Machines with eyes wide open

Seeing machines are revolutionizing our lives because they can take on more and more tasks. The technology that teaches machines to see is called machine vision or image processing. For many years, the technology has been developing dynamically and opening up completely new perspectives.

stock.adobe.com/borislav15

The human being – often copied, never reached

The human body is a miracle. Many of its functions and possibilities are so mature that even with the most modern technology it is hardly possible to reproduce the performance of individual body parts or senses. Examples such as artificial hearts or joints as well as robot hands do not come close to the human role models despite all engineering skills. This also applies to one of the most important senses of human beings: vision.

Easy as pie yet incredibly difficult

Just a few weeks after birth, a baby perceives different colours and shapes. At just a few months old, it recognizes the faces of its parents and can distinguish them from other people. Soon thereafter, it can pick up individual building blocks from a pile, even if they are partially covered. Apples and pears are quickly distinguished as naturally as dogs and cats. Decisive for these abilities is not only the eye, which perceives the environment, but also the brain, which processes, interprets, understands and triggers corresponding reactions.

What is simple for an infant can be a sophisticated task for a machine vision system. If a robot has to recognize and grab the correct part from a box with unsorted objects, this is a demanding task for the technology and shows how highly developed human vision and judgment are.

Reaching into the box – child’s play?! source: iStock/Steve Debenport
For machine vision systems still a challenge! source: i-mation GmbH

Human and machine vision

In machine vision, the tasks performed by the eye and human brain are taken over by technology. Machine vision systems consist of at least one camera or image sensor with suitable optics, illumination and a processor on which the image data is processed by software.

Both, human beings and technical image processing systems do not actually see the object under observation, but only the reflections of light reflected from the object onto the eye or onto the camera. The iris of the human eye increases or decreases the diameter of the pupil to control the brightness of the incident light. In the camera, this function is taken over by the aperture.

Source: Shutterstock/Aaron Amat
Source: iStock/seoterra

In humans, light passes through the lens of the eye, which bundles the light and focuses it.  In a technical system, this task is performed by one or more lenses in a camera. In humans, an image of the scene in question is created on the retina. Around 120 million rods are responsible for detecting brightness and around 6 million cones for detecting colors. An enormous performance, however, which industrial cameras achieve – and in some cases even exceed – in resolution as measured in pixels.

In humans, nerves transmit the image information from the eye to the brain, which processes and evaluates the data and, if necessary, triggers reactions from body parts or the storage of the recorded information. In machine vision systems, cables ensure data transmission between the camera, processing unit and the subsequent systems, which can then react on the basis of the calculated results.

I spy with my little eye

Machine vision works similar to human vision.

From seeing to understanding

Despite many similarities between human and machine vision, there are great differences between the two worlds. The biggest difference is in understanding and interpreting image data. In the course of their lives, people learn the meaning of objects and situations that they perceive daily through their eyes. Human beings understand other people’s facial expressions, recognize animals or plants and much more. We capture and filter the constant flow of images mostly intuitively. We then react to every new situation on the basis of the experiences stored in our brain.

In contrast, a machine vision system only correctly identifies objects if they have been previously programmed or trained. Software algorithms exist for specific tasks, e.g. for the recognition of objects, patterns, errors, characters, colors, dots, edges and much more. Developers of machine vision systems need to know in advance what this system is supposed to accomplish and design it accordingly.

Factories with eyes

Due to the breathtaking progress in the technology, machine vision is firmly established in production and is constantly conquering new fields of application.

Development and History

Machine vision has developed into a key technology and has recorded a high increase in sales in recent years:

VDMA Robotik + Automation

Milestones of machine vision

An overview of some of the most important development steps in machine vision technology:

1969

Development of sensors with CCD technology (Charge-Coupled Device) for data storage that can capture two-dimensional images. In 2009, Willard Boyle and George E. Smith received the Nobel Prize in Physics for this invention.

1990

The first CMOS sensors (Complementary Metal-Oxide Semiconductor) are developed. In contrast to CCD sensors, the charges generated by photons are converted into voltage in the pixel.

2006

Interoperability is key: New international industry standards such as GenICam and GigE Vision greatly facilitate the integration of image processing components and make a significant contribution to industry growth. Further standards will follow.

2015

The arm technology is finding its way into machine vision and enables cost and energy-saving applications which were previously PC-based. Embedded Vision expands machine vision’s market potential in the following years.

1979

The first industrial machine vision systems are available.

1996

The future begins in Germany: Vision Components launches the first smart camera. Image data is processed directly in the camera and only the measurement results are delivered as output.

2011

Machine vision becomes key enabling technology in the context of Industry 4.0 – the next level of production.

2015

According to VDMA, the German machine vision industry achieves a new sales record and surpasses the 2 billion mark € for the first time.

1987

VISION – the leading trade fair for machine vision and international meeting place for the industry ever since – takes place for the first time.

2004

The change from analog to digital camera technology in machine vision begins with the introduction of FireWire technology. In the years to come, the resolutions of digital cameras will increase and enable the implementation of higher quality standards in the customer industries.

2012

While machine vision was previously limited to 2D applications, industrial sensors are now available that reliably deliver 3D data in one single image.

2016

Machine vision goes deep learning. At VISION, the industry’s leading trade fair, standard software products demonstrate how this new technology can be used profitably for machine vision applications.

Which kind of “eye” is needed for a machine depends on its tasks: Microscopy, mechanical engineering, medicine, environmental protection, astronomy or football stadiums – the application possibilities are almost unlimited. But all systems work according to the same pattern: The captured images are analyzed, identified and evaluated with suitable software.

Vacuum cleaner with a sharp eye

Small and integrated everywhere – embedded vision is machine vision, which is directly embedded in all kinds of devices (mobile phones, cars, vacuum cleaners etc.). The devices can now “see” and are therefore smarter, better and safer.

Development and History

READ MORE

Focus on progress

Machines that see and understand are already in use everywhere – from the detection of food in supermarkets to gigantic construction projects for tunnels or sky-scrapers.

10 reasons to use machine vision

Machine vision and seeing machines have developed at an above-average rate in recent years and have become indispensable in the modern industrial production field and in many parts of daily life.

1. High savings potential

Machine vision systems reduce costs. They often pay off after only a few months.

2. Maximum product quality thanks to 100 % quality checks

Producing quality non-stop, 24 hours a day, 7 days a week – expensive product recalls, product liability claims and image damagecan be avoided thanks to machine vision.

3. Safe production, reliable products

Machine vision guarantees safety – in the production process as well as in the end product.

4. Sustainability

Improved environmental protection through optimized use of energy and resources , , more efficient recycling and smooth material flow – machine vision makes it possible.

5. Stable and optimized processes

Recognizing trends and irregularities in production processes early on – machine vision paves the way for realizing the smart factory of the future.

6. Flexibility in production

Modern machine vision systems are flexible. Even batch size 1 becomes feasible.

7. User-friendliness

Specific programming skills were perhaps previously required. Simple operation, easy setup and seamless integration into the production process have been commonplace for a long time now.

8. A technology serving people

With and for people, constantly striving for safety, quality and efficiency – in the factories and outside. The optimization of traffic flows, the perfect swing for golf, the training of doctors, the inspection of moles, waste separation and recycling and many other applications we all benefit from machine vision!

9. Higher productivity and competitiveness

Modern production is automated. Only with machine vision can companies sustainably secure their competitiveness, prevent the migration of key technologies, create qualified jobs and capture new markets.

10. Ergonomic workplaces

Monotonous and dull tasks are performed by a seeing machine – machine vision systems support workers, ensure a perfect human-machine interaction resulting in a more advanced and safe workplace.

The future becomes visible

Machine vision systems have benefited for years from the permanent performance improvements in sensor and processor technology, from improvements in the transmission of large data volumes and from innovations in algorithms. Due to the ongoing progress in these areas, the application possibilities for seeing machines are constantly expanding.

In addition to use in production, seeing machines will also enable innovative solutions in many other areas in the future. There is no shortage of ideas for future applications.

10 reasons to use machine vision

READ MORE

How seeing machines will change our lives

Intensive research is already being carried out on many visionary applications, some of which are already in the testing stage. Here are some examples of visions for the future:

Enabling cars to see

Traditionally, the automotive industry has always been one of the pioneers in the use of innovative technologies. This industry has recognized the possibilities of image processing and uses them accordingly in their products, where it is playing an ever-increasing role: driver assistance systems are already ensuring increased safety.

In the future, cars and other mobile machines will become increasingly linked to networks and autonomous. This is only possible with image processing!

In times of increasing automation in production, design and quality assurance, machines, algorithms and entire factories become intelligent. In addition, with regard to highly automatic driving, intelligent interior monitoring and advanced driving assistance and comfort functions, our cars are becoming “seeing machines”. High demands on sensor accuracy and the efficiency and safety of computer-aided and intelligent image processing play a central role in Volkswagen’s efforts to continue producing first-class mobility products for its customers.

Dr. Peter SchlichtVolkswagen Group Research – Automated Driving

Look at that!

Fascinating numbers from the world of vision and image processing:

1

million

pixels is the resolution of today’s standard industrial cameras. The maximum possible resolutions are even higher.
Source: VDMA Machine Vision group

90

percent

of all human sensory perceptions are optical stimuli.
Source: Wikipedia

10

times

per minute is the human eyelid closed. In the course of his life, an 80-year-old person blinks her eyelids more than 420 million times.
Source: Wikipedia

100

percent

increase in ten years – turnover of the German machine vision industry doubled within the years 2008 to 2017.
Source: VDMA Machine Vision Brochure 2017/18

100,000

pictures

per second and more are recorded by special high-speed cameras, e.g. to make extremely fast processes visible.
Source: Mikrotron GmbH

24

pictures

per second is currently the usual frame rate for cinema films. The human eye generally no longer perceives delays at this speed.
Source: Wikipedia

120

million

rods perceive the brightness in the human eye. Around 6 million cones recognize colors.
Source: Wikipedia

Machine vision as a driver of Industry 4.0 has become an indispensable part of the modern factory and has already conquered our everyday lives. Whether traffic management systems, autonomous driving, waste separation and recycling or health care – machine vision makes it possible.

Dr. Olaf MunkeltChairman of the Board of the VDMA Machine Vision division

Related Links

VDMA Machine Vision Group

More than 115 member companies are organized in the Machine Vision Group, which is part of the VDMA Robotics + Automation trade association.

Go

Video: Image processing in tomorrow's agriculture

Drones equipped with machine vision can help farmers to optimize agriculture.

Go

VISION: The world's leading trade fair for machine vision

THE meeting place for the machine vision industry and enthusiasts. Takes place bi-annually in Stuttgart. Next time 6-8 November 2018.

Go

automatica: machine vision in combination with robotics and smart automation

The automatica trade fair in Munich shows how machine vision and smart automation ensures optimised production.

Go

Authors

Anne Wendel

Director, VDMA Robotics + Automation

Anne Wendel is in charge of the VDMA Machine Vision group with around 115 member companies. Her work focuses on statistical analyses, standardization, marketing activities, public relations, trade fair policy as well as networking events and conferences.

Peter Stiefenhöfer

PS Marcom Services

Peter Stiefenhöfer is the owner of the press office PS Marcom Services, which supports companies and institutions in the field of machine vision in their press work. Based on his engineering studies (production engineering) he has about 25 years of professional experience in machine vision and during this time worked as an editor for a trade journal and as a press and marketing manager in a leading European company in this field.