Hybrid optoelectronic correlator architecture for shift-invariant target recognition

Area of Science:

• Optoelectronics
• Image Processing
• Computer Vision

Background:

• Conventional holographic correlators (CHCs) are effective for image recognition but can be complex.
• Existing optoelectronic correlators often rely on nonlinear materials, limiting performance.
• There is a need for high-speed, translation-invariant image recognition systems.

Purpose of the Study:

• To present the theoretical details and architecture of a novel hybrid optoelectronic correlator (HOC).
• To demonstrate a method for eliminating convolution terms in HOC output.
• To propose an integrated graphics processing unit for enhanced processing speed.

Main Methods:

• Utilizing spatial light modulators (SLMs), detector arrays, and field-programmable gate arrays (FPGAs).
• Employing plane wave interference to conserve phase information without nonlinear materials.
• Implementing a phase stabilization and scanning circuit to remove unwanted convolution signals.

Main Results:

• The proposed HOC architecture successfully correlates images using detectors, conserving phase information.
• Phase stabilization effectively eliminates convolution terms, making HOC behavior similar to CHC.
• Integration of a graphics processing unit enables parallel electrical processing for high-speed operation.

Conclusions:

• The developed HOC, with phase stabilization, offers a viable alternative to CHCs for high-speed, translation-invariant image recognition.
• This architecture bypasses the need for nonlinear materials, simplifying the system.
• The integrated GPU approach promises ultimate correlator speed.