Because processor and memory costs continue to decrease as their processing power increases, programmers can do more than just take an image and compare it to a corresponding reference model, whereas earlier large Some 2D visual images are based on this technology.


development of the visual system

The robot vision system was first applied to the complete set of body production robots of the automobile production line. Without a vision system, special tools are needed to machine apertures in the car body so that the robot can know the exact location of the car body.


However, after the development of the vision camera system, such expensive machining tools are no longer needed.

The robot is able to automatically determine the exact position of the car body and then mathematically calculate the positions of the four apertures.


Fanuc's Mr. Roney further explained: "From a very early age, we recognized that the adoption of vision systems was a great way to reduce tooling costs. Machining tools are very expensive and fixed. If a manufacturer needs to change a model the following year, then all the tooling has to be updated.” According to Mr. Roney, by adopting vision technology, automakers can save millions of dollars a year in mold and jig manufacturing.


How the Vision System Works

Vision systems use computing to recognize images, and then, through training, teach robots to recognize things and find what they need. The image is built on the basis of a large amount of pixel data, each pixel in the series has a gray level, and then the data is analyzed by arithmetic techniques.


The robot can determine where the image was taken, so it can identify the location of the object, and then determine its corresponding size, shape, and quality; it can also change the program based on the image and algorithm, for example, parts of different sizes can take different paths , part A will drop where part B should have fallen.


David Dechow, President of ApturaMachine Visioning in Lansing, Mich, explained: "Robots with vision systems will eventually be able to manipulate any type of part in any direction.


"Using the included camera system or remote control camera, it is possible to take a quick shot of an object and find its relative position to the robot.

"The FANUC robot can then use this positional data to determine where the object is, no matter where the robot is moved," said Mr. Jeremy Pennington, a control engineer at GuideEngineering, Inc., Ft. Wayne, Ind.


Application of 3D Vision System

3D vision systems can be used in many fields, among which the picking of parts in containers, the loading and unloading of workpieces on machine tools, and the fields of packaging and welding have achieved ideal results.


1. Applied to parts picking in containers

Until now, robots have mainly been used to pick randomly stacked parts in containers. To achieve this goal, three basic elements need to be met: vision system, container anti-collision and anti-collision monitoring system. The need for a vision system is a no-brainer, because the first thing you need to do is find where the part is. However, relying only on the container wall for inspection has certain limitations, as the robot's gripper will also penetrate further into the bottom of the container, so the picking of parts will become more and more difficult. The styling used to pick parts in a container fully understands the limitations of the part picking tools, sensors, and robotic arms themselves. Under such working conditions, once the position of the part is determined, the robot begins to perform automatic calculations to determine whether it can actually pick the part from the container. The third element is anti-collision detection. Eventually, the manipulator must touch the container wall, so it needs to distinguish between soft and hard contact, which could damage the robotic system.


"We've had great success using robots to pick parts both structurally and arbitrarily," says Fanuc's Mr. Roney. Structural picking means that every part in the container is face-up. , and the arbitrary picking of parts means that the parts are randomly stacked in the container. Mr. Roney believes that the latter part picking method is more challenging, but with the three elements mentioned above (vision system, container anti-collision and anti-collision detection system), everything is possible.


2. Applied to workpiece loading and unloading on machine tools

Mr. Burg of Ellison Technologies said: “In many applications, after a part is picked, it is directly mounted on the machine for processing. In most cases, the machine’s gripping system does not allow the robot’s part-mounting position to appear. Any errors." Therefore, the precise positioning of the part is very critical to the part's clamping system.

The robot is able to determine where the image was taken, so it can identify where an object is located, and then make a corresponding judgment on the size, type, and quality of that object.


Without these functions, the machine tool will likely suffer costly failures. Therefore, the use of vision technology facilitates the correct positioning of parts.


3. Applied in the field of packaging

Vision technology is also critical to packaging applications. For example, food often needs to be transported to the food picking area through a conveyor or slider device, so there is no problem of repeated positioning. But in fact, because the food is in a different location, there needs to be a process of reorientation and picking, and then putting it into the box. The vision system can help the robot find the corresponding food and complete the packaging task as required.


4. Applied in the field of welding

When welding, the robot can use the vision system to adapt to subtle changes between the two welding components. Even in spot welding applications, errors can be corrected using vision technology.

Application of the system in practice

Ellison Technologies Automation installed a robotic system with 3D vision technology to handle the sheet metal. The sheet is cut from the original slats, then dropped into a bucket and transported to a workshop equipped with robots, where welding techniques are used to weld a layer of hard material on the sheet to give it a long service life.


"The welding process can contaminate the clamp," explains Mr. Burg. "If the clamp hits a new part, that part is contaminated." Before vision technology, the necessary adjustments had to be made manually. Now, every location on every part can be verified. “The technology allows the robot to display the condition of the entire part through a 3D camera system,” Mr. Burg continued. “It allows operation without human intervention.”


In the 10 years prior to adopting vision technology, the company had four Fanuc robotic systems, at least one of which often required manual adjustments. Now, the company only needs three robotic systems to do the same amount of work, with very little human intervention.


The future of 3D vision systems

With further reductions in production costs and the adoption of more powerful systems, what will happen to the future of robotic 3D vision systems?

Fanuc's Mr. Roney replied: "Maybe there will be more interesting so-called visual servo systems. So far, the vision systems we have considered for applications only look for an object in time through a certain point." However, Parts tend to be in motion, and a single-point on-time fast camera may not work well at this point; a visual servo system can continue to obtain a variety of information about the object's position. This system allows the robot to steer and know where the object is at all times.


"Sooner or later, this system will find its way into assembly lines, where the assembly is always suspended from a constantly running drive chain or other material conveying device," suggested Mr. Roney. With further development, it is clear that the processing capacity of this visual servo system will continue to increase, and the production cost will also gradually decline. In the next few years, using this combination will make 3D vision technology more widely used.