(k) is matched with expectations. Almost whole trace was shadowed what resulted in estimation only of a torso of the true target track (Fig. 7(l)). Theoretical results from Fig. (c) completely coincides with the track estimated in Fig. 7(l). 5 Result Evaluation If we summarize the presented theoretical and experimental results and their analyses, it can be concluded that the obtained results confirm our explanation of the effect of the reduced ability to detect some targets within the multiple target tracking scenario by a single short-range handheld radar. The simple experiment with one moving and one shadowing person clearly showed the negative impact of mutual shadowing between targets. The experimental results deviate slightly from the theoretical analyses of the problem. In our opinion, this difference can be explained by the fact that the application of geometrical optics method is very rough for the problem under consideration. At the same time, we did not consider multipath electromagnetic wave propagation (especially in the vicinity of the targets). And finally, the human body cannot be considered to be a compact and homogeneous solid figure. His or her motion consists also of several different kinds of movement (e.g. movement legs, hands, head, etc.). The motion of the different parts of the human body can produce a number of different echos. Because the aim of the presented analyses has not been to provide an exact solution of the outline problem but to point out its origin, and taking into account the possible reasons of not perfectly matching of experimental results with the theoretical explanation of the problem, we believe the results of our preliminary qualitative analyses can be accepted. More exact representation and evaluation of the outlined problem of mutual shadowing between targets is the aim of our current research. At the present time we are evaluating experiments specialized on quantitave analysis of the shadowing effect. The measurements have been oriented mainly on enumerating of the values of additional attenuation within the shadow zone due to shadowing person, finding of dead zone shape and its borders. Conclusion In the case of multiple moving persons tracking, the mutual positions of the targets and radar antennas are changing and hence, the number, shapes and locations of the dead zones are time-variable. The creation of the dead zones results in the reduced ability to detect and localize multiple moving targets by the UWB handheld radar system. Because the person motion velocity is smaller than the radar measurement rate, the missing part of the persons tracks due to the dead zones cannot be usually substitute by their extrapolation within radar signal processing procedure. A person moving nearby the radar antennas can produce the effect of partial or total shadowing of remaining targets. In the case of partial shadowing, target detection can be improved by the application of methods of weak non-stationary signal component enhancement present at the radar response []. If the effect of total shadowing occurs, the person moving nearby the radar antennas creates a time variable dead zone in a monitored region. If a target is located within this zone, it is impossible to detect it by the same radar. Following the presented analyses, we have found out that the application of a UWB sensor network can provide the efficient solution for short range tracking of moving persons. The considered sensor network should consist of several simple UWB radar systems (sensors) connected by a communication network. The particular sensors have to be located in the monitored region in such a way as to cover the region of interest as good as possible. At the same it is expected that they are able to provide some diversity with regard to targets positioning. Each sensor is able to detect some persons from the total number of targets and at same time it is able to estimate their coordinates. The target coordinates provided by all sensors are transmitted to each sensor or a central unit and subsequently they are processed by data association and fusion methods. The detail solution of this problem is over this paper. However, our preliminary results in this research area show that the sensor network consisting though only from two nodes is able to suppress the negative impact of mutual shadowing between targets. Acknowledgement This work was supported by the Slovak Research and Development Agency APVV under contract No. LPP-- 9. References [] P. Withington, H. FLUHLER, S. NAG, Enhancing homeland security with advanced UWB sensors, IEEE

7 VOL., NO., DECEMBER Microwave Magazine, Sept. 3, vol., no. 3,p [] S. H. Chang, et al, People tracking with UWB radar using a multiple-hypothesis tracking of clusters (MHTC) method, in Springer: Int J Soc Robot., Jan., p. 3. [3] S. Z. Gurbuz, W. L. Melvin, D. B. Williams, Comparison of radarbased human detection techniques, in Proc. of Asilomar Conf. on Signals, Systems and Computers, 7, p [] S. S. Gauthier, W. CHamma, Surveillance through concrete walls, in Proc. of SPIE C3I Technologies for Homeland Security and Homeland Defense III, Sept., p [5] S. Nag, et al, Ultrawideband through-wall radar for detecting the motion of people in real time, in Proc. of SPIE-Radar Sensor Technology and Data Visualization, July, vol. 7, p. 57. [] S. S. Ram, H. Ling, Through-wall tracking of human movers using joint doppler and array processing, IEEE Geoscience and Remote Sensing Letters, July, vol. 5, no. 3, p [7] J. Rovňaková, Complete signal processing for through wall tracking of moving targets. Germany: LAP LAM- BERT Academic Publishing, Sept.. [] D. Daniels, Ground penetrating radar. nd ed., London, U.K.: The Institution of Electrical Engineers, July, ch. M-sequence radar. [9] J. Sachs, R. Herrmann, M. Kmec, and P. Peyerl, Modified M-sequence UWB-radar, in German Microwave Conference (GeMiC), Karlsruhe, Germany, March. [] J. Rovňaková, and D.Kocur, UWB radar signal processing for through wall tracking of multiple moving targets, in Proc. of EuRAD, Paris, France, Oct., p [] D. Andreuccetti, R. Fossi, and C. Petrucci, Calculation of the dielectric properties of human body tissues in the frequency range Hz- GHz, Italian National Research Council, Institute for Applied Physics, Florence, Italy, [Online]. Available: [] J. Rovňaková, and D.Kocur, Weak signal enhancement in radar signal processing, in Proc. of Radioelektronika, Brno, Czech Republic, Apr., p