Neural single-shot GHz FMCW correlation imaging.

Area of Science:

• Optics and Photonics
• Computer Vision
• Robotics and Autonomous Systems

Background:

• Depth sensing is critical for 3D environmental perception in applications like autonomous driving and AR/VR.
• Traditional time-of-flight (ToF) methods suffer from phase wrapping artifacts, limiting their effective depth range.
• Multi-frequency ToF methods require multiple measurements, increasing hardware complexity and imaging time.

Purpose of the Study:

• To develop a novel 3D imaging method for depth sensing.
• To overcome the limitations of existing ToF techniques, specifically phase wrapping artifacts and measurement time.
• To enable high-resolution depth mapping with a single per-point measurement.

Main Methods:

• Integration of frequency-modulated continuous wave (FMCW) operation with all-optical correlation ToF imaging.
• Development of a specialized frequency-decoding neural network for enhanced data processing.
• System designed for all-optical correlation imaging at GHz rates for rapid data acquisition.

Main Results:

• The proposed method successfully performs 3D imaging with a single per-point measurement.
• Achieved high-speed all-optical correlation imaging at GHz rates.
• Validation through simulations and real-world experiments demonstrated favorable comparisons with existing methods.

Conclusions:

• The novel FMCW-based all-optical correlation ToF imaging method offers a significant advancement in depth sensing.
• The system effectively mitigates phase wrapping artifacts and reduces measurement time.
• This technology holds promise for improved 3D environmental perception in various applications.