Artificial vision
ARTIFICIAL VISION AND REAL CONTROL
(AUTOMATIC SELECTION OF PARTICULAR PRECISION PRODUCTION PARTS)
Each time we heard of artificial vision, automatic control, computerized inspection systems & Co, we immediately think of something connected with new technologies, breakthroughs or of some complex electronic circuits…. But what is artificial vision really? It is not “RoboCop”, nor the bionic eye of the “Six Million Dollar Man”. It is a system based on a camera, which is the eye, and an image processor, which is the brain.
We can try to analyse this “machine” starting from the eye, that is the camera.
The images we “see” are perceived by millions of optic nerves and each of them “captures” a tiny fraction of the image and stores it.
As it occurs for a camera, the film quality is related to ASA and the more the number of “points”, the best the resolution of the picture.
In a camera the “frame” quality depends on the number of points used to store the image. These points are the pixels, which form the electronic film of the camera.
The CCDs (Charge Coupled Devices) are usually small in size, 5x7mm for example, and contain thousands, millions of pixels depending on the the instruments quality.
The pixel, hit by light intensity and by the image fraction’s colour received, generates electric tension which is transformed in a mathematical number.
OK, we did it! We have reconstructed a chart, a list of numbers, that are like bread for the computer’s teeth.
The other part of the machine is the “brain”, that is the processor. Specific or not, the processor is always connected with a software.
The last one is the interface between the image and the set of parameters/control tolerances.
Now that we have our machine, which can analyse images or, at least, transform a picture into numbers, it is easier to imagine a software capable of mathematical comparing, applying filters as Laplace, Fourier, average calculation, derivatives, standard deviation, trigonometric equation.
In order to have a clearer and completer view of the system, we should try to optimize our image, by highlighting what we need to check. It is at this point that lighting comes into play. We have spoken of pixel, the minimum photosensitive unit and that is why light becomes so important.
Different physics laws play a significant role, such as Snell, Cartesius, reflection, refraction, projection, etc. We are not having an illuminating engineering class, but we understand that we can use different techniques and each technique depends on the visual control objectives.