[Elphel-support] 答复: 答复: 答复: Re: Quote for 1 NC393-Stereo camera

[Elphel Support]

Hello Winson,

We do not have any C/C++ code for stereo matching and calibration. It is a very mature area and there is a lot of software that can be used for that purpose, the best is probably OpenCV. We could not possibly write some magic in-house code that would significantly outperform what many very good software developers have done over many years. Elphel binocular stereo cameras may be calibrated and processed with Open CV or any other existing programs. The lenses are good and have low geometric distortion so you should not have problems using such software.

We do make special cameras and calibrate them - all the programs are free software (GPLv3) and we post blog articles about how we do that. Our focus is on the area of combining precise hardware  (that can provide imaging data not yet possible with the cellphone cameras) and the software developed for such cameras. In this specific area we can make a difference even with or limited development resources.

The camera sensors we use for calibration have pre-selected lenses (we measure all and keep only the best ~50%), the sensor front ends (SFE) use precise thermal compensation (there are blog posts about it, all the mechanical parts are documented on our wiki). Then we adjust the SFE and glue them in two stages - UV glue first, then epoxy with low thermal expansion. And finally we calibrate these SFE (and the complete  multi-sensor systems) using our 7 meters by 3 meters pattern and Java software (https://git.elphel.com/Elphel/imagej-elphel).

For the distance measurements we use 4-sensor (not just binocular) cameras - you probably have seen the results, and current software uses the same Java code. It is far from real-time, and this is why I'm working on the FPGA code of what the Java program was just an emulation/prototype. Our current FPGA development is here - https://git.elphel.com/Elphel/x393/tree/dct/dsp , but it is far from being finished (major development started just a month ago). When done it will provide results that are currently possible only with active, LIDAR/ToF-based systems, and we plan to eventually implement such image processing systems with teh custom ASIC.

Andrey



---- On Wed, 20 Dec 2017 19:16:59 -0800 Winston Zhang<winston.zhang at blacksesame.com.cn> wrote ---- 

    Hi, Andrey
       Ok, I will resend the email to support mailing. I means that do you have some C/C++ source code for developer. I searched the wiki web, and could not find some cues or C/C++ source codes about how to complete the stereo calibration and stereo matching. 
  
    _______________________________________________ 




Support-list mailing list 
Support-list at support.elphel.com 
http://support.elphel.com/mailman/listinfo/support-list_support.elphel.com 





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://support.elphel.com/pipermail/support-list_support.elphel.com/attachments/20171220/023898cc/attachment.html>