|
With the advent of large arrays of IR detectors, it is possible to construct quasi-staring sensors whose function is to detect weakly radiating moving targets against an intensely cluttered background. The detection philosophy is based on the fact that as the target passes across a detector's footprint, the irradiance on the detector during the current frame is different than the irradiance during the previous frame when the target was outside the detector's field-of-view. By comparing (i.e., differencing) the detector's output signal from the current frame against the output from the previous frame, the presence of the moving target can be recognized while the effects of spatial variation in the background are completely nullified. If for some reason, however, the background scene moves even slightly relative to the detector during a frame, the comparison process is unable to completely reject background clutter-background spatial variations, i.e., clutter, now produce changes in the detector's output, and background clutter leaks through the signal processor. We have investigated the statistics of clutter leakage and options in signal processor design for a satellite-borne sensor whose goal is to detect moving aircraft against the earth background when the sensor has a slowly drifting line-of-sight. We have developed a statistical methodology and corresponding formulas to allow us to evaluate the peak signal-to-rms clutter (and peak signal-to-rms shot noise) for a general sig-nal processor based on digital filtering. Taking advantage of the diffraction limit in the optics transfer function, which insures that there is a (temporal frequency) region in the detector output spectrum in which there is no clutter noise, we found that we could configure a digital filter which let a significant portion of the target spectrum through its passband, but only allowed clutter through by leakage in its side lobes. The achieved signal-to-clutter ratio and the allowable line-of-sight drift velocity were impressively high for this signal processor-much better than for a differencing processor. Details of the signal processor designs and sample quantitative results will be presented, along with trade-off-considerations.
|
