David L. Flannery

Design elements of binary joint transform correlation and selected optimization techniques

A new power spectrum binarization technique is introduced that eases experimental implementation while significantly improving binary joint transform correlation performance. Impacts of fundamental design parameters on correlator performance have been evaluated using ncomputer simulations. Realistic images including multiple objects in challenging backgrounds were used to characterize the effects of parameter variations such as threshold levels and low-frequency blocks. The design aspects emphasized were those contributing to ease of experimental implementations, including practical dynamic range limitations and achievable threshold levels. Experimental results using the optimization techniques developed with simulations are given.

Optimal trade-off distortion-tolerant constrained-modulation correlation filters

A closed-form filter formulation that builds on the optimal trade-off filter formulation and uses the minimum-Euclidean-distance principle is presented. This new formulation provides filters that are optimal in a mean-square error sense relative to multiple criteria over a training set of images when they are implemented with limited modulation levels typically afforded by spatial light modulators. The relationship of the new formulation to several existing filter concepts is discussed.

Transform -ratio ternary phase -amplitude filter formulations for character recognition

Five types of discrete-valued correlation filters, including binary phase-only and ternary phase-amplitude types, were tested in simulations addressing specific character recognition problems. The filters were evaluated for classification accuracy and correlation efficiency. Transform-ratio filter ternary phase-amplitude filters provided the best classification performance but also the lowest correlation efficiency.

Experimental investigation of transform ratio ternary phase-amplitude filters for improved correlation discrimination

Transform ratio ternary phase-amplitude filters (TR-TPAFs) encodingnthe modulation states 1, 0, — 1 have been investigated in theorynand practice. Simulations have demonstrated that TR-TPAFs can be formulatednto provide increased discrimination compared with binarynphase-only filters. TPAFs have practical advantages over complex-valuednmatched filters—efficient electronic filter storage and real-time implementationnwith available devices, such as magneto-optic spatial light modulatorsn(MOSLMs).These are associated with the use of only three discretenmodulation levels. In an experimental study, significant increases in correlationndiscrimination were demonstrated for several test patterns, showingnsubstantial agreement with computer simulations. The effects of annimperfect MOSLM zero-modulation state, present in our experiment, werenmodeled and investigated. The performance of the TR-TPAFs for differentnthreshold line angles also was investigated.

TOPS pattern recognition algorithms and their extension to the IR

The TOPS pattern recognition program produced a successful demonstration of optical correlation applied to automatic target recognition (ATR) using visible (video) images. The TOPS correlator hardware will soon be tested performing ATR using IR images. In both cases, effective algorithms must be defined and implemented in three functional areas: (1) distortion-tolerant correlation filter design, (2) input image preprocessing, and (3) filter management strategy. The algorithms implemented for TOPS are reviewed and the modifications required for IR imagery are discussed.

Neural networks with optical-correlation inputs for recognizing rotated targets

Backpropagation-trained neural networks with optical correlation inputs are used to predict target rotation and to synthesize simplified optical correlation filters for rotated targets.

Basic design elements of binary joint transform correlation and selected optimization techniques

The performance of binary joint-transform correlation using realistic input scenes has been studied by simulations addressing the effects of variations in threshold level, low-frequency blocks, and spurious signals due to regularly spaced groups of multiple (identical) input targets. A design tradeoff between better correlation performance and easier implementation (higher thresholding levels) was observed for constant thresholding. A new adaptive thresholding technique is introduced which alleviates the problems encountered using constant thresholds and significantly improves performance. Results are given for input scenes consisting of challenging backgrounds with multiple targets.

Performance of mixed-metric optimized ternary correlation filters on realistic binary and gray-scale imagery

The ternary phase-amplitude filter (TPAF) is by definition restricted to the modulation values -1, 0, and 1, thus comprising a binary phase-only filter (BPOF) multiplied by a binary- amplitude pattern, i.e., a region of support. The TPAF offers an attractive combination of real-time implementation with available devices and good correlation performance. Smart (optimized distortion-invariant) TPAF formulations have been developed. The TPAF enables filter implementation with magneto-optic devices and these devices also can be used for image input if gray scale scenes can be binarized while preserving good correlation performance. We provide simulation results addressing the comparative performance of mixed-metric smart TPAFs using gray scale, edge-enhanced and binary images derived from identical original scenes. The variation of filter performance with training set background intensity level is examined.

Model-based ternary phase-amplitude feature correlation filters

Ternary phase-amplitude filters (encoding only -1, 0, and +1 modulation values) have been designed to respond to geometric features oftargetobjectsin scenes, forexample thewhecls on atruck. The filtersarebasedongenenc models involving only a few parameters, whose adjustment allows the model to be matched to different instances ofthe feature (e.g., different size wheels). The formulation ofthe model-based filtersand theircorrelation performance usingrealistic images ofvehicles in clutter backgrounds willbepresented. Candidatepost-processing approaches to exploitthe correlation results fortargetrecognition will be discussed.