So, here is what mine looks like. Sorry my program is a little messy; I haven’t cleaned it up since developing it and I’m not really a programmer to begin with 
The UR5 is watching a moving conveyor for a part, and when a part moves underneath the camera, the camera sends x and y coordinates for the part in mm. The UR receives this information in a vector:
conv_find := [3 (for data points to follow, 0 if no data), 1 (for part found, 0 if no part), x (mm), y (mm)].
The x and y are given relative to the center of the photograph being the origin. This is important because the camera is mounted on the end of arm and is defined as the default TCP, so my get_actual_tcp_pose() returns the location of the center of the camera’s lens, in base coordinates.
Then I run this next portion of code to convert my camera’s position from base coordinates to feature plane coordinates. The feature plane I am using is called “conveyor_belt” and is oriented with its x and y axes parallel to those in the camera’s field of view. Because of the direction of the z-axis in my conveyor_belt feature pointing downward by default, I have to reverse my x and y from the camera and flip their signs, as you can see in the code.
Then I define variable positions for the robot to navigate to in order to pick the part. I do this inside conveyor tracking so that the UR is already following the part while it calculates positions. All I have to do is add the x and y location of my part from the camera, to the x and y location of my TCP in my feature plane.
Note, you can see that I have my fancy coordinate rotations commented out and just ended up using the direct x and y values from my conv_pos_inv calculation. It got too complex doing those so it is just easier to make sure your camera’s field of view and your conveyor belt feature axes are lined up so they point the same way. There is some unnecessary code as a result of my playing with this; it would have been simpler to define:
conv_hover := p[ Pc_x + x_conv, Pc_y + y_conv, hover_z_conv, Rc_x, Rc_y, Rc_z]
where x_conv and y_conv are the part coordinates from my camera, and hover_z_conv is a specified z coordinate that I define at the beginning of the program when I initialize variables. I put the rotation coordinates I calculated in there, too, because if I don’t, the UR does some very weird and unpredictable moves.
When I perform my move to the variable waypoint I defined, I have the feature plane I want to navigate in selected, and I use my picking end of arm tool as the TCP for the move. This navigates my picking tool right on top of the part, where the camera lens was previously hovering.
One other thing to watch out for: Make sure your camera is perfectly vertical with respect to the conveyor belt. If it is at an angle relative to your conveyor belt, that will introduce position error.
I hope that helps.
