Three-dimensional and two-dimensional sequence spatial-frequency domain processing of speckled ladar images for automatic target recognition

摘要

Abstract: Direct-detection laser radars can measure the range and the intensity returns from a target, with or without clutter, for each part of the target resolved in angle by the optical system. Because the ladar's angular resolution is in micro- radians, there are generally at least a few angular pixels 'on target.' In addition, for narrow pulse (approximately 1 ns) ladar systems, there may be ten or so sequential intensity measurements in range per pixel as the laser pulse propagates down the target's surface. The output image is, therefore, potentially a three dimensional 'cube' of intensity measurements and quantized in the range axis by the range-bin size or 'voxel' size. This is known as 'range resolved angle- angle-intensity' ladar. In a previous paper we transformed this 3D-matrix image into the spatial-frequency domain using 3D-Fourier transforms and followed conventional 2D template correlation techniques to perform target recognition and identification. During this previous study, it was noted that the 2D range-bins could be placed in sequence and 2D filtering used on these synthetic images. Results of 3D and 2D-sequence target correlators using the 'joint transform correlator,' 'the inverse filter,' the 'phase-only matched-filter,' the 'binary phase-only filter,' and the classical 'matched filter' are presented here. Far-field test data using conical shaped targets are used to study the 3D and 2D correlators, and the effects of laser speckle are discussed. Recent developments in negative-binomial driven shot-noise effects in range-resolved direct-detection ladar are reviewed as well. These 3D or 2D- sequence template correlators may supplement or refine less computationally intensive algorithms such as total signal; range-extent; x-z, y-z, and x-y plane image centroid estimation; and image moments. !26