A Time-Sequenced Rotation Invariant Optical Correlator For Multiple Target Recognition

Statistical filtering of time-sequenced peak correlation responses for distortion-invariant recognition of multiple input objects

“…First order simply means that only data from the present input image are being used, as described in Eqs. (4) and (5). The filter collapses to first order when the new input image is different from the previous input image.…”

“…The DICPF consisting of a single distortion element, namely rotation, along with the target group in the distortion class has been shown in both simulation and experiment to successfully recognize multiple objects in a binary input image in the presence of both noise and clutter. 4 A vector of peak correlation plane irradiance values, one element for each filtering operation, resulting from the time sequencing process are input into the DICPF as shown in Fig. 1.…”

“…Mathematically, this is given as i=1,2 M-l . (4) A statistical hypothesis test, described in the next section, is used to detect a changing input and determine the a, coefficients. Following this test, the correlation plane vectors are concatenated, yielding the resultant data vector y, which is input to the DICPF:…”