[Canberrauav] Raspberry Pi Camera Lag Testing

Bruce Baxter brucebaxtersolutions at gmail.com
Thu Feb 1 00:59:37 AEDT 2018


Oh, to test it take a photo out the window of your car as you drive along the same piece of road in each direction.
No plane needed

Sent from my iPad

> On 1 Feb 2018, at 12:57 AM, Bruce Baxter <brucebaxtersolutions at gmail.com> wrote:
> 
> As a side question here, 
> Has anyone tried using an iPhone and triggering it with a servo on the volume button?
> It has its own gps and time tagging and can automatically upload to "cloud" storage via wifi or cell data. 
> It may be more accurate being all integrated. 
> 
> Also, the flirone clip on for a cellphone may function the same way.
> 
> Sent from my iPad
> 
>> On 31 Jan 2018, at 11:24 PM, Carlos Peco-Berrocal <carlos.peco at gmail.com> wrote:
>> 
>> Stephen,
>> 
>> What about the time it takes the "time" to appear in the video output
>> of the TV? Is that a constant ?
>> 
>>> On 1/31/18, Stephen Dade <stephen_dade at hotmail.com> wrote:
>>> Hi all,
>>> 
>>> I attempted some measurements of the Raspberry PiCam + cuavraw camera lag.
>>> It was fairly crude - I pointed the PiCam at my TV, which was displaying the
>>> "time" bash command from the Pi, with cuavraw running in a screen session in
>>> the background.
>>> 
>>> Attached is a (cropped to reduce file size) example image I captured. From
>>> the image, I did have some trouble with the terminal/screen refresh rate and
>>> the PiCam capturing. This meant that I couldn't use ~20% of the images I
>>> captured.
>>> 
>>> For the remaining images where I could pick out the latest timestamp my
>>> results were:
>>> Console time lagging the camera time by 600 - 800 milliseconds
>>> 
>>> Given that cuavraw tags the image timestamps after image capture (but prior
>>> to debayering and jpg encoding), this would indicate that there's a
>>> 600-800ms time lag between the camera shutter capturing the image and the
>>> Raspberry Pi downloading the image from the camera.
>>> 
>>> If we were flying at 30m/s, this would equate to a 18-24m error in our
>>> georeferencing.
>>> 
>>> 
>>> 
>>> Thanks
>>> Stephen
>> --------------------------------------
>> CanberraUAV Mailing List
>> Help support our work via our Patreon: https://www.patreon.com/canberrauav


More information about the Canberrauav mailing list