How to build a vision driven system in ROS

asked 2016-05-13 12:14:49 -0600

ami gravatar image

I am new to ROS, trying to build a vision driven robotic system (UR10). The application is to control the UR10 motion based on the images captured by a webcam. The webcam captures images of a certain object in the environment, go through image processing using Opencv, based on the results from image processing, UR10 will plan its path to follow/track the detected object.

I have tried to learn ROS by going through the ROS-I (Indigo) Training exercises. I believe my application is similar to the perception-driven manipulation demo in the training exercise, excluding using the gripper. But I still don't know how to build my application from scratch. Should I build a MoveIt package for UR10 using MoveIt setup Assistant first, then write a node to control the UR10 motion? But how to transfer the image processing results to the robot?

Is there anyone here can give me some guidelines? Thanks!

edit retag flag offensive close merge delete