Structured Light Experiments (LiDAR Attempt). / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

Structured Light Experiments (LiDAR Attempt).

previous 1 2 3 4 next last

Move Thread

LAN_403

Patrick

Sun Nov 25 2012, 09:08AM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Pinky's Brain wrote ...

There are a lot of cheap laser diodes both at Roithner and Ebay between 800 and 1000 as well.

you saved me with this Roithner link!! TY!!

EDIT: its 1:40am and im drunk, so i might be \wrong, but CAD is indicating i can get the laser width down to 12um at the sensor plane, so that means 3 or 4 pixels will be lit buy my 2mm wide line laser beam... thus about 1 inch of resolution should be possible.

Patrick

Sun Nov 25 2012, 07:36PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Added a animated .gif, will soon upload -- experiment 2 (long range, dark object and thin pixel lighting.)

First major problem, i dont know how to calculate the width of the laser light as it falls on the pixel array. At 2.9 meters distance, the laser is 1.9mm wide, but once it strikes an object and returns to the lens, how do i calculate its width? the pixels are 4um wide, and im hoping for about 8-12um beam width on the sensor.

Steve Conner

Mon Nov 26 2012, 01:05PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Care to explain what "structured light" means in this context and how your experiments indicate success?

Patrick

Mon Nov 26 2012, 10:12PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Steve Conner wrote ...

Care to explain what "structured light" means in this context and how your experiments indicate success?

i measured a 4 inch y-distance, with a 1.4 x-distance, thus prooving a long triangle can measure a long distance with a short side/ angle change.

Now i need to demonstrate that 4um pixels can be lit up, with a 10-14um beam...

Patrick

Mon Dec 10 2012, 02:25AM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

ok with the help from forum users here, ive been able to get enough math and mechanics, optics, and such to make a first attempt soon.

This first effort/device will be called : "prototype 1-A" (visible light tests)
then, "1-B, a full IR and pixel array, with no rotation.
After i learn enough, and if successfull ill move on to rotating in version "1-C"

pics to follow shortly...

iliasam

Mon Dec 10 2012, 06:54PM

Registered Member #8467 Joined: Mon Dec 03 2012, 05:36PM
Location: Moscow, Russia
Posts: 9

I am also trying to create triangulation based lidar.
The work principle of my lidar is the same like in Neato.
If somebody is interested, here some information about it: http://roboforum.ru/forum10/topic12095.html (in Russian)

As you can see, my lidar is already working, but it's resolution became low in big distances (> 3m). It's caused by low resolution of image sensor (TSL1401 - it's resolution 128 pixels).
So the question - where is it possible to buy image sensor with bigger resolution?
You wrote about DLIS-2k, but I can't find, where to get it (especially in small quantity).

Patrick

Mon Dec 10 2012, 08:38PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Very cool work !!!

iliasam, ive PM'ed you with some details.

I would like to work with others on this matter, if possible.
Next, how did you find out about my work?
and finally, your work is cool, it appears you have considered how we are to get a SLAM model working. should we use desktop/laptops, or embedded systems?

I see even with only 128 pixels, your still getting good data. youll have to see my thread here :

(there are two pages)
which documents the "range problem" and yes more pixels with a smaller pitch helps...

A few questions if you dont mind:

-how did you align your optics?
-are you using a line or dot laser? (one of your pics in the dark shows dots)
-are you using the STM32 MCU ?
-whats your RPM on your belt driven turning table ?
-How are you getting the data to the base robot? I plan to use IR .
-which pixel sensor are you using?
-what wavelength of light?

that tricylce bot is cool too!

iliasam

Tue Dec 11 2012, 10:21AM

Registered Member #8467 Joined: Mon Dec 03 2012, 05:36PM
Location: Moscow, Russia
Posts: 9

how did you find out about my work?

It's interesting, but I think that if you try to use line for scanning, image sensor wouldn't get enough light.

it appears you have considered how we are to get a SLAM model working

I have connected my lidar to ROS (running at PC): http://roboforum.ru/forum10/topic12095-60.html#p253548
But I can't normally start SLAM node - the resulting map was very bad (I think that it's caused by low lidar speed).

-how did you align your optics?

As you can see on photos, laser has a simple holder, which allow to change beam direction by rotating two screws. Also, laser holder can be rotated to the left/right.

-are you using a line or dot laser? (one of your pics in the dark shows dots)

Yes, it's simple dot laser.

-are you using the STM32 MCU ?

Yes, the controller is STM32F100C4. It's enough for 128 pixels senor, but for bigger resolution sensor more powerful controller would be necessary.
For DLIS-2K would be good STM32F4 controller, because it's has DCMI interface. GPIO is too slow for such data frequencies.

-whats your RPM on your belt driven turning table ?

It's slow - 3 rotation per second.

How are you getting the data to the base robot? I plan to use IR .

Now I'm using bluetooth module, but I'm thinking about slip ring (like in Neato).

-which pixel sensor are you using?

TSL1401

-what wavelength of light?

I'm using red laser from DVD drive, because usual 3 mw laser module doesn't give enough light. I cant measure power of my laser, but I think it's 20 mw (without modulation).
Laser is modulated with short pulses, so average power of laser is 7 mw.
I'm know that my laser is not safe for eyes.
Using image sensor with better sensitivity allow to decrease laser power.

Bot is Roomba 400. It's controlled from PC by me manually.

Patrick

Tue Dec 11 2012, 06:08PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

the STM32F4 is fast, i have one, but i dont think we will run into speed problems with the STM32 and DLIS-2k, the DCMI is really meant for full cameras... i would have thought the GPIO's would be fast enough... 10MHz transistion times on parrallel GPIO pins should be doable.

iliasam

Tue Dec 11 2012, 08:04PM

Registered Member #8467 Joined: Mon Dec 03 2012, 05:36PM
Location: Moscow, Russia
Posts: 9

With the lidar speed - 10 rps and laser modulation, data frequency would be:
360*10*2*2097=15 MHz.

Also there is another problem - DLIS-2k pixels a very small, so micron misalignment would case distance calculation error, so laser and image sensor holders must really solid and strong.
In my lidar I have found that the small pressure to one side of the lidar case error, and my sensor is only 128 pixel wide.
So now I am thinking about making lidar with stronger mechanics, and ELIS-1024 sensor.
ELIS-1024 can be switched to 128 pixels resolution, so the changes in microcontroller program will be small.

previous 1 2 3 4 next last

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob