IMAGE FORMATION THROUGH WALLS USING A DISTRIBUTED RADAR SENSOR NETWORK CIS Industrial Associates Meeting 12 May, 2004
THROUGH THE WALL SURVEILLANCE IS AN IMPORTANT PROBLEM Domestic law enforcement and military missions beginning to merge - active defense domestically - peacekeeping internationally - rules of engagement similar Better tactical surveillance reduces chance for violence - search, hostage, and barricade incidents - better options for intervention Operational effectiveness is improved if both interior features and locations of individuals are available - implies need for imaging system The nature of currently available technology and its implementation place operational constraints on its effectiveness There is a need for cost effective, reliable solutions for these missions
STANDOFF IMAGING Widely used in commercial and military applications - medicine, earth science, oceanography, surveillance Preferred technique - single frequency source - narrow beam - move the probing radiation source: wide aperture improves resolution - take data from multiple orientations For lower frequency sources with wide beams - synthesize narrower beam through movement: SAR, side scan SONAR for example
SENSOR SELECTION IS DRIVEN BY WALL PROPERTIES A radar sensor operating between 500 MHz and 2 GHz represents a good compromise between technical and practical performance Increase of attenuation with frequency makes higher resolution imaging techniques unsuitable for through the wall surveillance
LOW OPERATING FREQUENCY REQUIRES APERTURE SYNTHESIS Moving a single antenna is impractical for most through wall situations A multiple antenna array improves cross range resolution significantly - scanning the array further improves resolution by a factor of two
IMAGE FORMATION REQUIRES RESOLUTION IN TWO DIMENSIONS Range resolution is a function of bandwidth - R = c/2 BW - for 1500 MHz BW, range resolution is 0.1 m Cross range resolution varies with distance Fine cross range resolution at long distances requires large antennas and/or higher frequencies of operation - seeing that two people are standing next to one another matters 100 500 MHz 100 1000 MHz Array Aperture - meters 80 60 40 20 0.2 m 0.5 m 1 m 2 m 5 m 10 m Array Aperture - meters 80 60 40 20 0.1 m 0.2 m 0.5 m 1 m 2 m 5 m 10 m 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Detection Range - meters Detection Range - meters
ELECTROMAGNETIC WAVE PROPAGATION THROUGH WALLS Forming images through walls is complicated by the details of how electromagnetic energy propagates in the walls
IMAGE RECONSTRUCTION Stepped and swept frequency radars can resolve signals from different ranges All objects at a constant range, though, add to the radar return - to resolve more than one object at a constant range requires changing illumination - independent observations from different angles are needed to solve the equations - each bistatic antenna pair of the scanning array produces an independent observation
IMAGE AMBIGUITIES False objects can appear at a given range because of multiple reflections Observations from different antenna pairs help reduce the effect of these multiple bounce returns Continuous scanning helps increase signal to noise ratio and improves image formation
IMAGE RECONSTRUCTION USING FFTS The FFT of data from an antenna pair gives the signal return as a function of bistatic range Signals from each antenna pair are summed for each bistatic range - corrections can be made for signal attenuation as function of range - corrections for antenna pattern effects more difficult
IMAGE RECONSTRUCTION USING BACKPROPAGATION For each pixel of the image, the received data is phase adjusted - done a single frequency at a time - effects of range attenuation and antenna pattern are easily added Subimages at each frequency are created for each transmit/receive pair - the final image is created from summation of the subimages More computation intensive but better control over system parameters
BRASSBOARD IMAGING RADAR SYSTEM Antenna Array Antenna Switchbox Radar Portable system - 20 lb., 2.2 m long, 4 antenna, collapsible array - battery powered
IMAGE FORMATION Antennas used in bistatic configuration, each pair scanned in turn Frequency data from each pair Fourier transformed to create range profile Range to each pixel in image map calculated for each antenna pair Sub image for each pair formed using complex FFT values and range Sub images added and magnitude taken The scattered returns from objects in the field of view add coherently In phase and quadrature components at other locations in field of view average toward zero
TEST CONFIGURATIONS - THROUGH THE WALL
TEST RESULTS Wall at 4.57 m range, 0-2.44 m cross range Array at 0 m range, 0-2.21 m cross range Person at 10 m range, 2 m cross range
TEST RESULTS Wall at 4.57 m range, 0-2.44 m cross range Second wall at 7.19 m range Person at 10 m range
TEST RESULTS Two walls Person at 10 m range, 2 m cross range Second person at 5.88 m range,.6 m cross range
INTERIOR WALL RESULTS - LONG ARRAY Experimental data shows the effectiveness of increasing the aperture of the sensor system studs wall configuration All current through the wall sensors used fixed antennas However, the imaging algorithms do not require use of a fixed array Interior walls are relatively easy to see through buried concrete block tree three parallel walls Note : 0.05 m wide stud appears to be 0.2 m wide at a distance of 4.6 m from radar, illustrating effect of cross range resolution on image quality
REINFORCED CONCRETE WALL RESULTS Attenuation and dispersion complicate image formation Without correction for dispersion, images become defocused Only frequency domain sensing methods allow for frequency dependent dispersion correction person 2 12 exterior poured, steel reinforced concrete wall radar on inside of room pointing out person 1
REINFORCED CONCRETE WALL RESULTS corner reflections UCSB lecture hall Strong corner reflections from external wall supports seen in background image person 1 person 2 metal trash cans Background subtraction shows image of three trash cans in triangle and two people
MOTION DETECTION USES COHERENT SCENE SUBTRACTION Motion sensing works even in more difficult, cluttered environments Composite image is from motion detection experiment Trailer used for experiments is a working administrative area Interior office where motion test was performed
MAJOR AREAS IDENTIFIED FOR IMPROVEMENT Increase speed of radar - allows motion detection at realistic rates - allows better imaging through signal averaging Increase speed of display and processing - required to handle increased data rate of faster radar Antenna array may ultimately limit overall system performance - more robust array concept needed
A FIXED LINEAR ARRAY HAS LIMITATIONS Practical limitations - view limited by array placement - bulky and complicated Signal processing - regular spacing leads to ghost images - cross range resolution set by fixed length The imaging algorithms do not require use of a fixed array
RANDOM ANTENNA PLACEMENT IMPROVES IMAGES Additional advantages - variable aperture improves cross range resolution - faster setup/teardown - more operational flexibility
DISTRIBUTED IMAGING NETWORK CONCEPT Individuals and vehicles become sensor nodes Random array improves imaging by reducing ghosts Wireless network receives data and distributes information View of operational space changes as sensors move The network IS the sensor
DISTRIBUTED RADAR SYSTEM ELEMENTS Timing and control, sensor location, and processing speed are the major uncertainties
CURRENT DEVELOPMENT HARDWARE 500-2000 MHz 10 µsec per point software defined 50 mw power output ~ $1,000 parts cost Power conditioner Antenna connectors Radar interface 4 diameter 1.5 tall ~ 13 watts battery powered 2.25 lbs w/battery Radar Tx/Rx
8 ANTENNA INTERFACE BOARD Antenna switch network allows scanning of up to 8 antennas - uses only a single radar - wired operation - same operational concept as TTW brassboard - low cost, early option for field use Board mates to radar through two RF connectors - plug interface allows development of boards with special characteristics without requiring modification of radar
RECENT RESULTS Recent testing shows capability of the radar to detect breathing of a stationary person In the video a person walks from behind and to the left of the radar, past the wall, to the right around the chair, then sits in the chair and remains motionless for a while, finally getting up and walking back out past the wall and radar on the left. For large motion algorithm, person disappears from the image while sitting still Movie File (MPEG) Movie File (MPEG) Small motion algorithm Large motion algorithm Test Geometry
STATIC IMAGE person wall Image taken from single frame of breathing test
MORE RESULTS Tests recently completed on a series of 5 walls - drywall, brick faced, concrete block, adobe, reinforced concrete - all walls 12 thick Breathing response detected through each wall - drywall: at 19 from standoff of 12 - brick: at 21 from standoff of 14 - concrete block: at 28 from standoff of 21 - adobe: at 16 from standoff of 4 - reinforced concrete: at 22 from standoff of 10
CONCLUSION Physics and technology for through the wall imaging are well understood - no fundamental principles limit success - commercial technology exists for implementing a robust system Imaging and motion detection algorithms have been demonstrated with real data The random array concept offers attractive operational benefits