Proc. SPIE. 5076, Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XIV
KEYWORDS: Thermography, Signal to noise ratio, Imaging systems, Spatial frequencies, Image processing, Modulation transfer functions, Minimum resolvable temperature difference, Thermal modeling, Systems modeling, Visibility
Minimum Resolvable Temperature Difference (MRTD) has long been used to describe the performance of thermal imaging systems. The Visibility Model II developed for second generation thermal imaging systems includes sampling and aliasing issues without assumptions about the observer. As with the earlier reported Visibility Model, applicable for first generation imagers, both objective and subjective measurement schemes can be accommodated. The visibility concept has been demonstrated to be applicable in predicting the objective MRTD. The laboratory measurement of objective MRTD provided the data to evaluate the performance of Objective VISMODII model. Although the measurements are limited in scope, they do demonstrate that the VISMODII predictive model can also be applied to objective MRTD measurements.
Shore-based polarized infrared images of ship target and background were obtained in the MAPTIP measurement series, using vertical and horizontal polarization filters. These images have been analyzed to yield degree of polarization and polarization contrast enhancement for targets with sea and sky background Strong vertical polarization is observed in the sea surface emission near the Brewster Angle in the 8 - 12 micrometer (LWIR) band, whereas horizontal polarization due to reflection is usually dominant in the 3 - 5 micrometer (MWIR) band. Ship targets at broadside aspect show a degree of polarization less than 5%. An enhancement of contrast by up to 30% in the LWIR (less in the MWIR) is obtained by polarization suppression of sea background. The contrast enhancement in the MWIR is counteracted by the horizontal polarization of reflected sunlight.
Sea radiance in the mid and far infrared shows a considerable degree of polarization which affects observed target-to-background contrast. An improvement in contrast is achieved with horizontal (s-plane) polarization filtering to suppress sea surface emission. Visibility and range affect the contrast in both polarizations. Scenes recorded during the MAPTIP measurement series off the coast of the Netherlands with the oceanographic ship HrMs Tydeman show decrease in contrast with range and better contrast for horizontal polarization against sea background. A simple mathematical model is presented relating contrast to extinction and path radiance which increase with increasing path length or worsening visibility.
We present a new model for predicting the minimum resolvable temperature difference (MRTD) curve of thermal imaging systems. The analysis for the new model concentrates on contrast reduction due to spatial frequency limiting factors of subsystem components. Curves have been generated for this model for a system with typical component values. These results are compared with curves generated from the NVL's static performance model. The proposed visibility model leads to a relatively simpler development for a MRTD predictor which can readily account for artifacts due to a nonzero system phase transfer function. In addition the visibility model makes no assumptions regarding the recognition process and therefore is adaptable to the goal of modeling an objective MRTD measurement. The visibility model agrees with the static performance model except at very low and very high spatial frequencies where the proposed model appears to be in better agreement with observed trends in measured MRTDs.
Realistic inversions of lidar signals for extinction using the Klett technique must take into account both the theoretical limitations of the inversion and the experimental constraints of the hardware. A simple test of the Klett inversion algorithm s19 is performed with and without experimental limitations and uncertainties. The effect of limited accuracy in the digitization of the lidar return and limited dynamic range is presented. A simplified technique for detecting clouds in the presence of low visiblity is developed, and some limitations are presented.
In previous papers we have presented measurements of degree of polarization in the mid and far infrared both in near-horizon sun-glint and in the adjacent sea background radiance. The polarization has been related to the Fresnel reflection and emission coefficients and compared with predictions. These measurements have been extended to include target-to-background polarization contrast measurements. Target polarization was small, but sea surface emission showed moderate degree of polarization in the p-plane (vertical). Considerable ship- background contrast improvement was achieved by polarization filtering in the horizontal plane. This effect was greater in the far infrared. Values are given for the degree of polarization of target and background and the contrast improvement factor due to polarization filtering.
Measurements are reported of the infrared sunglint clutter channel caused by the direct solar reflection from the wave-perturbed sea surface at near-grazing angles of incidence. Apparent radiance has been measured over Monterey Bay as a function of azimuth and elevation angles relative to the sun direction using an AGA Thermovision 780 dual-band radiometric imaging system in the wavebands 2 to 5.6 (SW) and 8 to 12 micrometers (LW) with 7 degree(s) FOV. Time averaged profiles from multiframe averages show near-Gaussian angular distributions with half widths in the range 3 to 20 degrees (depending on solar angle) for look-down angles of 1 to 10 degrees below the horizon. The p- and s- polarized components of sea surface radiance have been obtained using an external wire-grid polarizing filter and compared with unpolarized measurements. The degree of polarization within the glint is shown to be horizontal and variable in the range 1% to 30%, depending on solar elevation, the higher degrees of polarization being found in the SW band. Significant vertical sea radiance polarization has been observed outside the solar glint in the 8 to 12 micrometers band, and is attributed to sea surface emission polarization.
Tactical Decision Aid codes provide field prediction of maximum range of FLIR use using simplified local environmental parameter input. A series of experimental comparisons at sea using airborne operational FLIRs with an instrumented ship target have shown poor correlation of observed range with prediction for detection and recognition. Classification and recognition range in UFLR are found to be highly insensitive to radiosonde atmospheric profile data input. Previous work has addressed modeling of the average target to background contrast temperature difference and atmospheric propagation of contrast. This paper addresses the implementation of the MDTD and MRTD algorithms in the code. Comparisons are presented of the prediction accuracy of the UFLR TDA using the standard Moser/Hepfer algorithm and an adaptation of the Johnson criterion used in the NVEOL Ratches code. For the limited data set of the study a reduction of RMS prediction error is achieved using the NVEOL algorithm.
This paper presents a sea radiance polarization model and experimental measurements of near- horizon sea glint polarization in the 3 - 5 micrometers and 8 - 12 micrometers spectral bands. The experimental measurements include the effects of polarization on the glint statistics, the degree of linear polarization and the polarization signal-to-noise ratio (SNR) improvement factor for both spectral bands in the presence of sea glint. The results indicate that the polarization in the 3 - 5 micrometers spectral band is dominated by the reflected solar and sky radiance and is polarized in the s plane. The polarization of intense sea glint in the 8 - 12 micrometers region is low and s polarized due to the weak solar spectrum in this band. In little or no glint, the radiation is weakly p polarized. Experimental data indicate that a polarizing filter can produce a significantly larger SNR improvement for the 3 - 5 micrometers spectral band than for the 8 - 12 micrometers band. Theoretical calculations using the polarization model show good agreement with the experimental data.
A narrow-beam laser altimeter was used to measure the reflected signal from the ocean surface as represented by the waters beneath the Golden Gate Bridge. This site allowed precise measurements as a function of angle from the vertical not possible from flying platforms. For short-wavelength water waves superimposed on swell, the signal amplitude probability distribution for the reflected signals showed periods of zero reflection, even for vertical incidence, apparently due to tipping of the water surface. The nonzero signals showed a distribution that could be fitted with an antilog-normal distribution. This is skewed toward higher signals than a normal (Gaussian) distribution. With incidence angle displaced from the vertical, the distribution shape was retained but with more frequent zero reflections. The decrease with angle of the average signal, including the zeroes, is well fitted with a Gram- Charlier distribution, as seen by earlier observers using photographic techniques which masked these details of the structure. For the simpler wave pattern due to a long sustained wind direction, the probability distribution is log-normal with no zero signal periods. At large angles from the vertical the log-normal distribution shifts toward exponential. For surface states intermediate between the above two extremes the distribution is often normal. The larger return signals resulting from the skew toward larger amplitudes from lognormal are more favorable for disposable laser altimeters than previously believed. Also, for an altimeter which may be swinging from a parachute or balloon, the return remains high at angles other than vertical. The presence of occasional zero return signal does somewhat degrade the accuracy of altitude measurement for a descending altimeter, but the signal available assures performance at larger altitudes than previously expected.
In support of a continuing program of evaluation and experimental validation of FLIR Tactical Decision Aid performance codes, a series of measurements has been made of ship radiance temperature distributions together with sea and sky backgrounds. The measurements have been made at ranges from one quarter to one mile off the coastline in Monterey Bay, using a land- mounted Agema 780 dual band Thermovision radiometric sensor, with computer data acquisition and storage. The target ship was the research vessel Point Sur carrying a full suite of meteorological instruments and an array of thermal sensors for ship surface temperature distribution. Rawinsonde balloons were released to obtain vertical temperature and humidity profiles for path correction using LOWTRAN. The normal skin emissivity was measured in a separate experiment. The current data band consists of 898 stored radiometric frames containing ship images including starboard, port, bow and stern aspects, together with sea and sky background frames with varied zenith angle. These files are available for false color display and analysis in a variety of formats.
UFLR is one of an evolving set of FLIR performance prediction programs used at sea to predict the ranges for detection, classification, and identification of target ships. One aid in the validation of such a program is a sensitivity analysis of the program parameters. Sensitivity analyses indicate that the ranges for detection, classification, and identification are strongly sensitive to target areas, target-to-background, temperature difference and atmospheric conditions such as windspeed, visibility, humidity, and vertical temperature, humidity, and pressure profiles. One uncontrollable parameter is the noncontiguity in space and time of the radiosonde and FLIR measurements. This problem was investigated by dithering the radiosonde data, input to UFLR, with a random number generator to generate variations in the pressure, temperature, and relative humidity in the atmospheric profile. Results indicate that noncontiguity of measurements can lead to 50% error in range predictions.
Commercial frame grabber technology in IBM PC compatible computers has been adapted to allow direct digital input of infrared search and track data at rates of up to 10 mbytes per second, permitting real-time processing and display of false color thermal images. Examples of single frame displays and background suppression by frame subtraction are shown. On- board processed Fourier spectra and Fourier power spectra of selected frame lines are shown. Curve-fitted representations are compared for clear air, cloud and land clutter, and a commercial aircraft at close range.
The NPS-IRST,a modification of the Advanced Demonstration Model of the Navy AN/SAR-8 InfraRed Search and Target Designation system, mounted in a rooftop location at the Naval Postgraduate School, Monterey, has been used to record background scene information. High data-rate tape-recording with reduced-rate playback is used for data processing and a 'framegrabber' board for image display and processing. Irradiance probability density function, pulse width probability density function, and autocorrelation function plots are displayed for five background types selected from false-color image displays.
The performance of the Tactical Environmental Support System FUR tactical decision aid program UFLR was reported upon previously. In using this code it was assumed that the temperature difference between ship and sea is fixed. In actuality the temperature difference changes with sea state, aspect, cloud cover, and visibility conditions. In this article a simple predictive temperature contrast taking these effects into account is described and validated against experimental data. A modified ThA code UFLRB including this input is compared with observed ranges. The modified procedure shows up to 36% improvement in detection range prediction at night.