Machine Vision Cameras
The purpose of the camera in a machine vision system is to capture an accurate representation of the object to be inspected and provide an interface through which the image data can be transmitted. The format of many cameras dates back a couple of decades to a time when image capture was governed by the entertainment industry and the Vidicon tube was the sensing element. In this section we will present an overview of the camera types currently available and the technologies used to implement them.
Fundamentally, cameras can be split into two categories – Area Scan and Line Scan. As their names suggest Area Scan cameras contain image sensors capable of scanning an area (multiple lines), historically with an aspect ratio of 4:3, while in line scan cameras the sensing devices only scan a single line before the image data is read out. Although many of the basic principles of the two cameras are similar we will treat each one separately because for all intents and purposes they are unique technologies.
Spectral Response Cameras are also available for applications where it is necessary to view Ultra Violet or Infa Red light spectrums.
Technologies and Processes
Camera sensor technologies occur in two forms – the CCD (Charge Coupled Device) and CMOS (Complimentary Metal Oxide Silicon) Sensors. Each have various advantages and disadvantages, so the choice of camera is driven by the application rather than any outright benefit enjoyed by a particular sensor type.
Colour information is obtained from a camera sensor through digital colour imaging and there are a multitude of methods for the transmission of video data.
The most common type of shutter on a machine vision camera tends to be a global shutter where all pixels are exposed at the same time and their values are retrieved in unison. Global shuttering therefore allows for accurate capture of moving objects.
Rolling shutters come in a number of different forms. CMOS cameras are lower in cost but each pixel value is retrieved one after another. This results in distortion when viewing moving objects as the image may have changed in the time it takes to read one pixel value to the next. For full frame or frame transfer CCD cameras, the charge from each pixel is passed to the storage or shift register by one pixel to the next. This can lead to distortion if the light is not blocked from the active pixels while their charge is being retrieved.
The ability of a camera to distinguish between subtle differences in a field of view is determined by its dynamic range. Dynamic range is the term used to describe how many grey levels can be translated from a pixel value. It is expressed as the saturation voltage to total RMS noise of the camera output. If the contrast between features in an image is vast, a camera with a low dynamic range would be suited. A camera with a high dynamic range could better distinguish between subtle differences in features. The dynamic range of a camera can be increased by raising the saturation voltage or full well capacity of each pixel. The higher the capacity, the greater the dynamic range. Modern vision cameras vary in pixel capacity from 15Ke to 200Ke. The higher the full well capacity of a sensors pixels, the greater the physical size of the sensor. Better dynamic range can also be achieved by reducing the total RMS noise of the camera by using better quality components and reducing the sensors clock speeds. This has obvious repercussions on cost, size and performance.
A wider dynamic range not only improves definition of features but also allows simultaneous viewing of light and dark areas. Logarithmic sensors can be used to achieve higher sensitivity to light variations due to their logarithmic response to light levels as compared to near-liner responses of conventional sensors. True logarithmic sensors do result in higher levels of noise and require rolling shutters. However, there is new CMOS sensor technology which can achieve logarithmic responses with global shutters.