The Smarter Vision demo taking place this week in the Xilinx booth at the SPS IPC Drives conference in Nuremberg, Germany illustrates the kind of end-to-end image processing and video analytics needed to create modern machine-vision applications. The demo pairs SiliconSoftware’s VisualApplets with MVTec’s HALCON, with both packages running on a Xilinx Zynq All Programmable SoC. The VisualApplets package provides real-time image-processing algorithms and high-level design entry and simulation tools. VisualApplets run on the Zynq SoC’s programmable logic. HALCON provides higher-level functions including image matching, blob analysis, morphology analysis, dimensional measurements, OCR, and bar-code decoding and runs on the Zynq SoC’s ARM Cortex-A9 MPCore processor. Together, these two packages supply a complete, end-to-end, vision-processing software chain for creating differentiated and intelligent machine-vision products.
Real-time image processing is required, for example, to compensate for the real-world lenses used in machine vision. The long list of lens aberrations includes spherical aberration, astigmatism, curvature of field, coma, and chromatic aberration. Cheaper lenses, preferred in cost-critical designs, tend to have more aberrations. The exciting news is that advanced processing software like SiliconSoftware’s VisualApplets can compensate for imperfect lenses and can do so in real time thanks to the processing power of programmable logic. Real-time image processing also performs other sorts of image manipulation such as color correction, vignetting correction, contrast adjustment, etc. SiliconSoftware is no newcomer to the use of FPGAs for real-time image processing—the company’s VisualApplets package has been available for several years now and has run on several Xilinx FPGA generations. This week’s demo using the Zynq SoC is merely the latest demonstration.
According to the “Machine Vision 4 Users” blog post “What can you do with HALCON?”, “pretty much any machine vision task can be done with HALCON from MVTec.” The package wraps an IDE around a large library of more than 1800 operators for low-level, mid-level, and high-level image processing in 1D (like bar codes), 2D, and 3D applications. There’s a book with example code based on HALCON available: “Machine Vision Algorithms and Applications.” There’s also a free downloadable version of HALCON on the MVTec Web site.
I expect to get video of the SPS IPC Drives conference demo soon. Until then, here’s a short 2-minute video from Stemmer Imaging demonstrating SiliconSoftware’s VisualApplets running on Xilinx-based, pre-Zynq hardware, recognizing objects in 0.02 msec. Now that’s industrial-class real time!
And here’s an even shorter 1-minute video with some details about the latest version of MVTec’s HALCON: