What is robot vision?

Robot vision refers to a robot's ability to see and sense the world around it. To do this, robots use cameras and image processing software to analyze their environment and navigate on their own.

The captured images and video are then processed by an onboard computer, which works analogously to a human brain. 

The computer uses machine learning algorithms to identify what’s going on around it and navigate without bumping into things or falling flat on its face.

 Some of the techniques include:

• Object recognition: Identifying specific objects like vehicles, traffic lights, people, etc.
• Motion detection: This allows the robot to “tell” what’s going on in the environment, which lets it sense and avoid obstacles.
• Depth perception: The bot utilizes stereo cameras or depth sensors to determine the distance of objects (something we’ve all struggled with more than once). This 3D sensing capability is important for navigation and manipulation.
• Scene understanding: Analyzing the relationships between objects in an environment for better navigation.

‍

How does robot vision work, exactly?

This is the fascinating, technical part. Robot vision systems typically consist of one or more digital cameras that capture images of the surrounding environment. 

Some systems even use stereoscopic vision, with two cameras mimicking human binocular vision to give robots depth perception.

Then, specialized software takes care of the magic and processes these images to detect, recognize, and track objects. 

What information does the software extract?

The software analyzes the images to extract all-important details such as:

• Color, shape, and texture of objects
• The location and orientation of objects in 3D space
• Motion and the speed and direction of moving objects
• The identity and classification of known objects

This information helps the robot understand what's happening in its field of view to interact with the world — and keep humans safe if it’s working collaboratively with them. These robots are called Cobots, precisely for their ability to join forces with people.

(Bear in mind, though, that not all Cobots have vision systems, but it does tend to make them safer.)

They can get better over time

“Machine learning” is not just an overhyped Silicon Valley term. It’s a very real thing. Many robot vision systems use machine learning algorithms to learn on the job.

As the robot explores the environment, its software logs images and uses them to teach itself. The more images the system sees, the better it gets at doing its thing: object/image classification and navigation.

Some robots allow you to manually train them by showing them images and telling them what’s in them. Then, after some time, they’ll be able to recognize these objects on their own.

‍

Is robot vision useful? 

Absolutely. In the same way that vision is useful for humans, robot vision opens up the door for myriad applications.

Check out some of the ways vision capabilities create new opportunities for robots to handle a variety of tasks:

Now they know what’s up

We’ve established that robots can identify and classify objects using computer vision and machine learning algorithms. 

But another impressive feature is that by being trained on massive datasets of images to learn the features of objects, they can spot them even in new situations. 

This type of object recognition is useful for tasks like sorting waste, detecting flaws in manufacturing, and identifying potential dangers.

Moving around

Robot vision lets bots build a 3D map of the space around them and use visual cues to determine their location and the best path to a destination while avoiding collisions.

This is doubly important when robots need to work in close proximity to humans.

Some real-world examples? Autonomous mobile robots with navigation capabilities are now being normalized in warehouses, hospitals, airports, and that Tesla you keep thinking about buying.

Dangerous inspection duties

Robot vision enables robots to handle visual inspections in hazardous environments or for highly repetitive tasks that would boggle the mind.

Think about inspection in confined spaces where not even the slimmest, most nimble humans can’t access. Examples include pipe inspection, turbine inspection, and underwater inspection.

Shepherding

Some robot vision systems are not for the robot itself — they help guide other robots or equipment. 

They can track the movement of parts or materials and provide real-time guidance and coordination to other parts of an assembly line or industrial process.

For example, vision-guided robotic arms are now normal in many manufacturing applications like electronic component assembly, precision welding, and automotive manufacturing. 

But let’s not get ahead of ourselves!

How is robot vision reshaping the industry?

To say that robot vision is useful for industrial applications would be the understatement of the century. 

In particular, manufacturing and warehouse environments benefit hugely from automated visual inspection and guidance.

Here are some applications for these googly-eyed robots:

QC, or quality control

Robot vision systems are way better than even the most eagle-eyed humans for inspecting products to ensure consistent quality standards. 

They can scan for minute flaws, check that components are properly assembled, and double-check that products are the correct size, shape, and color. 

For instance, Coca-Cola uses Cognex's DVC 500 vision systems to scan every bottle on its production line.

The result? Higher overall quality and reduced waste.

Warehouse automation

This is where it gets good. Warehouse robots equipped with vision are ideal for transporting and sorting inventory, machine tending tasks, pick-and-place, and even palletizing. 

They can scan items to identify them, then navigate to the proper storage location to retrieve or shelf them. 

As we mentioned above, vision lets bots avoid obstacles and safely operate around human workers, minimizing accidents and opening a universe of new collaborative possibilities.

As an example, Amazon relies on Kiva robots equipped with Amazon Vision Kit to navigate its massive fulfillment centers, automatically retrieving and delivering items.

Assembly guidance tasks

Robot vision is especially important for guiding robotic arms and manipulators during assembly. 

Through their image analysis capabilities, a vision system can determine the correct parts, sequence, and orientation needed to assemble a product. 

Then, the robot (typically a multi-axis robot arm) can then execute the necessary motions to put the components together. 

BMW, for instance, partners with Siemens to use its Simatic Robot Package with integrated vision for the precise assembly of electric vehicle batteries. The system guides robots to accurately place battery modules and connect them securely.

Moving around on their own

Sounds sci-fi, right? But already, some mobile robots are using vision to navigate autonomously. 

Autonomous navigation with robot vision is useful for transporting materials in factories, hospitals, and other structured settings. Oh, and Cybertrucks. 

And since we don’t want to run Tesla examples into the ground, we’ll mention that Mitsubishi Heavy Industries developed the AGV Mover, an autonomous robotic cart using vision sensors to navigate hospital corridors and deliver medical supplies and equipment to various departments. 

‍

Our top 3 robot vision winners

The mind boggles at the amount of systems with robot vision out there. Lucky for you, we’ve compiled a list of the best 3 on the market right now. 

Take a look!

FANUC CR-3iA: Don't let its size fool you. This compact dynamo packs a vision punch. Its iRVision system, a high-resolution camera, and brainy algorithms can spot any misplaced screw or faulty weld. One of the best picks for quality control.

Universal Robots UR10e: True to Universal Robots’ track record, UR10e packs a vision punch too. Its URVision software and built-in camera let it read barcodes, track objects, and even follow visual cues with minimal programming. 

Standard Bots' RO1: Don't underestimate the power of a good foundation. RO1's stereo camera head is one of the best in its class. RO1 tracks people, learns its environment, and skillfully tracks objects for pickup. The best part? It requires absolutely no coding to work its magic!

‍

Summing up

We hope you’ve enjoyed this quick primer on how robot vision works and what these advanced systems are capable of. 

As you can probably imagine, robot vision has enabled huge leaps forward in robotics and made many of the robots we interact with today possible.

Not only that, but robot vision systems are making huge strides forward. Not only are they helping assemble most of our day-to-day conveniences with better quality than ever, but they’re also paving the way for new, safer means of collaboration, transportation, and more.

Plus, they are now much more affordable and widely implemented, allowing small and medium businesses to compete with established industrial giants on an even footing for perhaps the first time in history. 

‍

Next steps

Interested in bringing an automated system with robot vision to your shop floor? RO1 by Standard Bots is the best choice for industrial businesses of any size.

• Affordable: RO1 is the most affordable robotic arm in its class, starting at almost half the price of its closest competitors. 
• Powerful: RO1 is faster and more precise than competitors, with high-grade robot vision systems and a best-in-class payload of 18 kg. 
• Easy to set up: RO1 is extremely easy to use straight out of the box, owing to a revolutionary no-code framework that requires zero coding expertise. 

Speak to our solutions team today to organize a free, 30-day onsite trial and get expert advice on everything you need to deploy your first robotic partner.