Detecting Moving Targets in SAR Via Keystoning and Phase Interferometry

Transcription

1 5 The MITRE Corporation. All rights reserved. Approved for Publi Release; Distribution Unlimited. Deteting Moving Targets in SAR Via Keystoning and Phase Interferometry Dr. P. K. Sanyal, Dr. D. M. Zasada, Mr. R. P. Perry The MITRE Corp., 6 Eletroni Parkway, Rome, NY 134 Ph: , Fax: psanyal@mitre.org Abstrat We require ontinuous and unambiguous radar traking of surfae moving targets for several minutes to target and engage moving targets. Conventional radar surfae moving target trakers typially drop or onfuse traks after only a short time. If we an ouple state of the art motion-ompensated Syntheti Aperture Radar (SAR) tehniques with advaned Surfae Moving Target Information (SMTI) tehniques, we may be far better able to automatially and ontinuously trak individual targets through zero radial veloity in diffiult environments. Without motion ompensation, moving targets within SAR images are generally blurred and diffiult to detet. MITRE has developed a tehnique alled the Keystone Formatting for motion ompensation of targets, the advantage of whih is that it an ompensate for several targets moving at different veloities simultaneously. Along with aeleration orretion, this produes sharp images. Complimentary to the Keystone d Range-Doppler-Intensity image, one an form a phase- interferometry image. In the phase image, where all points on the non-moving surfae nominally appear as a ontinuum of phase differenes while the moving targets appear as disontinuities. By judiious omparison of both the intensity image and the phase image, it is possible to detet and loate moving targets in the SAR. Key words: SAR, Keystone Formatting, Phase Interferometry Introdution For tatial purposes, we require ontinuous and unambiguous radar traking of surfae moving targets for several minutes to target and engage moving targets. Conventional radar surfae moving target trakers typially drop or onfuse traks after only a short time. This is aused by target detetion drop-outs due to target stops, starts, quik turns, low target radial veloity, terrain sreening, et. Hene ontinuous attention by a human operator is urrently required to stith these short trak segments bak together. Without motion ompensation, moving targets within SAR images are generally blurred. These effets are shown by the SAR images in Figures 1 and. MITRE has developed some novel tehniques for produing sharp, foused SAR images and then to detet the moving targets therein. In this paper, we present results of moving target detetion in multi-hannel SAR data olleted by Linoln Laboratories of Lexington, MA, USA. Figure 1. SAR With and Without Motion Compensation (Soure:

2 5 The MITRE Corporation. All rights reserved. Approved for Publi Release; Distribution Unlimited. beomes zero when we use the temporal transformation f t = ( ) t. f + f Figure. Moving Boat Appears Displaed From its Atual Position in a SAR Image (Soure: MITRE-developed Keystone Formatting MITRE has developed a tehnique alled Keystone Formatting for motion ompensation of targets, the advantage of whih is that it an ompensate for several targets moving at different radial veloities simultaneously. Keystone formatting an be derived by noting that the spetrum of a single reeived pulse is given by, f ) exp[ i ) R( t)] where P( f ) = spetrum of transmitted pulse B B f = baseband frequeny ( f < ), f arrier frequeny. = (1) With the above substitution, (3) an be written as, f )exp[ i ) R (4) f t i f Rt & π i ) R&& ( ) ]. f + f Sine the Keystone formatting does not solve the quadrati (or higher order) motion problem, let us also drop the quadrati term in (4) and simplify it as: f ) * (5) exp[ i ) R i f R & t ]. Notie that the substitution of t for t has removed the phase term that varied with both time and frequeny and this removes the range-walk. Thus no matter what veloity the target is moving at, it will remain in a given range ell determined by its position at the enter (t=) of the oherent proessing interval. Figure 3 shows the keystone nature of the transformation. Expanding R(t) in a Taylor series, we get: 1 R ( t) = R( t ) + R& ( t ) t + R&& ( t ) t + L. () Substituting () into (1) and dropping ubi and higher order terms, f )exp[ i ) R i ( f f ) Rt & π + i ) Rt && ]. The seond term in the brakets ontaining the produt f Rt & gives rise to range walk. This term (3) Figure 3. Keystone Formatting Performs Motion Compensation for Targets Moving at Different Veloities Figure 4 shows the effet of Keystoning on the range walk The left inset in Figure 4 shows the RTI with two targets moving at different speeds. The two targets go through different amounts of range walk. Coherent proessing of the data without any ompensation for target motion results in an integration loss and smearing of the target over multiple range ells. Standard motion ompensation will orret the range walk for one target at a time.

3 5 The MITRE Corporation. All rights reserved. Approved for Publi Release; Distribution Unlimited. Aeleration Corretion The Keystoning orrets only for the range walk due to the veloity. However, by itself, it annot orret for the image defousing that results from the aeleration that introdues quadrati phase error (QPE). Figure 6 shows the typial SAR geometry and the resulting aeleration. Given the look angle, one an pre-ompute the aeleration and apply an appropriate orretion to the data. Figure 4. Keystone Formatting Performs Motion Compensation for Targets Moving at Different Veloities Figure 5. With Keystone Formatting, Targets Moving at Different Veloities will Fous Simultaneously The Keystone proess is seen to ompensate for the motion of both the targets simultaneously. Figure 5 shows that they an now be oherently integrated without any signifiant loss due to range walk. Figure 7 presents SAR images reated from data olleted by Linoln Laboratory. The left inset is the range-doppler image without any aeleration orretion while the right inset is the image that results after an optimum aeleration orretion has been applied. Without the appropriate aeleration orretion, the image is pratially unreognizable. With the optimum aeleration orretion, the image is learly reognizable as an area with several buildings, roads, a ball park, et. To find the optimum aeleration orretion, we applied pre-seleted trial few values around the expeted value of 1.97 m/s until the image intensity peaked. This value was found to be.5 m/s. Moving targets tend to appear removed from their atual loations in SAR. For example, moving trains appear to be floating a onsiderable distane away from the stationary traks. v θ R -v Figure 6. The Typial SAR Geometry and the Resulting Aeleration

4 5 The MITRE Corporation. All rights reserved. Approved for Publi Release; Distribution Unlimited. a) No aeleration orretion b) Optimum aeleration orretion Figure 7. SAR Images of Ft. Huahua reated from LiMIT Data Colleted by Linoln Laboratory (a).9 Seond Dwell Time Image (b).9 Seond Dwell Time Image, Dwell 1 seond Later Figure 8. A Moving Target in the Ft. Huahua Image (Channel # 1) The Doppler, f D, of a point at θ radians from the normal to the veloity vetor for a small angle θ redues Vθ to f D =. λ If a moving target has the Doppler f targ, then it will appear shifted in ross-range by an angle θ suh that Vθ vt arg Vθ vt arg f t arg = or = or θ = λ λ λ V At a range R, this amounts to a linear shift of Rvt arg ross range shift = Rθ = V Figure 8 shows two images from the LiMIT data about.9 seonds apart. A areful inspetion of the two images reveals a moving target at the tip of the red arrow. Though the omparison of a sequene of SAR images an reveal the presene of moving target (i.e., oherent hange detetion), we have applied an in-line phase interferometry tehnique for deteting and loating moving targets in a multi-hannel SAR data. Figure 9a shows the same image shown in Figure 8 but with the range axis strethed; hene, the image is not reognizable as suh. However, one an now see

5 5 The MITRE Corporation. All rights reserved. Approved for Publi Release; Distribution Unlimited. (a) Amplitude Image from Channel #1 (b) Phase Image (Ch. #1 Ch. # 8) Figure 9. A Moving Target in the Ft. Huahua Image (Channel # 1) that there are ertain streaks, speifially the one identified with the arrow that seem to have a ant to their streth. This is an indiation of a moving objet. Figure 9b plots the pixel-by-pixel phase differenes between the images from hannel #1 and hannel # 8. The hannels or sub-arrays are loated at the two ends of the antenna array and eah point on the ground (and objets at the same spot on the ground) has slightly different path lengths to the arrays and thus there is a path-length differene or a phase differene. The path length differene varies very little as one moves out in range but there is a signifiant hange as one moves in ross-range aross the image. This produes the vertial striped nature of the phase image. Phase Interferometry for Moving Target Detetion At the same loation in the phase image where there was a feature with a different ant in the amplitude image, we notie a feature that appears to have different oloration and have a different phase differene value from its surrounding. Speifially, we see a light bluish feature in a red bakground. This signifies that the objet that generated the feature belongs in the image where the other bluish stripes are, but beause of the Doppler proessing, it appears at a loation onsistent with the rate of hange of the phase to any hannel. The presene of disontinuities in the phase image indiates the presene of moving targets and the value of the phase differene at the disontinuity indiates where they atually belong. Having loated a moving target, we hipped out the moving target and foused it further to produe the image shown in Figure 1. From an estimation of the length of the objet and it s slightly urved appearane, it appears highly likely that it is a trator-trailer. Unfortunately, we did not have ground truth information available to verify this. Summary For tatial purposes, we would require ontinuous and unambiguous radar traking of surfae moving targets for several minutes to target and engage moving targets. By oupling state-of-the-art motion-ompensated Syntheti Aperture Radar (SAR) tehniques with advaned Surfae Moving Target Indiation (SMTI)

6 5 The MITRE Corporation. All rights reserved. Approved for Publi Release; Distribution Unlimited. In this paper we have desribed the Keystone Formatting for removing range walk in SAR data and have shown, using real multi-hannel SAR data olleted by Linoln Laboratories of Lexington, MA, USA, how the Keystone formatting and aeleration orretion produes sharp SAR images. We have shown that moving targets appear in the SAR at positions displaed from their atual loation in the sene. This displaement is diretly related to their radial veloities. Figure 1. Aeleration Foused SAR Image of a probable Moving Trator-trailer Further, we have shown that phase interferometry using two hannels of the multi-hannel data reveals the presene of moving targets in the SAR image. tehniques, one may be far better able to automatially and ontinuously trak individual high value targets through zero radial veloity in diffiult environments. Referenes 1. Rihard P. Perry, Robert C. DiPietro and Ronald L. Fante, The MITRE Corporation, SAR Imaging of Moving Targets. IEEE Transations on Aerospae and Eletroni Systems, Vol. 35, No. 1, pp Stokburger, E. F., Held, D. N., Interferometri Moving Target Imaging, IEEE International Radar Conferene, 1995 Author Info Probal Sanyal got his B.Teh. (EE) and M. Teh. (Control Systems) from IIT, Kharagpur, India in 1967 and 1969 Respetively. He reeived his Ph. D. in Systems Eng. From RPI, Troy, NY in He has been with the MITRE Corp. for over 15 years and is urrently a lead engineer. He has worked on various radar signal proessing projets. David M. Zasada is a senior Prinipal Engineer at the MITRE Corporation. He reeived a BS in Physis from Georgetown University (1971), an MS in astrophysis from RPI (1973) and a Ph.D. in Eletrial Engineering from Syrause University (1995). He joined MITRE in Rihard P. Perry is a Senior Prinipal Engineer at The MITRE Corporation and has a BSEE degree from Lehigh University (1953) and a MSEE degree from the University of Pennsylvania (1966). He is urrently working on advaned Syntheti Aperture Radar proessing algorithms. He has been with MITRE for 15 years and has previously worked at RCA and GE.