In his opening keynote address of the Vision Summit of this year’s NIWeek in Austin, TX, Robert Tait of GE Global Research (Niskayuna, NY, USA;www.ge.com/research) focused on the subject of design for inspection. “Many of today’s products are designed for ease of assembly and test,” said Tait. “However, it is still difficult for systems designers to build machines that allow finished products to be inspected easily. To do so requires an understanding of the product being manufactured and how inspection tasks can improve the quality control process.
“Early adopters wishing to improve the quality of their products included those in the electronics and printing industries,” said Tait. For example, electronics manufacturers employ fiducial marks located on PCBs as reference marks to locate and identify the position of ICs; the printing industry uses color registration marks to ensure that each product is correctly printed. Embedding these marks onto a product allows the vision system designer to more easily develop systems to inspect the finished product for imperfections.
To build inspection systems for other types of products, however, using CAD to model the complete system can help developers visualize their systems before they are deployed. “Before this can occur,” said Tait, “OEM suppliers must provide CAD models that allow cameras, illumination, and lens systems to be used to develop a virtual system.”
Products such as Rhinoceros from McNeel (Seattle, WA, USA;www.rhino3d.com), Pro/Engineer from PTC (Needham, MA, USA; www.ptc.com), and POV-Ray from Persistence of Vision Raytracer (www.povray.org) all allow limited aspects of this task to be accomplished. For example, Rhinoceros allows the user to model the type of lighting and camera used in the system; Pro/Engineer lets the developer use specific sensor sizes to determine the working distance and angular field of view of the camera.
“Such CAD packages allow machine-vision systems to be incorporated as part of the initial CAD design, not as an afterthought,” said Tait. However, only a few lens suppliers currently provide software to determine which type of lens best fits a particular application. Of those that do, MachVis software from Linos (Göttingen, Germany;www.linos.com) is designed to assist lens selection using object size, working distance, sensor size, and camera mount.
In operation, MachVis calculates the optical parameters and then selects Linos lenses that meet these specifications. Similarly, Gaussoptik software from Schneider Kreuznach (Kreuznach, Germany;www.schneiderkreuznach.com) allows mechanical and electrical engineers to analyze, modify, and specify the optics and structure of their products at any stage of design or manufacturing, without direct assistance from an onsite optical engineer. “Unfortunately, none of these tools provide 3-D CAD models of the lens body,” said Tait, “leaving the developer to manually enter mechanical parameters into CAD packages.rdquo;
Although design for inspection may still be in its infancy, Tait sees a number of opportunities for both manufacturers of finished parts and those that build machine-vision systems to inspect them. First, parts vendors must incorporate identifiable markers such as barcodes, distinct colors, or shapes on their parts so machine-vision systems developers are presented with an easier development task. Providing systems developers with a CAD model of the part to be inspected would also allow the dimensions of the inspected part to be checked against its measured parameters.
Suppliers of OEM components such as lighting, lenses, and cameras could also provide optical models of their lighting products as well as 3-D models of their cameras and lens housings. These could then be incorporated into third-party CAD models. “In summary,” said Tait, “if an OEM supplier can supply me with a CAD model of their products, I am more likely to purchase them!”
