Abstract

We report the results of a multi-day diurnal study in which polarimetric and conventional thermal imagery is recorded in the mid- and long-wave IR to identify and compare the respective time periods in which minimum target contrast is achieved. The data shows that the chief factors affecting polarimetric contrast in both wavebands are the amount of thermal emission from the objects in the scene and the abundance of MWIR and LWIR sources in the optical background. In particular, it has been observed that the MWIR polarimetric contrast was positively correlated to the presence of MWIR sources in the optical background, while the LWIR polarimetric contrast was negatively correlated to the presence of LWIR sources in the optical background.

© 2010 OSA

1. Introduction

Thermal imagers have been established as the primary tool used in military and security activities that involve surveillance, targeting and tracking, and night-time operations. Unlike I2 devices, which depend on ambient light levels, thermal imagers exploit the fact that all objects with a temperature above 0 K emit thermal radiation by creating a pseudo-image of the scene based on this thermal emission. The two thermal imaging windows are the MWIR, 3-5 µm, and the LWIR, 8-14 µm, both chosen for the relatively low amounts of absorption from atmospheric species, such as CO2 and H2O.

Contrast between the objects within a thermal image is determined by their effective temperatures, which are a function of their true temperature and emissivity, a characteristic that describes how efficiently an object radiates absorbed energy as compared to a blackbody. If there is no thermal contrast between a target and its background, it cannot be seen in a thermal image. The diurnal cycles of the thermal properties of both manmade and natural objects tend to bring about periods of low contrast within thermal images, often referred to as thermal crossovers or inversion periods [1]. These crossover periods tend to occur during periods of rapidly changing temperatures, such as sunrise and sunset, but may occur at any time throughout the day, depending both on temperature differences between objects and their backgrounds, and environmental factors, such as solar loading.

1.1 Thermal polarimetric imaging

Thermal polarimetric imaging has been proposed as a method to enhance conventional thermal imaging [2]. It creates images of a scene that incorporate the states of polarization of the IR light emitted or reflected from the objects within the scene. An object’s polarimetric signature is a function of its surface geometry and roughness. Due to the different geometrical and roughness features of objects constituting natural backgrounds and manmade objects, the polarization states of the emitted and reflected thermal light can be used as a discriminator between objects of interest and background clutter. Using the polarization of the light, it is possible to obtain an image of a scene that has polarimetric contrast between objects and their backgrounds, despite the fact that there is no thermal contrast.

In addition to the environmental factors that affect thermal contrast, there are also environmental factors that affect polarimetric contrast. The most significant of these factors involves sources of IR radiation in what has been referred to as the optical background. The optical background is defined as the sources of IR radiation that are not necessarily visible within the scene but still illuminate the objects in the field of view of the camera [3]. Potential sources include vehicles, buildings, trees, clouds, water vapor, etc. The optical background can become partially polarized upon reflection off of the objects in the scene and act as a competing component to their emitted polarized light. Polarized emitted and reflected light are orthogonal to each other because the emitted light is polarized in the direction parallel to the surface normal while light polarized upon reflection is perpendicular to the surface normal. When the two components superpose, the effect can be a reduction in the magnitude of the polarimetric signature of an object.

This study compares the temporal occurrence of conventional thermal crossover periods to that of polarimetric and correlates these crossovers with environmental factors. Our goal is to quantify the periods of time in which contrast within polarimetric images is present, while conventional thermal contrast has been lost. Under these conditions, it can be concluded that polarimetric imaging is capable of enhancing conventional thermal imaging.

2. Experiment

2.1 Overview

We used two different polarimetric sensors, MWIR and LWIR, to view ground targets from a fixed, elevated observation point over a 48 h period. The sensors were operated close to continuously in an effort to observe all variations in the conventional thermal and polarimetric signatures. A very complete set of environmental/meteorological data was recorded. All data was recorded with reference to a common time standard. Most analysis took place after the completion of the test.

2.2 MWIR polarimetric sensor

The MWIR imaging polarimeter is based on a division-of-aperture (DOA) lens technology developed by Polaris Sensor Technologies [4]. The system employs a 2x2 array of mini-lenses that forms four identical images of the scene on four quadrants of the sensor focal plane array. Each mini-lens is followed by a linear polarizer at a different orientation. For this system, a set of four linear wire-grid polarizers are used, oriented at angles 0°, 90°, 45° and 135°. Thus, the focal plane array (FPA) captures four images simultaneously of the object at these four different polarization states. The images are precisely registered to within 1/10 pixel in software, and weighted subtractions are done to compute the Stokes parameter images of the scene. Table 1 gives the optical specifications for the DOA MWIR Imaging Polarimeter.

Tables Icon

Table 1. A summary of the MWIR sensor optical specs

2.3 LWIR polarimetric sensor

The LWIR imager is a microbolometer-based rotating retarder imaging polarimeter developed by Polaris Sensor Technologies, Inc., Huntsville, AL [5]. It operates by capturing up to 12 images sequentially in time, each at a different orientation of the rotating retarder. Together, the retarder and linear polarizer act as a polarization state analyzer for the light forming the image. Using the data reduction matrix method, the Stokes vectors are calculated, which completely characterizes the polarization states of the light from the scene. Table 2 lists the sensor specifications. It is worth mentioning that microbolometer FPAs are less sensitive than other IR FPAs and therefore may produce noisier polarimetric data during times of low polarization signal, e.g. low thermal emission and heavy ambient IR loading.

Tables Icon

Table 2. Specifications for the LWIR imaging polarimeter

2.4 Field test site

The test was conducted at the Precision Armaments Laboratory located at Picatinny Arsenal, NJ. The camera was situated on the sixth floor of a tower (approximately 200 ft high) looking out of the windows in the SSE direction towards the target site, which was at approximately 0.5 km in range and consisted of two military vehicles and natural backgrounds, including grass, brush, and trees (Fig. 1 ). In addition to the 200 ft elevation of the sixth floor of the tower, the tower itself was situated on top of a ridge approximately 175 ft above the target site.

 

Fig. 1 Target site consisting of two military vehicles and a natural background. The test was conducted on May 13-14, 2009.

Download Full Size | PPT Slide | PDF

Environmental measurements available at the target site include air temperature, relative humidity, ceilometer data, and pyrgeometer (precision infrared radiometer [PIR]) data. The ceilometer provides real-time reports of cloud bases and depths directly above the ceilometer and determines cloud cover by using a weighted average of 30-s cloud hit reports over a 30-min period. The pyrgeometer measured ambient downwelling IR radiation from 3 to 50 µm within a 2π steradian FOV.

The data acquisition clocks for the environmental data and the cameras were synced to ensure coincident data. The temporal resolutions of the data are as follows: air temperature, relative humidity, and pyrgeometer—2 seconds; ceilometer—10 seconds; and cameras—10 minutes. Data was acquired continuously between 0:00 on May 13, 2009 and 23:50 on May 14, 2009.

2.5 Image analysis

A contrast-based study was performed between a target (surrogate tank) and its immediate backgrounds. Data for the target directly facing the tower in Fig. 1 is used in the analysis. The exact scene used in the analysis is shown in Fig. 2 , and consists of the surrogate tank and the surrounding grass and trees/brush. This target was selected because most of its surfaces are oriented in the same direction relative to the FOV of the camera. Therefore, widely varying magnitudes of polarimetric signatures due to diversity of surface geometry minimally affect the target’s mean signatures. In addition, this surrogate tank was at ambient temperatures and is only the shell of a real tank; therefore, its response to temperature changes is different from that of a real tank due to its smaller mass. The size of the scene was 28x22 and 59x54 pixels for the LWIR and MWIR systems, respectively.

 

Fig. 2 The test target and its natural background. The target, grass, and trees regions of interest correspond to the blue, red, and green boxes, respectively.

Download Full Size | PPT Slide | PDF

The data products used in this study include S0 and S1 Stokes parameter images, where S0 is the total intensity (equivalent to conventional thermal image), and S1 is the horizontal minus the vertical components of polarization. The S2 Stokes parameter ( + 45° minus −45° polarized components) and degree of linear polarization are excluded because the positive 45° quadrant of the MWIR sensor was not functioning properly, and so this data is not available for comparison with the corresponding LWIR parameters. Because S0 and S1 are of different magnitudes, the analysis was performed on standardized versions of these images [6,7]. In other words, each image had its mean (μimage) subtracted and was normalized by its standard deviation (σimage):

xij=xijμimageσimage

In Eq. (1), xijand xij refer to the unstandardized and standardized images, respectively, and the i and j subscripts refer to the pixel coordinates. This makes it possible to directly compare the contrasts for both the S0 and S1 images. Using temporal sequences of standardized co-registered images, regions of interest were defined for the target and its backgrounds, and the mean values were calculated for both data products at each time in the series. The contrast is defined as the absolute value of the difference between the mean target value, μt, and the mean background value, μb, where the prime indicates that the means were calculated using standardized images:

contrast=|μtμb|

Because standardizing the images gives them unitless values, the contrast defined by Eq. (2) is also unitless.

Examples of contrast values calculated using Eq. (2) and their corresponding images are shown in Fig. 3 . The contrasts were calculated for the LWIR S0 images of the scene in Fig. 2 taken at three different times of the day. Because this scene has two different natural backgrounds, one is chosen at a time to calculate the contrast between the target and the chosen background. In the case of Fig. 3, the grass background is used to calculate the contrast.

 

Fig. 3 Example contrast values comparing the target to the grass and the corresponding S0 image taken at a. 07:00, b. 11:00, and c. 19:00.

Download Full Size | PPT Slide | PDF

3. Results

The results are presented in the form of diurnal contrast plots, calculated using Eq. (2). Included with each of these plots are the corresponding environmental data. The ceilometer data is omitted because the cloud information can be obtained from the pyrgeometer (PIR) data. A direct comparison of ceilometer and pyrgeometer data revealed that the baseline reading for a cloudless daytime sky was roughly 275–280 W/m2. Any value higher than this typically indicates the presence of clouds such that the higher the value, the greater the cloud cover. In addition, the relative humidity data is omitted because it displayed the usual inverse relationship to the temperature profile and appeared to have minimum affect on the Stokes data. During this study, sunrise and sunset occurred at roughly 05:00 and 20:00, respectively.

3.1 May 13, 2009

Figures 4 and 5 show the diurnal contrasts between the tank and its natural background (trees and grass, respectively) in the LWIR S0 and S1 and the MWIR S1. The MWIR S0 contrast is omitted because it is very similar to the contrast in the LWIR S0. In addition, both figures provide imagery for select times, allowing for comparisons with the contrast plots. For the majority of the day, contrast in the LWIR S1 is higher than in LWIR S0 or MidIR S1 and is reduced by the presence of clouds. The clouds affect the scene in two ways: first, during the day, they decrease the solar flux at all wavelengths that reach the objects in the scene and, therefore, decrease the rate at which these objects are heated; second, they always act as sources of IR light because they radiate IR light, as well as reflect the IR light emitted from the earth back down onto the scene [8]. This additional IR light is then reflected off of the more reflective objects in the scene (in this case, the tank) and is partially polarized. This reflected component of polarization is perpendicular to and therefore competes with the emitted component and effectively reduces the overall magnitude of the LWIR polarimetric signature of the tank.

 

Fig. 4 Diurnal contrast between the tank and its background of trees on May 13, 2009.

Download Full Size | PPT Slide | PDF

 

Fig. 5 Diurnal contrast between the tank and its background of grass on May 13, 2009.

Download Full Size | PPT Slide | PDF

The peak in the MWIR S1 contrast between 8:00 and 10:00 in Figs. 4 and 5 is due to polarized emission resulting from the heating of the tank. As clouds begin to enter the atmosphere at 10:00, the tank’s MWIR S1 signature begins to acquire a larger reflective component, resulting in the crossover at approximately 10:30 and the subsequent contrast modulation throughout the rest of the day [9,10]. The anticorrelation with the presence of clouds between the tank’s LWIR and MWIR S1 contrast is most readily observed between 12:00 and 16:00. Peaks in cloud cover at 13:00 and 14:30 are associated with decreases in LWIR S1 contrast and improvements in MWIR S1 contrast, while during the period of relatively clear skies between 15:00 and 16:00, the LWIR S1 contrast is at its maximum during the 24-h period, and the MWIR S1 contrast falls to zero.

The contrasts between the tank and the trees are similar to those between the tank and the grass with a few exceptions. The improved contrast between the tank and the trees in the LWIR S1 was consistently greater during the periods 0:00 to 10:00 and 20:00 to 22:00 than between the tank and the grass. The two thermal crossovers between the tank and the trees occurred at 8:00 and 20:30 while there were three thermal crossovers between the tank and the grass occurring at 7:00, 9:50, and 23:00. The MWIR S1 contrast between the tank and trees is very similar to the case between the tank and the grass. Most importantly, the inverse correlation of the relationship of LWIR S1 and MWIR S1 to the presence of clouds remains between the tank and trees.

3.2 May 14, 2009

Figures 6 and 7 show the diurnal contrasts between the tank and its natural background (grass and trees, respectively) in the LWIR S0 and S1 and the MWIR S1 on May 14. On this day, inclement weather moved into the Picatinny area in the early morning hours and there were intermittent periods of light rain up until the early afternoon. During this time, there was thick cloud cover, as reflected in the PIR plots. The LWIR S0 and S1 profiles in both figures display a considerable degree of variation. In the case of S0, this is primarily due to the periods of light rain during the first 14 h of the day. For S1, the optical background emitted by the overcast skies, along with the periods of rain, contributes to the noisy profiles. The last of the rain occurred at approximately 14:00, and a brief decrease in cloud cover occurred at 15:00. During this time, the inverse correlation between MWIR S1 and LWIR S1 with cloud cover is observed, where LWIR S0 and MWIR S1 contrast decreased while LWIR S1 increased. As soon as the cloud cover returned to its previous levels, the LWIR S0 and MWIR S1 contrast increased once again, and the LWIR S1 contrast returned to noise levels. In fact, the MWIR S1 contrast benefited from the overcast conditions throughout the whole day.

 

Fig. 6 Diurnal contrast between the tank and its background of grass on May 14, 2009.

Download Full Size | PPT Slide | PDF

 

Fig. 7 Contrast between the tank and its background of trees on May 14, 2009.

Download Full Size | PPT Slide | PDF

4. Conclusions

In this study, the temporal occurrence of conventional thermal and polarimetric crossovers was examined, as well as their correlations to environmental factors. Imagery was recorded with two polarimetric IR sensors. The first was a LWIR sensor employing a 324x256 microbolometer array using a spinning achromatic retarder to perform the polarimetric filtering. The second was a MWIR polarimetric sensor based on a DOA approach with a 640x512 InSb focal-plane array. The images used in this study included the S0 and S1 Stokes parameter images of a scene containing a military vehicle and the natural background. In addition, relevant meteorological parameters measured during the test periods included air temperature; ambient loading; relative humidity; and cloud cover, cloud height, and cloud density.

This study revealed that during most thermal crossover periods, there remains polarimetric contrast between at least some facet of the target—and in many cases, the whole target—and its background. In addition, the data shows that the chief factors affecting polarimetric contrast are the amount of thermal emission from the objects in the scene and the abundance of MWIR and LWIR sources in the optical background.

Furthermore, it has been observed that the optical background due to cloud cover decreased LWIR S1 contrast, while it increased the MWIR S1 contrast. This is consistent with the notion that MWIR polarimetric signatures tend to have a higher reflection component than LWIR polarimetric signatures.

Typical thermal crossovers can affect contrast for up to 2 h at a time, reducing the effectiveness of thermal imaging systems and jeopardizing the success of military and security missions. The findings of this study suggest that conventional thermal imagery can be enhanced by incorporating fused-MWIR and LWIR polarimetric information. The waveband that showed the most promise for polarimetrically enhancing conventional thermal imaging during this test was LWIR. The ability of LWIR polarimetric imaging to enhance conventional thermal imaging is limited by susceptibility to the optical background. If a LWIR polarmetrically enhanced thermal imager is being used during a mission under conditions in which there are significant optical background sources, then the performance of the system will reduce to that of a conventional thermal system. By fusing LWIR and MWIR polarimetric data, it may be possible to extend the polarimetric enhancement of conventional thermal imagery into the periods of high optical background by exploiting the reflective nature of MWIR polarimetric signatures.

While promising, these conclusions should be considered as preliminary and may be strengthened by extending the research to cover a broader range of environmental conditions and parameters, targets, and additional contrast metrics.

Acknowledgements

This work was accomplished using data from the Hyperspectral and Polarimetric Target Detection Program at the Precision Armament Laboratory (PAL) Tower at the Armament Research and Development Engineering Center (ARDEC), with the help of Mr. Joao Romano (Program Manager, joao.m.romano@us.army.mil), Mr. Mark Woolley (PAL Manager), and Zed Habte (test engineer).

References and links

1. D. L. Shumaker, J. T. Wood, and C. R. Thacker, Infrared Imaging Systems Analysis, (DCS Corporation, Alexandria, 1993), Chap. 2.

2. J. S. Tyo, D. L. Goldstein, D. B. Chenault, and J. A. Shaw, “Review of passive imaging polarimetry for remote sensing applications,” Appl. Opt. 45(22), 5453–5469 (2006). [CrossRef]   [PubMed]  

3. J. S. Tyo, B. M. Ratliff, J. K. Boger, W. T. Black, D. L. Bowers, and M. P. Fetrow, “The effects of thermal equilibrium and contrast in LWIR polarimetric images,” Opt. Express 15(23), 15161–15167 (2007). [CrossRef]   [PubMed]  

4. J. L. Pezzaniti and D. B. Chenault, “A division of aperture MWIR imaging polarimeter,” Proc. SPIE 5888, 239–245 (2005).

5. J. L. Pezzaniti, B. Hyatt, and J. Reinhardt, “Systems Users Manual: LWIR Rotating Retarder Imaging Polarimeter,” 2008.

6. W. R. Dillon, and M. Goldstein, Multivariate Analysis: Methods and Applications, (John Wiley & Sons, New York, Chichester, Brisbane, Toronto, and Singapore, 1984), Chap. 1.

7. M. Felton, K. P. Gurton, D. Ligon, and A. Raglin, “Discrimination of Objects Within Polarimetric Imagery Using Principle Component and Cluster Analysis,” ARL-TR-4216, 2007.

8. J. A. Shaw, P. W. Nugent, N. J. Pust, B. Thurairajah, and K. Mizutani, “Radiometric cloud imaging with an uncooled microbolometer thermal infrared camera,” Opt. Express 13(15), 5807–5817 (2005). [CrossRef]   [PubMed]  

9. M. L. Salby, Fundamentals of Atmospheric Physics, (Academic Press, San Diego, New York, Boston, London, Sydney, Tokyo, and Toronto, 1996), Chap. 1 and 8.

10. J. A. Shaw, “Degree of linear polarization in spectral radiances from water-viewing infrared radiometers,” Appl. Opt. 38(15), 3157–3165 (1999). [CrossRef]  

References

  • View by:
  • |
  • |
  • |

  1. D. L. Shumaker, J. T. Wood, and C. R. Thacker, Infrared Imaging Systems Analysis, (DCS Corporation, Alexandria, 1993), Chap. 2.
  2. J. S. Tyo, D. L. Goldstein, D. B. Chenault, and J. A. Shaw, “Review of passive imaging polarimetry for remote sensing applications,” Appl. Opt. 45(22), 5453–5469 (2006).
    [CrossRef] [PubMed]
  3. J. S. Tyo, B. M. Ratliff, J. K. Boger, W. T. Black, D. L. Bowers, and M. P. Fetrow, “The effects of thermal equilibrium and contrast in LWIR polarimetric images,” Opt. Express 15(23), 15161–15167 (2007).
    [CrossRef] [PubMed]
  4. J. L. Pezzaniti and D. B. Chenault, “A division of aperture MWIR imaging polarimeter,” Proc. SPIE 5888, 239–245 (2005).
  5. J. L. Pezzaniti, B. Hyatt, and J. Reinhardt, “Systems Users Manual: LWIR Rotating Retarder Imaging Polarimeter,” 2008.
  6. W. R. Dillon, and M. Goldstein, Multivariate Analysis: Methods and Applications, (John Wiley & Sons, New York, Chichester, Brisbane, Toronto, and Singapore, 1984), Chap. 1.
  7. M. Felton, K. P. Gurton, D. Ligon, and A. Raglin, “Discrimination of Objects Within Polarimetric Imagery Using Principle Component and Cluster Analysis,” ARL-TR-4216, 2007.
  8. J. A. Shaw, P. W. Nugent, N. J. Pust, B. Thurairajah, and K. Mizutani, “Radiometric cloud imaging with an uncooled microbolometer thermal infrared camera,” Opt. Express 13(15), 5807–5817 (2005).
    [CrossRef] [PubMed]
  9. M. L. Salby, Fundamentals of Atmospheric Physics, (Academic Press, San Diego, New York, Boston, London, Sydney, Tokyo, and Toronto, 1996), Chap. 1 and 8.
  10. J. A. Shaw, “Degree of linear polarization in spectral radiances from water-viewing infrared radiometers,” Appl. Opt. 38(15), 3157–3165 (1999).
    [CrossRef]

2007 (1)

2006 (1)

2005 (2)

1999 (1)

Black, W. T.

Boger, J. K.

Bowers, D. L.

Chenault, D. B.

Fetrow, M. P.

Goldstein, D. L.

Mizutani, K.

Nugent, P. W.

Pezzaniti, J. L.

J. L. Pezzaniti and D. B. Chenault, “A division of aperture MWIR imaging polarimeter,” Proc. SPIE 5888, 239–245 (2005).

Pust, N. J.

Ratliff, B. M.

Shaw, J. A.

Thurairajah, B.

Tyo, J. S.

Appl. Opt. (2)

Opt. Express (2)

Proc. SPIE (1)

J. L. Pezzaniti and D. B. Chenault, “A division of aperture MWIR imaging polarimeter,” Proc. SPIE 5888, 239–245 (2005).

Other (5)

J. L. Pezzaniti, B. Hyatt, and J. Reinhardt, “Systems Users Manual: LWIR Rotating Retarder Imaging Polarimeter,” 2008.

W. R. Dillon, and M. Goldstein, Multivariate Analysis: Methods and Applications, (John Wiley & Sons, New York, Chichester, Brisbane, Toronto, and Singapore, 1984), Chap. 1.

M. Felton, K. P. Gurton, D. Ligon, and A. Raglin, “Discrimination of Objects Within Polarimetric Imagery Using Principle Component and Cluster Analysis,” ARL-TR-4216, 2007.

M. L. Salby, Fundamentals of Atmospheric Physics, (Academic Press, San Diego, New York, Boston, London, Sydney, Tokyo, and Toronto, 1996), Chap. 1 and 8.

D. L. Shumaker, J. T. Wood, and C. R. Thacker, Infrared Imaging Systems Analysis, (DCS Corporation, Alexandria, 1993), Chap. 2.

Cited By

OSA participates in CrossRef's Cited-By Linking service. Citing articles from OSA journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (7)

Fig. 1
Fig. 1

Target site consisting of two military vehicles and a natural background. The test was conducted on May 13-14, 2009.

Fig. 2
Fig. 2

The test target and its natural background. The target, grass, and trees regions of interest correspond to the blue, red, and green boxes, respectively.

Fig. 3
Fig. 3

Example contrast values comparing the target to the grass and the corresponding S0 image taken at a. 07:00, b. 11:00, and c. 19:00.

Fig. 4
Fig. 4

Diurnal contrast between the tank and its background of trees on May 13, 2009.

Fig. 5
Fig. 5

Diurnal contrast between the tank and its background of grass on May 13, 2009.

Fig. 6
Fig. 6

Diurnal contrast between the tank and its background of grass on May 14, 2009.

Fig. 7
Fig. 7

Contrast between the tank and its background of trees on May 14, 2009.

Tables (2)

Tables Icon

Table 1 A summary of the MWIR sensor optical specs

Tables Icon

Table 2 Specifications for the LWIR imaging polarimeter

Equations (2)

Equations on this page are rendered with MathJax. Learn more.

x i j = x i j μ i m a g e σ i m a g e
c o n t r a s t = | μ t μ b |

Metrics