Those 100ms lightning motion detection models, which are they? - page 3 - General Help and Assistance on using CHDK stable releases - CHDK Forum  

Those 100ms lightning motion detection models, which are they?

  • 40 Replies
  • 21202 Views
*

Offline whoever

  • ****
  • 280
  • IXUS950
Re: Those 100ms lightning motion detection models, which are they?
« Reply #20 on: 10 / April / 2008, 05:05:31 »
Advertisements
Oops, sorry, corrected... Good that there's someone keeping an eye on us naughty kids...

Re: Those 100ms lightning motion detection models, which are they?
« Reply #21 on: 10 / April / 2008, 13:39:10 »
Hi all,

I have measured the time via scope to about 260-300 ms. Analog-Scope is set to 50 ms/Div.
Script is:

@title MD Lightning6
@param a Columns
@default a 6
@param b Rows
@default b 4
@param c Threshold (0-255)
@default c 10
@param d Compare Interval (millisecs)
@default d 1
@param e Trigger Delay (0.1 secs)
@default e 12
@param f Rows to Exclude
@default f 0
@param g Pix-Step(speed/accuracy adj)
@default g 8
@param h Reg Mode(0-no,1-incl,2-excl)
@default h 0
@param i Measure Mode(1-Y,0-U,2-V)
@default i 1
if a<1 then a=1
if b<1 then b=1
if c<0 then c=0
if g<1 then g=1
if f<1 then f=1
e=e*100
print ">[";a;",";b;"] threshold: ";c
x=30000
:md
press "shoot_half"
t=0
do
md_detect_motion a, b, i, x, d, c, 1, t, h, 1, 1, a, f, 0, g, e
until t>0
click "shoot_full"
goto "md"

*

Offline DataGhost

  • ****
  • 314
  • EOS 40D, S5IS
    • DataGhost.com
Re: Those 100ms lightning motion detection models, which are they?
« Reply #22 on: 10 / April / 2008, 14:01:50 »
If I recall correctly, shoot_half is released after you click shoot_full (since the last one is actually two keys). Your camera will refocus and measure the light again *after* detecting motion, so I suggest you move the press "shoot_half" inside your loop, just before md_detect_motion.

Re: Those 100ms lightning motion detection models, which are they?
« Reply #23 on: 10 / April / 2008, 14:16:40 »
Hi DataGhost,

I have tried the suggestion. The result is that the first shoot is not focused, otherwice nothing changes. Still 260 ms.

gueloh


Re: Those 100ms lightning motion detection models, which are they?
« Reply #24 on: 10 / April / 2008, 16:27:57 »
Deleted
« Last Edit: 22 / April / 2008, 16:19:27 by Barney Fife »
[acseven/admin commented out: please refrain from more direct offensive language to any user. FW complaints to me] I felt it imperative to withdraw my TOTAL participation. Nobody has my permission, nor the right, to reinstate MY posts. Make-do with my quoted text in others' replies only. Bye

Re: Those 100ms lightning motion detection models, which are they?
« Reply #25 on: 10 / April / 2008, 22:00:57 »
I've thought of another speed test which seems to be giving good results.

I set the camera up pointing towards the highway. We live on the "through-town" highway where the speed limit is 70km/h (19.4m/s). Most people do sit on the limit here. At a shutter speed of 1/400s a car doing the limit would only travel 5cm so I'm ignoring that small discrepancy for now.

It's interesting that most of the vehicles are captured at roughly the same position - about 2 - 2.5m of car in the shot. With a larger data sample I just took it seems most are closer towards 2m.

So, using an average distance and assuming the smallest bit of car coming into shot triggers the MD routine,

    2.25 / (19.4m/s) = 116ms

and,

    2.5m / (19.4m/s) = 129ms

and,
    2m / (19.4m/s) = 103ms

which are very similar to the range of results MD_Test2 is giving me with current script params :)

A lot of error here but with calibration of the horizontal frame distance and car length instead of guessing I think this has promise. There are problems in that different colored cars take different times, and the larger the vehicle the faster it triggers due to more pixel changes. But I still think better results could be had than with a monitor test program, plus it's less artificial.

At the very least it shows the MD routine is fairly constant with the detection times. Plus the resulting "time lapse" video is pretty cool too  8)

Re: Those 100ms lightning motion detection models, which are they?
« Reply #26 on: 11 / April / 2008, 02:06:15 »
Hi all,

I think my test set-up to evaluate the speed is proper done and is direkt calibrated readable. But my mistake was that I have used the end of the beam. This was wrong, because the shutter speed was 1/8 s (125 ms) and this is reflected in the width of the beam. So the right speed value when the motion detection function triggers the camera is the beginning of the beam: 160 ms delay with the "MD Lightning6" script and the A720.

Thank you all.

Re: Those 100ms lightning motion detection models, which are they?
« Reply #27 on: 11 / April / 2008, 06:23:20 »
Deleted
« Last Edit: 22 / April / 2008, 16:19:48 by Barney Fife »
[acseven/admin commented out: please refrain from more direct offensive language to any user. FW complaints to me] I felt it imperative to withdraw my TOTAL participation. Nobody has my permission, nor the right, to reinstate MY posts. Make-do with my quoted text in others' replies only. Bye


*

Offline whoever

  • ****
  • 280
  • IXUS950
Re: Those 100ms lightning motion detection models, which are they?
« Reply #28 on: 11 / April / 2008, 06:24:45 »
I get the occasional 45ms ones with most around the 60ms to 90ms marks.

You surely remember the discussion about tracking the correct frame buffer (there are three of them, each updated 10 times a second), and how it actually seemed to slow down the response, or at least to remove those occasional ultra-fast triggers? In my understanding, if one only looks at one buffer (which is the default for most cameras), then it sometimes happens that this buffer was the correct current one in the previous 30ms interval, and has just been updated to become (although is not yet) current at a later time. In this case, which occurs roughly with the probability 1/3 * 1/3 = 1/9, you get the fastest possible trigger. Do your "occasional 45ms ones" happen about once in ten shots?

So one would like to know not only the current buffer, but also the one scheduled to become current next! Perhaps DataGhost has an opinion in this respect, as he's been deep inside the ARM processor for quite a while.

Re: Those 100ms lightning motion detection models, which are they?
« Reply #29 on: 11 / April / 2008, 06:26:42 »
Deleted
« Last Edit: 22 / April / 2008, 16:20:08 by Barney Fife »
[acseven/admin commented out: please refrain from more direct offensive language to any user. FW complaints to me] I felt it imperative to withdraw my TOTAL participation. Nobody has my permission, nor the right, to reinstate MY posts. Make-do with my quoted text in others' replies only. Bye

 

Related Topics