Date: December 2012
Location: Oshawa, Ontario
Company: UOIT
Team: Cameron, Andrew, Santiago
Objective: Detecting, Sorting, and Stacking Disks
Several Disks will be placed on a conveyor belt. The work cell must detect the disks and stack the first 8 it sees. The stacked disks must all have the same orientation, and be sorted by pattern.
Solution:
This work cell had 3 components on the table.
- Conveyer belt
- Overhead camera
- Epson ProSix PS3
The conveyor was controlled through a computer to run until the IR sensor at the end detected a disk. Once the belt had stopped, the overhead camera processed a frame to find one of two learned patterns: soccer ball and basketball. The pattern learning and image processing was all handled natively with the PS3 controller.
Once the pattern was detected, the position and orientation were passed into the path planning process that picked up the disk. Each disk was then stacked by pattern on the table.
As an extra bonus, we also used a shape detection method to ignore all none-disk shapes. We extended this into a workflow where it would test for circularity, then test for pattern. If it wasn’t a circle, it gets passed off the end of the belt. If it doesn’t have a recognized pattern, it was flipped over and re-tested. If it still did not have a pattern it was passed off the end.
Further Work:
There was a limited number of disks shared amongst the groups, so we could not test large volumes of disks. However, we could have built an auto-palletizing script to prevent the stacks from getting too high. Also, I really want to have to learn new patterns on the fly, but the software did not support that.