I have been talking about adding
vision to robots these last two columns
and I will continue with that theme this
month. There are many camera options
to choose from that are small enough to
fit on just about any robot. The simple
camera modules often output their data
in ways that our lower cost controllers
cannot handle. Some output analog
levels need to be converted on a pixel-by-pixel basis, and we need to track
horizontal and vertical sync signals.
Some output digital values for each
pixel, but use a compressed video
format that our microcontrollers lack the
processing power (or memory space) to
deal with.
With that in mind, I went searching
for cameras that could be interfaced to
our simple microcontrollers that can be
used with rudimentary vision algorithms
to do either navigation or object
detection.
The Gameboy Camera Hack
When we consider using inexpensive vision systems
with our inexpensive robot controllers, we should mention
the first hacks which made this possibility a reality over 10
years ago. I speak, of course, about the Gameboy® camera
hacks that were able to give small robots monochrome
vision. The first hack that I know about is detailed in the
SRS Encoder (Seattle Robotics Society Newsletter) in 2002
( www.seattlerobotics.org/encoder/200205/gbcam.html).
The article was by Dafydd Walters
and used a 68332 robot board. The link
will give you the details of how he used
the camera.
The second source I know
personally. Daniel Harrington showed
me his Gameboy camera hack in action.
Daniel used a custom controller board
with an Atmel 8515 that he
demonstrated following a white ping-pong ball. A write-up of his design is
actually on the Atmel microcontroller
site.
I tried to include the direct link here,
but the URL is really ugly. So, I suggest
that you go to www.atmel.com and
search for “issue4_pg39_ 43_
robotics.pdf” to get Daniel’s project.
Figure 1 is a picture of his ping-pong
ball tracking robot. His article can also
be found in Circuit Cellar issue #151
(February 2003); you can find his code
files at www.dtweed.com/circuitcellar
/ caj00151.htm.
This camera is particularly interesting because it had
edge-detection features that simplified using it for object
tracking. Because it is monochrome, you can pull an image
from it with our newer, faster microcontroller boards very
quickly to analyze images in real time.
If you like hacking your own stuff, I recommend that
you check these articles out and try it yourself. You can still
get these cameras on eBay for $7 or so. The total cost of
this project will be time and probably less than $10 in parts
(not including the microcontroller board).
by Dennis Clark
Our resident expert on all things robotic is merely an email away.
roboto@servomagazine.com
Tap into the sum of allhumanknowledge and get your questions
answered here! From software algorithms to material selection,
Mr. Roboto strives to meet you where you are — and what more
would you expect from a complex service droid?
Ask Mr. Roboto
r. Roboto is limping back from crawlspace floodwaters this month. The "mile
high" land of Colorado seems more like a mile deep these days. My spirits are
high, however, and I have nearly beaten the kelpies back from taking over the
house. Things should be settling down soon! As I write this column, the
SparkFun 7th annual AVC competition is coming up in a week. You can bet Mr. Roboto will
be there to cheer everyone on and watch for innovations that we can all pick up on!
Anyway, let’s move onward to this month's topic: robot vision.
M
Figure 1.
Your robotic problems solved here.
SERVO 08.2015 7
Post comments on this article at www.servomagazine.com/
index.php/magazine/article/august2015_MrRoboto.