Embedded 3-D machine-vision system guides robotic cutting of beef carcasses
The stereo vision system used in this industrial-sized nail clipper is pretty cool. The manufacturer describes it this way:
TYZX’ DeepSea G2 Stereo Vision System is a small embedded camera with low power requirements that delivers high performance visual analysis (up to 60 fps) in variable lighting conditions. Designed for widespread deployment in applications such as security and navigation, the G2 provides real-time 3D vision processing without the use of workstations…
The DeepSea G2 system is based around Xilinx FPGAs that perform image rectification, 2D and 3D quantization and other compute-intensive tasks. In a technical white paper, the system designers describe how the FPGA is combined with an ADI Blackfin processor and a custom ASIC to strike the right balance between performance and flexibility. One interesting tidbit is found in the following table, where the first and second generation products are described:
Notice that the Pentium III processor that did much of the heavy lifting in their first-generation product completely disappears in the second generation. This example points again to the need for tools and design methods that support partitioning between many different types of processing elements.