A presentation at the recent A3 business conference reminded me of the confusion in terminology that existed before the words “machine vision” were coined to differentiate this technology from “computer vision”. The famous 1982 textbook by Ballard and Brown titled “Computer Vision” http://homepages.inf.ed.ac.uk/rbf/BOOKS/BANDB/bandb.htm was invaluable to many researchers and developers working on early machine vision algorithms that formed the basis for many of the fundamental tools in use in machine vision today. Now, computer vision seems less associated with what has become machine vision.
Listening to the presentation, I was struck by how these two technologies have separated even further, and also by how much progress has been made. In my opinion, computer vision has advanced solidly in the area of simulating human perception within a computer. The technology appears to be making great strides into scene recognition and prediction – things that are critical if, for example, we ever want a robot to navigate freely around our human physical domain. But do these capabilities apply to “machine vision”? Certainly, the research in computer vision will likely advance algorithms in industrial robot guidance and inspection. However, the need for specific and discrete analysis of features (for mundane things like presence/absence, measurement, reading, and so on), will always be the staple of machine vision technology.