Machine vision – what is it and where is it used


Machine vision (technical vision) is a technology and methods of obtaining information based on automated image analysis. The term “machine vision” encompasses many technologies, including software and hardware, system integration and forensics.

Machine vision cameras transmit data in an uncompressed format and have a high shooting speed. They use large-format sensors with a global shutter, high sensitivity and resolution, industrial data transmission interfaces, control and management of all sensor parameters, input/output lines for synchronization with external equipment, and a tool for developing an image analysis program.

The main task of a machine vision camera is to obtain the highest quality image for subsequent data analysis in a user-developed program.

What is machine vision

The hardware (hardware) part of machine vision is based on various technical vision cameras, with additional equipment for them, which ensures the shooting and transmission of the frame. Software part – algorithms based on classical mathematical transformations and/or neural networks. Elements of the system can be either separate or assembled into one housing – such devices are called smart cameras.

The following elements of the system are minimally required for shooting:
  1. Light is a source of information for the camera, so the object of control must be well illuminated, and most importantly, with constant brightness and no flicker. 
  2. Synchronization sensor – an optical sensor for the presence of an object at the shooting location, an encoder or a signal from a PLC.
  3. A suitable machine vision camera with a compatible lens that produces a frame of the required parameters (color, bit depth, resolution, frame size) at the desired speed.
  4. PC with installed software – responsible for receiving and processing images from the camera. And also transfers the received data to the executive elements or to the database.

Where is machine vision used?

In an effort to solve even the most inconceivable complexity of production problems, camera manufacturers have created a large line of functional and reliable cameras. Their use has gone far beyond the scope of production, now these cameras can be found in many sectors of the economy.

Being the main “generator” of tasks for machine vision, it has the largest number of possible applications. Moreover, production with machine vision can be, like an “automatic plant” from industry 4.0, or a small sawmill with control of the input volume of wood. World experience shows a huge number of solved production problems. These are just a few areas:
  • production lines
  • Sorting foods or minerals
  • Robotics
  • Diagnostics of infrastructure facilities
  • 3D scanning
  • Instrumentation
  • Video analytics
  • Inspection and automation of welding works
  • Marking and aggregation
  • Logistics (including track&trace)
  • Output control of products
  • Shooting in dangerous or inaccessible areas
  • High speed process control
Movement organization
Control of automobile traffic, ITS projects, infrastructure conditions (scanning of the asphalt concrete bed, railway, contact network). Thanks to the reliability and unpretentiousness of the machine vision camera, they successfully solve these problems while being on the street and working 24/7, which can significantly improve transport safety. Go for more here.
Entertainment and film industry
Volumetric shooting of VR / AR content (volumetric capture), interaction with visitors using augmented reality animation, and much more – cannot do without machine vision cameras that can synchronize with each other and see not only color, but also in the IR range.
The science
Many research centers, universities, special design bureaus and schools use machine vision for scientific and educational purposes. The market for machine vision cameras is rich in cameras with different characteristics, resolutions, etc. very often with the help of such cameras the tasks of science are successfully solved:
  • high speed cameras
  • high resolution cameras
  • Chambers protected from water and temperature changes
  • Monochrome cameras with a color depth of 12 bits (typically 8 bits – up to 256 color gradations)
  • Sensitivity to a large spectrum (cameras can shoot from NIR – to UV)
  • Ability to use spectral filters
  • Cameras that give a high-quality picture, which are much more compact than cameras
Photo finish systems, athlete tracking, re-play systems at football matches with 3D visualization.
Object security
Identify a person by face in a crowd, retina recognition for access to a room, create a Three-dimensional face recognition system with laser projection, anonymous counting of people in a room – all this can be easily and reliably implemented on machine vision cameras. 

machine vision requirements

Many mistakenly believe that conventional video surveillance or webcams can really solve industrial problems. In fact, there are a number of developed requirements for technical vision, which must be met by all equipment, incl. cameras to work stably in the production automation system: 

Camera Quality
The big difference between just “computer” vision and “technical” or “machine” vision is the requirement for reliability and stability of the result. No production can afford to simply temporarily replace a failed camera or reject the entire batch because the camera did not see it due to an autofocus failure.
The machine vision system must work reliably, like any other sensor – to provide the necessary information for many years without operator intervention and with only little maintenance. In machine vision, this is achieved through the use of an industrial-grade component base in all elements of the system, high-precision production and careful control of each element – all this leads to a significant increase in the cost of manufacturing a camera, PC, lens, cable, but guarantees a long service life in harsh industrial conditions.
Hardware Capabilities
Many image processing algorithms, and even more so neural networks, are applicable to both scientific/educational and machine vision. But for the stability of the processing result, a stable picture is required from the camera – the algorithm should not try to “adjust” to the shooting conditions – since in this case it risks becoming a weak element of the system.
To ensure high-quality images, machine vision cameras transmit data in an uncompressed format and have a high shooting speed. They use large-format global shutter sensors with high sensitivity and resolution, industrial data transfer interfaces, input/output lines for synchronization with external equipment. In the software shell of the camera, control and management of all sensor parameters is available,
With a machine vision camera, you definitely get the software development kit (SDK). Through it, you receive an image from the camera with high-precision timecode for subsequent synchronization, adjust camera settings, resolution, frame rate, exposure, etc.
The SDK contains examples of interactions with the camera in various programming languages.
Calibration capability
It is possible to make a measurement system out of a machine vision camera, thanks to the high-precision manufacturing of components, the ability to fix all shooting parameters, including lens focusing. Subsequent calibration of such a camera will ensure that real data is received from it, without discrepancies from vibrations or the influence of auto functions. This is especially useful when using the camera to control robotic arms.

Camera selection

When choosing a machine vision camera, you should be guided by the following parameters:

  • data interface
  • spectral range (visible, IR, SWIR, UV)
  • resolution and sensor size
  • required shooting speed
The lens for the camera must be selected based on the geometry of the problem being solved, the size of the camera matrix and the spectral range.


Please enter your comment!
Please enter your name here