Unfortunately, there is no such thing as an “image quality” specification. Instead, we have to understand what the basic specifications for a machine vision camera are and how they affect image quality.
The quantum efficiency of a camera/imaging device is simply its ability to convert light energy into electrons. The higher the quantum efficiency is, the more information will be captured by the image sensor. In bright lighting conditions, most cameras will deliver good images, but in poorly lit environments, cameras that exhibit high quantum efficiency will do much better.
Sensor Size and Resolution/Pixel Size
Generally speaking, a larger sensor can capture more light, and thus have better image quality. Image resolution on the other hand, has an inverse effect on picture quality in terms of dynamic range. Dynamic range is the ability for a camera to capture details in both the light and the dark areas of a scene. Larger pixels are better at capturing light and have better dynamic range. However, for the same sensors size, the higher the resolution, the smaller the pixels on a sensor, so while the image will be sharper, the dynamic range will be poorer.
A typical application where you may need a high dynamic range is a traffic camera. If someone speeds through an intersection, you would like to catch their license plate, as well as their face. While the license plate is very bright, the driver’s face inside the vehicle is often dimly lit. Only a camera with high dynamic range can capture both of these details in the same shot.
Besides dynamic range, cameras with larger pixel sizes will often also have better signal to noise ratios, which means they will produce cleaner images with clear details. These specifications are standardised through the EMVA1288 data sheet. Developed by the European Machine Vision Association, the goal of this data sheet is to provide a standardised way to compare machine vision camera performance.
CMOS vs CCD sensors
While CCD sensors used to be the gold standard, heavy investment and continued innovation in CMOS sensor technology, driven mainly by the massive smartphone camera market, has greatly improved their image quality. These days CMOS sensors will generally produce comparable image quality at a much lower price and with significantly lower power consumption than CCDs.
NIR for Poor Lighting Conditions
In low lighting conditions, it is impossible to capture high quality images with traditional cameras. In these situations, it is possible to use NIR cameras. NIR is the light spectrum just outside of visible range, and image sensing hardware supporting it will exhibit enhanced image capture capabilities. Until recently, this required expensive specialized CCD sensors. Now it can be accomplished with cost-efficient CMOS sensors, opening up a plethora of new machine vision applications.
In many cases, machine vision systems will need to deliver high speed capture and processing of images. On a factory assembly line, components must be imaged precisely as they arrive in position. The timing has to be precise and the camera needs to be able to respond quickly in real-time. This is demanding on both the imaging system’s processing performance plus the network throughput. The attributes that are expected for high bandwidth industrial cameras have been encapsulated within the GigE Vision standard. Prime examples of GigE Vision cameras are the Basler Ace series. These have high data rates of up to 100MB/s, support long cable distances of up to 100m, and use Gigabit Ethernet and USB 3.0 connectivity. Featuring CMOS sensors from ON Semiconductor, they offer up to 14Mpixel resolutions and support frame rates of as much as to 751fps.
Clarifying Machine Vision
While it is easy to understand what a high-quality image looks like, making sure a camera can really capture such images in the conditions you need it to work in is a different story. Understanding how to read an EMVA1288 data sheet, as well as important features such as sensor type, NIR, and real-time imaging capabilities can help make sure your machine vision application turns out picture-perfect.