Archived Viewpoints

About Machine-Vision Systems

September 2001

Machine-vision technology uses a computer to analyze the content of an image for a host of inspection and guidance applications, primarily in manufacturing. Vision systems can measure part size, verify the absence of gross defects, detect surface flaws, guide automated vehicles and robots, locate parts for assembly with other components, and recognize and match patterns for optical character recognition.

The young machine-vision market has developed more slowly than observers expected. Initially in use in the automotive industry, machine vision expanded to the electronics industry and is now beginning to find use in lumber mills, pharmaceutical factories, medical laboratories, and other industrial settings. Only recently have vendors begun to move from selling quirky, difficult-to-implement vision technology to selling embedded vision systems actually tailored to users' niche applications. Machine vision could find use in most manufacturing settings, but the technology must become more transparent to the end user to gain wider acceptance. Developments in computer hardware price/performance, sensor technology, user interfaces, and image preprocessing hardware and software will further the spread of vision systems.

Machine-vision systems are available off-the-shelf and stand to benefit most applications where a concern exists for high-end-product quality. However, implementing vision systems typically requires multiple experts to solve lighting, parts-handling, control-structure, and other problems. The payback is long in coming and comes not through job replacement but through improvements to quality, particularly in areas where defects are intolerable (such as in pharmaceutical production). Despite the long payback period, demand for vision systems is growing.

In the long term, machine-vision systems may find new applications beyond inspections and guidance in factory use. The applications will include intelligent vehicle/highway systems, security systems, multimedia, agriculture harvesting, and biometrics; characteristics of the applications include continuity, preciseness, quickness, and mass in identification and inspection. Advancement of technologies such as parallel processing, neural networks, fuzzy logic, data compression, and color vision will enhance the reliability and robustness of machine-vision systems.