Machine vision

from Wikipedia, the free encyclopedia
The articles computer vision and machine vision overlap thematically. Help me to better differentiate or merge the articles (→  instructions ) . To do this, take part in the relevant redundancy discussion . Please remove this module only after the redundancy has been completely processed and do not forget to include the relevant entry on the redundancy discussion page{{ Done | 1 = ~~~~}}to mark. Viennese charm Disk 12:26, 3 June 2020 (UTC)Wels wappen.jpg


Computer Vision (English: machine vision ) covers all industrial applications, in which, based on visual systems, automated processes are steered. Typical areas of application are industrial manufacturing processes, automation technology and quality assurance . Further areas of application can be found e.g. B. in traffic technology - from the simple radar trap to the "seeing vehicle" - and in security technology ( access control , automatic detection of dangerous situations ). Methods from the field of computer vision are used.

Computer vision is often equated with machine vision . Although technologies and methods from the field of computer vision are used here, they differ in the requirements that arise in the special industrial environment. Industrial visual systems require a high level of reliability, stability and must be particularly robust. In this respect, machine vision tries to apply and integrate existing technologies in new ways.

Tasks

The following tasks can currently be solved economically:

research

Only a comparatively small part of the current research projects is concerned with actually understanding the meaning or content of images ; Most of the time it is more about detecting objects in images, describing them, measuring their properties, classifying them , and using these results to make decisions or control processes. Machine vision uses methods from the field of computer vision and is therefore a branch of computer science that has strong links to photogrammetry , signal processing and artificial intelligence . The tools of machine vision mostly come from mathematics , in particular from geometry , linear algebra , statistics , operations research ( optimization ) and functional analysis . Typical tasks of machine vision are object recognition and the measurement of the geometric structure of objects as well as movements (foreign movement, proper movement). Use is made of algorithms from image processing , for example segmentation, and methods of pattern recognition , for example for classifying objects.

Methods

Image processing tools for automatic interpretation are:

Models are often used in more complex recognition tasks. These include prior knowledge that can be used to identify an object. For example, a face model describes that the nose must always be between the mouth and the eyes. A search algorithm thus roughly knows where to look for the mouth when it has already found the eyes and nose. Here are some modeling techniques:

developments

In 2001, two computer scientists developed the Viola Jones method for facial recognition, named after them . The process is based on machine learning and is widely used.

The methods described above determine properties and values ​​from images and assign them to the objects contained therein. This is based on the deductive application of known rules. Machine learning is an extension. It uses inductive methods to find these rules in the first place. In supervised learning, for example, a sufficiently large number of images is assigned to a class. After this learning phase, the implicitly existing rule or regularity can then be used to classify new images. Examples of industrial applications are character and text recognition (optical character recognition, LeCun), which is based on the use of neural networks.

Applications

In industrial environments , the techniques of machine vision nowadays successfully used. For example, computers support quality control and measure simple objects. The programmer largely determines the environmental conditions that are important for the error-free execution of his algorithms (camera position, lighting, speed of the assembly line, position of the objects, etc.).

Examples for use in industrial environments are:

In natural environments far more difficult demands are made on the techniques of machine vision. Here, the programmer has no influence on the ambient conditions, which makes it much more difficult to create a robust, error-free program. This problem can be illustrated using an example for recognizing automobiles: A black car stands out well against a white wall, but the contrast between a green car and a meadow is very low and it is not easy to distinguish between them.

Examples of use in natural environments are:

  • automatic detection of the lane and of pedestrians at the roadside
  • Recognition of human faces and their facial expressions
  • Recognition of people and their activities

Further applications can be found in a large number of different areas:

tasks

Machine vision systems are additional investments in production. Against these costs one calculates the advantages, which lie in the following potentials:

  • higher quality level
  • Analysis of disturbance variables and process improvement
  • less scrap
  • Securing the supply chain
  • Monitoring of highly dynamic production processes
  • Cost optimization.

See also

literature

  • Carsten Steger, Markus Ulrich, Christian Wiedemann: Machine Vision Algorithms and Applications . 2nd Edition. Wiley-VCH, Weinheim 2018, ISBN 978-3-527-41365-2 . link
  • Gottfried Gerstbach : Eye and sight - the long way to digital recognition . Sternenbote booklet 43/8, p. 142–157, Vienna 2000.

Web links

Individual evidence

  1. ↑ What is Machine Vision | Cognex. Retrieved June 11, 2020 .
  2. ↑ Vision Online. Retrieved June 11, 2020 .