To implement a vision system for object recognition, what do you recomend for me to take a look at? I am new to robotics and the learning curve is steep, fun and sometimes frustrating (but still fun). I read that for this I should use URCap. I followed the tutorial ‘My first URCap’ and understood this for the most part and also made my own UI to test if it works as expected. However I did not learn yet how to implement a camera and do (simple) object detection. My understanding is that this should also be done in the URCap but I am not sure if this is the right way or maybe there is a better way.
The goal is to to object detection with the camera all within the robot (so no external PC or server to process the camera image). What or where do you suggest I learn or read to accomplish this? Should I dive deeper into URCaps or is it possible with the URScript? I ask you because I already spent way to much time before finding out about the URScript and the articles that explained how it works. Since you have way more experiance you probably know what I should look at.
What you’ll want to look into is something called Daemons.
This is background task running in python or c++ (assumming polyscope 5 or 6?)
with a python daemon you could then connect to a USB camera and use feature matching in opencv (as one option).
you then need an installation node to create urscript functions that connects your URScript with the daemon and the vision code running on it.
cg1’s answer covers it already quite well. Maybe a couple of things to add here:
1.) What exactly is it you want to do? Are you working on a DIY Project and want to do everything yourself, or do you want to solve an Application that will work in a Productionline or something?
2.) While it is definetly possible to have everything running on the UR, it is important to keep in mind that the processing power of our ControlBoxes is quite limited. So this will give you some constraints down the road, especially regarding the time to process the image.
3.) There are ready to use Products out there that already have a Camera a URCap and everything you need to get going with your project. Products
Hi Florain,
Thank you for your response (and thank you cg1).
I am doing this project at a university for a grocery store. The application is a ‘proof of concept for a pick and place operation for shelve items within a grocery store.’ The proof of concept is just a grocery shelf with grocery items that is a mess. The robot needs to clean up the mess by re-allocating the products.
Tank you! based on this it might be better to let the vision system run on an external PC since a grocery store has a lot of products. I’ll convince them of this.
I am aware of this. Pickit is one of the options the grocery store was looking at as a solution however thier buget for this project is slim since it is only a proof of concept. If I succesfully finish this project they will do more research with more budget.