Dear Winston, I did not get email attachment, can you please resend it so I understand what is wrong in you setup.
I do not recommend to use white balance for image matching - you may set it once, then store settings so the camera will next time boot with the same settings - this is what we do for such applications. The autoexposure daemon is designed to operate independently - in panoramic applications cameras pointed in different directions may need very different exposure. If autoexposure is unstable you probably just need to modify settings, because the goal of the autoexposure algorithm may be different. In you case you probably have some bright electrical light that fools the camera. Here is generally how algorithm works: it maintains exposure so only a certain fraction of all pixels will have value above a certain value. For the normal good quality outdoor image the fraction of overexposed is set to a very small value (like 99.2% pixels should have <=250 of 255 value). Such combination provide good quality images if there are no bright lights (or sun reflections) in the FOV. If there are some, you may want to reduce required "good" (not overexposed) pixels to say 98% or 95%, possibly simultaneously reducing the required level. There are illustrations on the wiki: https://wiki.elphel.com/wiki/Autoexposure . That will make the camera ignore larger bright lights/reflection and eliminate exposure fluctuations. On the other hand exposure settings as well as gain settings are included in each frame Exif header, so the post-processing software can eliminate influence of the exposure/gains and gamma correction. When we process jp4 images in ImageJ plugin (https://git.elphel.com/Elphel/imagej-elphel/blob/master/src/main/java/JP46_Reader_camera.java ), the first step is that the JP4 image from the camera is converted to a floating-point data with linear intensity scale (gamma conversion is undone), and each color gain is reversed too, providing data close to what sensor would provide with the same gain settings. Exposure value is also available, so it is possible to divide these linear-scaled image by the exposure value to compensate variations - that will help image matching. Andrey ---- On Mon, 08 Jan 2018 00:54:12 -0800 Winston Zhang<winston.zh...@blacksesame.com.cn> wrote ---- Hello Thank you for your patient to answer my question. But I have some others as follow. 1) How should I make the left and right camera AWB change synchronously? The left camera image color is very different with right camera image color. Please look at the last email attachment. I followed your advices, and set the WB_EN=0 and GAINR/GAING/GAINB = some value, and made the left and right camera image WB very close. But it just applies only to one situation. If I change the ambient lighting brightness, the GAINR/GAING/GAINB value should be set again. 2) How should I make the left and right camera exposure change synchronously? The left and right camera exposure at different time and cause the image brightness different. So, if I capture the same timestamp frames, the frames brightness are so different. I want to do some stereo-match work, but all of these get me into trouble. Best Regrad! Yicheng-Zhang Email: winston.zh...@blacksesame.com.cn Teil: 18516291005 Addr: 707 room, 7th floor, building 4th, No.1761, Zhangdong road, Pudong area 发件人: andersonnotg...@gmail.com [mailto:andersonnotg...@gmail.com] 代表 Oleg 发送时间: 2017年12月28日 星期四 3:13 收件人: Winston Zhang <winston.zh...@blacksesame.com.cn> 抄送: Elphel Sales | Olga Filippova <sa...@elphel.com>; Dan Pan <ad...@blacksesame.com>; Tao Zhang <tao.zh...@blacksesame.com.cn>; Qun Gu <qun...@blacksesame.com>; elphel support-list <support-list@support.elphel.com> 主题: Re: [Elphel-support] 答复: 答复: 答复: 答复: 答复: Re: Quote for 1 NC393-Stereo camera Hello, 1) How could I get the raw data of camera. Because when I got the rgb data from multipart jpeg stream, the image had some color blocks if I amplified eight times. I thought it could be caused by image conversion compression.(please look at the attachments). You can switch to JP4 format - it's close to raw. What I see in the attached images is they have different white balance. And also there are overexposed areas which can affect the correct balance. The white balance is run independently for each sensor - you can try to set it so the images look the same and then turn off for both sensors: a) open the web interface for each sensor - click on the green circle OR b) set WB_EN=0 (also for both channels) in parameters - if you change it through a table (http://192.168.0.9/parsedit.php?sensor_port=2&title=Parameters+for+groups:+whiteBalance+&GTAB_R&GTAB_G&GTAB_GB&GTAB_B&WB_THRESH&WB_PERIOD&WB_MASK&WB_EN&WB_WHITELEV&WB_WHITEFRAC&WB_MAXWHITE&WB_SCALE_R&WB_SCALE_GB&WB_SCALE_B&GAIN_MIN&GAIN_MAX&GAIN_CTRL&GAIN_STEP&ANA_GAIN_ENABLE) then there is a broadcast mask to set the parameter for all sensors at the same time - set it to F. if I amplified eight times. What are the gains and exposure for each sensor? 2) The camera temperature has risen quickly, and I could not capture images. Because it was so slowly to run the video when the camera had been opened for four hours. But If I closed the camera and the temperature dropped down, the video was showed normally. What is the temperature? http://192.168.0.9 > Temperature monitor If it's around 70-80 it's ok (at 90 degrees the camera is forced to shut down) Also, the camera should reach the steady temperature pretty much in the the first 30-60 mins. - what is the image quality? - what are the exposure times? you can try limiting exposure time. The default fps is 10 - so, It will get slow if exposure times are >100ms. - is the camera and the pc the only devices in the network? camera connected directly to PC vie cable? - connect to the camera via ssh and check cpu load: $ ssh root@192.168.0.9 root@elphel393:~# htop then click on CPU% table header. - your PC's cpu load? If you are watching from the browser - it might not be able to handle the streams as well. Best regards, Oleg Dzhimiev Electronics Engineer phone: +1 801 783 5555 x124 Elphel, Inc. _______________________________________________ Support-list mailing list Support-list@support.elphel.com http://support.elphel.com/mailman/listinfo/support-list_support.elphel.com
_______________________________________________ Support-list mailing list Support-list@support.elphel.com http://support.elphel.com/mailman/listinfo/support-list_support.elphel.com