3D machine vision blog - Zivid

How to succeed with Industry 4.0 automation

Written by Øystein Skotheim | 2018-08-08

Evolving in Industry 4.0

Small and Medium-sized Enterprises account for 99% of the businesses in every main industry sector. To compete and evolve, an ever-increasing number of SMEs are automating their production and adopting robots for automation.


Small-scale manufacturing and rapid production changes are factors that require new, smart and flexible robot and automation systems. Without compromising reliability, production cost must go down. Quality, accuracy, and profits up. Safe and effective human and robot interaction is on the forefront, and smaller collaborative robots, cobots, are outpacing traditional industrial robots.

The next era of industrial automation going towards Industry 4.0 raises the need for sophisticated vision systems and robots, which can recognize, evaluate and act in relation to the ever-varying surroundings and tasks. Tasks like bin-picking, packing, sorting, and assembly of a wide array of objects are incredibly complex due to challenging physical object characteristics such as pliability, irregular shapes, varied surface materials and colors.

Why everything hasn’t been automated 

When humans pick tiny objects from a bin, a complex, highly sophisticated set of synchronized actions take place. Our brain masterfully controls these actions based on sensory input from our eyes and via tactile feedback.

For robots, four crucial enabling technologies are necessary to replicate human actions:

  • A versatile manipulator arm
  • An advanced gripper
  • A set of eyes, a 3D camera with human-like capabilities
  • Artificial Intelligence (AI) controlling it all, the robot brain
Because our world is three-dimensional, robots and associated tools need to relate to their surroundings in 3D and cannot be limited by traditional, flat 2D images regularly used in machine vision setups.

Limitations of robot vision

Although different 3D cameras and scanners have existed for some time, present solutions have limitations due to unwanted compromises. A high-speed camera almost always comes with sacrifices in resolution and resolution accuracy.

Time-of-Flight (TOF) cameras and existing stereo vision cameras are considered fast, but they typically have millimeter and centimeter resolution range. At the other end of the spectrum, cameras that offer high resolution and accuracy, are slow and expensive. High accuracy scanners, for example, are often used in the automotive and aerospace industries.

These scanners are costly and take several seconds to capture an image. Additionally, most 3D cameras and high accuracy scanners only provide depth data with no color information.

The problem with 3D cameras has almost always been that seemingly good specifications don't meet the promises during real-life qualification. The output either is missing data points, or the result looks significantly worse than promised.

This mismatch between specifications and the actual output is because in the real-world, industrial objects have challenging and unpredictable physical characteristics. Challenges include the type of object, the material used, size, and complexity of the object's features.

The final problem is the quality of the object's surface, which varies significantly from matte and Lambertian, to reflective and shiny, to dark and absorptive, and even transparent or partly translucent. 3D data from a flat, white Lambertian planar plate (which often is where specifications are determined) and a semi shiny reflective, or partly translucent or highly absorptive surface, are often entirely different. 

Robot vision requires Quality of Data 

Understanding real-world parameters and designing a 3D vision solution that renders a true 3D representation of real-life objects is what separates today's machine vision experts. 

Cutting-edge technology and innovations in both hardware and software are based on 15+ years of fundamental research at SINTEF (the largest independent research organization in Scandinavia) was brought over to the spin-off company, Zivid. With the Zivid One 3D camera, Zivid introduced several new and innovative features into a small and cost-effective, robust industrial form factor.

Quality of Data (QOD) is a term that separates bad, good, and excellent 3D cameras. QOD represents the significant innovations found in the Zivid One, and Zivid's high-resolution technology offer the best QOD in the industry.

The "quality" component of QOD is measured by the degree of uncertainty in each measurement. For every pixel in the imaging sensor, there is a 3D measurement (XYZ) which is a true-to-reality representation of the object in the camera's view.

Zivid defines the gold standard for QoD by providing:

  • High absolute accuracy across the whole imaging sensor.
  • Accurate representation of shapes and forms, surface details, colors and texture.
  • Short baseline, which minimizes shadows and occlusions for both the camera and the projector.
  • Per-pixel processing: all analysis is done per pixel, without the need for analyzing spatial neighborhoods, eliminating block noise and providing true hd resolution.
  • High resolution data on any kind of surface, including shiny, reflective, translucent, dark and absorptive materials.
  • Smart patterns and software filters that eliminate noise and false points due to reflections and scattering

With the Zivid One 3D portfolio of cameras, Zivid delivers world class QOD plus all the innovative features described above. The results are clear - exceptional quality 3D in full HD and color at rapid rates (100 ms per image, including both acquisition and 3D processing time), and true-to-reality accuracy at 100 um.

With support for several High Dynamic Range (HDR) modes, combined with smart and adaptive projection technology and adaptive noise filtering, Zivid One can handle difficult objects of almost any material, whether they are dark and absorbing or shiny and reflective.

Successfully implementing robot and automation systems

As the market adopts the benefit of 3D vision, machine vision providers are faced with several challenges to be successful when implementing improved robot and automation systems. System integrators and their end-customers require 3D cameras that can withstand harsh work environments, varying light conditions and changing temperatures. Weight and size are also crucial elements, as heavy and bulky cameras cannot be mounted on robot arms without degrading load capacity or making maneuvering in complex environments difficult relative to a handling or interaction tasks.

The market has taken notice of Zivid, and recognized industry experts have awarded the Zivid One with multiple accolades. Independent juries from recognized industry magazines in Germany and US, with a global an unbiased view of the machine vision market, have chosen Zivid One 3D camera as a "Top Innovation" and "Gold Award Winner" in 2018. Red Dot picked the Zivid One as the best of product design.

The Zivid One 3D camera enables robots to see the world in three dimensions with the same, or better, visual acuity as humans. This radically increases the robustness and performance of existing automation systems, and high accuracy 3D enables entirely new possibilities for automation.

Market reception is best summarized by a few quotes from our customers, which is why we feel we have something special here at Zivid:

  • "It's too good to be true"
  • "This is the coolest 3D camera I have ever seen"
  • "I've been in the machine vision industry for 20 years, and I've never seen anything like this"