Machine vision is an interdisciplinary subfield of engineering. It has to do with engineering the necessary hardware and software for a machine vision system that has useful industrial applications. The most frequent applications are quality control and product counting. As machine vision advances as a field, it automates some of the most boring industrial jobs, permitting human employees to serve in more interesting positions.
Machine vision is considered a subfield of computer vision in general, which includes neurocomputational investigations into human vision and the development of flexible vision systems for autonomous robotics. “Machine vision” has the connotation of application in industrial settings. Machine vision is used in any industrial setting with a significant degree of automation, where product specs are well-defined and products are mass produced. This includes the automobile industry, the semiconductor industry, and the electronics industry in general. Sometimes, a machine vision system is coupled to a robot arm that discards defective products or even actively participates in manufacturing the product.
A machine vision system generally consists of a pressure or optical sensor, a camera, a lighting system, a central processing unit (CPU), associated software for processing images, and an I/O system for connecting to a larger network. If a machine vision system is located adjacent to a conveyor belt, as they often are, the sequence of events goes something like this: a pressure or optical sensor is activated when the product moves in front of the camera, a pulse of light is activated to illuminate the target, the camera takes an exposure, and the image is processed and fed to a decision tree, which returns an output that is then either acted upon by an automatic element or displayed to human operators. This simple process can automate the classification or inspection of thousands of products daily, saving millions of man-hours of tedium.
