Experimental use of iteratively designed rotation invariant correlation filters

Area of Science:

• Optical Engineering
• Image Processing
• Pattern Recognition

Background:

• Traditional optical correlators struggle with variations in target position, rotation, and intensity.
• Developing filters with comprehensive target information is crucial for robust recognition.

Purpose of the Study:

• To incorporate iteratively designed filters into an optical correlator for invariant target recognition.
• To achieve high discrimination capability for target images under challenging conditions.

Main Methods:

• Iterative filter design incorporating full target image information.
• Encoding complex-valued filters into computer-generated holograms (CGHs).
• Fabrication of CGHs using electron-beam lithography.
• Experimental validation using a liquid crystal spatial light modulator (SLM).

Main Results:

• Demonstrated position, rotation, and intensity invariant recognition.
• Successful detection of targets with low signal-to-noise ratios (SNR ≈ 0.5).
• Robust recognition of partially obscured targets.
• Experimental results closely matched numerical simulations and analytical predictions.

Conclusions:

• Iteratively designed filters offer excellent discrimination for invariant optical recognition.
• The developed system is effective even in the presence of significant noise and occlusion.
• Electron-beam fabricated CGHs and SLM-based systems provide a viable platform for advanced optical correlators.