Real-time deformation invariant optical pattern recognition using coordinate transformations

Area of Science:

• Optics and Photonics
• Computer Vision
• Image Processing

Background:

• Pattern recognition often struggles with in-plane distortions like scale and rotation.
• Traditional methods require complex algorithms or hardware for distortion invariance.

Purpose of the Study:

• To develop a distortion-invariant pattern recognition system using a polar-logarithmic coordinate transformation.
• To enable real-time processing of transformed patterns for optical correlators.

Main Methods:

• A computer-generated hologram, printed via laser, performs the polar-logarithmic coordinate transformation.
• Weighting terms in the output are analyzed for their impact on resolution and pixel data.
• A pocket liquid crystal TV interfaces the transformed pattern to a correlator for real-time analysis.

Main Results:

• The system successfully achieves in-plane distortion invariance.
• Analysis of weighting terms demonstrates their effect on output resolution and data reduction.
• Experimental results validate the real-time processing capability.

Conclusions:

• The polar-logarithmic coordinate transformation is effective for distortion-invariant pattern recognition.
• Computer-generated holograms and liquid crystal TVs provide a viable pathway for real-time optical pattern recognition.