Machine vision is a scientific discipline that includes methods for acquiring, processing, analysing and understanding real-world images in order to produce numerical or symbolic information. This technology, whose origins date back to the 1950s, has experienced a boom in the last decade, coinciding with the arrival of Industry 4.0.
A wide range of possibilities
It is a disruptive technology that is revolutionising areas as diverse as the management of photographic archives, industrial processes, quality control processes and even social networks. Facial recognition systems, massively used in applications such as Facebook or Instagram, are a good example of this. Counting people in crowds is another problem that machine vision is managing to refine, with error margins of between 10% and 20%.
The range of industrial applications for this type of system is wide. The automation of all kinds of processes, especially those that are involved in changing environments that require active interaction, is a good example of this. From the control of autonomous vehicles, load movement systems to robotic arms capable of discriminating characteristics of the environment that make their behaviour more flexible. Quality control of industrial production process chains is also becoming progressively faster and more precise thanks to these technologies.
Machine vision in waste management
In this context, machine vision represents a tool for change on the road to a circular economy, an area in which it already has applications. There are automatic waste sorting systems based on image analysis. These can be taken using a camera (visible spectrum) or using other bands of the electromagnetic spectrum, from infrared to X-rays, from which more complete characteristics of the typologies and contents of the waste analysed can be extracted.
Machine Learning and Deep Learning: Foundations for the development of artificial vision
Machine learning algorithmic techniques are closely related to computer vision. These techniques seek to model abstractions from large volumes of data, so that the machine assimilates these representations and is able to identify them in subsequent iterations (e.g. “is this image a plastic bottle?”). These techniques include computational models such as neural networks, which mimic the behaviour of biological brains (they do not follow explicit programming, but are based on mathematical nodes that are interconnected and work like “artificial neurons”). The development of these technologies represents the basis for the increased performance of machine vision systems, which will be a driver of change on the road to the circular economy.