Vision systems are for small and complex medical parts such as these dental implants.

Accurately validating the dimensional specifications of medical parts is a crucial aspect of design and product development. Often, however, medical parts can be too difficult to measure with conventional tactile measurement systems. Part features may be too small for, or out of the reach of, hard touch probes. Soft and pliable materials can deform from contact measurement. Contact probes might scratch highly finished orthopedic products.

Fortunately, vision and other non-contact probing technologies let users rapidly collect vast amounts of dimensional information for design analyses and subsequent part validation. The technologies use powerful CAD-based programs that let measurement systems operate in 2D, 2.5D, and 3D modes while collecting data that can be used both to validate dimensions and analyze designs and manufacturing processes.

These advancements were hard-won as vision-system software developers spent decades creating proprietary algorithms for accurately capturing images and transforming them into discrete points of data that could be automatically compared to nominals in the CAD model — and that was just the beginning. Current refinements to the software used with vision and multisensor metrology systems should be of great interest to medical-device developers.

Algorithms augment optics

A perceived barrier to using vision and multisensor measurement systems as primary equipment for medical-parts metrology has been the idea that adjusting the systems for appropriate lighting, contrast, and edge-detection sensitivity required too much specialized knowledge for the average user in the metrology lab. This may have been true at one time, but it is no longer the case. Many powerful new algorithms in vision-measurement software effectively automate these adjustments, allowing for consistent measurements from part to part and vision machine to vision machine.

Advanced software, including feature-based measurement with target splitting, lets users easily select areas of interest.

The subjectivity of making manual adjustments to optimize contrast is a legitimate concern. Optimizing contrast substantially improves measurement accuracy by improving the vision system's capability to detect edges and compensating for the tendency of light to bend around the edges of cylindrical surfaces, thereby shortening measured distances. Today, special algorithms automate the adjustment of contrast levels. At the touch of a button, the algorithm makes a series of rapid iterative adjustments until an optimal and stable level of contrast is gained.

Another potential source of vision measurement variability is caused by the differences in light sources (halogen or LED) used to illuminate the parts and the ambient lighting conditions in different plant locations. It is easy to correct for these variations when the measurement software includes algorithms that let user compensate for these factors just as they would calibrate a probe on a CMM.

Unlike tactile probes, cameras do not touch the edge they measure. Therefore, edge detection must rely on the accurate interpretation of data the vision software receives from the camera. Fortunately, advanced vision measurement software can tune its edge-detection algorithms to account for both the part's surface and the illumination conditions.

Generally, the software uses a dominant edge algorithm to select the edge of a part when using sub-stage illumination. Measuring top-lit parts with a high surface finish is more of a problem. In this case, users can select a specific edge algorithm, which allows the detection of the feature in question based not only on contrast but also its shape and location. If there are grind marks on the part, which might confuse the camera when top lighting is used, the software can apply another type of algorithm that chooses the most dominant edge out of possible candidates in the camera's field of view.

Continue to next page