No abstract

No abstract

The Air Force program, Smart Tactical Autonomous Guidance (STAG), has as its central concept, the use of passive millimeter wave imagery (PMMW) to enable an autonomous vehicle to perform its own smart guidance and attack. The algorithms on board the vehicle use image flow to derive the necessary range information for obtaining real time navigation updates. The results of a natural-imagery feasibility program will be reported, aimed at validating the STAG approach. PMMW imagery will be taken from land-based vantage points that mimic the geometry of an airbome, down looking, sensor. Essentially, hardware-in-the-loop simulation will be performed, where PMMW data will be real, and the ioop extends over many miles ofoutdoor terrain. The only departure from an actual mission will be that it is not real-time. All imagery will be gathered using frame times consistent with existing camera capabilities. Image flow and other data processing will be done off-line. The key ingredients will be the sequences of imagery and the computer processing ofthat imagery. The means for accomplishing both have been developed under the STAG program. The camera to be used operates at W-band and consists of an f/i , refractive, telecentric, image forming system with a 30 cm diameter input aperture. Image flow involves a model whose parameters are determined via automated pixel tracking from frame to frame. Passive range maps are then generated, and navigation is accomplished through the subsequent correlation ofthese maps with reference elevation maps. Automatic target recognition is also addressed.

Passive millimeter wave (PMMW) imaging systems have many important applications, both military and commercial. However, due to the longer wavelength, resolution is limited when compared with shorter wavelength imaging systems of comparable aperture size. One approach to this problem is super resolution. Over sampling in the focal plane supports super resolution techniques that utilize maximum likelihood and constrained least squares methods, while preserving the field of view. In order to test super resolution algorithms with real image data, a versatile PMMW test bed constructed for the Air Force Smart Tactical Autonomous Guidance (STAG) program was utilized to obtain 16x over sampled images of three test targets. First, a Gunn diode oscillator (GDO) source was imaged in order to accurately measure the point spread function (PSF) of the imaging system. A special source pattern was then imaged to measure the system's response to differently oriented step functions, and to determine the system time constant. An M48 tank was then imaged as an example of a real world military target. In each case the experimental images were obtained by mechanically scanning a single TRF receiver module in a two-dimensional raster pattern in the focal plane of a 94 GHz, f/i, two element, refractive, telecentric imaging system. Imagery was obtained at horizontal scan velocities ranging from 1.27 cmlsec to 12.7 cmlsec with 1270 or 1 344 samples per horizontal scan line, and a horizontal sample spacing of 0. 1 mm. For comparison, Nyquist sampling was achieved with a 64x84 image size and 1 .6 mm sample spacing. In order to achieve the large amount of over sampling, and be able to handle the resulting large quantities of image data, hardware and software modifications had to be made to the existing STAG test bed. These changes will be reported along with the sampling details and results for all three targets.

For objects of "large" vibration size such as waves on the sea surface, the choice of measurement method can create different understandings of system behavior. In one case, laser vibrometry measurements of a vibrating bar in a controlled laboratory setting, variation in probe spot size can omit or uncover crucial structural vibration mode coupling data. In another case, a finite element simulation of laser vibrometry measures a nonlinearly clattering armor plate system of a ground vehicle. The simulation shows that sensing the system dynamics simultaneously over the entire structure reveals more vibration data than point measurements using a small diameter laser beam spot, regardless of the variation of footprint (coverage) boundaries. Furthermore, a simulation method described herein allows calculation of transition probabilities between modes (change-of-state). Wideband results of both cases demonstrate the 1/f trend explained within-that the energy of discrete structural vibration modes tends to decrease with increasing mode number (and frequency), and why. These results quantify the use of less expensive non-imaging classification systems for vehicle identification using the remote sensing of surface vibrations while mitigating spectral response distortion due to coverage variation on the order of the structural wavelength (spectral reduction or elimination).