A play of colors:
Lighting with a variety of colors produces optimal contrasts for image processing. In the green light (above) the pressure regulator gaskets are tested to ensure they will seat perfectly. The connecting flange is inspected under the red light (below).
The Untiring Eye
Automated image processing is the method of choice for visual inspection of production operations. It’s fast, precise and – in contrast to humans – untiring. In addition to 2D image processing with camera technology, Bosch researchers are using white-light interferometry to establish a testing process for optically measuring components down to the nanometer scale.
Production lines are fast. At a Bosch factory, for example, a transmission pressure regulator comes off the conveyor belt every three seconds. A human being trying to visually inspect this production would be overwhelmed. But a sophisticated image-processing system is up to the job: It uses a CCD camera and various types of illumination to monitor surfaces, O-rings and component geometries. The trick is to use the captured images and the characteristics contained in them to ensure that the product conforms to production standards and is not damaged. This is only possible with automated image processing, because many products are manufactured in dozens or even hundreds of customer-specific variants.
At Bosch Research, the image-processing systems for production use are tested and tuned for their specific areas of application —whether for pressure regulators, injectors, generators, lambda sensors or spark plugs.
A spark plug races by the camera every 0.8 seconds, so the instrument has to be fast — and 100 percent error-free.
That is why the entire system must be optimized. This optimization begins with the lighting, which might seem of secondary importance, though it’s really a strength of industrial image processing: The measurement conditions are precisely standardized. The illumination is what makes it possible, for example, to produce optimal contrast for the subsequent image processing — whether it uses transmitted or reflected light, or light or dark-field illumination.
If the image has been captured with a CCD camera, the image-processing system first separates the important data in the image from the irrelevant information (for example backgrounds). Then characteristics such as geometric dimensions are extracted and compared with the specifications.
The researchers are also developing new testing systems and putting them through trials. One such system is what is known as a white-light interferometer, which is used to automatically make measurements of the parallelism, flatness, thickness and roughness of surfaces down to the nanometer scale. With the special interferometer technology, a surface-height measurement can be made for every pixel of the product’s image recorded by the CCD camera. This provides a 3D image of the surface that can then undergo further automatic processing and evaluation.
At Bosch Research, the image-processing systems for production use are tested and tuned for their specific areas of application —whether for pressure regulators, injectors, generators, lambda sensors or spark plugs.
A spark plug races by the camera every 0.8 seconds, so the instrument has to be fast — and 100 percent error-free.
That is why the entire system must be optimized. This optimization begins with the lighting, which might seem of secondary importance, though it’s really a strength of industrial image processing: The measurement conditions are precisely standardized. The illumination is what makes it possible, for example, to produce optimal contrast for the subsequent image processing — whether it uses transmitted or reflected light, or light or dark-field illumination.
If the image has been captured with a CCD camera, the image-processing system first separates the important data in the image from the irrelevant information (for example backgrounds). Then characteristics such as geometric dimensions are extracted and compared with the specifications.
The researchers are also developing new testing systems and putting them through trials. One such system is what is known as a white-light interferometer, which is used to automatically make measurements of the parallelism, flatness, thickness and roughness of surfaces down to the nanometer scale. With the special interferometer technology, a surface-height measurement can be made for every pixel of the product’s image recorded by the CCD camera. This provides a 3D image of the surface that can then undergo further automatic processing and evaluation.
