Vector disparity sensor with vergence control for active vision systems.

Area of Science:

• Robotics
• Computer Vision
• Hardware Acceleration

Background:

• Active vision systems require efficient disparity computation for dynamic environments.
• Existing static camera methods are limited to 1-D disparity after rectification.
• Binocular systems need generalized disparity computation for vergence control.

Purpose of the Study:

• To develop a real-time architecture for vector disparity computation on FPGA devices.
• To implement and evaluate gradient-based and phase-based vector disparity engines.
• To integrate disparity computation with active vision control modules (vergence, version, tilt).

Main Methods:

• Designed a real-time architecture for vector disparity computation using Field-Programmable Gate Arrays (FPGAs).
• Implemented disparity units and control modules for fixation point determination.
• Developed and compared two on-chip vector disparity engines: gradient-based and phase-based.
• Evaluated multiscale engine performance on VGA resolution images up to 32 fps using benchmark sequences.

Main Results:

• Achieved high accuracy in vector disparity computation around the fixation point.
• Demonstrated real-time performance (up to 32 fps) for VGA resolution images.
• The gradient-based approach showed superior performance and accuracy compared to the phase-based method.
• FPGA implementation provided efficient resource utilization.

Conclusions:

• The developed FPGA architecture enables efficient real-time vector disparity computation for active vision robotics.
• The gradient-based approach is recommended for its optimal trade-off between performance, accuracy, and resource usage.
• This work advances the capabilities of robotic vision systems in dynamic environments.