A New Look At Vision Measurement
May 28, 2009
Multisensor measurement systems are capable of collecting geometric data from 3-D parts using various combinations of measurement technologies. For the most part, such systems are based on vision systems made more versatile by the addition of tactile probes as well as lasers, white light and other devices. This allows confident measurement of parts using the most appropriate combination of probes for the job at hand, without compromising accuracy.
There has been a tremendous improvement in the multisensor system’s primary sensor: the camera. Some of this owes to the enhancement of mechanical equipment and optics, but the bulk is the result of advancement in computer-aided design (CAD)-based measurement software.
CAD IntegrationVision became a mainstream measurement technology when equipment and software vendors made the transition from text-based programming and operating environments to CAD-based systems. Parts are designed in CAD, and computer numerical controlled (CNC) tool paths are generated in computer-aided manufacturing (CAM). It only made sense to complete the manufacturing cycle using CAD as well.
Now that many vendors have made the initial transition, they are quickly implementing improvements to their systems. For example, they are adding intelligence to their software that automatically adjusts parameters such as lighting, focus and magnification based on settings in the software and real-time feedback from the camera.
With CAD, vision systems are moving away from conventional approaches to defining geometries. Older software required operators to take measurements by first designating the feature type to be measured and then collecting the associated data according to pre-defined rules. New software is eliminating those restrictions. Now, instead of having to tell the software the type of feature to measure beforehand, the operator simply collects vector points on the geometry of interest. The software automatically recognizes the feature type and then constructs and evaluates it. This works for an ever-growing list of feature types, including points, lines, ellipses, circles, squares notches and slots. Because of these improvements, programming and operating a vision system has become significantly easier and more comparable to using a coordinate measuring machine (CMM).
Lighting, Focus and MagnificationVision measurement has a reputation for being fussy. Magnification, focus and lighting requirements vary not only in relation to each other but also as changes occur in types of geometries and materials measured, not to mention variations in the measurement equipment itself. Until recently, operators had to take these into account when measuring and adjust for them. By automating the vision measurement process, some vendors have nearly eliminated the need for manual adjustments.
For example, software now lets operators calibrate the illumination of vision machines so that extremely linear and accurate measurements are provided for a range of illumination intensities and material reflectivity. Additional functions automatically adjust lighting intensity on-the-fly to improve contrast. This works by having algorithms that evaluate contrast in the area of interest and then make adjustments if needed. Some measurement software can even adjust lux values to compensate for the gradual loss of illumination intensity over time. This is a particularly useful feature with LED type systems.
Controlling focus is another area of improvement. Again, algorithms evaluate edge strength and decide if it is necessary to refocus to improve accuracy. This functionality improves measurement cycles by selectively eliminating the need to refocus every time the camera pauses over a new area of interest. In addition, some software allows for calibration of focus on individual machines for optimal measurement speed and accuracy. This can be particularly useful when the same parts may be measured on more than one vision system, including different models.
Edge Strength. Hard probes know an edge when they meet one. Traditional vision systems have numerous filters that operators have to adjust to boost edge strength. These require specialized knowledge in determining when and where to apply which type of filter.
A newer and much easier approach is to let the software decide which filters to use at the time of measurement. The software evaluates the condition and lighting of the feature it is about to measure. Then, based on an operator-defined measurement objective, it uses built-in intelligence to decide which edge-strength algorithm will provide the best results for the task. There is no operator intervention, and the result is faster and more accurate measurement.
Measurement Efficiency. Some of latest vision software releases take advantage of dual and quad core central processing units (CPUs) so that imaging and image processing can take place simultaneously.
Accuracy. Conventional wisdom might deem accuracy to be solely dependent on optics and equipment characteristics. In truth, well conceived and well executed measurement software makes a significant contribution to accurate, repeatable vision results. And as the mathematics gets better so will the measurement results.
CAD-based software has moved vision measurement into the manufacturing mainstream. Whether in the quality lab or on the shop floor, manufacturers are finding it to be at least as useful and flexible as any other measurement tool. Because of both the integration of CAD into vision measurement and ongoing software improvements, vision is no longer an esoteric measurement technology best left to highly trained and experienced specialists.