Optical pattern recognition with a binary phase-only filter for image tracking

摘要

The use of binary phase information results in a degradation of performance due to the coarseness of the quantisation of the phase data. Furthermore, binary data results in two correlation peaks being formed in the output plane of the Fourier Transforming lens.Since it is often required to track a spot on a moving sought object present in the scene. It is convenient to make this spot the centre of the reference image (used to create the reference pattern) for each scale/orientation of the reference object sought. In this case when the spot on the reference object is in the centre of the input plane, the output correlation peak is at the centre of the output plane. Thus, in the tracking process the objective is to keep the correlation peak at the centre of the output plane when the desired region of the scene object is always centrally aligned with a tracking device (e.g. a camera or drill). In the case of the pattern recognition system of the present invention employing binary phase SLMs, this has the added advantage that in these circumstances the output peak has full intensity, the two symmetrical peaks discussed above merging into one strong central peak. This fact can be used as an aid to reference object identification and tracking and exploits the fact that the output peak is not split when the location of the scene reference object matches that of the reference perfectly. Of course, implicit in this statement is the fact that the reference object is identified in the processor.