Re: [Elphel-support] Ordering stereo camera

[Olga Filippova]

Dear Elham,

we can send you the Inline camera but the waterproof case is not going to
be ready. It will take several weeks to machine some parts that can not be
3D-printed, after the design is finished.
We can sponsor this project provided that your software is free and open
source and all documentation is available on Github or similar repository.
It would be also nice if you can write an article on Elphel blog about your
project, like other sponsored clients did:
http://blog.elphel.com/category/success-story/

Best Regards,
Olga


On Sat, Sep 3, 2016 at 8:41 AM, Elham Omrani  wrote:

> Dear Andrey,
> Thank you. You’re so helpful. I apologize for late reply.The robot is an
> industrial type that will be used in outdoor conditions. There fore, it is
> very important to detect obstacles with high accuracy  since the accuracy
> parameter is more critical than ambiguity. In addition, we must finish
> this project in the next month and we don't have enough time to code new
> software.
> According to above statements, we prefer to use Inline model.
> I hope we use H-model in future projects.
> what is difference of 2 stereo cameras(that has 4 lenses) with a 4 lenses
> stereo system?
> I'm hurry in my project please quote as soon as possible.
> I looking forward hearing from you.
> Best Regards
> Elham
>
>
>
>
>
> On Tuesday, August 30, 2016 12:41 AM, Elphel Support <
> support-list@support.elphel.com> wrote:
>
>
> Elpham,
>
> Using dual 2-lens system does not give you any advantage compared to two
> separate systems, and for research project I would rather use 2 separate,
> for simplicity.
>
> You know that there are many ways to process multi-view images, most use
> feature extraction and keypoints before calculating disparity and so the
> distance.
>
> Alternatively you can start from correlation (in most cases a
> superposition of the correlation and phase correlation) of the (usually
> just 2) images trying to get distance to objects that do not have easily
> identifiable feature but still have many high-frequency components.
>
> Both paths have some advantages and disadvantages, correlation provides
> better Z-resolution (you can easily operate with disparities being a small
> fraction of a pixel - like 0.1), but it is easy to get into ambiguous
> results, especially when the objects have some periodic structures.
>
> Using more than 2 view simultaneously can reduce ambiguity significantly,
> and finally get high precision distance measurements before the image is
> analyzed for feature extraction. Right now we do not have this code in
> FPGA, but it is one of our close targets that we'll work on when the base
> software for the cameras will be finished, so correlation will be done at
> "hardware" speeds in the same NC393-based cameras running new code.
> Meanwhile you could post-process images on a regular PC.
>
> I've seen much less works on the distance measurements with multiple-view
> systems compared to the standard 2-sensor systems, so that would give you
> an opportunity to do an interesting research.
>
> You could play (just online or download sources for analysis) with our
> "triclops" images that are linked on our web site to get an idea of the
> advantage of the multiple view systems.
>
> Andrey
>
>
>  On Mon, 29 Aug 2016 12:27:28 -0700 *Elham
> Omrani>* wrote 
>
> Dear Olga,
> Thank you for  providing the requested information. Yes, we want to use
> Inline camera as 2 stereo cameras. Inner lenses are a stereo vision system
> with smaller baseline and focal length (suitable for near distance) and
> outer lenses are a stereo system with bigger baseline and focal length (for
> far distance). As you know, stereo vision error increases with increasing
> distance. Because of this, researchers uses 2 stereo camera in a unified
> system to decrease disparity error.
> If you use same baseline and focal lenses, What is your purpose of
> designing H- model or inline model with same baseline and focal lenses?
> About lens, In addition to 4.5 mm lens, what lens do you recommend (we
> prefer bigger than 4.5 focal lens)?
>  I'm looking forward hearing from you.
> Elham
>
>
>
> On Monday, August 29, 2016 9:53 PM, Olga Filippova <
> support-list@support.elphel.com> wrote:
>
>
> Dear Elham,
>
> can you please explain the technical reasons for ordering inline camera?
> It looks like you plan to use Inline camera as 2 stereo cameras with 2
> different bases. This is fine, but it has been done before, while the
> H-camera presents interesting research opportunities.
>
> Practically speaking we offer 4.5mm lenses (custom adjustment has been
> made to fit this lens to the camera, using 3 different adapter rings, as
> can be seen here: http://www3.elphel.com/importw
> iki?title=File:N125B04530W_adapter.jpeg) But we do not offer any 9 mm
> lenses - we have not tested any of 9mm lenses and we don't have these in
> stock, however you are welcome 

Re: [Elphel-support] Ordering stereo camera

[Elham Omrani]

Dear Andrey,Thank you. You’re so helpful. I apologize for late reply.The robot 
is an industrial type that will be used in outdoor conditions. There fore, it 
is very important to detect obstacles with high accuracy  since the accuracy 
parameter is more critical than ambiguity. In addition, we must finish this 
project in the next month and we don't have enough time to code new 
software.According to above statements, we prefer to use Inline model. I hope 
we use H-model in future projects.
what is difference of 2 stereo cameras(that has 4 lenses) with a 4 lenses 
stereo system?I'm hurry in my project please quote as soon as possible.I 
looking forward hearing from you.Best RegardsElham



   

 On Tuesday, August 30, 2016 12:41 AM, Elphel Support 
 wrote:
 

 Elpham,

Using dual 2-lens system does not give you any advantage compared to two 
separate systems, and for research project I would rather use 2 separate, for 
simplicity.

You know that there are many ways to process multi-view images, most use 
feature extraction and keypoints before calculating disparity and so the 
distance. 

Alternatively you can start from correlation (in most cases a superposition of 
the correlation and phase correlation) of the (usually just 2) images trying to 
get distance to objects that do not have easily identifiable feature but still 
have many high-frequency components.

Both paths have some advantages and disadvantages, correlation provides better 
Z-resolution (you can easily operate with disparities being a small fraction of 
a pixel - like 0.1), but it is easy to get into ambiguous results, especially 
when the objects have some periodic structures.

Using more than 2 view simultaneously can reduce ambiguity significantly, and 
finally get high precision distance measurements before the image is analyzed 
for feature extraction. Right now we do not have this code in FPGA, but it is 
one of our close targets that we'll work on when the base software for the 
cameras will be finished, so correlation will be done at "hardware" speeds in 
the same NC393-based cameras running new code. Meanwhile you could post-process 
images on a regular PC.

I've seen much less works on the distance measurements with multiple-view 
systems compared to the standard 2-sensor systems, so that would give you an 
opportunity to do an interesting research.

You could play (just online or download sources for analysis) with our 
"triclops" images that are linked on our web site to get an idea of the 
advantage of the multiple view systems.

Andrey


 On Mon, 29 Aug 2016 12:27:28 -0700 Elham Omrani 
wrote  

Dear Olga,Thank you for  providing the requested information. Yes, we want to 
use Inline camera as 2 stereo cameras. Inner lenses are a stereo vision system 
with smaller baseline and focal length (suitable for near distance) and outer 
lenses are a stereo system with bigger baseline and focal length (for far 
distance). As you know, stereo vision error increases with increasing distance. 
Because of this, researchers uses 2 stereo camera in a unified system to 
decrease disparity error.If you use same baseline and focal lenses, What is 
your purpose of designing H- model or inline model with same baseline and focal 
lenses?About lens, In addition to 4.5 mm lens, what lens do you recommend (we 
prefer bigger than 4.5 focal lens)?  I'm looking forward hearing from you.Elham
 

On Monday, August 29, 2016 9:53 PM, Olga Filippova 
 wrote:
 

 Dear Elham,

can you please explain the technical reasons for ordering inline camera? It 
looks like you plan to use Inline camera as 2 stereo cameras with 2 different 
bases. This is fine, but it has been done before, while the H-camera presents 
interesting research opportunities. 

Practically speaking we offer 4.5mm lenses (custom adjustment has been made to 
fit this lens to the camera, using 3 different adapter rings, as can be seen 
here: http://www3.elphel.com/importwiki?title=File:N125B04530W_adapter.jpeg) 
But we do not offer any 9 mm lenses - we have not tested any of 9mm lenses and 
we don't have these in stock, however you are welcome to purchase such lenses 
separately and install them on the camera.

I am researching the exporting regulations to Iran. I think we can ship our 
camera products to Iran, but I need to receive answers from the exporting 
agency, to make sure.

We offer standard 48V power supply that converts AC power from wall outlet to 
48V DC power. The camera uses 3.3V, but the power supply has to be minimum 12V, 
or 18V better, especially for camera with 4 image sensors. You can power the 
camera from the boat power supply. For development it is nice to have a wall 
adapter, like the 48 V AC-DC converter we offer (or you can use the one you 
already have or can purchase locally).

Best Regards,


On Sun, Aug 28, 2016 at 12:55 AM, Elham Omrani