Things to know about machine vision

by lyonOP

Machine vision has been advancing rapidly for several years. Neural networks and deep learning open up numerous application possibilities that make our everyday lives easier and more enriching. In this article, we will give you an overview of the concept and application of machine vision.

The concept of machine vision

Machine vision refers to the ability of computers to perceive visually. With machine-aided vision, analog image information is captured and converted into digital signals. Video cameras and signal processing technologies are used for this purpose.

How does machine vision work?

The basis of machine image recognition is, on the one hand, extensive image databases that enable the machine to compare objects with objects in the outside world (training data) and provide the necessary artificial neural networks. The resulting image models learn basic structures (e.g. curves, edges), patterns, colors and objects.

Image models therefore contain the necessary prior knowledge that is used by the computer to identify objects. This prior knowledge is used in a complex training process to develop existing models forto use new problems. This is also called transfer learning.

The deep learning algorithms used to classify objects compare and classify the individual images to be examined. The deep learning algorithm breaks down the image into a grid (numerous smallest squares) and extracts the image information, each of which is examined for a specific image property.

By automatically comparing several images, the system recognizes patterns and then calculates the probability of whether the respective image is a specific object. With computer-aided vision, the visual content is recognized using deep learning and neural networks.

What are the typical tasks of computer vision?

The possibilities of digital vision are used, for example, for the recognition and assignment of:

  • optical signs,
  • Patterns and
  • Objects.

The typical tasks of computer vision include in particular

  • Classification of objects,
  • Localization of objects,
  • Searching large amounts of data,
  • Motion analysis,
  • Description of images,
  • Construction of 3D images from individual 2D representations and
  • Reconstruction of image content.

Application examples for machine-assisted vision

Machine vision is used, for example, for:

  • Identification of signatures or manuscripts,
  • Checking banknotes,
  • Material testing,
  • Recognition and assignment of plants or plant seedlings
  • Analysis of medical images (example: identification of individual cell nuclei, detection of skin cancer).

In natural environments, machine vision is used, for example, in the following cases:

  • Identification of persons based on biometric data,
  • Recognition of facial expressions or gestures of people or
  • Detection of lanes and pedestrians outside a lane.

Industrial application areas are among others

  • Automation technology (example: control of welding robots in a desired working position),
  • Quality assurance (e.g. checking the product quality at the end of a manufacturing process or during the useful life of a workpiece or machine)
    • Surface inspection,
    • Measurement of layer thicknesses,
    • Detection of defects also under the surface,
    • Checking of position, dimensions and shapes,
    • Check for completeness,

• Security technology (access controls and recognition of dangerous situations ) and

• Traffic engineering (quality assurance and autonomous driving of “seeing” vehicles).

Machine vision in the example of face recognition

Based on a given face model, the machine uses a search algorithm to know

  • that there are always noses between the mouth and eyes and
  • where the mouth must be roughly based on the eyes and nose that have already been identified.

Face recognition can also be done on social networks after uploading images. Image classifications can also be combined with language functions (as practiced on Facebook, for example).

  • If a recognized image is spoken by an output device of the machine, it is possible for the visually impaired to understand the visual content.
  • The image recognition process also enables lips to be read and then converted into speech signals.

The technologies of computer-aided vision are also the basis of the Google Photos app and the Google Photo search.

For more information, check now on

Related Posts