Generic FPGA Pre-Processing Image Library for Industrial Vision Systems

Area of Science:

• Computer Engineering
• Image Processing
• Hardware Acceleration

Background:

• Increasing demand for diverse, high-quality products necessitates advancements in industrial machine vision.
• Growing complexity in industrial systems requires enhanced processing efficiency for vision-based algorithms.
• Hardware acceleration is critical for meeting the demands of future industrial vision systems.

Purpose of the Study:

• To present a generic library of pre-processing filters for Field-Programmable Gate Arrays (FPGAs).
• To reduce overall image processing time in machine vision systems through hardware acceleration.
• To evaluate the performance and industrial applicability of FPGA-based image pre-processing.

Main Methods:

• Developed a generic library of pre-processing filters implemented in VHDL (VHSIC hardware description language).
• Utilized an experimental setup based on the Zybo Z7 Pcam 5C Demo project for filter validation.
• Compared execution times against Graphics Processing Unit (GPU) and Central Processing Unit (CPU) platforms.

Main Results:

• Achieved a significant reduction in image pre-processing time, decreasing it from milliseconds to nanoseconds using FPGAs.
• Demonstrated the effectiveness of the VHDL-based filter library on the experimental hardware.
• Evaluated the integration potential of the developed solution in industrial settings.

Conclusions:

• FPGA-based pre-processing offers substantial speed improvements for machine vision tasks.
• The generic filter library provides a viable solution for accelerating industrial image processing.
• This approach addresses the need for efficient hardware acceleration in evolving industrial systems.