3D cameras revolutionize companion robots – David Chen of Orbbec [Interview]

3D camera

Imagine a friend that can diagnosis your illnesses, play with your children, and follows you around to provide a cold drink whenever you need it. These kinds of friends exist – they just aren’t humans.

From Heineken’s beer dispensing droid to the AI-enabled physician’s helper, companion robots seamlessly fulfill specific human needs and functions. How? 3D cameras.

David Chen, director of engineering at Orbbec, says 3D cameras are what allow these digital sidekicks to analyze, navigate and monitor both their environment and the humans around them.

In an exclusive interview with RoboticsBiz, David Chen predicted that while companion robots are currently playing a specific and important role in assisting the elderly and those with disabilities, we are not off out from robots providing multifunctional roles in most households. Chen believes that the advancements in 3D cameras will continue to enhance the capabilities of companion robots.

Read the complete interview below:

1. 3D imaging is one of the hot topics in robotics today. Unlike 2D, 3D vision allows a robot to detect the orientation and recognize an object with complex geometries and reflective properties in low light conditions. Can you give us an idea about the new breakthroughs/advancements happening in 3D sensing technology today?

Depth-sensing solutions have already been widely used in robotics technology, but typically people are referencing LiDAR technology. This is a point or line-based technology which has been used for some time combined with RGB/2D vision systems. This combination provides robots with the measurements needed to recognize and navigate their environment.

The downside of LiDAR is always related to the performance-cost ratio. Full-field high-performance LiDAR usually has a higher price compared to the low-cost LiDARS, which are usually only capable of single point (scanning) measurement.

What we are seeing right now is 3D cameras gradually taking the place of conventional RGB/2D cameras and LiDAR combination models. 3D cameras provide a much more detailed and reliable vision system with the ability to gather three-dimensional full-field information, which robots can then use to analyze with a high level of accuracy. The imaging ability of 3D cameras is constantly improving – with the ability to detect and analyze further distances and capture accurate images with fewer concerns for environmental conditions. On top of that, 3D cameras are more competitive in cost while providing similar functions to the current RGB/2D camera and LiDAR combination.

Now for the robotics industry, there are specific 3D advancements being made that are very exciting. One of the most important advancements is deep learning-based 3D reconstruction. Unlike conventional 3D reconstruction methods, deep learning-based 3D reconstruction can provide very detailed results with an acceptable accuracy rate – while conventional 3D reconstruction can provide higher accuracy but less detailed 3D data. Such detailed 3D data for tracking and recognizing are urgently needed by robotic developers.

2. In May, Orbbec announced a collaboration with Microsoft to introduce new cameras in 2022 to satisfy the burgeoning demand for 3D sensing technology. Can you tell us about this collaboration?

This is a very exciting time at Orbbec as we begin a great partnership with Microsoft. We are collaborating with them to develop a new series of high-performance 3D cameras for developers and solution providers worldwide.

Encompassing Microsoft’s market-leading 3D sensor technology and Orbbec’s manufacturing and unique embedded computing design capabilities, we will introduce new ToF cameras next year. The camera will securely connect to the Microsoft Azure cloud platform and leverage device management, data streaming, and AI analytics capabilities that can run advanced depth vision algorithms and use onboard computing to convert raw data into precise depth images. Designed for advanced human/machine interface, robotics, 3D scanning, and surveillance use, as well as gaming and other consumer applications.

3. What role does Orbbec play in developing companion and service robots, especially in human-robotic interaction?

Orbbec’s main task in this industry is to provide early-stage imaging for robotic vision systems. We are creating scenario-oriented cameras that are specifically designed to help a robot complete its programmed task. What makes our solutions unique is that we can also incorporate embedded computing solutions in our camera hardware, such as single-board computers, to help the customer build a customized vision system.

We are currently working on a series of accessories that can accompany these systems. For example, precisely manufactured optical frames/mounts will help the robot builder or manufacturer set up the vision system more easily. And, we are also able to provide some manufacturing services to those in the robot companion and service industries – especially for smaller companies lacking supply chain capabilities.

4. Can you tell us about your new Time-of-Flight (ToF) camera product line? 

Time of Flight (ToF) depth technology has evolved over the last two decades and is commonly being used for SLAM, navigation, inspection, tracking, object identification, and obstacle avoidance. It is a powerful vision technology for assisting in the evolution of the robotics industry.
Earlier this year, we introduced a new ToF sensor, Femto, that can scan objects with high accuracy at a depth-of-field ranging from 0.2 to 5 meters. Its state-of-the-art depth algorithm makes it perfect for all kinds of applications. This expertly extends the use of 3D imaging for a wide range of environmental requirements, such as high-temperature environments and complete darkness.

Orbbec is unique in the marketplace for its ability to deliver ToF cameras with embedded computing capability. Unlike many rival ToF cameras, Orbbec Femto can output high-quality depth data without other external computing capabilities, giving users significantly greater design and application flexibility.

5. Do you have any predictions about the future of 3D sensing technology? What will the future look like in the next five years? 

In five years, 30 percent of RGB/2D cameras will be replaced with 3D cameras. 3D depth cameras provide one more dimension of image details for robots to “see,” which is particularly critical for the industry. Additionally, 3D systems address growing privacy concerns. Anyone can read or view a 2D image. But 3D images rely on a point cloud system which makes personal information (images) no longer required. Therefore, 3D vision systems provide an extra layer of privacy that people will be looking for in their imaging solutions.

As more depth cameras begin to replace traditional RGB cameras, the cost and barrier for advanced technology solutions, like companion or service robots, will drop. This gives more industries, at every level, the opportunity to incorporate these solutions into their business.