New updates to improve your Wrist Camera’s speed and accuracy

JeanPhilippe_Jobin 2017-06-06

The Robotiq Wrist Camera is about to celebrate its one year anniversary. Almost 12 months have passed and we have learned so much about how to improve our vision system.

This week, we’re releasing a software update that will improve speed and accuracy for the Wrist Camera URCap. You can download the update now on Robotiq’s support portal.

Here are the new features:

A Parametric Teaching Method: Allowing the user to input the dimensions of a 2D shape (circle, rectangle, square and ring). The vision system then locates your parts by finding this 2D shape on it without any scanning, which is way faster. With a simple input of the shape’s dimensions, the system is able to recognize parts with few distinctive features.

Faster cycle time thanks to a fixed lighting setting: This allows to skip light adjustments between each iteration, which saves a lot of time when the illumination conditions are constant in the working environment.

Improved accuracy: By improving the calibration procedure, we’ve brought pick and place accuracy to a new level, allowing the vision system to move objects a lot more precisely.

You can download the software update now from Robotiq’s support website.

Try it out and let us know how you like it!

Chad354 2017-06-15

Hi Jean-Philippe,

With the software update and improved calibration what would the cameras new accuracy be rated at?

While I have you could you tell me at what height the camera would need to be from the work plane to get the maximum FOV?
The manual gives the height for the minimum FOV but I didn't see where it gave the max.

Thanks in advance,
Chad

David_Levasseur 2017-06-15

Hi @Chad354, theoretically, there is no height limit for a FOV. Realistically though, the limit will be the reach of the robot.

JeanPhilippe_Jobin 2017-06-15

@Chad354
The system accuracy values are presented in our Product Manual. Note that :

Those values represent the Typical Maximum Errors
Those values are not related to the camera only but the complete vision system, including the robot. So for instance, it includes the accuracy of the robot;
All measurements have been taken for a 180 degrees rotation where the error is typically more important;
All measurements have been taken inside the calibrated region (region defined by the four colored circles of the calibration plate);
Measurements have been taken at two different heights relative to the calibration plane (-3 and 80 mm);
Typical maximum standard deviation is 0,00009 mm
Note that with perfectly planar workplane and calibration grid as well as with a robot that has no excessive backlash in the joints, we can get better accuracy than the typical maximum values presented in our manual.

In order to have the best accuracy, we recommend to have a field of view which is the size of the calibration grid. If the accuracy is less important you can increase the FOV almost without limitation. If you need to cover a wider surface but with high accuracy, you can use many snapshots positions in your program.

mhowe 2017-06-16

Is it possible to have variable that lets the robot know there is no more parts to pick?

Robocase 2017-06-16

We updated our software and trained a part with the parametric teaching method. The identification of the part is now very fast at about 1.5 to 2s!

Etienne_Samson 2017-06-17

@mhowe add a counter variable to your program, put an if / else logic, the camera locate is in the if, the counter is in the else. Thus you will count how many time you do not see a part. You could use this so that after 5 or 10 photos without objects you can trigger your part feeding system.