How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
The complexities of CPUs, says MathLab’s Tannenbaum, are making it harder to write imaging software that makes optimal use of the architecture. “The engineers and scientists who have been willing to write their own software in the past are going to be less likely to do so in the future. The learning curve to write efficient, multi-threaded software will increasingly seem less worthwhile when software vendors provide solutions with substantial performance advantages.”
MathLab’s latest products, including its Parallel Computing Toolbox, add parallel programming constructs for explicit parallelization of its code. The company’s Image Processing Toolbox recently added capabilities for efficient display and navigation of large images as well as multi-threading in specific functions.
Silicon Software’s Noffz sees the role of software becoming more important in deciding on a system solution. “The interface will become more intuitive,” he says. Self-learning methods for recognition tasks and automatic adaptations to varying temporal or spatial situations will let the software react anticipatorily. Even complex software for robot vision applications will be affected.
“Parallel imaging processing also will become very important,” he adds. “Parts of the image processing are swapped to the most efficient processor architecture, like CPU, FPGA or GPU, if available. Different processor technologies will be seamlessly combined to achieve the highest performance for the inspection tasks.”
Silicon Software’s latest product is VisualApplets (Version 1.3), a graphical programming tool for FPGA-based image processing. Its new features include a blob analysis library for 1D and 2D images and a high-speed compression library that supports the JPEG format.