Wednesday, March 18, 2015

Some Ideas

I've had a few ideas. The bad news is, this is going to be pretty difficult. The good news is, it may be possible.

The main operation of the machine would take place on a circular table, maybe three feet or so in diameter. Basically, it has to be able to hold any article of clothing at any orientation. Somehow (still working on how) a single item of clothing is put on the table. The article of clothing will, presumably, be in a rumpled ball. The first task is to flatten it out.

This can be done as follows: A roller moves along the table and detects when it changes height. The first change of height indicates it has hit the article of clothing. Any further change in height indicates the clothing is not flat.

Suspended above the first table will be a matching table. This table, however, can move side to side, up and down, and rotate. When a "not flat" area is detected, the second table will press down on all previously scanned areas of the piece of clothing, sandwiching it between the two tables. It will then move forward, pushing the clothing to the edge of the bottom table, until the rumpled part falls off. Then, it moves backward, pulling the clothing back onto the bottom table. The roller then resumes it's task. This process may have to be done iteratively.

Once the clothing is flattened out, it needs to be identified. This can be done with some (I hope!) fairly simple computer vision. If a camera has a field of view composed only of the table, it should be reasonably simple to differentiated between "table" and "not table." Just subtract a previously recorded image of the table from the current image. The "not table" section can then be reduced to an outline and that outline can be centered. From there, it is (again, hopefully!) a straightforward thing to compare the given outline with the known possible outlines. Rotations are not a problem here--the given outline and all its rotations can be compared with the known outlines and just the best match kept. I don't know how long this will take to run, but the laundry folding machine really doesn't have anything else to do, so it doesn't matter if it's slow. (My intuition says it won't be, at least on a human scale.)

Once the piece of clothing is identified, it needs to be correctly oriented. The bottom table will have one half that can fold over onto the other half. Therefore, for any folding that needs to happen, the clothing will be oriented so the desired location of the fold is lined up with the middle of the table. The orientation can be done with the top table. Half of the bottom table then swings upward, completing the fold.

Once the folding is done, any required hanging can also be done. I'm still working on this, but at least I will have the clothing in a known position. Also, the folded and/or hung clothing then needs to be removed from the table. Again, still working on this, but I think that will be part of the hanging apparatus. Right now, I have a vague notion of some sort of clamp that can move through the hanger, grab the clothing, and pull, so that the clothing also goes through the hanger for pants. Not really sure about shirts. An opposite system might work--grab the hanger through the shirt.

Also, I'm not sure about that top table. It's an awfully cumbersome way of doing things. The roller might be able to move the clothes to the edge of the table itself, in which case the top table would only have to rotate the flattened clothing. This would be a significant simplification.

Anyway, I'm still working on it, but I hope this sort of made sense. I wish I had pictures to include, but the whole thing is too complicated and too vague (at this point) for my very limited artistic abilities. I will keep you posted.


No comments:

Post a Comment