Scenario Automated Assembly with Zivid Two 3D camera - White Goods

PLAIF

Plaif enables new possibilities in appliance manufacturing with AI and 3D machine vision

Quick facts

Company
Plaif

Application
Consumer white goods manufacturing

Features
Bin-picking
Assembly
Pick and place

3D camera
Zivid 2 M70 - industrial 3D color camera

At a glance 

The manufacturing industry is seen as an early adopter of robotic automation, but there are a number of areas where the technical challenges were still too hard to overcome. Plaif has leveraged advances in AI and 3D machine vision to bring a high-reliability solution to these hard-to-fix challenges.

Challenge

Manufacturing often involves the picking and manipulation of small, complex objects. Until recently machine vision systems have not been of sufficient resolution and quality to reliably perform these tasks. Especially, small complex metallic parts present many vision-related challenges, and it is essential that they are seen with clarity and as they really are.

Solution

After trying out a number of 3D sensors, Plaif was unsure that there was a 3D camera that could deliver on their needs. Then they tried the Zivid 2 3D camera. The excellence of the point clouds and its range of filters ensured that their AI software was always able to work with the highest quality data.

Result

Plaif is currently working with a major Korean white goods manufacturer on deploying this application that uses the Zivid 2 L100. This system delivers highly reliable results and is due for full deployment across their manufacturing facility over the next couple of years.

 

We were looking for a true industrial 3D vision solution partner we could trust and develop together. I was impressed by the quality of the Zivid 3D cameras and how the team takes all the reliability and calibration tests seriously to deliver consistent performance” 

Tae-Young Jeong, CEO of Plaif

 

Schedule a Zivid demo

01 Story

Plaif is a Korean start-up focusing on AI for robots. The company aims to create a world where AI robots can handle simple and repetitive tasks so that humans can focus on creative and valuable work. Using Zivid’s 3D machine vision, Plaif enables a dynamic pick-and-place solution to reduce the cost of generating learning data and improve productivity through object recognition AI and robot behavior AI.

Plaif's bin-picking demo using the Zivid 3D camera (recorded 6 times faster)

 

Plaif’s solution is seeing adoption in large manufacturing companies in South Korea as it has proven its ability to solve the technical limitations that previously prevented automation adoption by using its innovative and advanced 3D and AI technologies. The company chose the Zivid 2 L100 industrial 3D color camera to solve bin-picking challenges in white goods assembly lines. Zivid’s superb-quality point clouds, excellent noise and artifact reduction filters, and true industrial-grade long-term quality helped Plaif deliver the 3D vision required for a range of robotic applications in demanding industrial environments. 

 

02 Challenges

Picking complex-shaped, hard-to-manipulate, and highly reflective metals is still a challenge for robots due to poor vision. Plaif’s customer was dealing with small brackets used in electronic white goods assembly, some of which were inconsistent in shape. They evaluated a selection of 3D cameras to see if the project was feasible but had little success. The quality of the point clouds was not good enough to capture a full scene with a true-to-reality representation.

Detecting these parts when densely stacked was not easy since the robot couldn’t distinguish details to pick and place each part safely. Low occlusion performance is another challenge. The shadow from a bin causes occlusion and prevents the camera from seeking parts in the corner, making it impossible for a robot to pick objects in that region. Lastly, the performance of the 3D camera had to stay consistent in rapidly changing thermal environments. The consumer-grade cameras were not reliable enough to be used in manufacturing factories, as various factors might affect their performance.

Scenario Automated Assembly with Zivid Two 3D camera - White Goods

“We were looking for a true industrial 3D vision solution partner we could trust and develop together. I was impressed by the quality of the Zivid 3D cameras and how the team takes all the reliability and calibration tests seriously to deliver consistent performance.”

Tae-Young Jeong
CEO of Plaif

03 Solution

Plaif chose the Zivid 2 L100 3D vision camera for their bin-picking application. They initially used the Zivid 2 M70 but switched to L100 to benefit from an extended working distance and larger working volume suited for bin-picking.  

Zivid 2 L100 was superior to other alternatives Plaif tested regarding high-resolution point clouds. It enabled minimal occlusion, capturing all the fine details for reliable object recognition and separation of boundaries for picking. 3D HDR and Artifact Reduction Technology (ART) also ensures excellent suppression of imaging artifacts from reflections, inter-reflections, specular highlights, and high-contrast distortions.

Based on Zivid’s accurate point clouds, Plaif can apply deep learning to detect object's exact 6D Poses and Reinforcement Learning to robot behavior AI so that it can recognize, judge, and move on its own. Reinforcement Learning is a machine learning technique that allows a robot to find the optimal method through trial and error. For example, if the light reflection is so strong that the point cloud data cannot be obtained, the robot can move the object to avoid it by itself. Plaif's AI solution includes simulation learning methods and cloud-based AI learning to reduce the cost of learning data. 

Plaif needed a vision partner they could trust for reliable high-performance over a long time. They learned that Zivid is designed for harsh industrial environments and takes long-term testing seriously by undergoing 100+ hours of arduous performance, reliability, and calibration tests in typical industrial settings. The robustness and reliability helped Plaif decide to replace previously tested 3D sensors with Zivid industrial-grade cameras.

 

04 Result

Plaif’s pick and place system will be installed in an automation system for one of the largest electronic companies in South Korea. The plan is to deploy the solution across multiple lines from 2024 once the pilot version proves successful.

At the same time, its 3D vision-based AI solution will be used this year in picking automotive parts for another manufacturing project. Plaif expects to solve a broad range of challenging automation problems by adopting its advanced AI and 3D technologies.

Schedule a Zivid demo

 

About Plaif

Plaif is a Korean start-up focusing on AI for robots. The company aims to create a world where AI robots can handle simple and repetitive tasks so that humans can focus on creative and valuable work. Using Zivid’s 3D machine vision, Plaif enables a dynamic pick-and-place solution to reduce the cost of generating learning data and improve productivity through object recognition AI and robot behavior AI.

→ plaif.com

Zivid 2 cameras

Contact Sales

Read more

Case Studies →

Watch more

Videos →

Browse

Point Clouds →

Share the story!