An integrated method for detection and discrimination of a signal pulse in two-dimensional cluttered background noise is presented in this paper. The method is used to estimate the magnitude (detection) and the two-dimensional position (centroid) of the signal using a statistical model for the correlated background noise. The optimal discriminant can use measurements from multiple spectral bands and prior information about the signal shape as well as derived discriminants such as the velocity of the signal pulse. Preprocessing techniques include background suppression in multiple spectral bands with a generalization of temporal and spatial filtering. The coefficients of the optimal filter used for detection and discrimination depend on the statistical characteristics of the signal and the background (averages and cross correlations). Illustrative two-dimensional scenes and representative numerical results are presented to show how this integrated approach (which uses all discriminants) compares with a piece-meal approach to discrimination.
Statistical nonrecursive spatial filters are used to process noisy infrared mosaic sensor images for background clutter suppression and target detection. It is assumed that the clutter can be modeled as an additive, spatially correlated noise and described by its auto-correlation function. The filter is designed to estimate the targets and also to suppress the clutter based on "a priori" knowledge of the statistical property of the noise. Square shaped spatial filters are studied. Signals of all pixels in the filter are linearly weighted and summed to estimate the signal at the center of the filter. The weighting coefficients are designed by a minimization of mean squared error procedure. Target detection is accomplished by thresholding after the spatial filtering. Furthermore, this filter can also be designed to accomplish discrimination of point and line targets simultaneously with the enhancement of target to clutter ratio. Filter results based on computer simulation are shown.
In order to evaluate infrared detection and remote sensing systems, it is necessary to know the characteristics of the observational environment. For both scanning and staring sensors, the spatial characteristics of the background may be more of a limitation to the performance of a remote sensor than system noise. This limitation is the so-called spatial clutter limit and may be important for systems design of many earth application and surveillance sensors. The data used in this study is two dimensional radiometric data obtained as part of continuing NASA remote sensing programs. Typical data sources are the LANDSAT Multi-Spectral Scanner (1.1 micrometers), the airborne Heat Capacity Mapping Radiometer (10.5 - 12.5 micrometers) and various infrared data sets acquired by low altitude aircraft. Techniques used for the statistical analysis of one dimensional infrared data, such as Power Spectral Density (PSD), exceedance statistics, etc. are investigated for two dimensional applicability. Also treated are two dimensional extensions of these techniques (2D PSD, etc.), and special techniques developed for the analysis of 2D data.
To achieve sufficient sensitivity, background noise suppression is required because of the relatively weak target signal. Background noise suppression techniques, first order, second order temporal differencing, spatial and temporal differencing, are analyzed and compared. Background noise due to both background drift and system jitter effects are calculated. Pixel-to-pixel offset-induced boresight errors is also evaluated.
Infrared astronomy is often done with rocket probes or orbiting satellite telescopes to escape the limitations of atmospheric absorption. The data returned from such missions is a highly abstracted digital representation of the measurements made by the analog detectors. The ability and methods needed to locate and measure stars and other infrared emmision from these data streams depends on a thorough understanding of the information flow from the telescope aperature to the computer center. This paper surveys the salient components of this end-to-end concept and the impact each has on the data processing algorithms, including the division between on-board and ground processing for scientific measurements. The most important step in this chain is the detection of sources in the noisy data. The various detection algorithms which have been used in rocket probe and satellite programs are reviewed and their performances compared. Specific algorithms, including thres-holding, peak detection, and matched filtering are shown to be equivalent to an optimal correlation algorithm under different simplifying assumptions. The performance of this optimal algorithm on data from an infrared satellite sky survey is presented. Once detection is completed, the scientific value of the measurements hinges on the methods used to eliminate redundancy and spurious sources, and to measure the positions and brightness of the sources. The impact of reliability and completeness criteria on the choice of these algorithms is presented and compared to previous programs. The importance of the end-to-end concept in meeting these criteria becomes obvious from a discussion of the successes and failures in previous efforts.
This paper describes the philosophy and implementation of a two-dimensional transversal (finite impulse response) filter. The filter will provide automatic background cancellation and image enhancement in real time at rates in excess of 4 megapixels per second. The paper begins by discussing the philosophy of the filtering process as applied to typical electro-optic imagery together with computer simulated images illustrating its performance. A hardware architecture is described that causes a 15- x 15-pixel scanning window to pass over the image, performing a two dimensional convolution in real time with a weighting function that is programmable under the control of a microprocessor based console. To simplify the hardware design, certain restrictions are imposed on the kernel weighting functionnamely that it must constitute a rectangularly symmetric function.
This paper reviews the concept of a spatial spectral interferometer which measures object functions to fine spatial and fine spectral resolutions, using an aperture typically two orders of magnitude smaller than ordinarily expected, while the number of detectors used is one order of magnitude less than ordinarily expected. The modulation of the component spatial frequency signals is described. It is then shown that a sparse detector array with spacings larger than the apparent Nyquist limit can be used to reconstruct the image with the alias spectra cancelled.
Some of the familiar Rinds of optical designs for the visual region of the spectrum take on new meaning when applied to infra-red problems. The availability of high indices of refraction, for example, accentuates by-ways of optical design in the visual into quite u -usual results in the infrared. As a first example, the Rbsch combination of catadioptric system derived from the Schmidt telescope leads to a kind of wide-field system of low a perture-ratio and low obscuration, although at a sacrifice of compactness. In another such instance, it is feasible to design an optical system in the I-R that can illuminate some appropriate form of detector from a major fraction of a 4-pi solid angle over a field of view of appreciable size. Similarly, it is possible to design a form of retro-reflector for the I-R that has a 4-pi field of view, as for the Luneberg sphere. Further examples of special design for the infrared are discussed, including all-refractive, all-reflective and catadioptric combinations. The author reviews some of these useful facets of optical design where appropriate infrared materials can be employed and suggests how some of these forms may be applied to infra-red problems.
Progress in wide-field spaceborne infrared telescope systems design over the last decade has been driven by parallel developments in focal plane technology. The scanned linear detector array of 1970 is yielding to the two-dimensional mosaic array of 1980, which permits consideration of systems for continuous full-earth coverage from geosynchronous altitude. Secondly, the successful development of alternative detector materials and associated cryogenic systems for use in the new focal planes has made it possible to consider multiband missions at selected wavelengths over the 2- to 14-micrometer region.
More than 40,000 infrared measurements of stellar sources have been obtained since November, 1976 during the ongoing process of compiling an Equatorial Infrared Catalogue. Because of the problem of eliminating spurious sources, which has affected earlier space surveys, we are making an extensive effort to verify the sources by means of (a) repetitive observations by satellite sensors, (b) crosscorrelation with a large data base developed from ground-based and space surveys at other wavelengths, and (c) investigation of a significant subset of the sources with a ground-based infrared telescope. As sources are verified, they are transferred from a working list to a screened preliminary version of the catalogue. The catalogue comprises the only survey of a significant area of the sky that has been accomplished (or is presently planned) with positional accuracies of a few arc seconds at a wavelength of ≥ 2 µm.
In high-sensitivity, deepspace surveillance missions, detection of point-source target satellites must be accomplished amid a background of thousands of stars. These stars must be removed from the data stream to permit target detection and to accommodate the data stream to the capacity of the downlink. Two different star elimination techniques, color elimination and moving target indicator (MTI) are analyzed and compared. It is shown that neither of two color discrimination schemes is adequate to meet the data downlink capacity goal. Star leakage is reduced by two orders of magnitude using MTI rather than color discrimination. Thus, the downlink data rate is reduced dramatically along with the probability of false target occurrence. Therefore, the MTI technique is a promising scheme for rejecting inertially stationary objects.
An optical background suppression technique is described in which the lower spatial frequencies of a background field are effectively suppressed, thereby enhancing the detectability of point-like targets. The technique is particularly applicable for use in image-forming double-beam interferometers in which the individual detector elements are matched to the diffraction-limited resolution of the instrument. The technique utilizes pairs of optical systems whose modulation transfer functions are suitably matched and a number of these pairs are described.
Current earth radiance models in the 14-16 micrometer band assume that the temperature of stratospheric CO2 alone controls horizon sensor output and that radiance decreases monotonically from the hot pole to the cold pole. Data are presented which demonstrate the existence of unpredicted variable output from horizon sensors as the sensor traverses the earth from pole to pole. It is concluded that horizon sensor output is influenced by phenomena not included in current models. The primary contributors to variable output are believed to be high-altitude clouds and/or high-altitude jet streams. It is hypothesized that a systematic study of actual horizon sensor output could ultimately lead to improved understanding of the lower stratosphere. Such an understanding might provide an early warning system for active weather fronts as well as an improved interpretation of horizon sensor output with resultant reduction in attitude uncertainty for satellites dependent on horizon sensor output.
The modulation required for the detection of sources on the AFGL infrared sky survey is produced by scanning the infrared telescope and bandlimiting the frequency response of the sensor system,a technique which attenuates the low frequency contents of the extended sources. The survey data is being processed in order to obtain information on these extended sources. The high pass filtering of the system is rectified by applying its inverse to the data stream. The details of this procedure, the uncertainties along with some results are presented.
The Multispectral Measurements Program (MSMP) is being conducted by the Air Force Geophysics Laboratory (AFGL) for the Space and Missile Systems Organization (SAMSO). The MSMP will acquire spectral, spatial and total radiant intensity data on the high-altitude (<150 km) plume characteristics of low-thrust rocket engines (315 lbs. and 1150 lbs. ) in the long-wavelength infrared (LWIR), medium-wavelength infrared (MWIR), short-wavelength (SWIR), vacuum ultra-violet (VUV), and ultraviolet (UV) regions of the optical spectrum. The experimental technique consists of launching both a target engine and a highly instrumented sensor module on a single launch vehicle. Upon separation, an X-Band tracker, in conjunction with an attitude control system in the sensor module, is used to acquire and track an X-Band beacon located on the target engine module to point the fixed sensors. A series of six flights are planned - three test flights from the White Sands Missile Range (WSMR) in New Mexico in which the payloads will be recovered, and three high velocity (6 km/sec) flights from Vandenburg Air Force Base (VAFB), California. This paper describes the results of the first systems test flight from WSMR on 10 November, 1977 and the present status of the program.
We describe the operation and performance of two Schottky IRCCD staring sensors. Both sensors are mono-lithic and are fabricated from conventional integrated circuit grade silicon. The devices include a 256 element linear array and a 25X50 element area array of PtSi Schottky photodiodes which are sensitive from 1.1 to 4.6 μm. The 1250 element mosaic incorporates an interline transfer architecture. Signal readout for both devices is via a 4-phase buried channel CCD network and an on-chip amplifier. The focal plane may be operated between 20°K and 100°K; optimum performance is observed between 50°K and 90°K. We present examples of human thermal imaging against ambient backgrounds as well as photoresponse, thermal transfer, and resolution characteristics. Measured results are compared to theoretical predictions of performance and future developments for this class of sensor are projected.
Lincoln Laboratory has developed and is operating a Ground Electro-Optical Deep Space Surveillance (GEODSS) Experimental Test System (ETS) for detecting and measuring the position of artificial satellites'. Based on this work a GEODSS system is being deployed worldwide by TRW under contract to the Air Force. Described is a GaAs single cell radiometer which is based on an operating radiometer at the ETS. The calculated performance of the radiometer is compared with that expected using CCD's and intensified CCD's. The CCD promises improved performance over the GaAs design under bright sky conditions and for faint objects against a dark sky but at the expense of more complexity in data recording and reduction. The intensified CCD improves the performance of the CCD in the proposed system only if the CCD has an electronic readout noise in excess of 30 electrons and then only under dark sky conditions.
In many sensor systems, simultaneous sightings with two independent focal planes are required. A two-color experiment is an example. When the sensor has a mosaic focal plane with thousands or millions of detectors, precise alignment between each pair of corresponding detectors is impossible. The individual chips on each focal plane will be offset from the desired position. These offsets can be in two linear directions, a rotation, or a combination of these. The effect of these offsets depends both on the distribution of the energy in the observed scene and on the parameters of the sensor. In this paper, these effects are analyzed. A specific example is shown to demonstrate the effects and method of evaluation. No discussion is included about how to align the focal planes, how to measure the boresight errors, or the sensor rigidity requirements.
Performance in area step-stare detection is measured primarily by the sensitivity and revisit time. The various steps of the detection process are modeled by simple equations, resulting in approximations for the performance figures of a sensor as a function of the design parameters. A second more difficult problem is to arrive at the basic sensor design parameters given the performance figures of detectable target size and required revisit time. The procedure outlined for doing this is to consider the footprint size as a parametric variable. A family of sensor designs described by simple formulas is the result. A trade-off study can then be used to arrive at the optimum footprint size and optimum design. This process must be repeated for each of the candidate wavebands with a final trade-off selection between these optimum designs.
In addition to the one-dimensional, narrowband infrared data generated by the DARPA Background Mea-surements Program (BMP), collateral two-dimensional visible light data were also obtained. System designs require two-dimensional background models, but in the absence of the requisite data, isotropy has been assumed to exist in the backgrounds, so that the one-dimensional BMP infrared data can be utilized directly. We test the validity of this critical assumption for a variety of background types. The scenes along the aircraft flight paths in several visible light data frames have been scanned and digitized for comparison with the infrared radiance scans. To the extent that the visible light and infrared data correlate along one dimension, the degree of anisotropy measured in the visible light data is inferred to apply to the infrared data also. Sample scenes exhibiting two-dimensional anisotropy are presented.
The Defense Meteorological Satellite Program satellites carry two IR sensing instruments for collecting information of meteorological significance on a daily basis. One of these instruments, designated the Primary Sensor, is a high resolution wide-angle scanning radiometer which provides both visual ("albedo") and infrared (temperature) information to the Air Force Global Weather Central data base. The second is an infrared sounder which measures the IR energy in sixteen narrow spectral bands within a well defined field-of-view for discrete time intervals. The data from these sensors are computer processed to extract meteorological information such as cloud cover, cloud height, vertical temperature profiles, vertical relative humidity profiles and atmospheric ozone content measurements. This paper describes some of the infrared technology incorporated in the Thermal Imager and how it is combined into the Primary Sensor system which routinely collects the vast amounts of information which form a significant portion of the military global weather data base.
An image plane scanner concept has been evaluated for application to a medium resolution visible and infrared multispectral imaging instrument (MRVIR) that was originally planned to be flown on the french Earth Observation satellite SPOT-1. The instrument design in-cludes several distinctive features, in particular (1) an all-reflective Schmidt Telescope having 17° fov, (2) a 18 - objectives scan wheel supported on a magnetic bearing, (3) dichroic beam splitters, (4) cooled focal plane array for IR detectors, (5) CFRP structure , (6) semi-active thermal control. MRVIR is designed to provide 70 m ground resolution in bands 1-6 and 140 m in band 7 (thermal infrared). A relatively large swath width of 240 kms improves the repetition rate of the ground coverage.
Cooled infrared detective assemblies are usually packaged in a high-vacuum dewar which provides good thermal insulation and reduces cooling power input requirements. This paper describes the development and evaluation of a nonevacuated infrared detector package which may be an attractive alternate in some system. applications. The nonevacuated package is insulated with polyurethane foam and where mounting space permits can often directly replace the more expensive conventional vacuum package. The paper also describes a generic thermal mathematical model which has been developed as a design tool for these non-evacuated packages.
Substantial progress has been made toward the development and application of fast instrumentation for multispectral plume radiation measurements in high altitude simulation facilities. Fast instrumentation permits short duration testing which in turn reduces the heat load and minimizes the chamber recovery time without the addition of costly pumps. Required measurements include the spatial and spectral distribution of plume radiation from the vacuum UV through the IR. Although several instruments are still in the development stage, at this time we are reporting on the successful application of four instruments: an electronographic camera was used to obtain vacuum UV plume images for the first time; mid UV spectral emission data were obtained with a 1/8-meter Ebert-Fastie spectrometer; a thermal imaging camera provided IR spatial images of the plume in four different spectral bands: and a rapid scan Michelson interferometer acquired IR emission spectra. All these measurements were accomplished within a 1-2 sec engine firing time.
Criteria and laboratory techniques used to measure the performance of integrating image sensors are described. The integrating image sensor has a sensing layer that continually monitors the field of view and an electronic mechanism that sequentially reads out the integrated signal on an elemental basis. Test procedures are given for determining transfer characteristics, spatial response, spectral response, response uniformity, and image retention of these integrating image sensors.
Landsat-3 carries the first Multispectral Scanner (MSS) to include sensing of a fifth spectral band for the thermal emission of the earth scene in addition to the first four bands, which sense reflected solar irradiance. Unique design features of this band are described: a thermal reference level provided by the detector viewing its cold surroundings in a mirror, and an incremental gain adjustment used to maintain dynamic range without penalizing the signal-to-noise ratio. Ground calibration of the thermal band was performed using a large-area blackbody source. During that testing the detector and instrument temperatures were varied. The body of data thus generated was analyzed using calibration algorithms which were derived by a weighted least-squares technique. The resulting values were then applied to the remainder of the thermal. vacuum test data as verification of the adequacy of the calibration. The data obtained from space would be used to identify the appropriate function to find signal radiance from scanner signal in a manner shown.
The calibration of a cryogenic infrared spectrometer is described. The calibration data are obtained in two facilities that simulate the vacuum and cold background of the upper atmosphere in an attempt to duplicate the conditions under which the measurements are to be made. These facilities provide for a distant small-area source and an extended-area source calibration. The sensor has a free spectral range of 7.3 to 22.7 µm, 2-3% resolution, a noise equivalent sterance (radiance) of about 4.5 x 10-11 W cm-2sr-1 µm-1 at 15 µm and obtains 2-scans/s. The measurement equation is derived from the calibration data as L(λ) = C(λ)[3.36 x 10-9 V + 1.1 x 10-12 V2] W cm-2 sr-1 µm-1 where C(λ) varies from unity upward, and the precision and accuracy are estimated at ± 7 to 12% and 13 to 16% (depending upon the wavelength) respectively.
An AGA Thermovision® System 680 was used in studying the IR signature from an axisymmetric nozzle after-body model at transonic Mach numbers (M = 0.0, 0.6, 0.9, 1.3). Radiance measurements from the exhaust plume in the CO2 emission band were observed through a narrow spectral filter (4.21-4.41 µm). Results are presented in the form of axial and radial plume radiance profiles. Such spatially resolved plume radiance measurements serve as a flow visualization to reveal shock structure and as a baseline for plume radiation model prediction.
A method of using the SIMS (the selective modulation interferometric spectrometer) to measure the difference between the spectral content of two optical beams is given. The differencing is done optically; that is, the modulated detector signal is directly proportional to the difference between the two spectra being compared. This optical differencing minimizes the dynamic-range requirements of the electronics and requires only a simple modification of the basic cyclic SIMS spectrometer. This technique can be used to suppress background radiation for the enhancement of target detection and tracking. Laboratory measurements demonstrating the application of this technique are reported.
This paper describes the design and performance of a serial scan thermal imaging system with emphasis on the design of scanners. The sensor with a single-element HgCdTe detector employs a oscillating mirror and a truncated hexagonal pyramid-type mirror as vertical and horizontal scanners, respectively. The size of the scanners was optimized to be as compact as possible by disposing each axis of scanners at the real and horizontally scanned position of exit pupil of the afocal system respectively. This condition also reduces the amount of migration of the scanned parallel beam, which has made easy to design a diffraction limited focusing lens. The image distortion and timing error of horizontal sweep caused by the rotating scanner are compensated by signal processing techniques which have been proven to give satisfactory system performances. The system has the following parameters: field of view; 40 x 53 mrad, angular resolution; 0.4 mrad, diameter of afocal telescope; 100 mm, frame rate; 20 sec -1, and 121 TV lines with a 2 to 1 interlace. The results of system performance test have shown that NETD is 0.6°C and MRTD is 0.6°C at 1 cy/mrad which are close to the design values.
A technique for synthetically extending the upper limit of the dynamic range of sampled IR sensor systems was developed. The method utilizes the secondary overshoot, commonly found in the response of tuned amplifiers, to estimate the magnitude of the primary overshoot. This "Second Overshoot Technique" may be used in the saturated and unsaturated operating regions of the amplifiers. In both operating regions, it is possible to estimate the magnitude of an amplifier's primary overshoot by careful observation of the magnitude and relative occurrence time of the secondary overshoot. The net effect is an artificial extension of the top end of the sensor's dynamic range. This technique can be useful in either planned or existing systems as a means of extending the dynamic range. This extension method is presently in use on an orbiting satellite sensor system.
A method for calculating the probability of detection of generalized adaptive threshold systems is applied to sensors employing pulsewidth-adaptive-threshold signal processing concepts. These sensors typically use the signal as input to a guard channel that controls the threshold against which the signal is itself com-pared. The variation of probability of detection with signal-to-noise ratio and with the form of the input signal plus clutter is presented. For the specific sensor analyzed, in which Gaussian statistics hold, the probability of detection curves branch off the cumulative normal distribution curves as a function of guard channel gain. There is also a range of guard channel gains for which the detection performance of pulse-width adaptive sensors will suffer essentially zero insertion loss due to the adaptive threshold itself.