Patent Grant
8542347
The present invention relates to telescopes and in particular to Fourier telescopic imaging systems.
Fourier telescopy is an imaging technique that uses multiple beams from spatially separated transmitters to illuminate a distant object. A large number of wide aperture telescopes (referred to as “light buckets”) are used to detect the light reflected from the object with sufficient signal-to-noise ratio, but no attempt is made to form an image directly from the collected light. Each beam is shifted relative to the other beams to produce interference between beams at the target. Thus interference patterns are produced on the target. These interference patterns are encoded in the time dependent reflections from the target. The reflected light collected by the light buckets is Fourier transformed into a frequency spectrum as a function of time and this spectrum is correlated with known positions and frequencies of the transmitted beams as a function of time to derive a two dimensional image of the target.
This imaging technique has been studied extensively for use in imaging deep space objects. In prior art system designs, for example, three beams would be transmitted simultaneously in pulses to image a geosynchronous object. It would take many hours to transmit the tens of thousands of pulses needed to construct all of the spatial frequencies needed to form an image of the object. Because the position and orientation of the object would remain essentially constant, this approach seemed feasible. Three illuminating apertures were used in order to eliminate the degrading atmospheric phase aberrations using the well known technique of phase closure, and then the closure phases used to reconstruct the illuminated target image. Previous experiments in both the lab and field have verified that this implementation of the Fourier Telescopy technique to imaging geostationary targets is both viable and robust.
The AEOS (short for Advanced Electro-Optical System) telescope is a 3.67-meter 75 ton telescope situated at the crest of the dormant volcano Haleakala on the Hawaiian Island of Maui. It is owned by the United States Department of Defense and is used for tracking and viewing satellites as well as many scientific purposes. Virtually year-round viewing conditions are possible due to the relatively stable climate. Dry, clean air and minimal scattered light from surface sources enable visibility exceeding 150 km. Based on double star observations, seeing is typically on the order of one second of arc. AEOS can be used simultaneously by many groups or institutions because its light can be channeled through a series of mirrors to seven independent coudé rooms below the telescope. Employing sophisticated sensors that include an adaptive optics system, radiometer, spectrograph, and long-wave infrared imager, the telescope tracks man-made objects in deep space and performs space object identification data collection. AEOS is equipped with an adaptive optics system, the heart of which is a 941-actuator deformable mirror that can change its shape to remove the atmosphere's distorting effects. Other associated equipment includes a 1.6-meter telescope, two 1.2-meter telescopes on a common mount, a 0.8-meter beam director/tracker, and a 0.6-meter laser beam director. The telescopes accommodate a wide variety of sensor systems, including imaging systems, conventional and contrast mode photometers, infrared radiometers, low light level video systems, and acquisition telescopes.
A rough estimate of for diffraction limit for a telescope is:
Rs=Raλ/D
where Rs is the resolution, Ra is the range, λ is the wavelength and D is the aperture diameter. For the AEOS telescope, a range of 106 m, λ, of 0.532×10−6 m and an aperture diameter of 3.67 m, the resolution is about 0.14 m.
Like many high quality telescopes the AEOS telescope provides nearly diffraction limited resolution. What is needed is techniques and equipment for improving the resolution of telescopes like AEOS and other telescopes to better than diffraction limited resolution.
The present invention provides a super-resolution Fourier telescope. A target is illuminated with at least three laser beams, each beam having a slightly different frequency so as to produce an illumination pattern comprised of several sets of straight interference fringes which sweep across the target. The frequencies of the illumination beams are chosen so that each pair of beams provides a unique beat frequency, and the corresponding fringe pattern for each pair sweeps over the target at a unique speed. By collecting a series of images, and demodulating them at the various beat frequencies, the downshifted spatial frequencies can be identified, correctly up-shifted, and fitted together with a set of special Fourier transform type algorithms to reconstruct high-resolution images. Applicants have performed laboratory experiments and simulations that show that this invention can provide resolution substantially better than diffraction limited resolution.
Preferred embodiment include receiver detector components defining an array of high speed detectors adapted to monitor and record spatial intensity data at frequencies at least twice the highest beat frequency of the laser light reflected from the target; and at least one computer processor programmed with a set of algorithms for producing, from the recorded spatial intensity data, images of the target which are better than diffraction limited images corresponding to the telescope aperture. In a specific preferred embodiment the array of high speed detectors is an array of 32 ×32 avalanche photodiode detectors and the set of algorithm is adapted to determine phases of the uplink beam by comparing data in a spatial frequency overlap region and to combine redundant elements of the target's special frequencies in a noise optimized way. In this preferred embodiment the target illuminating laser system comprises a laser adapted to provide a series of very short pulses within bursts having durations of less than one millisecond. In a preferred configuration, the bursts are in the range of about 250 microseconds and the very short pulses are in the several microsecond range.
One application of this technique is the imaging of low-earth-orbit (LEO) satellites using land-based telescopes. The advantage of this method is that the resolution of existing telescopes (such as the AEOS telescope) or other optical imaging instrument can be increased without increasing the size and weight of the optical components. This resolution advantage has wide applicability in many military and commercial areas such as optical surveillance, satellite reconnaissance, and target identification in that high resolution imaging is obtained with less complicated and costly optical systems.
a shows a concept for the new insertion mirror that would be used to replace the existing AEOS insertion mirror for super resolution.
The receive aperture for laser light reflected from a target is the receive aperture of the existing AEOS telescope as shown in
The Receiver detector should be located downstream of the adaptive optic (AO) deformable mirror so that receive laser light reflected from the target may benefit from atmospheric compensation. The receiver detector for this preferred embodiment should be located after the deformable mirror and should meet the requirements listed in Table 2.
A laser of 100 watts total power provides adequate signal strength on target for detectable reception through the AEOS telescope according to simulations performed by Applicants. The chosen wavelength is 532 nm. The laser meets the requirements. The laser beam is separated into three legs of equal power (30 W each). Each leg passes through an acoustic-optic modulator (AOM) which puts a unique frequency shift onto each beam. As indicated above the three beams respectively have the following frequency shifts: 0, 10, and 30 Hz. An example layout (demonstrated in a laboratory experiment as described below) for the laser, beam splitting, and implementation of the AEOS is shown in
The preparation of the three laser beams should be accomplished on an optical bench which is elevated on pneumatic isolation legs. The beam diameter of each of the three transmit beams upon exit from the laser transmitter system will be determined based on the injection location into the AEOS beam path. The beam size should scale such that the magnification through the AEOS system yields a 20 cm diameter upon exit from the telescope. The three beams will be equally spaced in an equilateral triangular configuration. Similar to the beam sizes, the beam separation (center-to-center of sub-apertures) will also scale such that the magnification through the AEOS system yields a 305 cm separation upon exit from the telescope.
The transmit laser shall be injected into the beam path at the location of the current AO Injection Mirror. The insertion method will involve moving the AO Injection Mirror to the available “OUT” position and installing a new mirror with three pass-through holes sized and aligned to the three CFT transmit laser paths. This mirror will be called the “CFT Injection Mirror” and will be in a mount with kinematic feet that are compatible with the existing AO Injection Mirror mount kinematic foot pads. The “OUT” position for the AO Injection Mirror is normally used in operations if light from the mt36 is desired to by-pass the AO system completely. The action of moving the AO Injection Mirror to the IN and OUT positions is a normal and routine activity. Thus, the insertion of the CFT transmit laser should have minimal impact on the AO system.
The beam size at the injection mirror is about 23 cm based on the scaling referred to above. The injection mirror shall be oriented at approximately a 45 degree angle to the beams. Thus, the mirror shall have a physical size to accommodate an elliptic beam footprint of at least 23 cm by 32.6 cm as shown in
The injection mirror pass-through holes for the transmit laser shall be spaced to accommodate the projection of the three 19.33 cm spaced beams (length of legs of the equilateral triangle) onto the optic at an angle of 45 degrees. For the condition in which two transmit beams are near the bottom and the third is near the top of the injection mirror, the hole spacing in the mirror (in the plane of the surface of the mirror) shall be 19.33 cm between the bottom two holes and 23.67 cm between the top hole and each of the bottom holes, as shown in
The injected laser light will not encounter either the AEOS deformable mirror or the AEOS tracker. Thus, the outgoing laser light cannot be compensated nor utilize the AO Fast Steering Mirror (FSM, aka Tracker Control Mirror (TCM)) to achieve track/point-ahead. A separate fast steering mirror shall be located on the transmitter table and shall be slaved to the AO tracker system to achieve point-ahead angles. The track system will utilize the AO Track Camera as it closes track loop on the return light from the target.
A preferred algorithm for converting collected data reflected from the target is described in this section. As described above the information is obtained from pairs of beating beams which apply a sinusoidal intensity modulation to a target (sometime referred to herein as (an “object”). The sinusoidal modulation offsets the spatial frequencies measured by the camera by in effect “mixing” the image with the modulation, thereby “down-converting” the high spatial frequencies of the object down to spatial frequencies resolvable with the telescope.
There are two components to the algorithm:
We will consider here that the illuminating lasers are of sufficient bandwidth so that speckle effects are insignificant.
Consider then an object intensity profile given by Iob(x,y), with a depth function z(x,y). When we illuminate with short pulse lasers, and use a time resolved detector, we are in effect dividing the image into depth slices. We therefore more conveniently describe the intensity function as a series of slices Iob(x,y,z), where each value of z used represents a time bin.
This intensity pattern is illuminated by the multiple illumination lasers. For this case we will assume 3 lasers situated at the periphery of the aperture. We will assume that the aberrations for these 3 beams will be insignificant (for now), but the relative phases between the beams is unknown. We will call the single beam uplink pattern Iup(x,y), and the laser positions x1,y1 . . . x3,y3. Then the projected pattern at the object is
The reflected intensity Iref=Iill(x,y)·Iob(x,y,z). Notice that while Iob is a function of z, Iill is not. This is very important for noise properties, because the phasing information from all of the slices is the same so can be combined.
The breakthrough of the present approach was the realization that use of slices greatly facilitated the phasing; this is because phasing requires high spatial frequency features to “key off” of, and the slices have a lot more high spatial frequency content than the flat image. In this case we will likewise combine the data from the slices to get robust estimates of the unknown phase.
For now, we will assume a diffraction-limited receiver aperture. The telescope point spread function PSF(x,y) is then the standard form for a circular aperture. The measurement is then
M(x,y,z)=[Iill(x,y)·Iob(x,y,z)]◯PSF(x,y),
where the operator “◯” denotes convolution. Here we have assumed that the detector has sampled the data out to the Nyquist limit, at least two pixels per λ/D. When we extract the DC component and 3 beat frequencies from the measurement, we obtain the components
MDC(x,y,z)=[3·Iup(x,y)·Iob(x,y,z)]◯PSF(x,y)
M12(x,y,z)=[Iup(x,y)·Iob(x,y,z)·exp(i(kx1−kx2)·x+i(ky1−ky2)·y+i(φ1−φ2))]◯PSF(x,y)
M23(x,y,z)=[Iup(x,y)·Iob(x,y,z)·exp(i(kx2−kx3)·x+i(ky2−ky3)·y+i(φ2−φ3))]◯PSF(x,y)
M31(x,y,z)=[Iup(x,y)·Iob(x,y,z)·exp(i(kx3−kx1)·x+i(ky3−ky1)·y+i(φ3−φ1))]◯PSF(x,y)
where kxi=2πxi/λR, kyi=2πiyi/λR.
The Poisson noise in the DC component is a real Gaussian random variable with variance (when normalized to photocounts)
σ2DCMDC(x,y,z).
The AC components have complex Gaussian random noise with complex variance
σ212=σ223=σ231=MDC(x,y,z).
Detector noise would be added to this variance, and should be the same increase for each of these quantities.
Next we will define new terms to simplify the discussion. For the DC term, we will just redefine its PSF-like term:
TDC(x,y):=3·PSF(x,y)
M′DC(x,y,z)=MDC(x,y,z)=[Iup(x,y)·Iob(x,y,z)]◯TDC(x,y)
For the AC terms, we will introduce a phase slope onto the measurements, which does not affect the noise properties:
M′12(x,y,z)=M12(x,y,z)·exp(−i(kx1−kx2)·x+−i(ky1−ky2)·y)
so that
M′12(x,y,z)=exp(i((φ1−φ2))·[Iup(x,y)·Iob(x,y,z)]◯T12(x,y)
T12(x,y):=PSF(x,y)·exp(i(kx1−kx2)·x+i(ky1−ky2)y)
and similarly for the other AC components. These expressions are in an especially convenient form to consider in Fourier space.
As a further abstraction to simplify the terminology, just call
I:=Iup(x,y)·Iob(x,y,z)
M′12:=M′12(x,y,z), etc.
exp(iφ12):=exp(i((φ1−φ2)), etc.
and
T12 will represent the linear operator “T12(x,y)”, etc. Then we have
M′DC=TDCI+nDC
M′12=exp(iφ12)T12I+n12
M′23=exp(iφ23)T23I+n23
M′31=exp(iφ31)T31I+n31
Since the noise terms all have the same variance, which we will call σ2(x,y,z) or just σ2, the optimum values for the unknowns is found by minimizing w.r.t. I and φ the average over x,y,z of:
{|M′DC−TDCI|2+|M′12−exp(iφ12)T12I|2+|M′23−exp(iφ23)T23I|2+|M′31−exp(iφ31)T31I|2}/σ2
Variation w.r.t. I gives
In order to (greatly) simplify the mathematics, we are going to let the T operators commute with the 1/σ2 operation. This is reasonably justified since 1/σ2 is considerably more slowly varying that T. This allows us to solve for I:
I={TDC†TDC+T12†T12+T23†T23+T31†T31}−1 . . . {TDC†M′DC+exp(−iφ12)T12† M′12+exp(−iφ23)T23† M′32+exp(−iφ31)T31† M′31}
This is the expression we use to find the image once we have solved for the phase. Since the T operators are just convolutions, the entire evaluation becomes a few convolutions. The inverse operation only needs to be computed once for a given configuration.
To solve for the phases, we take this solution for I and plug back into the minimization function. Since we only are solving for phase, we first throw out terms which have no phase dependence:
maximize average of
2·Re{M*′DCTDCI+M*′12exp(iφ12)T12I+M*′23exp(iφ23)T23I+M*′31exp(iφ31)T31I}/σ2
which can be written
2·{[M*′DC+M*′12exp(iφ12)T12+M*′23exp(iφ23)T23+M*′31exp(iφ31)T31]. . . [TDC†TDC+T12†T12+T23†T23+T31†T31]−1 . . . [TDC†M′DC+exp(−iφ12)Tas†M′12+exp(−iφ23)T23† M′23+exp(−iφ31)T31554 M′31]}/σ2
In order to carry out this computation, a straightforward implementation is to form a figure-of-merit image FOM(x,y,z,φ12,φ23,φ31)
FOM=(1/σ) {TDC†TDC+T12†T12+T23†T23+T31†T31}−1/2 . . . {TDC†M′DC+exp(−iφ12)T12† M′12+exp(−iφ23)T23† M′23+exp(−iφ31)T31† M′31}
and then find the maximum over the two unknown phase parameters of FOM2 averaged over x,y,z. Since T is just a convolution, the half power merely requires performing a square root in the Fourier domain. This phase finder implementation is nice because it has a number of terms in common with the image computation expression. Since in this case we only have 2 parameters to optimize, we might as well just brute force the search. If more beams are used, then a SAINT-style search algorithm can be included. (A SAINT-style algorithm is provided in companion patent application filed concurrently with this application entitled “Fourier Telescopic Imaging System and Method” which is incorporated by reference herein.)
Once we have found the phases, then we can go back to the above expression for the optimum image and compute the image slices.
When we perform inverse operations above, it should be noted that regularization parameters are added to prevent excessive noise amplification. The values of the parameters are truly a matter of taste, as it is really just a means of getting data from the computer into the reviewer's eye. Typically 10% of the max value is a reasonable way to pick a value. Changing the value only changes the degree to which the image looks filtered, it does not really change any information content.
To help explain the operation of the operation of the reconstruction algorithm, Applicants have included
The validity of Applicants' approach to super-resolution has been examined by means of numerical simulation and a laboratory experiment. This section describes the laboratory experiment.
The maximum angular resolution of an image formed by a conventional imaging system is determined by the fundamental diffraction limit ˜λ/D where λ, is the wavelength used for imaging and D is the diameter of the aperture stop of the imaging system. Two examples of this aperture are the variable iris in a camera lens and the unobstructed part of the primary mirror of a reflective telescope. “Super-resolution” refers to various techniques for exceeding this fundamental resolution limit. The effect is similar to having a conventional optical system with a larger aperture, as we demonstrate below.
This super-resolution method consists of illuminating the object with a pattern comprised of several sets of straight interference fringes formed by pairs of laser beams. The effect of the patterned illumination is to down shift spatial frequencies (moire effect) that would otherwise be blocked by the aperture so that they pass through the optical system. The frequencies of the illumination beam are chosen so that each pair of beams has a unique frequency difference, and the corresponding fringe pattern sweeps over the target at a unique speed. By collecting a series of images, and demodulating them at the various beat frequencies, the downshifted spatial frequencies can be identified, correctly up-shifted, and fitted together to reconstruct the higher-resolution image.
A laboratory experiment demonstrating the capability of Applicants' supper-resolution technique has been completed. A schematic drawing of the experimental setup is shown in
The frequency of each beam is shifted by reflection from a traveling sound wave in an acousto-optic modulator (AOM). Amplified sine waves from Programmed. Test Systems (PTS) Model ×10 frequency synthesizers produce the signals that drive the AOMs' sound waves. As an example of choosing the beam frequencies properly, in one instance the AOMs were driven by the frequency synthesizers at 40 MHz, 40 MHz+50 Hz, and 40 MHz+150 Hz, so that the beat frequencies measured at a point in the interference pattern contained 50, 100, and 150 Hz frequency components.
The desired fringe illumination pattern is generated in the far-field by focusing the beams toward the target from near the periphery of a large lens. A second collimating lens then forms a scaled copy of the fringe pattern that would result if the beams had been projected by a telescope toward a LEO satellite. Photographic transparencies for testing the imaging concept are illuminated by the fringe pattern. The dimensions of the target images (which were about 3 mm), the spacing of the interference fringes (which were about 4/mm), and the diameter of the illumination beams are correctly scaled with respect to the optical system to be approximately physically equivalent to the planned field imaging system (the AEOS modified system described above).
A rapidly rotating diffusing plate sits directly behind the target transparencies to mimic the diffuse-reflection characteristics of real specular targets and also the shorter coherence length of the illumination lasers that would be used in a fielded system. A pair of lenses images the target on a high-speed charge-coupled-device (CCD) camera with 128×128 pixels from DALSA operating at 830 Hz frame rate as the time-resolved focal plane detector in this laboratory setup. An iris was placed directly in front of the target imaging lens to function as a variable aperture stop for the system. By adjusting the size of this iris, we could obtain various resolution levels of the target on the DALSA camera.
The CCD collects a sequence of images, which are saved in a file for further processing.
The laboratory experiment is controlled by a single main computer that contains a frame grabber so that the high speed CCD camera data can be recorded and stored. These sequential images of the recorded modulated fringe pattern on the target are then processed and reconstructed using a developed algorithm to obtain the super-resolution information of the target. This experiment computer also uses a secondary camera as a diagnostic to verify the beams are overlapped on the target image. An auxiliary phase control computer was also utilized to control the beam shutters and frequency modulators within the experimental setup.
Similarly
Applicants have planned a ground demonstration of the super-resolution technique utilizing the 2.4 meter Magdalena Ridge Observatory (MRO) telescope. The anticipated layout of the super-resolution ground demonstration would be to place a small target at a range of 0.5-1 km away from the MRO telescope. At these ranges, the integrated atmospheric turbulence is expected to be sufficiently small such that near diffraction limited imaging can be obtained utilizing only tip/tilt correction to provide for image stabilization on the receiver detector. Also, the target ranges are sufficiently distant enough to allow use of reasonable size targets of a few millimeters in diameter. The three laser illumination output beams could either be placed at the perimeter of the 2.4 meter telescope on the primary mount truss or injected through the telescope optics and mapped to the perimeter telescope aperture.
As suggested above many high power lasers are available meeting the laser requirements set forth in Table I. Applicants have however have developed a specially designed a laser system for the AEOS and MRO telescope applications. This design is described below with reference to
The laser specifications for the laser system is provided in
The reader should understand that the present invention is not limited to the specific embodiments described above and that many modifications and additions or deletions could be made to those described embodiments. For example the invention could be applied to much larger and much smaller telescopes than the ones referred to. It could also be applied to microscopes. New telescopes could incorporate the features of this invention. Other high speed detectors could be used in place of the avalanche photodiode detectors. Larger or smaller pixel arrays should be tailored to specific telescopes. Changes to the specific algorithm described herein are possible. Therefore, the scope of the invention should be determined by the appended claims and their legal equivalents.
The present invention was made in the course of work performed under Contracts No. FA9451-04-D-0300 with the Defense Advanced Research Projects Agency and the United States Air Force and the United States Government has rights in the invention.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5969334 | Hutchin | Oct 1999 | A |
| 8058598 | Sandler et al. | Nov 2011 | B2 |
| 20080218851 | Chen et al. | Sep 2008 | A1 |
| 20080259447 | Oppenheimer | Oct 2008 | A1 |
| 20100110208 | Gerwe | May 2010 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20120105822 A1 | May 2012 | US |
