SITIS Archives - Topic Details
Program:  SBIR
Topic Num:  AF071-239 (AirForce)
Title:  Spatially Registered Multispectral Polarimetric Sensor
Research & Technical Areas:  Sensors, Electronics

  Objective:  Investigate sensor concepts that combine multispectral and polarimetric exploitation of targets as a basis for target discrimination and classification.
  Description:  The ability to correctly classify target type can greatly enhance the situational awareness of the battlespace; however, detection and classification of military targets using broadband optical sensors can be extremely difficult. To be effective in a wide range of viewing conditions, imaging sensors used in field or airborne situations must exploit multiple characteristics of potential targets to discriminate between target classes and/or between threat targets and clutter. Promising discrimination techniques used today employ spectral differences, temporal differences, or polarimetric differences in target characteristics as the basis for classification. Sensor technology has progressed sufficiently that imaging systems when combined with currently available processing power and multiple classification techniques may be able to observe a greater range of target characteristics that can be used as inputs to powerful discrimination techniques, such as fuzzy logic, optical/temporal flow, pattern matching, neural nets, etc. Target characteristics including spectral and polarimetric differences would support the use of these powerful identification techniques. The focus of this effort is to explore novel ways in which a single sensor can exploit spectral, temporal, and polarimetric characteristics of military targets. Concepts are desired that provide spatially registered and temporally registered snapshot images of a scene that capture spectral and polarimetric data in the visible, NIR, SWIR, MWIR, and/or LWIR bands. The instrument should simultaneously provide imaging data in at least two different spectral bands, and polarimetric data with at least two different polarization vectors. The system should also show growth path potential and/or scalability toward a fieldable unit, such as would be suitable for a portable instrument or airborne platform.

  PHASE I: Develop a system design for a high frame rate imaging sensor with multiple spatially co-registered images, each with a different polarization or spectral band. Technical analysis should be provided that supports the proposed design.
  
  PHASE II: Construct and test a prototype sensor to demonstrate the concept feasibility for the spatially registered multispectral polarimetric sensor. Identify and reduce the risk of component technologies needed to perfect the design and demonstrate the approaches needed toward a field-ready portable system or flight-capable instrument.

  DUAL USE COMMERCIALIZATION: Military application: Target discrimination and classification Commercial application: Remote environmental sensing

  References:  1. Howard Barnes, Paul Burke, and Art Caneer, “Mid-Wave and Long-Wave Infrared Polarimetric Phenomenology Sensors,” Proc. IEEE, Vol. 4, p. 87-93, 1999. 2. Firoz A. Sadjadi Cornell S.L. Chun, “Remote sensing using passive infrared Stokes parameters,” Optical Engineering, Vol. 43, Issue 10, pp. 2283-2291, October 2004. 3. Stephanie H. Sposato and Matthew P. Fetrow, “Two long-wave infrared spectral polarimeters for use in understanding polarization phenomenology,” Optical Engineering, Vol. 41, Issue 5, pp. 1055-1064, May 2002.

Keywords:  spectral/temporal, multispectral, polarimetric, passive sensors, target discrimination, classification, situational awareness

Questions and Answers:
Q: For gathering Polarimetric data of non-stationary scene, multiple focal plane arrays may be required. Does this topic require only single focal plane array to gather all the polarization vectors?
A: No, multiple focal planes may be used. Image registration among the focal planes should be adequately addressed.
Q: Is a software implementation for the registration of multispectral and polarimetric images sufficient for the system design and proof of concept in early phases of the SBIR? Included would be a design for a hardware implementation of the registration algorithm (e.g. FPGA, ASIC). The registration algorithm would later be implemented to hardware to meet DOD performance requirements for the sensor system.
A: I'm presuming your proposal will address the sensor design, and the registration software/processing hardware is being addressed as part of the overall system.

Yes, for the purposes of the SBIR Phase 1, describing the registration software and processing hardware is sufficient. If a Phase 2 is awarded, the specific implementation of the software/hardware would be addressed during the Phase 2 kickoff, and the design finalized during that phase.
Q: Is there some preferred field of view and/or aperture diameter?
A: There is no specific field of view. Typically, this device would be used initially in a laboratory, then tested in our tower facility. The tower has a line of sight to the target area of a few hundred meters out to 1 km. Fixed targets in the scene would be panels which are 2 meters on a side, or could be a typical automobile. For material detection, we'd want 5 to 10 pixels on the target. To image a fast temporal event (such as a flash), we'd want a few pixels on target up to 10 pixels.
Q: The topics calls for a high frame rate imaging sensor. How high a frame rate and how many pixels in the frame?
A: The sensor ought to have a capability to exploit spectral, temporal and polarimetric characteristics of a military target. A temporal event would be something like a muzzle flash.

Temporal events are expected to be fast events, where some portion of the FPA is read out at rates greater than 200 Hz. It isn't anticipated that the entire array would be read out at that fast rate. If the sensor is imaging a large scene to detect an object (a non-temporal event), then the entire array would be read out at video rates (15 to 30 Hz).
Q: 1. Do the polarimetric images have to have perfect spatial registration?

2. Is a focal plane array where, say, the odd pixel columns provide an image with one polarization and the even columns provide an image with a different polarization, of interest?
A: 1. Polarimetric images need to be registered to minimize aliasing or edge effects. Polarimetric images should be registered to 1/10th of a pixel or less. Spectral images should be registered to a 1/2 pixel or less.

2. Yes, being able to mount/deposit polarizers onto the FPA itself is of interest. Realize there are sampling issues (Q >1) with the two polarizers on adjacent pixels/columns.
Q: 1. Does "at least two different spectral bands" mean multiple spectral bins in say the VIS along with multiple spectral bins in say the NIR? Are there "extra points" given for certain bands, or is any combo of 2 okay?

2. What is the minimum number of spectral bins in each band? What is the maximum FWHM spectral width of such a bin?

3. Must the image snapshots produce 2D imagery, or is 1D line imagery (say from a pushbroom configuration) acceptable? Any "extra points" for one versus the other?

4. Minimum framerate?

5. Is the technique of so-called "channeled spectropolarimetry" specifically excluded from consideration?
A: 1. Two different spectral bands in a given wavelength region -- two Visible/NIR bands, or two SWIR bands, etc. The requirement is not for two wavelength regions (Visible and SWIR, for example). The choice/suggestion of spectral bands is left to the proposer, based on his/her experience with instrument design and data collections. This doesn't exclude a design that might be based on a two color focal plane array (such as a MWIR/LWIR FPA), but a single color FPA (a Vis/NIR FPA, for example) with two selected spectral bands is considered responsive to the SBIR topic.

2. The SBIR description states a minimum of two spectral bands, within a spectral region. The spectral width of the band can be fairly broad for material detection, not narrow as would be required for gas detection or laser line detection. As in the above question on the wavelength region, the spectral width would be left to the proposer, based on experience with sensor design and data collections.

3. The sensor should be a 2D (framing) camera. A 1D scanning system would have a disadvantage if used to capture fast temporal events (such as a flash).

4. The SBIR description describes how the sensor ought to have a capability to capture spectral, polarimetric and temporal scene data. If used to image a fairly static scene for material detection (predominantly spectral and polarimetric), the entire FPA would be read out at video rates (15 to 30 Hz). If used to capture a fast temporal event, then framerates would have to be on the order of a few hundred Hertz. In this fast temporal, few hundred Hz mode, the entire FPA would not need to be read out, but only some portion(s) which image the event (only a portion of the entire FOV of the sensor).

5. No, channel spectropolarimetry technique is not specifically excluded from consideration. From the literature, the channel SP can simultaneously capture the spectral and polarimetric information in a scene. The ability to design and construct the sensor such that it can operate to collect the three regimes (spectral, polarimetric and temporal) at sufficient framerate (question 4 above) will be a deciding factor. The SBIR description states two polarizations are required; the channel SP captures polarizations to allow calculation of the 4 Stokes vectors, which brings some benefit to how this device would be used in collections, but not necessarily a great benefit.
Q: The topic calls for two different polarization vectors. Does that mean atleast two different polarization vector components (e.g. Q and U)?
A: Two polarizations, sufficient to calculate S0 and S1.
Q: For gathering Polarimetric data of non-stationary scene, multiple focal plane arrays may be required. Does this topic require only single focal plane array to gather all the polarization vectors?
A: No, multiple focal planes may be used. Image registration among the focal planes should be adequately addressed.
Q: Is a software implementation for the registration of multispectral and polarimetric images sufficient for the system design and proof of concept in early phases of the SBIR? Included would be a design for a hardware implementation of the registration algorithm (e.g. FPGA, ASIC). The registration algorithm would later be implemented to hardware to meet DOD performance requirements for the sensor system.
A: I'm presuming your proposal will address the sensor design, and the registration software/processing hardware is being addressed as part of the overall system.

Yes, for the purposes of the SBIR Phase 1, describing the registration software and processing hardware is sufficient. If a Phase 2 is awarded, the specific implementation of the software/hardware would be addressed during the Phase 2 kickoff, and the design finalized during that phase.
Q: Is there some preferred field of view and/or aperture diameter?
A: There is no specific field of view. Typically, this device would be used initially in a laboratory, then tested in our tower facility. The tower has a line of sight to the target area of a few hundred meters out to 1 km. Fixed targets in the scene would be panels which are 2 meters on a side, or could be a typical automobile. For material detection, we'd want 5 to 10 pixels on the target. To image a fast temporal event (such as a flash), we'd want a few pixels on target up to 10 pixels.
Q: The topics calls for a high frame rate imaging sensor. How high a frame rate and how many pixels in the frame?
A: The sensor ought to have a capability to exploit spectral, temporal and polarimetric characteristics of a military target. A temporal event would be something like a muzzle flash.

Temporal events are expected to be fast events, where some portion of the FPA is read out at rates greater than 200 Hz. It isn't anticipated that the entire array would be read out at that fast rate. If the sensor is imaging a large scene to detect an object (a non-temporal event), then the entire array would be read out at video rates (15 to 30 Hz).
Q: 1. Do the polarimetric images have to have perfect spatial registration?

2. Is a focal plane array where, say, the odd pixel columns provide an image with one polarization and the even columns provide an image with a different polarization, of interest?
A: 1. Polarimetric images need to be registered to minimize aliasing or edge effects. Polarimetric images should be registered to 1/10th of a pixel or less. Spectral images should be registered to a 1/2 pixel or less.

2. Yes, being able to mount/deposit polarizers onto the FPA itself is of interest. Realize there are sampling issues (Q >1) with the two polarizers on adjacent pixels/columns.
Q: 1. Does "at least two different spectral bands" mean multiple spectral bins in say the VIS along with multiple spectral bins in say the NIR? Are there "extra points" given for certain bands, or is any combo of 2 okay?

2. What is the minimum number of spectral bins in each band? What is the maximum FWHM spectral width of such a bin?

3. Must the image snapshots produce 2D imagery, or is 1D line imagery (say from a pushbroom configuration) acceptable? Any "extra points" for one versus the other?

4. Minimum framerate?

5. Is the technique of so-called "channeled spectropolarimetry" specifically excluded from consideration?
A: 1. Two different spectral bands in a given wavelength region -- two Visible/NIR bands, or two SWIR bands, etc. The requirement is not for two wavelength regions (Visible and SWIR, for example). The choice/suggestion of spectral bands is left to the proposer, based on his/her experience with instrument design and data collections. This doesn't exclude a design that might be based on a two color focal plane array (such as a MWIR/LWIR FPA), but a single color FPA (a Vis/NIR FPA, for example) with two selected spectral bands is considered responsive to the SBIR topic.

2. The SBIR description states a minimum of two spectral bands, within a spectral region. The spectral width of the band can be fairly broad for material detection, not narrow as would be required for gas detection or laser line detection. As in the above question on the wavelength region, the spectral width would be left to the proposer, based on experience with sensor design and data collections.

3. The sensor should be a 2D (framing) camera. A 1D scanning system would have a disadvantage if used to capture fast temporal events (such as a flash).

4. The SBIR description describes how the sensor ought to have a capability to capture spectral, polarimetric and temporal scene data. If used to image a fairly static scene for material detection (predominantly spectral and polarimetric), the entire FPA would be read out at video rates (15 to 30 Hz). If used to capture a fast temporal event, then framerates would have to be on the order of a few hundred Hertz. In this fast temporal, few hundred Hz mode, the entire FPA would not need to be read out, but only some portion(s) which image the event (only a portion of the entire FOV of the sensor).

5. No, channel spectropolarimetry technique is not specifically excluded from consideration. From the literature, the channel SP can simultaneously capture the spectral and polarimetric information in a scene. The ability to design and construct the sensor such that it can operate to collect the three regimes (spectral, polarimetric and temporal) at sufficient framerate (question 4 above) will be a deciding factor. The SBIR description states two polarizations are required; the channel SP captures polarizations to allow calculation of the 4 Stokes vectors, which brings some benefit to how this device would be used in collections, but not necessarily a great benefit.
Q: The topic calls for two different polarization vectors. Does that mean atleast two different polarization vector components (e.g. Q and U)?
A: Two polarizations, sufficient to calculate S0 and S1.

Record: of