Inside Vision: Novel methods improve vision inspection
Not content with just visualization, users of imaging inspection systems seek to check parts for much more than the mundane pass/fail decision.
Not content with just visualization, users of imaging inspection systems seek to check parts for much more than the mundane pass/fail decision. They want a thorough analysis of the pass and fail imaging data and are demanding that inspection systems work automatically and faster. To meet ever-tighter parts inspection requirements, system integrators are adapting computed- tomography (CT) technology, dual-DSP architectures, reconfigurable imaging, and advanced cameras into their platforms.
Originally designed for medical diagnosis, CT imaging technology has found use in industrial applications such as checking large automobile engines. With this approach, says contributing editor R. Winn Hardin, a CT image initially reconstructs a 2-D cross section of the engine and then a 3-D density model that locates and quantifies defects (see p. 23).
Although microprocessors have dominated general-purpose computing for many years, certain image-processing functions are more efficiently performed in custom or programmable logic devices. Reconfigurable building blocks offer superiority over general-purpose processors. According to editor at large Andrew Wilson, they permit designers to equate specific imaging-application requirements in terms of cost, accuracy, and control so that working platforms are produced faster and easier (see p. 37).
Until recently, the visual inspection of wafers and die at the back end of a semiconductor production line was a manual process, performed by inspectors with microscopes who made a few thousand pass/fail decisions a day. To overcome inspector speed, fatigue, and skill limitations, an automated vision-based defect-inspection system was installed by the semiconductor company. The results, reports contributing editor Larry Curran, equate to an inspection rate of 25,000 a day and no inspection problems (see p. 31).
By eliminating the need for frame grabbers, host-based CPUs, and host-based image-processing software, smart cameras offer systems integrators several ways to add intelligence to machine-vision systems and still reduce costs. Embedded chips in advanced cameras can now handle resolution, interfacing, host-PC operations, real-time operating systems, and system software functions. By incorporating built-in processors, interfaces, and software, says Andrew Wilson, advanced cameras are improving overall system performance at less cost (see p. 45).