automatic sensing & steering application

Archiver

Hello all,

I was recently approached by an old friend who has a problem that may be
able to be solved with a Stamp. My friend has a sod farm. In the
harvesting of the sod, they drive a machine that stripps up a swath of
sod to a depth of about 1.5 inches. On the next pass, they must hug the
line that the previously made by within 0.25 inches. He was wondering if
I could help him develop an automatic steering system for him. They
currently have one, but it is not very reliable/easy to use. From what I
understand, the current system is a 'sled' that rides on the depth
difference created between the unharvested sod and the bare earth. The
sled sometimes gets lost (can't find its way back to the proper place),
it gets muddy, and/or must be raised/lowered at the end of every pass.
He was thinking a non-contact sensing and control arragement could be
very benefical.

I have had two thoughts.
1. Some sort of optical sensors placed about 0.25" or less from each
other, pointing down towards the ground. One sensor should constantly
detect the green sod, the other, the brown/black dirt. If either sensor
starts 'seeing' the wrong thing, a steering correction could then be
made. Obviously some over-ride method would be necessary for the turns.
Does anyone know of appropriate sensors that can do this?

2. Same basic idea as above, but using ultrasonic rangefinding modules
to pick up the depth difference. Is it possible to have two ultrasonic
sensors working in this close of proximity to each other? How
susceptible are they to dust and machinery noise? Are the Polaroid 6500
modules the way to go? Do they have appropriate resolution if placed
about 12 inches above the ground? Is it possible to focus the beam
pattern tightly enough?

Any comments, feedback, and/or other ideas are greatly apprecaited.

Regards,
Aaron Garber

Archiver

Wouldn't various shaped rods connected to heavy duty switches stand
up to the environment better? I would think you'd get a lot of dust
and grass interference that would keep you from getting the .25"
tolerance you need.

Archiver

I have seen a truck setup with a weed sprayer that automatically sprayed
weeds beside the road by sensing the color green they said they could
adjust it to trigger on any color they wanted. When I asked how they
sensed the color the guy just said it was a trade secret. So a search on
automatic weed sprayers may get the info you need.

Craig

Archiver

I could be wrong and i've never done it, but i seem to
recall seeing something to that effect on discovery
channel a year or so back (funny eh?)...Where they
could tell the state of health of plants and thereby
determine richness of soil via overhead flights. I
believe they used some sort of infrared readings, the
colors of the vegetation being the determining
factors.

Each different plant type would have it's own color.
Might be something to look into. good luck

--- Craig Gardner <cgardner@r...> wrote:
> I have seen a truck setup with a weed sprayer that
> automatically sprayed
> weeds beside the road by sensing the color green
> they said they could
> adjust it to trigger on any color they wanted. When
> I asked how they
> sensed the color the guy just said it was a trade
> secret. So a search on
> automatic weed sprayers may get the info you need.
>
> Craig
>
> To UNSUBSCRIBE, just send mail to:
> basicstamps-unsubscribe@yahoogroups.com
> from the same email address that you subscribed.
> Text in the Subject and Body of the message will be
> ignored.
>
>
> Your use of Yahoo! Groups is subject to
> http://docs.yahoo.com/info/terms/
>
>


______________________________________________________________________
Web-hosting solutions for home and business! http://website.yahoo.ca

Archiver

The technique is called spectral discrimination and sometimes hyperspectral
discrimination.
It works like this:
Take a picture/view a scene using a number of filters, of specific colors.
say green, yellow and red.
If you calibrate what the sensing apparatus generates as a signal for each
color you can discriminate
what you are sensing by the ratio of the responses.
A brick has lots of red and not much green or yellow
A dry leaf has little red lots of yellow and some green
A green leaf has little red, lots of green and somewhat less yellow.
some work from satellites uses hundred of colors for this work.
Early stuff from LANDSAT used multiple colors spanning the visble and
infrared.
I've done it using CDS cells in three arrays
the cells are set into arcs :

O O O O O O O O O O O O RED FILTERED CELLS
O O O O O O O O O O O O GREEN FILTERED CELLS
O O O O O O O O O O O O YELOW FILTERED CELLS

they stare out , cell for cell at the scene. which in my case was/is from a
low heigth rc car.
I'm looking for tennis balls on a tennis court. The balls are usually a
nice flourescent green.
Which makes them show up well in the green /yellow channels and poorly in
the red.
by ratioing the channels as the car moves it "looks" for and steers too
the highest contrast signal.

NOTE CDS cells work well for this type of stuff because they match the
response of the human eye well.

Archiver

Hi what you need is two photodiodes and op-amps arranged as IV
converters. Filter one to 650nm and the other 750nm. The chlorophyll
in the grass absorbs most of the red (650nm) light and reflects most
of the near-infra-red light (750nm). Soil on the other hand reflects
both wavelengths by approx the same amount. Stick the output of the
IV converter into an ADC then into the stamp. Take the ratio between
the two signals to determine if over soil or grass (ratio about 1 for
soil). Use two lots of these, and check that one is over soil, and
one over grass.

This is the way that automatic crop sprayers work. If you are in a
nice sunny place then it should be fine. If (like me) you are
somewhere that isn't so sunny (UK)it can be a bit of a hassle getting
the gains set correctly.

Look up NDVI (normalised difference vegetation index) on google or
another search engine to learn more.

Andy

--- In basicstamps@y..., agarb@j... wrote:
> Hello all,
>
> I was recently approached by an old friend who has a problem that
may be
> able to be solved with a Stamp. My friend has a sod farm. In the
> harvesting of the sod, they drive a machine that stripps up a swath
of
> sod to a depth of about 1.5 inches. On the next pass, they must
hug the
> line that the previously made by within 0.25 inches. He was
wondering if
> I could help him develop an automatic steering system for him. They
> currently have one, but it is not very reliable/easy to use. From
what I
> understand, the current system is a 'sled' that rides on the depth
> difference created between the unharvested sod and the bare earth.
The
> sled sometimes gets lost (can't find its way back to the proper
place),
> it gets muddy, and/or must be raised/lowered at the end of every
pass.
> He was thinking a non-contact sensing and control arragement could
be
> very benefical.
>
> I have had two thoughts.
> 1. Some sort of optical sensors placed about 0.25" or less from
each
> other, pointing down towards the ground. One sensor should
constantly
> detect the green sod, the other, the brown/black dirt. If either
sensor
> starts 'seeing' the wrong thing, a steering correction could then be
> made. Obviously some over-ride method would be necessary for the
turns.
> Does anyone know of appropriate sensors that can do this?
>
> 2. Same basic idea as above, but using ultrasonic rangefinding
modules
> to pick up the depth difference. Is it possible to have two
ultrasonic
> sensors working in this close of proximity to each other? How
> susceptible are they to dust and machinery noise? Are the Polaroid
6500
> modules the way to go? Do they have appropriate resolution if
placed
> about 12 inches above the ground? Is it possible to focus the beam
> pattern tightly enough?
>
> Any comments, feedback, and/or other ideas are greatly apprecaited.
>
> Regards,
> Aaron Garber