Author Topic: Live vision  (Read 22016 times)

0 Members and 1 Guest are viewing this topic.

Offline JesseWelling

  • Expert Roboticist
  • Supreme Robot
  • *****
  • Posts: 707
  • Helpful? 0
  • Only You Can Build A Robot!
Re: Live vision
« Reply #30 on: December 17, 2006, 01:55:07 AM »
my 3d sonar was just a hypothetical "how would I do this"

If I were really going to do it I would go with a line  with a camera above the middle sensor and of course I would use
a pulse. The sensors I was looking at have a about a 15 angle so it would kind of be hard to get them to work well
at various ranges but I would probably want such a sensor for the 20cm to 150cm range so it might work out OK...
But I'm going to leave this for a side project some time when I get board and have some free time (like that will happen ::))

Offline dunk

  • Expert Roboticist
  • Supreme Robot
  • *****
  • Posts: 1,086
  • Helpful? 21
    • dunk's robot
Re: Live vision
« Reply #31 on: December 17, 2006, 05:33:27 AM »
so if you mounted all the receivers in a line you would not be able to tell how far above or below the line sensed objects are.
an object 1 meter above the sensors would appear the same distance away as an object 1 meter in front of them. (ok, the sonar unit's don't have that wide a field of view. i'm just demonstrating a point....)

to build the sort of 3d map we are talking about i think a triangle of sensors would be needed.

like Jesse, i have no plans to build anything like this. i'm just throwing ideas about.
i currently have the prototype of my laser ranging/mapping solution working quite well so don't really need sonar on my current project.

dunk.
(with freekin lasars!)

Offline Cognaut

  • Full Member
  • ***
  • Posts: 58
  • Helpful? 0
  • How do you know that I'm not a robot?
Re: Live vision
« Reply #32 on: December 17, 2006, 11:47:58 AM »
It's a big challenge to get a 3 axis system out of an array of 3 UNIdirectional transducers.  There's no doubt that it will take some careful engineering.  Part selection will have a big effect on the design.
« Last Edit: December 17, 2006, 03:29:26 PM by Cognaut »

Offline dunk

  • Expert Roboticist
  • Supreme Robot
  • *****
  • Posts: 1,086
  • Helpful? 21
    • dunk's robot
Re: Live vision
« Reply #33 on: December 17, 2006, 03:20:05 PM »
i don't agree.
while a sonar gives you very vague information on which direction an object is in it gives very accurate information on distance.
as a result i would ignore the directional componet of the results returned and concentrate on the distance. this reduces each sonar sensor to being a 1 directional transducer, not 3 dimentional like you suggest.
with 3 cheap sonar sensors returning accurate distance information it turns this into a software problem rather than a hardware problem.

dunk.

Offline Cognaut

  • Full Member
  • ***
  • Posts: 58
  • Helpful? 0
  • How do you know that I'm not a robot?
Re: Live vision
« Reply #34 on: December 17, 2006, 03:25:04 PM »
By directional, I mean that the sensor is only receptive to sound coming from a given direction.  Microphones of all sorts have charts describing their directional sensitivity.

If you rotate your whole array, the directional limitation can actually be an advantage.  It can help you eliminate secondary echoes.

With omni directional mics, you wouldn't need to rotate the array.  Your transmitter would still need to rotate so that the echoes would have a known angle of incident.  An omni directional transmitter would not give you the result you are looking for.

I would place a transmitter with a rotating reflector in the center of an equilateral trangle formed be the three fixed, omni directional mics.  One way to make the three mics omni directional is with cone shaped reflectors.

This is still only a two dimensional result.  To get the third dimension will take some doing.  The transmitter would have to rotate on two axis.  Or you can use an additional receiver to form an equilateral, triangular pyramid.  Or you can orient your three receivers on the three axis, rather than on a plane.  That would still require omni directional mics or rotation.
« Last Edit: December 17, 2006, 04:31:07 PM by Cognaut »

Offline Cognaut

  • Full Member
  • ***
  • Posts: 58
  • Helpful? 0
  • How do you know that I'm not a robot?
Re: Live vision
« Reply #35 on: December 17, 2006, 04:42:46 PM »
You can use two mics, with "ear lobes," for spacially aware listening in a passive system or add a transmitter and make it "active."

I've actually done this.  I took two mics, each with a reflector, amplified them and listened through headphones.  It's really interesting to hear the psychoaccoustic effect of repositioning them.
« Last Edit: December 17, 2006, 04:46:55 PM by Cognaut »

Offline dunk

  • Expert Roboticist
  • Supreme Robot
  • *****
  • Posts: 1,086
  • Helpful? 21
    • dunk's robot
Re: Live vision
« Reply #36 on: December 17, 2006, 05:14:06 PM »
hey Cognaut,
your edit clears things up.
i thought you meant "3 dimensional" rather than "3 (uni) directional".
funny how a single syllable can change the meaning of a sentence so much....
(*insert plan to build robots with better communications than puny human languages here.)

i maintain this could be achieved with fairly low quality sensors backed up by a lot of processing power though.
the algorithm you would need to design would have to match one echo on all 3 sensors as being reflected from the same object in a similar way that stereo vision algorithms need to identify that an edge in each picture is the edge of the same physical object.
definitely not easy but possible.

dunk.

Offline Cognaut

  • Full Member
  • ***
  • Posts: 58
  • Helpful? 0
  • How do you know that I'm not a robot?
Re: Live vision
« Reply #37 on: December 17, 2006, 05:26:49 PM »
Yep.  I try to be precise, but I'm only human.   ;D  (If I told you that I'm not really a robot, would you believe me?)

Omni directional mics are not necessarily higher quality or more expensive.  I'm just saying that the part selection will effect the design.
« Last Edit: December 17, 2006, 05:34:11 PM by Cognaut »

 


data_list